The first time you see a Japanese bow, two things stand out: it is very long, and the hand does not hold it in the middle. The kyudo yumi has a base length of about 220 centimetres and the grip sits below the centre, so the upper limb is much longer than the lower one. It is made of bamboo and wood glued in layers, unstrung it bends the opposite way to the shape it takes once strung, and the hand that holds it does more than push: through tenouchi it twists the grip. To anyone coming from the Olympic recurve it looks like a bow built against everything they know, and the question follows naturally: how does it work, and why is it asymmetric?
Tradition has its own answers, handed down from workshop to workshop and from master to pupil. Since the late 1990s a small group of Japanese laboratories in mechanical engineering and materials science has been trying to turn them into measurements, using bending rigs, high-speed cameras, strain gauges and computer models. What comes out is interesting but fragile: almost every result comes from a single bow, a single workshop or a model only partly checked, and none of the registry claims this page rests on goes beyond reported. Two stand on calculations alone.
How a yumi is made
The fullest description of the materials comes from a 2017 review written by a group that works with a Kyoto workshop, that of a master bowyer who makes kyudo bows. The bow's cross-section has three layers: bamboo on the outer face, bamboo on the inner face and, between them, a core of bamboo and wood strips. The grip sits below the middle so that the upper limb deforms much more than the lower, at least as the bowyer intends it. The group turned into numbers three choices the bowyer makes by eye and by hand: the bamboo for the two faces, the core woods and the glue.
The bamboo on the two faces
The researchers tested in bending the bamboo that the master had already set aside for the two faces of a bow. Over a short span of 80 millimetres the two pieces could not be told apart; over a span of 400 millimetres the piece for the inner face showed a higher elastic modulus, which means it was stiffer. The cross-section explains the difference: hard fibre made up 18% of the inner piece and 9% of the outer one, and the modulus rose in proportion to the fibre. A vibration test pointed the same way: the inner piece vibrated with a shorter period, while damping was the same in both, so the difference lay in stiffness rather than in how the material dissipates vibration. Of each cane, about six centimetres across, only the layer near the skin is used, where the fibre is densest, planed down to a 4.5 millimetre strip.
The registry rates the claim that in the Japanese bow the bamboo chosen for the inner face is stiffer than that for the outer as reported, and the limit is plain to see: one bowyer, one batch of bamboo, and modulus values published only in graphs. The measurement shows that this master, for this bow, put the more fibrous bamboo on the inside; whether that is a rule of the craft or his own choice, it cannot say.
The core woods and the bowyer's words
The same group also put the master's vocabulary to the test. To describe the core materials the bowyer uses inherited words, hard, brittle, tough, and on that basis decides how to build up the bow. The researchers loaded 5 millimetre specimens of three materials, higo, hazenoki and maple, in three-point bending over an 80 millimetre span at 50 millimetres per minute. Hazenoki rose linearly until it broke at a very small strain, like a brittle wood, and maple reached larger strains at failure, like a tough one: on these two the curves bore the bowyer out. Not on higo, though, which the master describes as hard and brittle like hazenoki, while in testing it reached large strains, like maple.
The claim that what the bowyer says about the core materials of the Japanese bow only partly holds up in mechanical testing is reported. It is a single test, with one curve per material and no stated repeats, and it sets words that are not measurable quantities against curves that are. Its value lies mainly in the method, because it shows that workshop knowledge can be translated into laboratory tests and that the translation does not always match the original.
Animal glue
The third material is the glue. Traditional bows are glued with deer-hide glue, which workshops today often replace with synthetic glues, and the question is whether the difference reaches the arrow. The Kyoto group filmed kyudo archers at high speed shooting one bow made with animal glue and one with synthetic glue, and worked out the arrow's speed. Peak speed, for each shooter, was almost the same with both bows. What changed was how it got there: with the animal glue the mean acceleration, calculated in hundredth-of-a-second intervals, rose sharply from a hundredth of a second after release and from then on stayed above that of the synthetic bow, which showed no such surge, and the arrow reached the other bow's speed 27 milliseconds after release.
The authors conclude that the animal-glue bow gives more push just before the arrow leaves the string, and that is how the registry claim is worded: a Japanese bow glued with traditional animal glue pushes the arrow harder in the final stretch, at the same peak speed. Its status is reported, with a caveat that weighs more than usual here: the study does not say how many shooters or how many shots, and it says nothing about the bows beyond the glue. It is an observation whose basis is unknown, and it concerns the Japanese bow alone.
The two limbs of an asymmetric bow
Now to the asymmetry. On a standard-length yumi the section above the grip is about twice the length of the section below, and bowyers build it that way so that the longer upper limb deforms more than the lower. With a tool this unbalanced, kyudo tradition asks the bow hand for a technique of its own to send the arrow where it is aimed. Two bench tests have looked at how the two limbs deform, each on a single bow.
The first, in 2013, took a carbon-fibre yumi to about 700 millimetres of draw, with a horizontal force of 300 newtons, roughly 31 kilograms. The upper tip moved about 240 millimetres and the lower about 180. A model of rigid bodies joined by torsion springs reproduced the upper limb well, 230 millimetres calculated against 240 measured, but not the lower, 220 against 180. After release the two tips oscillated with amplitudes of the same order, despite starting from different deflections, with the upper one lagging in phase, and the motion died away within about three seconds.
The second test, in 2019, used a commercial 4.2 kg shihan-yumi, 164 centimetres long with a 157 centimetre string, divided into nine sections with measured curvature and stiffness and loaded with weights over a pulley. At the tips it found the opposite: the long limb rotated and bent along its whole length, and its tip moved relatively little, while the short limb rotated with little bending and its tip moved more. The tips also did almost all their moving in the first part of the draw, and after that moved little and almost only horizontally. Measurements and model agreed well.
The two tests therefore agree on the main point, that the two limbs do not deform in the same way, and disagree on which tip moves more. One possible explanation lies in the bows, since the 2019 bow had a ratio between the two sections of about 6 to 4, less asymmetric than a standard yumi. That is why the claim that on the asymmetric Japanese bow the two limbs deform differently is reported rather than corroborated: two different bows, two different rigs, and no repeat on the same kind of bow.
The draw force curve
For the Kyoto master the decisive moment in making a bow is when it is bound with cords and wedges and given its curvature. Whether curvature also decides how the bow draws is what the authors of the 2019 bench test examined by calculation. Two years earlier they had treated the bow as a flexible beam under large deformation and checked the model on a PVC strip 300 millimetres long, with strings between 210 and 450 millimetres: deflection at the nocking point grew quickly at first and then almost stopped, even as the force increased. On the nine-section shihan-yumi the calculated force curve came in three stages: it rose quickly at first, slowed in the middle and climbed again towards the end, and the pattern grew stronger the more the central sections curved outwards.
The word that matters is calculated. On the real bow the check covered its shape under load, which the model reproduced well, while the force curve came from the calculation and not from a dynamometer. That is why the claim that in the Japanese bow draw force does not rise steadily through the draw is transferred: it rests on models, and nobody has yet measured the full curve on a real yumi.
The direct measurements are peripheral: the 2013 carbon yumi gives a single point, 300 newtons at 700 millimetres of draw. The comparison with Western bows is clearer: three replicas, a recurve horse bow, a flatbow and an English longbow, taken on a dynamometer to 50 centimetres, give almost straight curves, and that is the basis on which the registry states that on the longbow, the flatbow and the horse bow, draw force rises almost in a straight line up to 50 centimetres of draw. If the models are right, the yumi behaves differently.
Curvature at rest and draw weight
The same model gives a second answer, about how heavy the bow is to draw: changing the curvature the sections have at rest in the model, the load on the string at equal draw came out almost proportional to the curvature, with an effect that faded as the draw went on. Not all curvature counts in the same direction, though. Increasing the curvature of the three central sections raised the load, while bending the two sections between the grip and the tips further inwards lowered it. In the model, then, draw weight depends on which part of the limb is curved and in which direction.
The claim that the more a Japanese bow's limbs curve outwards at rest, the heavier it is to draw is also transferred, for the same reason: it is a calculation, checked statically against a commercial bow, and nobody has compared on a dynamometer real bows that differ only in their curvature.
Tenouchi: the hand that twists the bow
The last part concerns the archer. Kyudo teaches that the bow hand must twist the grip through tenouchi, because without that rotation the arrow leaves to the side, and that getting the amount right is one of the hardest things to learn. The first work on the subject is a 1997 model, calibrated on a launching rig, in which three twist profiles passed the same total rotation to the bow. In the model the timing of the twist matters, because changing only its initial value altered the trajectory completely, while the arrow's speed varied by less than 1%.
In 2008 a test rig took the archer out of the picture. Bows of 13 and 15 kg mounted on a mechanical launcher, a single duralumin arrow, a target at 28 m outdoors and a torque on the grip ranging from zero to about 14 newton metres: varying that alone moved the landing point from one side of the target to the other, from about minus 40 to plus 80 on the scale of the published graph. The relationship was not linear, and the authors infer from it that too little twist sends the arrow further off line than too much. That is their reading of data disturbed by crosswind, presented in a graph and without any statistics.
On this basis the claim that in kyudo the twist the hand gives the grip shifts where the arrow lands is reported. For a claim about equipment a bench test is direct evidence, but here it is a single test, outdoors, and there is a gap the graph does not show. Nineteen university practitioners, measured in 2007 with an instrumented bow while shooting eight arrows at 28 m, twisted the grip about the bow's long axis by about six tenths of a newton metre, between 0.57 and 0.61 in most of them. On the launcher the torque about the same axis reached more than twenty times that, so most of the tested curve lies outside what archers actually do; and in the study of the archers the point of impact of their shots was recorded but did not enter the analysis.
For the recurve the registry has a more general claim, twisting the bow with the hand moves the arrow on the target, also reported. There a model calculates that on a 166 N bow thirteen hundredths of a degree of difference from one shot to the next moves the arrow by more than one ring at 70 m, and there too nobody has brought together, on the same shots by an archer, the twist and the point of impact.
Where the research stops
Set side by side, the seven claims about the yumi share the same profile. Five are reported and two transferred; every measurement comes from one bow, one workshop or one rig, and the two parts closest to what the archer feels, the force curve and the effect of curvature, are still calculations. Nobody has measured the full force-draw curve on a real yumi, compared with a stated sample bows that differ only in curvature or only in glue, or recorded on the same shot the twist a kyudo archer applies and where the arrow lands.
What I take from it
For those who shoot the yumi these measurements do not change technique, and they do not set out to. They do, however, give words to things the practitioner feels in the hand. If a bow seems to stiffen at the end of the draw, the models say it may come from how the bowyer curved the central sections; if two bows open differently, the calculation points to the curvature of individual sections, not just the overall curve, as a plausible reason. These are model-based hypotheses, not rules for choosing a bow.
On tenouchi the measurements call for more caution. The rig shows that twisting the grip is enough on its own to move the arrow, and the 1997 model adds that the timing of the twist matters too; the archers measured, however, twist with a far smaller torque than the one tested on the launcher, and nobody has followed an archer's arrow alongside their twist. I draw a single point from it, the same one the models suggest for the bow hand on the recurve: more than the amount of twist, what counts is that it repeats from one arrow to the next, because on the rig it is a change in torque that moves the landing point. That is my reading of the models and the rig, not a result measured on kyudo archers.
On materials, finally, anyone choosing between a bamboo bow and a fibre one, or between animal and synthetic glue, will find more curiosity than guidance in this work. The difference in stiffness between the two bamboos and the late push of the animal glue were seen in a single workshop, the second over a number of shots we do not know. For recurve archers, these studies describe how a traditional bow is conceived, not how to choose your own.