When the Shimazu Masamune reappeared in the historical record, a remarkably compact entry in an old catalogue came back into play. The sword was named; its shortening was noted; its length was given as two shaku, two sun and seven bu; an appraisal followed. Written across the page in English, the information looks almost like a modern object label. It was doing something rather different. The entry kept distinct the identity of a celebrated possession, the state in which it was known and the value attached to it. A length was one of several ways in which a reader might recognize the object.
The Japanese cultural heritage record for this unsigned sword explains why the identification commands attention. A book of outlines older than the celebrated Kyōhō inventory preserves evidence for comparison. The museum account points to particular correspondences in the visible pattern near the point and elsewhere on the blade. No amount of arithmetic would recover those correspondences from the written dimensions alone. Yet the number has a history worth recovering in its own right. To translate it faithfully we need to know how a unit was divided, which standard the translator has chosen and what exactly the old writer was describing.
A red ruler in the Hōryūji collection

A good place to begin is a small red object in the Tokyo National Museum. Its collection number is N-83. Viewed across a screen, it can look like a decorated strip of indeterminate scale, long enough perhaps to span a writing desk. The national e-Museum specifies a length of 29.6 centimetres, a width of 2.2 centimetres and a thickness of 0.8 centimetres. Those last two figures give the strip back its proportions. The red surface and the pale designs might catch the eye first, but the length of the object is itself part of its historical subject.
The ruler belongs to the Hōryūji treasures. The exhibit record compares it to rulers in the Shōsō-in repository and suggests that it was made in Japan after an imported example. The inference belongs to the curatorial account. It depends on a comparison of workmanship, and carries a different degree of certainty from the inventory number. A useful label can preserve both the stable identifying number and the qualified attribution without losing either. The object's length, the source says, corresponds to the Tenpyō standard. This makes the artefact particularly instructive for anyone accustomed to seeing a shaku equated automatically with just over thirty centimetres.
Nara National Museum's exhibition glossary gives 29.6 centimetres for the Tenpyō shaku and relates it to the Tang Chinese chi. It also explains why a richly decorated example belongs in an exhibition of treasures. Many rulers in the repository were ceremonial implements. A ruler could embody an accepted order of measurement and be worthy of decoration, presentation and preservation. It need never have belonged to the workshop of the maker whose work one hopes to understand. No catalogue evidence attaches the Tokyo ruler to any of the swords illustrated here. Its contribution is independent and in some respects more valuable. It witnesses a historically situated length for the unit itself.
The survival of several rulers also discourages an excessively tidy picture. The Imperial Household Agency's record for a white ivory example gives the same 29.6 centimetres in length but a substantially greater width. Another record, for a green-stained example, gives a length of 30.45 centimetres. The measurements describe the surviving objects as catalogued. Without evidence for where a graduated interval starts and ends, subtracting their total lengths cannot tell us the precise difference between two official standards. A museum can preserve a ruler of a certain size without making the claim that every part of it constituted one calibrated interval. That distinction matters as soon as a photograph is used as evidence for an ancient system.
An original, a reproduction and two very different dates

The Nara National Museum also preserves a modern reproduction of a red-stained ruler, catalogued as 1441-0. In the photographs shown here one instrument runs horizontally across its frame and the other stands upright. A reader who simply compared their spans on a phone would be comparing decisions made by photographers and page designers. Looking up the two collection records changes the question. The photographs represent separately identified objects kept by different institutions. A description of the reproduction belongs to the history of how such treasures have been studied and represented. Its accession cannot be substituted for the Tokyo accession in a citation to the ancient object.
There is a further trap in putting the two red strips side by side. A reproduction of a type of ruler need not have been copied from the particular ruler printed next to it. Calling the Nara example a reproduction of N-83 would introduce a relationship for which the juxtaposition itself supplies no evidence. The distinction sounds small until the claim is repeated in a classroom or a catalogue. From then on a newly invented chain of descent can find its way into descriptions of both objects. Keeping separate labels and linking the two records lets readers investigate that question without accidentally supplying the answer for them.
The same care applies to an older Japanese unit printed next to a modern metric figure. Sometimes both appear in the historical record. Sometimes an editor has supplied one. Sometimes the translator has made a conversion solely for the present reader. These are three different events. Retaining the Japanese wording before a parenthetical metric equivalent costs little space and preserves all three possibilities for later examination. It also keeps a modern calculation from being mistaken for an instrument reading made centuries ago. An editor can legitimately choose an approximate number for readability and still say exactly where it came from.
Two shaku, two sun and seven bu
The arithmetic of the Shimazu entry can be reproduced on the back of an envelope. A sun is a tenth of a shaku; a bu in this series is a tenth of a sun. Two whole shaku and two sun make 2.2 shaku. The remaining seven bu add 0.07. The result is 2.27 shaku. The relational definitions tell us how to combine the written quantities even before we decide on a metric equivalent. A reader who moves directly from the written numbers two, two and seven to a modern figure may miss that there are two distinct stages to the calculation.
The text of Japan's 1951 statute implementing its measurement law, reproduced by Nagoya University, gives an exact relationship for the ordinary shaku of ten thirty-thirds of a metre. Applied as an explicitly stated conversion convention, that makes 2.27 shaku about 68.8 centimetres. The result is an editorial equivalent for the published inventory entry. It does not establish that an Edo-period cataloguer consulted the later statute, and it does not establish the current measured length of the object. The expression in the historical entry can be reported directly. The centimetres are conditional on the stated standard. A reader should be able to quote either without silently inheriting a different claim.
For an illustration of why the choice matters, substitute the Nara glossary's 29.6 centimetres in the same calculation. One obtains about 67.2 centimetres. The two answers differ by roughly 1.6 centimetres before any uncertainty about the object itself enters the discussion. That second figure is a demonstration of sensitivity to the assumed unit. Nothing in this exercise establishes that a Nara standard governed the Kyōhō entry. Indeed, the time separating the Nara court from the compilation of the inventory is precisely the reason to label the demonstration. Its purpose is to disclose the assumption that made the first answer possible.
The arithmetic also explains an occasional surprise when translating a range. In its historical survey the Japanese Sword Museum prints the equivalent of two shaku four or five sun for a group of lengths. Read the smaller figure as 2.4 shaku and the larger as 2.5. Under the ordinary convention those equal approximately 72.7 and 75.8 centimetres, the interval the museum gives. Reading the two numerals as successive decimal places would create a single value of 2.45. An apparently more precise translation would have destroyed the range. Here the two languages printed on the same line provide a check on the interpretation of the syntax.
The dimension belongs to a particular part

The Metropolitan Museum of Art has a mounted tachi under number 10.187.26a, b. The catalogue distinguishes a sixteenth-century blade from a mounting of the mid-nineteenth century. A photograph lets a reader see more than one component in the same frame. The measurement line is even more revealing. It assigns 101.2 centimetres to the general length, 89.7 to the blade and 69.8 to the cutting edge. These are three measurements with three descriptions in one record. An entry in another institution's catalogue headed simply “length” could correspond to any one of several conventions. Matching the unit would do nothing to resolve the choice.
Imagine transcribing only the middle figure, under a heading that a reader understands as the length of the cutting edge. The transcription would preserve each digit printed by the museum and still introduce an error of almost twenty centimetres. Converting it to sun, or reporting it with an additional decimal place, would make the result look considered. Neither operation would repair the missing field name. A responsible comparison carries the wording of the parent record alongside the value. It then explains which two measurements appear comparable and leaves an open question where a source does not specify the extent that was measured.
This is also why a photograph chosen as a catalogue illustration cannot normally settle a disagreement between measurements. The visible outline may include the fittings, while the published number refers to an element concealed inside them. A page may display the whole photograph or a cropped detail according to the viewer's device. Knowing the number of pixels in an image gives the size of a digital representation. Recovering a physical measurement from an image requires additional scale and imaging information. In the absence of that information, the illustrated curves and fittings still tell a reader what sort of object the record concerns. They cannot provide an independent ruler.
A discrepancy that translation ought to preserve

A second Met record provides a particularly instructive exercise in checking a bilingual or dual-unit label. The blade signed by Masazane, accession 2001.574, is dated 1526. The record gives a length of 91.8 centimetres and a blade-edge length of 75.1 centimetres. The corresponding inch figures for those two measurements are consistent with the metric figures at the displayed precision. The curvature line, however, pairs three thirty-seconds of an inch with 2.4 centimetres. Both figures appear in the catalogue record. They do not describe the same length under the ordinary inch-to-centimetre relationship.
Three thirty-seconds of an inch multiplied by 2.54 equals about 0.238 centimetres, or about 2.38 millimetres. A difference close to a factor of ten can arise in several ways in transcription, conversion or typesetting. The numerical check identifies an inconsistency; it tells us little about how it arose. It also does not identify which printed figure corresponds to the curator's original measurement. Quietly changing one of the units in a translation would substitute an editor's inference for the museum's record. The appropriate object-specific statement gives the accession, identifies the field and reports the discrepancy. The photographs remain useful while the conflicting values await clarification.
A nineteenth-century inventory cannot always be queried as easily as a modern database. Nevertheless the same habit helps when a handwritten character, a printed fraction or a smudged digit seems implausible. Transcribe what can actually be read. If a digit is uncertain, specify the uncertainty in that position. If two editions differ, identify each edition and give its reading. An explanatory conversion belongs alongside this evidence, not in its place. Recovering the exact reading of a damaged page can change a numerical result. Changing the numerical result until it looks plausible will never recover the damaged page.
What a modern exhibition checklist makes visible
The Japanese Sword Museum's English checklist for its exhibition of Important Art Objects supplies a different kind of evidence. Its first row lists a tachi signed Kuninaga and a length of 69.4 centimetres. The entry also mentions a mounting and an accompanying Honami document. Several culturally connected things can thus be described together while the table offers one numerical field. A reader reusing the number should recover the column heading and the scope of the row. There is no reason to assign 69.4 centimetres to every item mentioned in the description, or to assume that an accompanying document originally gave the value now printed in the English checklist.
In another checklist, prepared for the 2015 Bizen exhibition at the Sano Art Museum, blade length and curvature have separate headings. The labels do some of the scholarly work that would otherwise have to be repeated in every entry. Downloading a PDF page or clipping out a single row may detach the figures from those headings. A transcription for publication should keep the page or row locator and reproduce the relevant column names in its own words. That makes later checks possible even if the institution redesigns its website or changes how an individual work is described online.
A particularly economical way of publishing the Shimazu example would be to give the inventory's named object and its condition, transcribe the units in sequence, then add a short note about the metric convention. A separate sentence could identify the older illustrated record on which the modern museum's argument depends. Such an entry gives a reader a number to use, shows exactly how it was obtained and preserves the evidence that singled out this particular unsigned work. It also makes it possible for a future editor to revise the interpretation without rewriting the primary evidence.
There is a quiet pleasure in arriving back at the Tokyo ruler after working through the catalogues. Its pale decoration and narrow red ground have not changed. What has changed is the set of questions a viewer can bring to it. A physical object can stand for a system without exhausting its history. A range printed in two sets of units can reveal how a sentence was meant to be read. A disagreement in a museum record can expose an editorial problem to a reader far from the gallery. And a few handwritten units can survive the disappearance of an object from public view, ready to enter the discussion when the object returns.
A final check concerns the units in the transcription itself. In a catalogue that separates length and appraisal, two and seven may be components of one measurement while two hundred belongs to a valuation on the same line. Treating every number as a candidate dimension loses the vocabulary that tied it to a material fact. In the Shimazu example the term for shortening precedes the length. The appraisal follows. A translation that preserves that order still reads naturally, and it lets a reader see that the length refers to an object whose alteration was already part of the record. It would be misleading to back-calculate an original length from it. That would require a separately documented earlier measurement. The word recording a change tells us something happened without telling us how many centimetres the change involved. What looks in isolation like an unfilled gap becomes, in context, a quite precise limit on what can be recovered.
Another small editorial choice concerns the last decimal place. Printing 68.787878 centimetres in a museum caption would faithfully reproduce the behaviour of a calculator. It would poorly represent the character of the surviving evidence. The last written unit in the entry is a bu, equivalent under the stated convention to about three millimetres. Giving the converted result to a tenth of a centimetre alongside the original lets a visitor follow the calculation. Giving it to a millionth would suggest an observation far finer than the precision expressed on the page. The restraint belongs to the historian as much as to the mathematician.
Questions readers bring to the catalogues
How many sun and bu are in a shaku?
One shaku comprises ten sun or one hundred bu in this system of length. The numerical relationship can be read before choosing a metric equivalent.
How long is the Shimazu Masamune in the Kyōhō entry?
The inventory cited by the Japanese cultural heritage record gives two shaku, two sun and seven bu following a notice of shortening. Using ten thirty-thirds of a metre per shaku gives about 68.8 centimetres.
Why does the Tokyo ruler measure 29.6 centimetres?
The e-Museum identifies that length with the Tenpyō standard. The ordinary later kane-shaku was longer. The collection number of the Tokyo example is N-83.
What do the three lengths for Met 10.187.26a, b describe?
The record distinguishes 101.2 centimetres overall, 89.7 for the blade and 69.8 for the cutting edge. The accompanying field names are essential for comparing the figures.
Can the conflicting curvature figures for the Masazane blade be reconciled?
The printed pair does not convert consistently. Three thirty-seconds of an inch is about 0.238 centimetres, while the record gives 2.4 centimetres alongside it. The record alone does not settle which figure reflects the original observation.
Did the length alone identify the Shimazu Masamune?
The museum account also compares the surviving work to a drawn record predating the inventory. Correspondences of the visible pattern were important to the identification.
Museum films
The Met — Art of the Samurai—An Introduction
The Met — Conserving Japanese Works of Art in Foreign Collections
Collection records and further reading
- Agency for Cultural Affairs: the Shimazu Masamune and the earlier outline
- Nara National Museum: Shōsō-in exhibition glossary
- National Institutes for Cultural Heritage: ruler N-83
- Imperial Household Agency: white ivory ruler
- Imperial Household Agency: green-stained ruler
- Nagoya University legal database: 1951 statute defining the ordinary shaku
- Japanese Sword Museum: dimensions in its historical survey
- Japanese Sword Museum: checklist of Important Art Objects
- Sano Art Museum: Bizen exhibition checklist
- How Long Is a Katana? Nagasa, Sori, and the Museum Method for Measuring Japanese Swords
