This is an experiment I did as part of a prototyping project, and was intended to satisfy my curiosity on the performance of sokui (続飯) or rice paste glue. The natural glue contains nothing but delicious Japanese rice and a little bit of water. Usually sokui is used as part of a system that also includes the mechanical strength of a guard, fuchi, or rawhide wrapping, or leather lace. One of the reasons rice glue was chosen for this purpose traditionally is that it is non acidic and does not degrade either the steel or the wood over time. Another is that, while relatively strong, it is not stronger than the wood itself. This allows a scabbard or handle to be split open for cleaning or repair without damaging the wood. These tests are designed to help me understand where the performance of rice glue ends when it is the sole source of structural integrity of a knife handle.
In this case I wanted to test the worst case scenario, a softwood tsuka core on its own strength. Similar to a traditional futokoro-gatana/kaiken tanto mount without any horn parts, really a shirasaya (Rupin Sansei no Goemon, anyone?). The test handle had no seppa or fuchi, no cord ferrule or wrapping, not even a mekugi peg to hold the blade in and mitigate stress to the top and bottom of the nakago-ana. In addition, Nootka Cypress is slightly softer and more compressible than Hounoki (Japanese Magnolia) which is the traditional handle core wood for nihonto.
I expected very little from the rice glue under these circumstances and continue to hold the position that it should only be used in conjunction with at least one physical component for proper performance, however I was very surprised with the amount of stress the glue withstood before failing.
The three use tests were whittling (pressure from the work via the edge of the blade), batoning (pressure from mallet force to the back of the blade), and splitting (pressure between the tip of the blade and the mallet on the pommel). An additional test involved water resistance strength of the bond by immersion of the extra piece cut off from the handle.
Making the Test Handle
I took a piece of Nootka Cypress split from a block I was using for another project and split it into two pieces, carved out the opening for the tang, and glued it back together with sokui (rice paste glue). I clamped it overnight, then carved it into a rough handle shape and cut off the excess wood from the end of the block (to remove the wood with deep checks in it…this is worst case scenario, remember). Once it had dried for another 24 hours, I put it to the test. The test blade measures 3″ long and 5/32″ thick at the munemachi, the handle just over 3.5″ long and 1.5″ by 19/32″ thick.
Test 1: Whittling
This test is about reproducing reasonable use type stress for this size of knife. The pressure of the edge against the work is transferred back to the nakago-mune and nakago-no-gata, creating a levered force on the inside of the handle. I hogged out some very large chunks at first and then shaved the wood into a round profile. A very dry scrap of 7/8″ square oak was the subject of the test, there were no issues to report. This stage alone makes the test a success with this construction, but I wanted to continue for harder use case scenarios.
Test 2: Batoning
This test is about creating the opposite leverage force as the whittling test, and introducing much higher forces and shock to the handle. A good 3″ hardwood limb was used to hammer on the spine of the blade and drive it through the same piece of 7/8″ oak, across the grain. I expected the handle to fail in the early stages of this test, but there were no issues to report. My confidence in the glue is far higher than before and I see it as a much more versatile solution than previously thought.
One feature of the nihonto style takedown system is that the handle can be removed easily and the blade used on its own for other purposes, particularly in emergency or survival situations (detail work, emergency spear, sterilized in boiling water, cord wrapped for skinning work, etc). This would also be one potential way to baton with no danger of damage to the handle core.
Test 3: Splitting
The first part of this test takes less force than the batoning above, but because the strike point is not directly above the working edge, it can put far more torque on the handle. A 3″ hardwood limb was used to strike the tip area of the blade and drive the middle of the blade through the oak along the grain. There were no issues with this process and the handle was still sound.
I was quite sure the second part of the splitting test would cause the handle to fail immediately as it was the first operation involving striking the handle directly. The tip of the blade was hammered straight through the oak near a knot by using the 3″ hardwood limb to pound on the pommel. After this test, the handle joint was still intact, though because there was no seppa this step did cause some internal damage to the wood that was revealed later.
At this point I decided to try some lateral and random stress and threw the knife (never throw a knife not designed specifically for throwing!) from about 6 feet against a knotty stump. After the second impact the handle popped and the test was over. On inspection, the cause of failure revealed itself as the second part of the splitting test. Hammering on the back of the handle had caused the blade to compress some of the wood near the machi (where a guard would normally prevent this movement) and put the core under strong outwards pressure from the tip of the tang. The angular force of being thrown against the stump caused the stress to exceed the limit of the rice glue and released the joint.
This damage would have been easily prevented by the use of a seppa or guard to spread out the force from the machi over the front of the handle.
Test 4: Water Resistance
This test would be better performed on a full handle and blade, but I used the 1″ by 1 3/4″ block that was cut from the end of the handle. This truly was worst case scenario as the original block had two deep checks in it running almost dividing it into three. I snapped one off by hand and used it as the control, to see what kind of shear force it would withstand when dry, and then submerged the rest for the water test.
I placed the dry block with a corner on the anvil and struck the other corner smartly with a wooden mallet. It did not shear even when the wood had compressed to the point of the anvil face reaching the glue joint on the other side. Then I placed it in the vise on the same angle and got it to fail.
The used the mallet and a couple of sharp blows to test the other block, then submerged it for half an hour. Using the same test and approximate force with the mallet it did not fail so I submerged it for 10 more minutes. The same results again so another 10 minutes of submersion. At the 50 minute mark the block did not resist much at all and the joint popped open. Because of the check down the centre, the maximum distance from the water for any part of the joint was less than 1/4″ and it held well for more than 40 minutes.
At some point I would like to repeat this test using intermittent soakings rather than continuous immersion to understand the rate that water moves into the joint under more realistic circumstances. Another area for study is the effect of high humidity, though Japan itself is the original humidity test and many antique tanto survive to this day in those conditions.
Conclusions
Some of the lessons learned are that rice glue is quite strong when mixed and applied in the right way, and that a slightly softer core can contribute substantially to the shock absorbing power of a handle. My understanding of the reason behind some of the nuances and details of nihonto tang shape and proportion has increased and I cannot recommend enough the value of a careful study of traditional forms for those interested in making excellent knives. A thousand years of research, design, and field testing is a good foundation to build upon.
When blades are to be subjected to strong parallel forces (tip first), the shape of the tip of the tang and some sort of guard or seppa are critical components for strength. In this case the final test would not have been the end if there was even a thin seppa at the front of the handle or perhaps even if the tip of the tang had a flat face and sharp corners and the wood was carved precisely to match it.
When blades are to be subjected to strong perpendicular forces (edge first), the mekugi and fuchi or some sort of wrap are critical components for strength. With a longer and heavier blade than this one, the chopping test may have failed unless there was a mekugi and a ferrule, fuchi, or even a simple cord wrap near the front of the handle.
I am quite happy with the results of the test and am now more able to advise clients on their choices for handle construction. I am far more likely to recommend the traditional rice glue for working knives than I was in the past, but will usually require some additional component (seppa, lacquered cord wrap, fuchi) to add strength to the system.
Addendum: The remains of a handle core (quite possibly the very same one) recovered from the charcoal kiln, after being turned into charcoal, dumped out of the kiln, and bagged. The rice glue appears to be still somewhat intact several years later, though in the form of mostly carbon at this point…
Read about making sokui here: Making Sokui (Rice Paste Glue)