A monk steps off a ship at a Chinese port carrying a rolled manuscript. The roll holds a diagram — a mandala, a map of the enlightened cosmos with the Buddha Mahavairocana at its center. He memorized which master gave it to him, the year, the temple. Decades later another monk carries a different version of the same diagram off a different ship, and insists his is the older line. Trace these carriers back far enough and you do not arrive at one original mandala. You arrive at four, each with a named human chain, each with a date, each disputing the others' claim to the source.1 Dukes lays the genealogy out like a customs ledger, and the ledger tells a harder truth than any single mandala does: there was never one Vajra-realm mandala. There were rival imports.
The Vajra World mandala — in Japanese, the Kongokai Mandara, the "thunderbolt-realm" diagram — is the esoteric Buddhist map of reality organized around Mahavairocana. Dukes's claim, stated flatly, is that "there are at least four different forms" of it.2 This page is the catalog of those four, the people who carried each one to Japan, and the dates that anchor the chain. It is a history page, not a doctrine page: the content is who brought what, from whom, when.
The whole import system rests on two Indian masters credited with bringing esoteric teaching from India into China on a systematic scale — Subhakarasimha (Chinese: Shan Wu Wei) and Amoghavajra (Chinese: Pu Kung Chin Kang).3 Their explanations of mandala iconography were said to rest on descriptions "given to Nagarjuna" and recorded in the Mahavairocana Sutra and a section of the Vajrasekhara ("thunderbolt crown") Sutra called the Tattvasamgraha.4 Hold that "given to Nagarjuna" phrase — it is the legendary root of the whole tree, and it does not bear the same weight as the dated links further up the chain.
The four forms, with their carriers:
Running underneath the four-form catalog is a second, sharper transmission story — the textual chain of the explanatory treatise. The Jogenshinjoshakkyumokuroku records that in A.D. 742 Silamegha, the monarch of Sri Lanka, gave a collection of the "Eighteen Assemblies" of the Vajrasekhara-yoga Sutra to Amoghavajra.9 Amoghavajra returned to China in 746 and began translating it, but completed only the first section of the first assembly before he died; he also compiled a short synopsis, the Juhachieshiki.10 For many years that synopsis was the only explanation Japanese monks had. Around the ninth century Danapala and Dharmabhadra translated the 1st, 6th, and 15th Assemblies. And the synopsis itself did not reach Japan until 1073, when the monk Jojin brought back a copy obtained from the Inkyo Temple in China.11
Here is the mechanism, and it is what makes this a history page rather than an art-appreciation note. A mandala is not just a picture; it is a receipt. Its specific form — nine panels or one, lotus thrones or animal thrones, square center or round — encodes which master's line you received it through. Change the throne and you have changed your ancestry. So when Enchin brings home a mandala where the Buddhas sit on animals, he is not making an aesthetic choice; he is asserting descent from Subhakarasimha's particular transmission, distinct from the line Kukai carried.7 The iconographic detail is the lineage claim. Read the picture and you can read the family tree.
This is why "at least four forms" is a charged phrase, not a neutral count. Each form is a competing answer to the same question — what is the true diagram of the enlightened cosmos? — and each carrier staked a sect's authority on his answer. Kukai's Shingon and the Tendai line of Ennin and Enchin were not just transmitting; they were differentiating. Dukes closes the section by noting all the mandalas used in Japanese Mikkyo were brought back by Kobo Daishi (Kukai) of Shingon, Dengyo Daishi of Tendai, or their students — "these vary in some respects."12 That mild "vary in some respects" is doing diplomatic work over a real institutional rivalry: two sects, two import lines, two slightly different maps of ultimate reality, each claiming the better pedigree.
And the dated textual chain underneath shows why the variation was inevitable. The full explanatory treatise was a bottleneck from the start: Amoghavajra translated only a fragment before dying, the synopsis stood alone for generations, and the source text did not reach Japan until 1073 — more than three centuries after Silamegha handed the collection over in 742.911 When the authoritative explanation arrives that late and that incomplete, the carriers in between are working from partial maps. Divergence is not corruption here; it is what a slow, fragmentary transmission produces.
This page hands the rest of the corpus a dated skeleton for the esoteric import. Wandering Acarya Transmission Mobility describes the mobile teacher as the unit of transmission; this page names six of them and pins them to years. Chuan Fa Transmission Lineage gets a parallel case: the same ship-borne, master-to-disciple mechanism that carried body-methods carried mandalas, and the dating logic transfers. Guru-Student Transmission as Consciousness Transfer supplies the inner side of what this page tracks from the outside — the mandala is the documentable husk of a transfer the tradition understood as mind-to-mind.
Look at the one detail Dukes singles out as distinguishing the third form — in the Gobushinkan, "the Buddhas and Bodhisattvas sit upon thrones formed from various animals, whereas in the others they sit only upon lotuses."7 On the surface this is iconography trivia. Read as a historian, it is a fingerprint.
Three things are riding on those animal thrones. First, attribution: the form is "said to be the form taught by the Indian master Subhakarasimha,"7 so the animal throne is a claim of descent from one specific Indian source, not the other. Second, carrier: it came to Japan through Enchin — and Dukes adds that Enchin was Kukai's nephew.7 That family detail matters. The nephew of Shingon's founder is carrying a Tendai mandala with a different throne convention. Kinship and sectarian lineage cut across each other; the blood relative imports the rival map. Third, divergence preserved: nobody harmonized the thrones. The animal-throne form and the lotus-throne form both survived, side by side, in Japanese Mikkyo, each marking its own line. That non-resolution is the historian's gold — it means the difference was load-bearing for identity, important enough to keep rather than smooth away. A throne shaped like an animal is a sentence in an argument about who has the real teaching, and the argument was never closed. Decompress that single furniture detail and you get the whole structure of the esoteric import: rival Indian sources, named carriers, family lines crossing sect lines, and differences kept on purpose.
You are checking a footnote for a draft on Japanese esoteric Buddhism, and the sentence under your pencil reads: "the Vajra-realm mandala was brought to Japan by Kukai." It smells too clean. You pull Dukes's ledger. No — there are at least four forms, and Kukai carried only one of them, the Nine Assemblies.5 You write "one of four" in the margin. Then you check the others: Ennin's Eighty-one Bhagavats, Enchin's animal-throne Gobushinkan, the Tibetan square-center variants — three more carriers, two more sects.678 You cross out "the mandala." Then the deeper date: the explanatory text didn't even reach Japan until 1073, three centuries after Silamegha gave the collection to Amoghavajra in 742.911 So the carriers were working from partial maps the whole time. You rewrite the sentence: "multiple rival forms of the Vajra-realm mandala reached Japan through competing Shingon and Tendai lines over three centuries, the full explanatory treatise arriving last." Less quotable. Also true. You flag one more thing in red — the claim that it all goes back to descriptions "given to Nagarjuna" — and write beside it: legendary root, do not date.
The source is Dukes's mandala-transmission section.13 Its strength is that it is unusually checkable for this book: it cites Taisho canon numbers (the Jogenshinjoshakkyumokuroku at Taisho Vol. 2157; the Juhachieshiki at Taisho 18 No. 869) and gives hard dates — 742, 746, the ninth-century translations, 1073.91011 Where Dukes anchors to the Taisho and to datable carriers, this page stands on firm ground, and the named-carrier chain (Kukai, Ennin, Enchin, Jojin) matches the broad shape of the standard account of Japanese Mikkyo's importation.
The load-bearing quarantine is the origin claim. The attribution of the iconography to descriptions "given to Nagarjuna" is the tradition's own legendary root, not a dated link, and must not be read with the confidence of the 742-to-1073 chain. [PLAUSIBLE — needs corroboration] on the Subhakarasimha-taught-the-Gobushinkan attribution, which Dukes himself hedges with "said to be."7 [DO NOT FILE as fact — document-don't-endorse] on the Nagarjuna-received-it origin: record it as what the tradition claims, never as event.4
Open: Dukes says "at least four" forms. Is four a real count or a floor — were there forms that did not survive transmission, and would a complete census change the picture from "four rival lines" to "many partial ones"? And: why did the explanatory treatise lag the diagrams by three centuries — was it harder to carry, harder to translate, or guarded more closely than the images?
This is one of the pages where Dukes is at his most checkable and least dependent on his temple's private cupboards — he is citing the open Taisho canon and public sectarian history, so the usual Kongoryuji warrant is light here. That said, the seam to watch is the one Dukes himself half-marks: he reports the Nagarjuna origin and the Subhakarasimha attribution in the tradition's own voice ("said to be," "descriptions given to Nagarjuna") while reporting the 742–1073 dates in his own. A Samuel-style critical historian would press exactly there: the dated carrier-chain is the kind of evidence the academy can audit, while the Indian-master-to-Nagarjuna root is the kind of lineage-claim the academy treats as institutional self-authorization. Dukes converges with that critical instinct more than he usually does — he keeps the legendary and the dated material in different grammatical registers, which is itself a quiet act of source-criticism. Where he diverges is in tone: he reports the rivalry between Shingon and Tendai lines as "vary in some respects," a practitioner's diplomatic understatement over what an outside historian would frame as a contest for authority. The mild phrasing is the inside view; the contest is visible anyway in the four forms he is forced to count.
Version Control and the Canonical Source — Modern knowledge-work runs on the problem of forks: a master document gets copied, edited by distant hands, and diverges into incompatible versions, and the hard question becomes which branch is canonical? The Kongokai transmission is that problem running over three centuries with no merge tool. Four forms of one diagram, each branched from a partial copy, each carrier unable to reconcile against a source that hadn't arrived yet. The structural parallel is exact: divergence is driven not by malice but by latency — the authoritative explanation (the Juhachieshiki source text) reached the editors last, in 1073, long after the forks had hardened.11 The insight neither domain produces alone: version-control practice frames forks as a defect to be merged back to a single trunk. The mandala history shows a system where the forks became the institutions — the divergence was not a bug to fix but the very thing that let Shingon and Tendai exist as distinct lineages. Read together they expose a rule operators miss: when the canonical source arrives late enough, the branches stop being branches and become species. You can no longer merge them, because the difference has been recruited into identity. The lesson cuts both ways — sometimes you want the merge, and sometimes the fork is the org chart.
Provenance as Authority Signal — Influence and marketing both know that a thing's origin story confers authority independent of the thing itself: the named source, the unbroken chain, the famous founder all make a claim more persuasive than its content warrants. The mandala carriers were running this exact play — the value of a form rode on its pedigree (taught by Subhakarasimha, descended from descriptions given to Nagarjuna), and the iconographic fingerprint was the credential.47 The history page supplies what the behavioral frame lacks: a centuries-long case showing provenance claims stratifying by reliability — the dated link (Jojin, 1073) is checkable, the legendary link (Nagarjuna) is unfalsifiable, and the tradition itself kept them in different registers. The behavioral frame supplies what the historian lacks: the function of the unfalsifiable root — the Nagarjuna attribution does work precisely because it cannot be audited; it anchors authority at a point no rival can reach. Together they yield a diagnostic the vault's history domain can reuse: in any transmission claim, separate the provenance you can date from the provenance you cannot, and notice that the undatable root is usually the one doing the most authority-work. The legendary origin is not weak evidence poorly kept; it is strong rhetoric well placed.
Sharpest implication: there was never "the" Vajra-realm mandala. There were rival imports, branched from partial copies over three centuries, and the differences between them — nine panels versus one, animal thrones versus lotus — are not variants of one thing but the fossil record of competing lineage claims. The diagram is a receipt for an argument.
Generative questions: