A blade associated with the Shimazu family disappeared from view for about a century and a half. When a candidate emerged, an old drawing gave investigators something more exact than a famous name. Near the tip they could compare the course of a boundary in the steel. Farther down they found two places where that boundary divided and crossed. The Japanese cultural heritage register puts these correspondences at the centre of its account of the Shimazu Masamune. A sheet made to carry information away from an object had helped people find their way back to it.
What sort of sheet could do that? In English, oshigata often becomes “rubbing.” The translation captures an important connection between paper and a surface, particularly an inscribed tang. Open an album, though, and you may also find a free line, written judgements, copied names and an arrangement designed for comparison. Some features could register through contact. Others required someone to look, select and draw. The space between the two tells us as much about the history of observation as it does about the object observed.
Our four illustrations come from a book published in 1928 by Hazawa Bunko, now digitized by the National Diet Library. They show the title leaf and three different openings or pages of its contents. The library supplies a persistent identifier for the entire book; each figure here also identifies its digital frame. The distinction matters. A reader who follows the first reference can establish what the volume is. A reader who follows the second can inspect the particular marks on which an argument rests.
Inside a book of several observers

The title includes Tsuguhira and also refers to sets of drawings associated with Hon’ami Kōtoku and Kōon. Its bibliographic date belongs to the edition in front of us. It cannot simply be transferred to the original drawings or to the steel they describe. A page may bring together a medieval blade, a much later act of observation, a later still act of editing and a modern digital image. All four can have secure dates without having the same date. The paper book on the library shelf is itself an object with a publisher, a size and a sequence.
Consider the title leaf as evidence in its own right. The large vertical characters occupy a narrow column, with smaller writing alongside and ample unprinted space around them. A screen can make this leaf look as densely worked as a later drawing by cropping out its margins. In the complete frame the hierarchy is quite different. The eye goes first to a title and then to subordinate information. Nothing about the visual prominence of a name proves that the person named saw every item represented between the covers.
The library lists Hazawa Bunko both as editor and publisher. Those roles explain why the 1928 volume has its particular unity. They leave open how its sources were arranged before publication. Even a scrupulous copy changes a drawing when it supplies a new neighbour, moves a note to fit a block or groups works under a single heading. The result can be excellent testimony to a tradition of comparison while being an imperfect witness to the appearance of an individual original sheet. Keeping both possibilities in view makes the edition much more interesting to read.
The difference between a hollow and a letter
An inscription cut into metal presents a problem that ordinary writing on paper does not. The mark occupies depth as well as area. Light falling from different sides can reverse the apparent relation between hollow and raised surface. Paper testimony can preserve a configuration of strokes with unusual clarity, yet its black and white describe a paper surface. They are not measurements of the depth of an incision. Two reproductions with different contrast can appear to disagree about a stroke even when the underlying outline is substantially the same.
A written transcription performs another change. It offers a reading of a character, often in an entirely different hand or typeface. That is invaluable to someone seeking a name in an index. It can dispense with the very thing an investigator wants to compare, such as a doubtful terminal stroke or the position of a character relative to a hole. An excellent catalogue may therefore reproduce an inscription and print a transcription nearby. One gives access to a spatial relationship. The other makes a linguistic claim that can be searched, translated or contested.
The margin provides a third kind of testimony. A note might identify a smith, preserve an owner’s name, give an old measurement or merely label which portion is illustrated. Its presence beside an image does not turn it into part of the inscription on the object. Look for the difference between writing enclosed by the outline and writing standing clear of it. Where the distinction remains ambiguous, “a note beside the drawing reads” is a more informative description than silently turning the text into a signature. It also leaves room for a later reader to improve the transcription.

Frame ten of the Hazawa volume makes a good exercise in looking before reading. The page offers several narrow forms, columns of writing and conspicuously empty paper. Some marks occupy a bounded shape. Others form orderly columns outside it. Even a reader who cannot decipher a single name can describe this arrangement and ask whether the same section of the object appears more than once. The image establishes the arrangement. Deciding which face is shown and assigning names to the different hands take further evidence from the text and the rest of the book.
The line that contact alone cannot report
An area of patterned steel can be visible to an observer while offering paper no raised lettering to register. In an oshigata, the hamon, the visible pattern associated with hardening, may therefore appear as a drawn interpretation. A modern call for oshigata art issued by the Nihonto Bunka Shinko Kyokai distinguishes work on the tang from a drawn account of the pattern. It is evidence for how practitioners of that exhibition understood the art. It also suggests a question to put to older sheets: where does a mechanical-looking mark give way to an observer’s line?
An observer can choose to follow a boundary through an area of low contrast, interrupt it where the original becomes uncertain, or record a small island inside a larger form. The three decisions produce quite different kinds of information. A continuous contour makes direction easy to follow. An interruption preserves the limits of what could be seen. An isolated mark can guide future comparison to a particular locality. A printed page showing only the result does not disclose how long the observer spent looking or how often the blade had to be viewed under changed light.
This is one reason drawings remain useful alongside photographs. A photograph saves the distribution of light that reached a camera at one time. A drawing may combine observations made at several times into a legible whole. The advantage comes with a cost. It is possible for an attentive observer to bring two features into an apparently continuous relation that another observer would hesitate to assert. An independent photograph, a second drawing or a direct examination can test the connection. The strongest case often depends on the agreement of features whose positions can be described individually.
The Shimazu comparison, read closely
The entry for the unsigned Shimazu Masamune supplies an unusually clear account of such a test. It locates a full view and a depiction of the hamon in the Tsuguhira compilation, which it places before the Kyōhō register of famous blades. It then compares the form at the tip and two separated crossings farther down with the visible object. This is a richer comparison than saying that two outlines look generally similar. A reader can tell where the correspondences occur, how many local points the writer singles out and which source contains the picture.
There are also two distinct records in that account. The list of famous objects gives a name, dimensions and a historical valuation. The illustrated source preserves a distribution of features. An entry in a list might allow several candidates to remain plausible. A picture that survives independently can narrow the field. Conversely, a very close visual match would tell us little about the name by which a household once knew an object if the diagram had lost every heading and reference. The paper and the register answer different parts of the identification question.
The cultural heritage entry attributes the identification to this comparison and reports a later history involving the Konoe family and a donor. It does not turn the interval of uncertainty into a fully documented itinerary. Knowing what the object is and knowing where it was in a particular decade are separate historical achievements. A museum caption can tell the identification story accurately in a few lines if it names the diagnostic evidence and retains the gap. Filling that gap with a plausible journey would weaken a case already strong enough to stand on its stated terms.

Frame forty-five, illustrated here, is a comparative example from the same printed compilation. The wavy interior contours, the outlined extremities and the neighbouring columns make the conventions readily visible. We have not identified this frame as the sheet used in the Shimazu investigation. That distinction is important in a digital article where a reader scrolling quickly can mistake a nearby picture for a portrait of the named object. The image documents a form of graphic testimony; the published heritage entry supplies the argument about a particular named blade.
Four famous names in a library gallery
The National Diet Library’s Japanese Sword gallery offers another way into the material. Its selections bring named examples associated with Yoshimoto Samonji, Fudō Kuniyuki, Yagen Tōshirō and Sayo Samonji into a common reading space. Elsewhere in the same presentation the library introduces editions of an annotated register and identifies an enlarged edition. A modern gallery can reunite subjects scattered across books, letting someone follow names more quickly than the pagination of any single volume allows. Its modern order is therefore a research aid with its own editor.
Yoshimoto Samonji is a particularly instructive heading because a name can preserve a historical association that the metal itself voices differently. Japan’s heritage collections and Kyoto National Museum record alternative names associated with this blade. An editor looking under one form of the name can miss material indexed under another. At the same time, combining every occurrence of a shared personal or workshop name into a single citation would join things that may have very different histories. A cross-reference should explain why two names identify the same object in the particular source being used.
The gallery’s juxtaposition of images and written appraisals brings the role of attribution into focus. A named drawing tells us what its maker or compiler thought was worth illustrating under that heading. A paper of appraisal embodies a judgement issued in a particular setting. A later museum entry can accept, qualify or revise a received attribution. Even when all three use the same name, their agreement should not be counted automatically as three independent identifications. A later compiler may have copied both the name and the drawing from the same earlier witness.
What a digital frame keeps and loses

Frame sixty-eight offers a different reading experience from frame ten. A broad interior form dominates one side while writing and a more densely worked strip run alongside. At book size the contrast between the economical outline and the busy area is immediate. Zooming in can reveal a small junction of lines. It can also remove every cue about which portion of the drawing one is examining. The frame number and an uncropped view make those local discoveries reproducible. A detail with no route back to its full page is surprisingly difficult to use as evidence.
The colour of the scanned page deserves similar care. A warm paper ground in a digital reproduction is not by itself a paper identification or a date. Lighting, digitization, the colour balance of a display and the age of the printed leaf can all affect the image. The dark gutter and the partly visible edge tell us that the camera recorded a page in a bound object. They do not tell us how an earlier loose sheet was mounted when an observer first viewed it. Here the physical trace that looks least important can be the easiest one to interpret securely.
The work also has several plausible systems of numbering. A digital service counts frames, including covers and preliminary matter. A printed volume may number leaves, openings or individual entries. A historical group of drawings may preserve a further sequence of its own. A reference such as “plate ten” can consequently lead to three different places. Giving the book’s persistent identifier and specifying digitized frame ten takes only a few extra words. It lets the next reader find the same image even when the online viewer changes its controls or opens at another page.
Reading disagreement as historical evidence
Imagine two editions of a famous-blade register that place a small mark on opposite sides of an outline. A modern reader might immediately prefer the cleaner print. The difference could have entered when a drawing was copied, set beside a different face, reduced for the page or presented as an enlarged detail. The first question is which way the associated lettering reads and whether the rest of the features also reverse. A single misplaced mark calls for a different explanation from an entire reversed image. An edition can be handsome and still be several steps removed from its exemplar.
A similar distinction applies to absence. An unmarked patch of paper could stand for a plain-looking area, an observer’s decision to leave out a difficult detail, or a passage that printed too lightly to survive in the scan. The surrounding page sometimes helps: if fine writing nearby is also weak, reproduction becomes a plausible factor. When the other lines remain sharp, omission merits consideration. Neither observation authorizes the conclusion that the original object lacked the feature. The surviving page can establish what a particular edition allows us to see today.
There is a human achievement behind all this caution. Someone had to decide that an elusive visual event was worth preserving and that another observer, perhaps generations later, would understand how to look for it. Compilers, printers, librarians and museum cataloguers kept that act of attention available, each in a different medium. The crossing lines in the Shimazu account give the history a wonderfully concrete end. The most valuable information carried by a drawing was something another person could find again, even after the name had ceased to point to a publicly known object.
A final clue sits in the spelling of the names on a modern catalogue page. Macrons, variant readings and translated titles make English prose easier to follow, yet none has to occur on the old sheet. Recording the Japanese title as well as the modern rendering lets someone who reads another language return to the same item. The library identifier is particularly valuable here because it remains constant across changes of romanization. Together, image, title and identifier keep a historical disagreement open to inspection. A reader can disagree with a proposed reading without losing sight of which strokes or which leaf prompted it in the first place.
Questions about paper records
Is every part of an oshigata a contact impression?
An oshigata may combine an impression of an inscribed surface with drawn features and written commentary. Each sort of mark records a different observation.
Which drawing helped identify the Shimazu Masamune?
The heritage entry cites the Tsuguhira compilation. It compares the tip and two places farther down where parts of the visible pattern divide and cross.
Does 1928 date the original observations in the Hazawa volume?
It dates the published edition held by the National Diet Library. The sources associated with Tsuguhira and members of the Hon’ami family are earlier.
Can an inscription transcription replace an image of the inscription?
A transcription gives a proposed reading. An image can also preserve placement, spacing and doubtful strokes that a normalized reading omits.
Why do the illustrations identify a digital frame?
A frame counts images in the library viewer, including preliminary matter. It can be located even when a leaf or entry carries a different number.
Is frame forty-five presented as the Shimazu identification sheet?
It illustrates the graphic conventions in the printed compilation. The evidence for the specific identification comes from the cultural heritage register.
Museum films
The Met — Art of the Samurai—An Introduction
The Met — Sunday at the Met: Brush Writing in the Arts of Japan
Collection records and further reading
- National Diet Library, Hazawa Bunko edition, 1928
- Cultural Heritage Online, unsigned Shimazu Masamune
- National Diet Library, Japanese Sword gallery
- Nihonto Bunka Shinko Kyokai, guidelines for the oshigata exhibition
- Kyoto National Museum, named blades in the museum exhibition
- How a Blade Acquires a Famous Name: Meibutsu in Albums and Inventories
