How to date a Japanese Buddhist painting

Dating a Japanese Buddhist painting is convergent, not a single test. Five evidence lines (ground and format, pigment and gold technique, iconographic recension, inscriptions and box records, and scientific methods with their wide error bars) are read against each other. A colophon-dated 1107 zuzō at the Met shows what an inscription-anchored benchmark does, and what it cannot do for the undated painting beside it.

The question is never answered by one test

Most Japanese Buddhist paintings carry no date. A hanging scroll of Amida descending, a Kasuga deer, a wrathful Aizen on a red ground: the painter did not sign it, the temple did not minute it, and the box it travelled in was made three centuries after the silk. What the museum label gives as “Kamakura period, 13th century” or “Nanbokuchō, 14th century” is not a reading off a single feature. It is a convergence: several independent lines of evidence pushed until they overlap, with the overlap reported as the date and the width of the overlap quietly suppressed.

This article walks the five lines a connoisseur and a conservator actually use, and it walks them on an object that does not need any of them. Met 2015.300.3 is an iconographic drawing of the Aizen Myōō mandala that carries its own date in a brushed colophon: the fifth day of the third month of Kashō 2, which is 1107.¹ The Met’s record reads the colophon further. The drawing was copied from a model inherited by a monk, Sanmai Ajari Ryōyu, from his teacher Chōen (1016–1081), a Tendai esoteric scholar at Ōhara, and was transcribed at Shōren’in in Kyoto, where the copyist left small notes on which deities take white flesh and which take red.¹² A painting that can answer the date question from its own surface is the right place to learn what the question costs when the surface is silent.

Ground and format: what the painting is made on

The first line is the support and the format, because they move slowly and they bound everything else. A Japanese Buddhist painting is on silk (kenpon) or on paper (shihon), and the choice tracks function: the prestige devotional icon on sized silk, the working drawing and the handscroll on paper. That contrast, and what sized silk does that paper cannot, is the subject of its own study and is not re-argued here.³ What matters for dating is narrower. Silk weave is measurable (warp and weft thread count per centimetre, single versus paired warps), and the high-status Buddhist silks were woven to a specification, in some traditions a doubled warp reserved for the purpose.⁴ A loosely pieced ground of mixed cloth is not the Heian–Kamakura prestige register; it is the signature of a later, thinner patronage that could not command a dedicated loom.⁴ The format carries its own coarse clock: the single hanging scroll, the set of three or five, the handscroll, the votive panel. None of this dates a painting to a decade. It tells you which centuries are even in play before the finer lines are read, and it sets the terminus that no later evidence can override: a painting cannot be older than the cloth it sits on.

The worked example sidesteps this line cleanly, which is itself instructive. Met 2015.300.3 is ink on paper, a zuzō, an iconographic drawing, not a devotional icon. The sheet is foxed, brown stain-blooms scattered unevenly across it, and a horizontal crease runs through the middle register where it was once folded or rolled tight. Behind museum glass it hangs vertical and lit evenly, framed like the devotional paintings it is read against; that framing is the opposite of how it worked. A zuzō was made to lie flat on a copyist’s low desk, close under a lamp at reading distance, a second sheet laid over it for tracing, the crease and the foxing the marks of a tool handled rather than an icon enshrined. Paper is the working medium, the medium of transmission rather than enshrinement, and that the dated benchmark survives on paper rather than silk is not incidental: the paper drawing is the carrier of the iconography, and the iconography is the next line.

Pigment and the gold: technique markers move in time

The second line is how the colour and the gold were put down. Mineral pigments (azurite, malachite, vermilion, lead white) age differently from organic lakes and dyes, and a conservator reading a paint cross-section can separate a body colour laid thick from a thin wash sunk into the weave. The diagnostic that matters most for date is the gold. Two techniques compete: kindei, gold ground to a powder and painted on as a fluid, and kirikane, gold leaf cut into hair-fine lines, lozenges, and triangles and laid onto the surface to build pattern.⁵ They are not alternatives so much as a sequence and a horizon. Kirikane was an imported Tang technique that flourished in Japan from the Late Heian period and carried the great Kamakura Pure Land and esoteric paintings, then thinned through the fourteenth century and all but disappeared as the patronage that paid for it collapsed.⁵ A robe whose gold pattern is built from cut leaf, dense and ruled, sits in a different and earlier band than one whose gold is brushed on as powder. The reading is real, and it is soft at both ends: kirikane does not begin on a known day and does not end on one, and a later workshop reviving an older manner can lay cut gold long after the horizon has formally passed.

The style horizon works the same way and is named, not measured. The Takuma school (founded by Takuma Tametō, active 1132–74, continued by his son and through Kyoto and Kamakura branches into the fourteenth century) is one such named band: thick sumi ink line, cool restrained colour keyed to the drawing, set against the warm wire-line Fujiwara manner before it and the gold-saturated register beside it.⁶⁷ To say a painting “is Takuma” or “is An’amiyō” is to assign it to a horizon a generation or two wide, on the judgement of an eye trained on the securely dated members of that horizon. It is the line that does the most work and the one that can least be checked. It is also the line the worked example anchors, because the eye that calls a painting Takuma is calling it against drawings exactly like Met 2015.300.3, the dated members the undated ones are measured by.

Iconographic recension: which version of the image

The third line is the one the dated drawing exists to serve. Esoteric Buddhist iconography in Japan did not drift anonymously. It was transmitted as named recensions through teacher-to-disciple lineages, each copy nominally faithful to a model, and the great compendia (the Zuzōshō of the 1130s, the later Kakuzenshō and Besson zakki) fixed those recensions as text-and-image references.² A composition that follows a recension first attested in the 1130s cannot, on iconographic grounds, predate the 1130s; that is a terminus post quem, the firmest kind of date because it is a logical floor rather than a stylistic guess. The method does not give an upper bound, since a faithful copy can be made at any later date, but it cuts away everything earlier, and on a tradition as recension-disciplined as Mikkyō painting the floor is often sharp.

This is exactly what Met 2015.300.3 is. Read the sheet: nine deities in three rows, several enclosed in ruled circles, the upper-right Aizen with his flaming hair-mane and bow drawn boldly outside any disc, a six-armed seated figure centred on a lotus, and beside several figures a small brushed caption giving the iconographic name, written so the next copyist cannot mistake the deity. There is no colour. The whole sheet is ink line and annotation, and that is the point of the object: it is not the painting, it is the carrier of the painting’s recension, the thing a later devotional Aizen on red silk would be measured against. The colophon names the line, Chōen to Ryōyu to the 1107 copyist at Shōren’in, and so the recension this drawing transmits is not merely datable, it is doctrinally owned.¹ When an undated Aizen painting matches a recension whose earliest dated witness is a sheet like this one, the drawing has supplied the floor the painting could not supply itself.

Inscriptions, colophons, seals, and the box

The fourth line is writing: dedicatory inscriptions on the painting, colophons on the mounting, seals, and the box inscription (hakogaki) that travels with the scroll. When it is present and trustworthy it is the strongest evidence of all, and most of the time it is neither present nor fully trustworthy. A dedicatory inscription naming a year and a patron, integral to the original paint layer, is the gold case, and it is rare. A colophon on a later remounting dates the remounting, not the image. A hakogaki attributing the scroll to a famous hand, brushed by a connoisseur centuries after the fact, is a document about reception, not about manufacture, and the literature treats box attributions as evidence of how a work was valued, not when it was made. Seals can be added, removed, and forged. The discipline of the fourth line is mostly the discipline of asking which layer a given piece of writing belongs to.

Met 2015.300.3 is the clean case precisely because its writing is integral and specific. The colophon is not a later box label; it is a transcription note left by the copyist on the sheet, giving the era date, the lineage it descends through, and working instructions for colour.¹ The Met API reports the object’s begin-date and end-date as the same single year, 1107: the rare certainty where a painting is fixed not to a century or a reign but to one year, because it says so itself and the saying is part of the object.⁸ This is the benchmark function in its purest form. The few inscription-dated works in the corpus are not interesting for themselves; they are interesting because every undated painting near them in style or recension is dated, in the end, by reference to them.

Scientific methods and the width of the answer

The fifth line is the laboratory, and the honest account of it is mostly an account of its limits. Radiocarbon can be run on the silk or the paper, but it dates the harvest of the fibre, not the act of painting, and a bolt of silk could be woven years before it was painted and a painting repaired with later patch silk centuries after.⁹ The error bar is the deeper problem. The calibration curve has plateaus where many calendar years map to the same radiocarbon age, and two of them, roughly AD 1240–1306 and AD 1396–1532, sit precisely across the Kamakura, Nanbokuchō, and Muromachi windows where most Japanese Buddhist painting needs dating.¹⁰ A radiocarbon result on a fourteenth-century scroll can return a span a century and a half wide, which is wider than the stylistic reading it was meant to check. X-ray fluorescence and other pigment analysis answer a different and narrower question: they identify what the gold and the colours are, which can exclude a date (a synthetic pigment not available before a known year rules out everything earlier) but rarely pin one. The laboratory’s real contribution is exclusionary and corroborative. It removes impossibilities and it confirms or unsettles the stylistic reading; it does not, for the medieval centuries that matter most here, deliver a date narrower than the connoisseur’s, and a method whose error bar is wider than the eye it is auditing cannot overrule the eye.

What dating actually is, and where the disagreement sits

The five lines do not vote. They constrain. The ground sets the centuries and the absolute floor; pigment and gold technique narrow the band; the recension supplies the sharpest terminus post quem; an integral inscription, when it exists, collapses the band to a year; the laboratory removes impossibilities and, on the medieval plateaus, widens rather than narrows the calendar answer. A responsible date is the overlap of the four non-laboratory lines, checked against the fifth, reported with the overlap’s true width, and the width is usually a generation, not a decade, even when the label prints a single century with the confidence of arithmetic.

The disagreement in the field is not about the lines but about which one leads. The connoisseurship tradition, descending from the named-style method, leads with the trained eye and the style horizon, treating recension and inscription as confirmation and the laboratory as a tiebreaker. The technical-art-history position leads with the material (weave count, pigment cross-section, back-work, the physical sequence of layers) and treats the named style as a hypothesis the material tests. The Aizen drawing favours the second order without settling the argument: it is a case where the writing on the object dates it absolutely and the style would only ever have approximated, which is the technical position’s strongest kind of example. But the worked example is also a paper drawing that carries its own colophon, and that is the easy case. The hard case is the undated silk Amida with no inscription, no integral seal, a remounted colophon, and a radiocarbon span that brackets two reign-eras — and there the honest answer is the convergence of style and recension, held to a generation, with the residual uncertainty stated rather than rounded away. bodhi reads with the material-led order, but only because the dated benchmarks the eye is trained on are themselves material objects with writing on them; the eye and the lab are not rivals, and the open question is not which to trust but how wide to admit the answer really is when the object, unlike Met 2015.300.3, refuses to date itself.

Related

• How bodhi reads an image — the method behind every reading here
• Silk or paper: the ground, and why it bounds every later reading
• The eshi workshop — how the painting was actually made
• Gyokugan — the sculpture-side technique-as-terminus parallel
• The Unkei attribution problem — the named-hand version of this question
• Kaikei and the An Ami signature — the signed-benchmark case in sculpture
• Funpon — the iconographic working drawing (entity)
• Kirikane — cut-gold-leaf decoration (entity)
• Aizen Myōō (entity)

Footnotes

1. Met museum record, “Iconographical Drawing of the Mandala of the Wisdom King Aizen,” acc. 2015.300.3 (object 53163), reading the colophon as the fifth day of the third month of Kashō 2 (1107), copied from a model inherited by Sanmai Ajari Ryōyu from his teacher Chōen (1016–1081), transcribed at Shōren’in with white/red flesh notes left by the copyist. The museum’s catalogue HTML returned HTTP 429 on direct fetch this pass; the colophon reading is from the museum’s own indexed catalogue text via the collection search excerpt and is corroborated by the API object record. The internal catalogue page is NOT independently pinned and is on the watch list. ↩ ↩² ↩³ ↩⁴

2. Met museum record, “Iconographic Drawings of the Five Kings of Wisdom (Myōō-bu shoson),” acc. 1975.268.6 (Heian period): the Met’s framing of Heian iconographic drawings as transmitted through teacher-student lineages with dated colophons, and of the Zuzōshō (compiled in the 1130s) as a ten-scroll encyclopedia fixing esoteric recensions in text and image. Used for the recension/terminus-post-quem mechanism; internal catalogue page not pinned. ↩ ↩²

3. The silk-versus-paper ground contrast, sizing, and back-work are the subject of the bodhi technique-history piece silk-vs-paper-ground-buddhist-painting and are deliberately not re-argued here; only the dating-relevant fraction (weave specification as a coarse clock, the absolute floor) is used. ↩

4. Chi-sun Park, “Goryeo Buddhist Painting: Materials, Techniques, and Mountings,” Smithsonian National Museum of Asian Art (Freer/Sackler) online publication: high-status Buddhist silk woven to specification (in the Goryeo case a doubled warp reserved for the purpose), later thin-patronage grounds pieced from mixed cloth, gold powder vs orpiment, and the near-universal later remounting that complicates original-technique reading. Applied to the Japanese case by close analogy; the Goryeo-to-Japan transfer is load-bearing and is flagged on the watch list. Web essay; internal pages N/A. ↩ ↩²

5. “Kirikane,” Wikipedia, reference-tier: cut-gold-leaf (and silver/platinum) technique imported from Tang China, flourishing in Japan from the Late Heian period and continuing to the 13th–14th century, after which it largely disappeared with the decline of Buddhist-art patronage. Used here only for the coarse chronological horizon, with the soft-edge caveat (no fixed start/end, later revival possible) stated in-body. Not independently pinned to a primary materials study. ↩ ↩²

6. “Takuma School,” Encyclopædia Britannica: a Japanese Buddhist painting school flourishing 12th–14th century, founded by Takuma Tametō (active 1132–74), continued by his son Tametatsu and through Kyoto and Kamakura branches with named 13th–14th-century members. Reference-tier; the founder dates are Britannica’s, claim-clause attached, not independently pinned to a monograph. ↩

7. Nomura Kakejiku art blog, “Takuma School / Schools of Nihonga”: the Takuma manner characterised as thick sumi ink line with cool, restrained colour keyed to the drawing, set against the warm wire-line Fujiwara manner and the gold-rich kirikane register. Practitioner/reference-tier; used only for the stylistic characterisation, not for any date. ↩

8. Met Collection API, object 53163 — objectBeginDate 1107 and objectEndDate 1107 (a single-year span), medium “Hanging scroll; ink on paper,” image 58.4 × 53.3 cm, Mary Griggs Burke Collection 2015, public domain (CC0), re-confirmed live 2026-05-16. The API is authoritative for object facts. ↩

9. “Radiocarbon dating of textile components from historical silk costumes and other cloth products in the Ryukyu Islands, Japan,” Radiocarbon (Cambridge University Press): the ¹⁴C date of a textile reflects the time of fibre production, not the time of use, since cloth can be used or repaired generations after manufacture. Reference-tier (abstract); the production-vs-use caveat is the load-bearing point and is general to the method. ↩

10. “Wiggle matching,” Wikipedia, and the supporting Historic England Bayesian-modelling and Hallstatt-plateau references: calibration plateaus map many calendar years to a single radiocarbon age, and wiggle-matching is noted not to recover precision in the windows AD 1240–1306 and AD 1396–1532 — the windows that bracket the Japanese medieval painting corpus. Reference-tier; the plateau windows are the specific, citable point and are stated as such. ↩