140 Blade lengths (with relevant historic laws)

[Gabriel L]

Hi everyone,

 

On mobile so I can't write much. I whipped this up:

 

 

EDIT: I notice two errors already, haha: it should say last *muzori* (not uchizori) blade, and also the academic lower limit for chu-wakizashi is listed correctly but marked 1cm too low. It is clear the academic wakizashi classifications are based on the historical laws.

 

Thought this was surprisingly more interesting than expected. Sample size is low, but it certainly seems to explain wakizashi well.

 

My biggest question: why the sudden inflection at 65 cm???

 

Regards,

 

—G.

[Rich S]

G -

 

Very interesting. Would also be interesting to see same y-axis vs timre (era and/or year).

To see what lengths were popular during what eras.

 

Rich S

[ROKUJURO]

Gabriel,

 

the blade length is certainly related to the physical dimensions of the people who used them. A KATANA, worn in the traditional way in the OBI, has a useful maximal length when drawn, if the SAMURAI had an average height of, let's say, 165 cm, and a respective arm length. Of course, there are always exceptions, but we talk about statisics.

 

The other side is the physical dynamics. The shorter a blade, the faster it can be moved. A longer blader is slower, but has wider reach. A TACHI, wielded from horseback, will have other requirements than a sword of an ASHIGARU. Fighting techniques and the develoment of armour will certainly play another role. So there will always be a consensus from practice in blade length if all factors are to be considered, and this comes obviously down to an average of around 65 cm.

[Gabriel L]

Jean,

 

But the data does not strongly support that.

 

If the falloff was purely the result of natural considerations for size based on practical use, you would certainly expect to see a change around there (and to be sure it does contribute to the grouping). But it would likely be a smoothly accelerating falloff, reflecting myriad differing opinions about where precisely the lower bound for a usable katana lies, and the myriad body sizes and strengths of users — all of which lie on smooth distributions. For example, the upper bound does seem to obey such a limit: interest in long blades is not constrained by the legal limit, but rather practical limits, resulting in a smoothly-accelerating curve after about 73 cm.

 

Instead, there is an extremely well-defined point at *exactly* 65 cm (or, if you prefer, 2 shaku 1 sun 5 bu). People are willing to go down to 67 cm, 66 cm, 65.5 cm, 65.25 cm, 65.0 cm… but the moment you cross that threshold, interest plummets.

 

That kind of sharp change is much more parsimonious with a specific, *explicit* limit. Examples might be: 1) some sort of law regarding minimum katana length, 2) some sort of widely-accepted and well-published specific idea about what is appropriate or practical, 3) similarly, some sort of universal superstition / numerology.

 

Of course I have to admit that the sample size is not quite large enough to be definitive, but on the other hand the effect is dramatic enough to be suspicious. EDIT: another possibility is that, with the explicit legal definitions for wakizashi, there is an indirect pressure on katana length to be at least a “significant-number” greater. For example, that lower bound is close to exactly 5 sun above the Edo Castle norm for wakizashi length (1.65 shaku + 0.50 shaku = 2.15 shaku or 65.1 cm) and 3 sun above the 1645 limit. But this feels weak to me.

 

------

 

Rich, I actually tried a few date-series graphs, but very little came out of it except a slight increase in the linear fits for moto/sakihaba, kasane, length, etc., and a very slight decrease in the linear fit for sori. A moving average might have worked better, but really what it comes down to is that 140 examples isn’t a very good sample size to reveal chronological trends. I may try a few more graphs based on subsets of the data though.

 

------

 

I actually have some more explorations of this information but I wasn’t able to post them this morning thanks to a faulty internet connection. I’ll be back in a bit with some more!

 

—G.

[BIG]

Hi,

 

don't See an average length of all blades. Comparing Pears with apples.

 

Average katana length from this very small Data is about 70 Cent.

 

More Sense we knew total Number in defined aera.

[ROKUJURO]

Thank you, Gabriel!

 

There is a point I am curious about: are the data based upon blades in general or were they chosen with regard to SURIAGE or UBU NAKAGO?

[Gabriel L]

Hi Peter,

 

…don't See an average length of all blades. Comparing Pears with apples.

 

Of course it would be, if I did that, but I didn’t give any kind of average, just plotted the raw data and identified groupings. Or are you referring to the time series? For those I limited the samples to specific regions (the definitive/characteristic groupings on the chart above). Still not very clear results, would probably need a couple thousand blades to be more revealing.

 

Average katana length from this very small Data is about 70 Cent.

 

To be specific, the average of the subjectively-defined “definitive sample range” for daito (everything from 65.0 to 73.0 cm) was 69.4 cm, with a standard deviation of 2.22 cm, for a ±sigma range of 67.2–71.6 cm.

[Surfson]

This is a wonderful approach to gathering data on nihonto Gabriel! i wonder if Aoi art would be willing to give you their historical data. I assume that the "market" decided the 65CM cutoff, and that this length was the minimum at which the fact that a sword was a katana became obvious across the room. Since merchants couldn't carry katana, this cutoff must have had social and status significance. Keep up the good work! Cheers, Bob

[Gabriel L]

Thank you, Gabriel!

 

There is a point I am curious about: are the data based upon blades in general or were they chosen with regard to SURIAGE or UBU NAKAGO?

 

An excellent question! Aoi doesn't explicitly list nakago status in a simple way (sometimes Tsuruta-san mentions if a blade is suriage, sometimes not). I did in fact mark down my very quick impression of whether the nagasa had been shortened and distilled that to a very simple binary 1 or 0 for "Ubu."

 

I then tried a few graphs with the entire sample and with just the "ubu" blades. Clear trends did not emerge, except that the groupings were less distinct, not more. In a way that is itself an interesting result. Unfortunately there is the problem that restricting the sample size makes the results that much less valid, and also since my "ubu" column was suspect (literally I spent about 5 seconds guessing each time, except when explicitly included in the item description) I elected not to produce a graph to post here. However I think I will go back and do so anyway, in the interest of discussion.

[ROKUJURO]

.......don't see an average length of all blades. Comparing pears with apples. Average katana length from this very small data is about 70 Cent. More sense we knew total number in defined aera.

Peter,

 

it seems, the x-axis is the number. I don't see an average length of blades, but in the range of the measured blades there is a significant number with a typical WAKIZASHI length and a typical DAITO length between 65 and 73 cm. Nothing extraordinary.

 

I think the objective was not to find the average length but to explain why a certain length was predominant.

Just my personal interpretation.

[Gabriel L]

.......don't see an average length of all blades. Comparing pears with apples. Average katana length from this very small data is about 70 Cent. More sense we knew total number in defined aera.

Peter,

 

it seems, the x-axis is the number. I don't see an average length of blades, but in the range of the measured blades there is a significant number with a typical WAKIZASHI length and a typical DAITO length between 65 and 73 cm. Nothing extraordinary.

 

I think the objective was not to find the average length but to explain why a certain length was predominant.

Just my personal interpretation.

 

This is correct. This isn't a plot of one variable against another, the x-axis is simply a way of stretching out the data to make it easier to see the groupings. Putting it another way, this is like if you lines up all the swords in a row. So the x-axis is, strictly speaking, the ordinal number of sword in the sample when arranged by height (i.e. x = 1 = sword #1, x = 2 = sword #2, etc.). I saw fit not to include an x-axis label because it seemed intuitive to me to the point that rigorously defining the x-axis would be likely to confuse more than help.

 

An alternative way of displaying this data would be a heat map. But I didn't have a quick-'n'-dirty way of producing one via Excel.

[hxv]

Gabriel,

 

Nice work! I think a histogram would most illuminating and it would take me under two minutes to do with Interactive Data Language (IDL) if I had access to raw data. I would be glad to do the plot for you if you could PM me or email me at vu.hoanh@gmail.com.

 

Regards,

Hoanh

[Gabriel L]

More data!

 

 

This is a little messy — I could have split it into several graphs, but it didn't seem entirely necessary.

 

The main thing to notice is that the Motohaba is fairly well-correlated with nagasa until about 40cm, after which point it is less consistent and less correlated (more constant). That is also when sakihaba starts appearing (i.e. a yokote).

 

The sakihaba & degree of taper (motohaba minus sakihaba) don't trend dramatically with nagasa (slight increase), and there is a lot of sword-to-sword variation. Same thing with kasane.

 

Uchi/muzori ends around 31 cm, giving another indicator of when tantō end and ko-wakizashi begin. After that, sori is so variable that I used a moving average rather than artificially impose a best-fit line.

[Gabriel L]

Gabriel,

 

Nice work! I think a histogram would most illuminating and it would take me under two minutes to do with Interactive Data Language (IDL) if I had access to raw data. I would be glad to do the plot for you if you could PM me or email me at vu.hoanh@gmail.com.

 

Regards,

Hoanh

 

Anyone wanting to examine or analyze the data is welcome. Here is the excel file.

[BIG]

Hi Gabriel,

 

See it, only reply to Other reply...

 

Great so, there is a CREED for more, including a Time axis...

 

Best regards

[Gabriel L]

Here are the ubu lengths (98 swords).

 

Interesting that the bottom end of daitō is still sharply defined, but at a slightly higher point (~66.6 cm) than before.

 

Again, please note that the categorization of whether or not a blade was ubu was not rigorously determined, i.e. this data should be treated as investigational and not conclusive by any means.

 

[hxv]

Gabriel,

 

Thank you for the Excel file. Here are the histogram plots - same data set with two different bin sizes (1cm and 2cm). It clearly shows the mode (most common occurrence) at about 70cm. There is a significant secondary mode at about 65cm. Thanks for starting this very informative thread. I'll work on similar histograms for wakizashi and tanto when I get a little time later today.

 

Regards,

Hoanh

[b.hennick]

Great thread! Keep the analysis coming. l Love it :D

[hxv]

Here are the histograms for wakizashi and tanto, which modes are 46-49cm and 26-28cm, respectively.

 

Regards,

Hoanh

[Gabriel L]

Thanks Hoanh.

 

I played around a bit more. Here are the time series graphs, but please keep in mind that like the categorization for ubu, these dates are somewhat iffy. Aoi Art often only lists "early/middle/late (period)", so there are spikes at, for instance, 1550 ("late Muromachi"). The biggest spike of all is Kanbun (1661).

 

I think maybe the most interesting thing is the lack of tantō in the middle Edo period.

 

You can also see that ubu blades before 1500 are quite rare, and most blades in the later Edo period are ubu. Also, "wakizashi" from before 1500 are probably all shortened daitō. Comparatively few tantō are ever shortened.

 

This is all kind of basic stuff, but it's nice to see it formally presented.

 

[Ken-Hawaii]

You can also see that ubu blades before 1500 are quite rare, and most blades in the later Edo period are ubu.

 

Gabriel, this doesn't seem intuitively obvious to me - in fact, just the opposite should be true. Early blades wouldn't have been cut down until the various governmental regulations started mandating specific lengths as a maximum.

 

Or am I missing something here...?

 

Ken

[Ken-Hawaii]

I just thought of one way to eliminate a big variable. The Bizen-Osafune Sword Village has a huge collection of blades - I saw about 150 the last time my wife & I were there - & I'm wondering if Piers could ask them for lengths, time period, ubu-ness, etc.? And as they would all be from a single school, I think the correlations would be more meaningful.

 

Piers, what do you think?

 

Ken

[Gabriel L]

You can also see that ubu blades before 1500 are quite rare, and most blades in the later Edo period are ubu.

 

Gabriel, this doesn't seem intuitively obvious to me - in fact, just the opposite should be true. Early blades wouldn't have been cut down until the various governmental regulations started mandating specific lengths as a maximum.

 

Or am I missing something here...?

 

I think something is getting lost in translation, or flipped around, yes. When is the last time you saw an ubu blade from the Nambokucho period? That's what this is revealing; old blades are most likely to have been shortened at some point (probably the 1500s–1600s), whereas Edo-period blades are less likely to have ever been shortened. The chart isn't showing when they were shortened, it's showing when they were made and IF they were (later) shortened. There's no way to actually tell when the shortening happened, of course.

[Jean]

Change in war technics is IMO much more responsible for sword shortening than edicts (till beginning of Shinto). Take uchigatana length, and change of war from cavalry to infantery, introduction of Teppo, some interesting reading to be made in the Connoisseur's book of Japanese swords. Compare the length of the uchigatana to Kamakura/Nambokucho swords.

[BIG]

Hi all,

 

Super Chance, what's about all member nihonto ( round about 5000 to 10000 I think)

Gives us One History?

 

Best regards

[O koumori]

Gabriel,

Fine work!

If you can find it, Thomas Temple Hoopes did his PhD dissertation on Nihonto in 1931 at New York University. He did a complete dimensional analysis of 103 blades, with accompanying descriptions.

I think the Smithsonian Institution might have a copy in its library holdings holdings, or maybe you can obtain a copy from the university...

Dan K.

[runagmc]

I'm surprised there weren't more ko-wakizashi in the 36-38cm range...

 

Anyway, I also wonder how much sales (what kind of swords aoi-art likes to buy and sell) influences all this? Could that have someting to do with the lack of daito between 60-65cm?... or not?... :D

 

Anyway, pretty interesting, even if not always surprising...

 

Thanks for all the hard work...

[Gabriel L]

I'm surprised there weren't more ko-wakizashi in the 36-38cm range...

 

Anyway, I also wonder how much sales (what kind of swords aoi-art likes to buy and sell) influences all this? Could that have someting to do with the lack of daito between 60-65cm?... or not?... :D

 

Anyway, pretty interesting, even if not always surprising...

 

Thanks for all the hard work...

 

You touched on a critical subject which is sample bias. What a specific dealer chooses to buy and list will surely influence the results. For instance, I think the huge spike of "Kanbun" 1661 blades in the Aoi Art subset is almost likely a reflection of the tendency to attribute any mumei low-sori Shintō katana as being from ca. that era, rather than a preponderance of blades actually made during Kanbun proper.

 

I am actually massively expanding the sample size (already at 1,600+ swords now, in no small part thanks to a huge data delivery by Stan of NihontoClub) and source diversity (different dealers, books, etc.) to both reduce the sample bias, and increase the statistical validity/range of results, allowing for greater confidence and also some new interesting conclusions.

 

It takes me a while to produce a round of graphs each time, and at the moment I am more concerned with adding data than analyzing it. When I am satisfied with the diversity, size, completeness and validity of the sample, I will produce a few more graphs. Also, I am going to write a more formal article on the final results.

[runagmc]

Sounds good...

 

It would also be interesting to see more type specific studies, like say, the percentage of tanto out of a sample group that are shobu, hira, kanmuri zukuri, etc....

 

For something like that it may be better (or more interesting) to collect blades only from a specific period...

 

Many things COULD be done along this line, and I'm sure some have been already...

[Jean]

Shobu zukuri tanto must be very rare, Adam, I don't remember having seen one (under 30 cm nagasa). Generally, we see them on wakizashi. Have you got examples in picture?