Barabar Caves
Bihar, India · not visited
Seven caves cut into granite hills north of Gaya, polished inside to a mirror finish in the 3rd century BCE. The viral version adds scans and tilting walls that I could not trace to any published survey, and says nobody knows what the caves were for. The inscriptions say.
- Latitude
- 25.0050° N
- Longitude
- 85.0631° E
- Dated inscriptions
- Ashoka, years 12 and 19; Dasaratha, c. 230 BCE
- Caves
- 7 (4 + 3)
- Open questions
- 4 logged
The short version that is going around
A passage circulating from a video or podcast describes the Barabar caves in India like this: caves cut into big granite outcrops, interiors polished like mirrors, circular rooms with a second round chamber behind, walls that tilt inward by about 1.5 degrees on both sides, scans that show near-perfect symmetry and flatness, and nothing else quite like it anywhere. And nobody knows.
Part of that holds up well. Part of it I could not trace to any source. And the last line is close to the opposite of what the stone itself says. This dossier separates the three.
What is actually there
The caves lie in the Makhdumpur region of Jehanabad district in Bihar, about 24 kilometres north of Gaya. There are seven, in two groups. Four are on Barabar Hill: Karan Chaupar, Lomas Rishi, Sudama and Visvakarma. Three are on Nagarjuni Hill, about 1.6 kilometres away: Gopika, Vadathika and Vapiyaka. Wikipedia dates the group to the Maurya period, 322 to 185 BCE, and the inscriptions narrow that, as the next sections show. Every source I read calls the rock granite. I found no geological study of it, so I cannot tell you more about the stone itself.
The shapes are less uniform than the passage suggests. The two best known, Sudama and Lomas Rishi, each have a rectangular hall and, behind it, a smaller chamber with a domed roof, joined by a narrow passage. Sudama's hall is given as 9.98 by 5.94 metres and its round chamber as about 6 metres across. Lomas Rishi's hall is 9.86 by 5.18 metres and its round chamber about 5 metres. Karan Chaupar is a single rectangular room of 10.2 by 4.27 metres. Visvakarma has a small rectangular room and an unfinished circular one about 2.8 metres across. At Nagarjuni, Gopika is the largest, an oblong room of 13.95 by 5.84 metres with circular ends. Vadathika and Vapiyaka are single rooms of about 5.1 by 3.4 metres. Those figures come from Wikipedia and are given there without a stated source, method or tolerance.
The Lomas Rishi facade is the one people photograph. Its doorway is an ogee-shaped arch, a stone copy of a thatched hut with curved eaves and timber struts, and a row of carved elephants advances along the curved beam toward stupa symbols. Wikipedia calls it the earliest surviving example of that arch form. The cave was never finished: it has no Ashoka inscription, the floor still has bumps of rock left in place, and Wikipedia mentions two explanations, a rockfall that stopped work or a later completion. The archaeologist S. P. Gupta argues that it was begun under the last Maurya emperor, Brihadratha, and halted in 185 BCE when Pushyamitra Sunga seized power. That is one scholar’s argument.
The finish has a name: the Mauryan polish.
The same glassy surface appears on many of the pillars Ashoka erected, mostly in sandstone, and on a few other objects. The Barabar caves are the first dated example, which is why they are cited so often. After the Nagarjuni caves the polish was abandoned in cave architecture: later sites such as Karla and Ajanta are not polished. The process is poorly documented. The polishing was done after carving, and inscriptions were usually cut after polishing, which chipped the stone around the letters. The abrasives are known only indirectly, from polishing grains found at Persepolis and from Greek use of emery from Naxos. Wikipedia’s page on the polish says that the technique has generally been seen as a Near Eastern import, with Achaemenid influence suggested, while S. P. Gupta argues for a local origin in Neolithic stone-tool polishing.
Several sources mention a strong echo. One popular article puts it at about three seconds in Sudama. Far longer figures circulate online, and I found no published acoustic measurement to check any of them.
The novelist E. M. Forster used these caves as the model for the “Marabar” caves in A Passage to India. That is where many English readers first heard of them.
The sky here
None of the sources I read gives an orientation or azimuth for any of the caves. This dossier therefore gives no figure instead of a guess. If you know of a survey, tell me and it goes in with the source.
Checking the claims
Six claims from the passage, each against what I could read. My sources are mostly Wikipedia, a news report of the Archaeological Survey of India’s heritage proposal, and popular articles. I found no scholarly survey of the caves that I could open, and I say so where it matters.
| Claim | What the sources I read say | Status |
|---|---|---|
| Caves cut into granite outcrops, interiors polished like mirrors | Wikipedia, its page on the Mauryan polish and the ASI proposal reported in 2022 all describe granite caves with highly polished interiors and a mirror-like finish. | Supported |
| Circular rooms, with a second round chamber behind | Sudama and Lomas Rishi have a rectangular hall and a domed round chamber behind it. Karan Chaupar is a single rectangular room. Gopika is oblong with circular ends. | Partly right: rectangular halls, not circular rooms |
| Walls tilt inward about 1.5 degrees on both sides | The Wikipedia pages for Sudama and Lomas Rishi give dimensions and call the walls flat, but give no wall angle. No source I read gives 1.5 degrees or any other figure. | Not verified |
| Scans show near-perfect symmetry and flatness | Popular pages repeat this. None names a scanner, an institution, a date or a report. The one article that gives tolerances calls them estimates and says comprehensive laser-scanning datasets are still limited. | Not verified |
| Nothing else quite like it anywhere | The caves are the oldest surviving rock-cut architecture in India, and no later cave I read about carries the polish. But Wikipedia lists the Sitamarhi cave as having similar structure and polish, and the Son Bhandar caves as reminiscent of Barabar. The polish itself is also on Ashokan pillars. | Partly right: rare, not unique |
| Nobody knows | The inscriptions name the patrons and the recipients. What is not documented is the tools, the method and the reason the polish stopped. | Overstated |
Status as of 16 October 2026. “Not verified” means I could not find a source, not that the claim is false.
Where the scan story comes from
One blog post says that “recent 3D scans” and “laser mapping projects” measured the caves in millimetre detail, with long wall surfaces deviating by less than a centimetre. It names no scanner, institution or date. Another popular article gives tolerances of about 2 to 5 centimetres of axial deviation over ten metres and 1 to 3 millimetres of surface flatness, but labels them estimates, cites no measurement behind them and says itself that comprehensive laser-scanning datasets are limited. Its reference list includes a ResearchGate paper whose listed title begins “Mirror-Polished Granite Caves”. I could not open it, so I do not know what, if anything, it measures.
The most specific version I found is a caption on a short video, which I saw only as a search-result snippet. It says a company founded by former stonemasons laser-scanned the caves, that the angles are perfect to within 0.10 degrees, and that the walls have a slight slope. I found no report from that company, and the caption gives none. A caption is not a source.
Can the 1.5 degree tilt be true?
It could be. A tilt that small is exactly what a plain description would leave out, so the lack of a mention in Wikipedia is weak evidence either way. But a claim with a precise number and no named measurement is a claim I cannot confirm, so I list it as not verified. If someone has the scan data or a survey drawing, it is easy to settle.
Who the caves were cut for
This is where “nobody knows” fails. Inscriptions in Brahmi record that Ashoka, under his title Priyadarsin, dedicated the Visvakarma and Sudama caves in the 12th year of his reign, and Karan Chaupar in the 19th year, about 250 BCE. His grandson Dasaratha dedicated the three Nagarjuni caves to the Ajivikas on his accession, about 230 BCE. The Ajivikas were an ascetic order that flourished alongside Buddhism and Jainism, traditionally traced to Makkhali Gosala. Their exact relation to the other two is unclear.
The reading of the Karan Chaupar inscription has itself changed. Eugen Hultzsch translated it in 1925 as a Buddhist-style rain retreat. Harry Falk’s 2007 reading has the cave given to the Ajivikas. In three of the seven caves with Ashoka or Dasaratha inscriptions the word “Ajivikehi” was later chiselled away. Wikipedia reports that fact; I did not find who did it or when. A later Sanskrit inscription of the king Anantavarman, from the 5th or 6th century CE, sits above the Lomas Rishi entrance.
Sources disagree about the exact BCE years. Wikipedia’s page on Sudama gives 261 BCE in one place and 257 BCE in another, and a regional heritage site gives 261 and 245 BCE. This page therefore uses the regnal years in the inscriptions, and keeps Wikipedia’s “about 250 BCE” for Karan Chaupar.
What if the polish was a state craft that existed only while an emperor paid for it? The dated caves end with Dasaratha, and so does the polish in cave architecture.
Why it is worth asking: because a technique this demanding appears in a short window and then disappears from cave architecture, which needs a reason. What supports it: the polish is on Ashokan pillars as well as in the caves, the dated caves stop with Dasaratha, and Wikipedia suggests later caves relied on individual donors, not imperial funding. What speaks against it: that funding explanation is a suggestion, not a finding, and Wikipedia’s page on the polish also says the technique continued on occasion until the first or second century CE. Other explanations, such as changing taste or lost knowledge, are offered in popular articles without evidence. What would tell the readings apart: tool-mark and abrasive analysis of the walls, and a firmly dated set of polished objects after the Mauryas. What stays unknown is the method itself, and I am speculating about why it stopped, not about who built the caves.
What is still open
How was the polish made?
No definitive record of the tools or process exists. A popular article says that experts generally agree on repeated grinding, abrasion and polishing with natural materials, but names none of them. The evidence for abrasives is indirect, from other sites. What would settle it: analysis of tool marks and abrasive residue on unpolished patches, and an experiment that reproduces the finish on similar granite. I found no published replication.
Where did the technique come from?
The common view is that it came from the Near East, with Achaemenid stonework as the model. Gupta argues for a local development out of Neolithic stone-tool polishing, with the Jain Son Bhandar caves as a possible intermediate stage. Son Bhandar’s date is itself disputed: the Mauryan polish page puts it generally in the 2nd to 4th centuries CE, while the page on Indian rock-cut architecture says it may be contemporary with or even earlier than Barabar. Until that is dated, the local route cannot be tested.
Why did the polish stop, and why was Lomas Rishi left unfinished?
Wikipedia offers a rockfall or a later completion for Lomas Rishi, and Gupta’s political halt in 185 BCE. For the end of the polish it suggests the move from imperial to private funding. A July 2026 popular article lists labour cost, lost knowledge, a shift toward sculpture and changing taste, and offers them without evidence. These explanations are not mutually exclusive, and none has dated evidence behind it that I could read.
Has anyone published a measured survey of the caves?
I found none I could read. In September 2022 the Archaeological Survey of India said it would propose the Barabar and Nagarjuni caves for UNESCO’s tentative list, with its Patna zone preparing the detailed proposal. I do not know how that ended. A nomination file usually carries documentation, but I have not seen this one. Until a survey with named methods and tolerances is published, the precision claims stay unchecked.
Sources
- Wikipedia, Barabar Caves: location, the seven caves, dimensions, inscriptions, the Ajivikas, Hultzsch and Falk, Gupta on Lomas Rishi. Sudama cave and Lomas Rishi Cave: walls called flat, no wall angle given, the facade, the unfinished work. The Sudama page gives two dates for the dedication, 261 and 257 BCE, which is why this page uses regnal years.
- Wikipedia, Mauryan polish: objects, sequence of carving, polishing and inscribing, indirect evidence of abrasives, the Near Eastern and local origin arguments, the end of the polish. Indian rock-cut architecture: oldest surviving rock-cut architecture, Sitamarhi and Son Bhandar.
- EastMojo, “ASI to seek world heritage status for Barabar and Nagarjuni caves in Bihar”, 29 September 2022: the proposal. A news report. I did not find a later one.
- Organiser, “Barabar Caves: The mystery behind India’s oldest rock-cut caves”, 25 July 2026: unnamed experts on the polish, unverified explanations for its end. Pragyata, “The Ancient Barabar Caves near Gaya”: the three-second echo at Sudama, granite, no measurements. Both popular sources.
- Sources of the precision claims, all popular and none citing a primary survey: Ancient360, “Barabar Caves Symmetry” (tolerances labelled as estimates, limited scanning data); Ahoy Matey, “The Barabar Caves Mystery” (“recent 3D scans”, unnamed); Historic Mysteries, “Ancient Perfection” (unnamed historians); CPREEC, “Barabar Caves” (a promotional comparison to laser cutting, and dates that differ from Wikipedia’s); a short video whose caption I saw only as a search-result snippet. The ResearchGate paper whose title begins “Mirror-Polished Granite Caves” is listed in one article’s references and could not be opened.
- Anandajoti Bhikkhu, “061 Lomas Rishi Cave”, Flickr, taken 17 March 2013, CC BY 2.0. The photographer’s description reads “Lomas Rishi Cave at Barabar, Bihar”. I could not open the image itself.
Not visited in person. Written from published sources. Page opened and last reviewed 16 October 2026. The plan drawing is an original illustration. If something here is wrong, tell me and it gets corrected on the page with a note saying what changed.