Chasing Transcendence

Several years ago I became aware of the work of Kintaro Yazawa, who, according to a 1986 article in FWW (issue 61), had taught furniture maker Alan Peters about the twisted dovetail joint, or nejiri ari gata. The name of the joint in Japanese, I might add, terms what we call a ‘dovetail’ in the west an ‘ant-shape’. The ‘twisted ant-form’ joint then.

A while later, Yazawa had re-established himself in Japan as a specialist in joinery and makes small presentation boxes and pieces of furniture incorporating technically difficult or ‘impossible’ looking joints. A few years ago there was a second article in FWW by Yazawa illustrating a couple of his methods and this seems to be the point where a bunch more people, in North America at least, became aware of the craftsman and his marvelous work.

I’ve had pictures of a couple of his pieces on my computer, in a special folder, for several years now. Like Yazawa, I love joinery and also share a liking for puzzle joints of all kinds.

There are various basic forms of ‘puzzle joints’. Some are made to be puzzles which just so happen to be wooden, though they could be in other materials. The other is a form of wooden joint which is assembled by apparently impossible means.

Of the former type of puzzle, there are what are known as karakuri. These have long been known in the west, as examples were brought over from Japan since the late part of the 19th century. Here’s a very simple form of karakuri puzzle box:

There are boxes like this with hundreds of moves required to get them open – one that I’ve come across requires 1536 moves!

Another popular type of puzzle involves interlocking sticks – these are termed kumiki in Japanese, and ’burr puzzles’ in English-speaking countries. Here’s a typical example sitting on my desk:

In the above types of puzzles, while the mechanism by which the puzzle goes together may be mysterious at first – or for a long while, as the case may be – with puzzle joints in woodworking, the joints cannot often be readily separated, either because they are incorporated in a larger structure, or because they are glued together, or because the assembly of the joint is such that it will not permit dis-assembly. There are also a small number of joints in which dis-assembly is possible, but difficult, in which case they are not significantly different than the wooden puzzles above.

The key difference, it appears to me, is that puzzles are usually readily recognized as puzzles – like the burr above, or are quickly found to be puzzles, as one might readily discover when trying to open one of those puzzle boxes. You know it’s a box, and there might be something inside, but how does it open?

With wooden joints which are puzzles, I would say that unless one is a woodworker who does joinery in solid wood (a minority of woodworkers) and knew how such joinery goes together, one likely wouldn’t notice anything odd about a puzzle joint. It’s the ultimate in stealth I suppose, however the message, as such, is more or less lost on most who see the construction.

A great case in point would be the Ōtemon, a gate at Ōsaka Castle:

Here’s a view from the inside looking out:

Both gateposts have unusual ‘rot joints’ – these are spliced in replacements for the bottom few feet of a posts that have rotted out.

A closer look at the joint on the post seen to the left in the above picture:

Here’s one side:

And the other:

Ignoring the other stuff going on at the connection, such as the missing plug part way into the front dovetail (to cover a metal bolt anchoring the joint to a lateral tie to the rear), and the end of a diagonal brace sticking out above the joint, we see a splice in which two opposing faces are dovetails, and two opposing faces are like a sword tip, or mountain peak if you prefer. If your response is ‘so what?’ perhaps take a moment to consider how such a connection might assemble in wood….

I would guess several million people walk past that gatepost every year, but only a tiny fraction notice anything unusual about that joint. For carpentry geeks of all stripes however, this is a delight! The Ōtemon-splice is famous in Japanese carpentry, and also goes by the name of basara-tsugi. Lots of Japanese carpenters know about it and have incorporated it into their work. Here’s an example of a freshly-installed one:

I just finished cutting one for a client, as on of several joinery models I am making. I might add that there are many other forms of rot-splice joints, and several of them incorporate a ‘how’d they do that?’ aspect. I’ll save that for another blog entry some day. I do detail the mechanism of the basara joint in the Volume III essay of the TAJCD series, BTW.

Getting back to Yazawa, some of his joints are mysterious from the aspect of being incredibly hard to execute, like this one:

The assembled joint:


Other joints he makes however feature a means of assembly which is, to put it mildly, hard to determine. Case in point is this box, which employs what Yazawa calls a ‘key hall’ joint on the Japanese portion of his site (I’m sure he means ‘key-hole’ joint):

Another view of the corner of the above box:

A view from the inside reveals nothing special, which is itself special in this case:

That joint is but a warm up of sorts, however, for this one, a joint that Yazawa terms the ‘transcendence joint’:

A look at an ‘impossible’ corner:

Please be clear that all these joints are done in solid wood without recourse to veneering or gluing pieces together. I’ve read some online discussions where people put forth theories involving such ideas, and they are false.

I spent some time working on the joint for the ‘key-hole’ joint above, thinking it was an easier puzzle to crack than the transcendence joint, and thinking they may have some commonalities in their mechanism.

I emailed Yazawa several years back asking him for a hint, and he replied that the secrets were going with him to the grave. Okay, well, then, I see how it’s going to be….

So, I left the puzzle on the digital ‘shelf’ of my computer, occasionally bringing it down to look at it again. Recently I got a bit more fired up about it and managed to solve the key-hole joint, working at it with a 3D drawing. I figured out an initial solution involving pivoting moves. I shared my drawings with Yazawa, and he gave me all sorts of encouragement (indicating that I had it wrong, but keep trying). No hints of course, or even a nod that I might be on the right track, which I don’t mind. I don’t want the solution simply handed to me on a platter. When I found a second solution to the problem, and the only solution (probably) which can work for a 4 sided box, I sent him those drawings. He hasn’t replied and it has been a week, so not quite sure what to make of that. Maybe he’s annoyed. I won’t presume to reveal the secret of that joint here – if it intrigues you, put some time in and try and solve it yourself.

The transcendence joint is next in my sights, and it is proving to be quite tough to solve so far. I’ll keep plugging away at it however and hopefully will also transcend the problem at some point myself.

I think that the study of karakuri puzzle mechanisms of all kinds will be very helpful for solving such mysterious joints, and for creating new ones.

Thanks for coming by the Carpentry Way.

24 thoughts on “Chasing Transcendence

  1. I'll go out on a limb to say that Yazawa's joints appear to be possible (though clearly created with the utmost skill) with the same angle of assembly as the twisted dovetail. The minimum requirement is that all mating surfaces share an axis of projection, which is also the axis of assembly.

    I had fun creating my own take on an impossible dovetail joint. I puzzled that one out in sketchup before making the model.

    Joinery tricks are a little like magic tricks. There's fun in the riddle, concealing certain details for others to speculate over and solve.
    -Mike S.

  2. Dang! (in reference to the key hole & transcendence joint) I'm not sure I want to aspire to get that good. I do love the diamond finger joint with mitered ends. That I think think is within my ability to try & master someday.

    Wow, I'm still aw'ed at the transcendence joint.

    Thank you very much for this post!

  3. Hi Mike,

    thanks for the comment and for sharing your work. In the case of the key hole and transcendence joints, I don't believe the angle of assembly is the same as the twisted dovetail. I initially considered the twisted dovetail as the mechanism type for those joints, but if you really look at them, it doesn't seem likely, especially given the inside faces of the box are clean. You've very welcome to prove me wrong with a SketchUp drawing however! The assembly I figured out moves together differently than the diagonal seen with the twisted dovetail.

    Like you, I enjoy the riddle of trick joints.


  4. Chris,

    thanks for your comment. If you think that key hole joint goes together at a 45˚ angle, well, as for the comment above from Mike (who is saying much the same thing), prove it! I think you'll find it cannot assemble that way – how would the fingers on one side get past the bulbs on the other?


  5. I love that you chose to show the work of Yawaza. I have his work as my screne saver at work. I have tried to explain why his work is so exciting, but my expainations seem to fall on confused and disinterested ears. I show the work of others (including yours) but if people have never tried making something like this, they don't even know what to look for. Many see dove-tails as the “ultimate show of woodworking skill”. I like well cut dovetails and even simple box joints can be impressive when done right, but work like this is almost only for woodworkers.

  6. Hi Paul,

    very nice to receive your comment. I well understand the story you relate about non-woodworkers and their uncomprehending stares when you try and share something like this with them. I have been on the other end of that mind you, as when a friend of mine who is really into jazz trying to share with me pieces he finds impressive, and I really didn't get it either!


  7. Chris,
    I can visual in my mind that the key hole joint could be formed on a radius which would take the inside of the keyhole inside of the joint. That would produce a clean inside corner. Without actually cutting this in wood right now I can't quiet get my head around the difficulty of cutting this in wood. What do you think about this theory?

    Jack Ervin

  8. Jack,

    thanks for your comment. If I understand you correctly…

    If you believe the keyhole rotates into position on a pivot point, then you are taking the same approach I did initially. While I found a way for the joint to assemble by pivoting (and requiring 2 pivot points at that), it is not possible to put a 4-cornered box together by such means unless the joints on different pairs of opposed corners pivot in opposite directions. And yes, a pivoting joint would be tough to cut.


  9. After posting the above and ponder the issues I started to rethinking the assembly and now think that the assembly is in fact at 45 degrees but not coming into the corner but across the corner. The keys would not be round in cross section to the assembly path but elliptical in cross section in a direction that would form a circle at 45 degrees to the assembly direction. What think you on this?

    Jack Ervin

  10. Jack,

    what do I think? Well, I think I'd need to see a drawing or a model to show what you are proposing, and you may wish to re-consider elliptical cross sections for the pins – look closely at Yazawa's joint – do the 'bulbs' of the joint look round?

    Have fun!


  11. Hello Chris

    I am spending my time during holidays as a woodworker making staircase in home,where I must connect two parts of vertical post-ash (3m and 2m parts should be connected together what makes approx.5 m high post).The way how joint works is similar as in Your picture of temple gate post.Joint must be as rigid as possible to carry some side loads.Post has square cross-section 130*130 mm. Do You have any suggestion (simple drawing or something, measures I can calculate myself). I have been looking in internet already many days but still no success. This joint should look beatiful also as it is exposed-therefore no mistakes are allowed.

    Best regards

  12. Priit,

    hello again. For the application you describe, the Otemon gate post joint would not be a good choice as it is not especially rigid in resisting bending loads. The type of splice selected for your job needs to reflect the length of wood available for the splice and the assembly direction. Without knowing those particulars I can't make a suggestion.

    I can suggest however that you consider purchasing the TAJCD Volume III essay on splicing joints which covers 288 pages. You won't find a more detailed look at splicing joints in any language. Excuse the shameless plug for my book!


  13. re the key hall joint i was drawing it in skethup when lighting stuck my brain. i think he put it together same as and box joint. only differance is what he did before putting the joints together. i thinking that this can be done by taking the cut ends, soaking them in water then compressing the rounds flat. letting that dry then putting the joint together and resoaking the completed joint/box to get the rounds to expand bact to original round shape. if i get time this next week ill try it out.

  14. Not reading the comments, I will just shot here:
    The walls change angles forming a diamond shape; oposite walls move in oposite inner-outer directions; opposite walls slide vertically together.
    This is what makes some sense intuitively, but I don't grasp the necessary directions but for the smaller walls sliding downwards.
    Is my intuition in the right direction?

  15. Bruna,

    you just went through something very common for a lot of folks looking at these joinery puzzles. At first you think you have it figured and then with a little more thought and you realize you don't…

  16. Intriguing. I would think that the joints are put together following an arc. So the groove for the bulb starts near the opposite shoulder, a millimeter or so in to allow for the straight finger appearance and follows a curve like path to its final position.

  17. Intriguing. I believe the keyhole joint is put together following an arc. So the groove for the bulb starts at the opposite shoulder set a few mm in to create the straight finger appearance and follows a rounded groove to its final position. The course of the arc corresponds to both sets of bulbs on all the joints and when assembled simultaneously it slowly slides together.
    The other one which looks ridiculously hard, at 1st I want to believe it may be just slid in at a 90° angle with one set of the flowers flying through and the other cut over the surface, but then the ones cut over the surface would be quite weak. Maybe it follows the same arc pattern of the keyhole.. I think the jury is still out on this one..

  18. Yes Tom,

    I also thought the curved assembly path might be the answer, but how to cut it out? I asked him if that was the answer and he said not. I also worked out a V-shaped assembly direction, also tough to cut out, but he said that wasn't it either.


Leave a Reply