How Japanese Masters Build Earthquake-Proof Joints Without Nails or Glue

You know what used to eat at me more than almost anything else in the shop? I'd put real weeks into a structural piece. A long scarf joint connecting two heavy beams or a mortise and tenon carrying genuine load in a timber frame. I'd cut it carefully, fit it, glue it up, drive a couple of hardwood pegs through, sometimes put a lag bolt in there, just to feel completely settled about the whole thing. And two or three years later, I'd walk past it and notice it had moved, just a hairline gap where there hadn't been one before. Just enough to tell me that whatever I'd done, it wasn't going to hold the way I'd wanted it to hold. And I'd stand there trying to figure out what I was missing, because I'd done everything the way I was taught. Then, about six years ago, I had the chance to spend a few days in Kyoto, Japan, watching a restoration crew work on a temple complex that had been standing in some form since the 12th century. These men were carefully disassembling joinery that was 800 years old. No glue anywhere, no nails, no metal fasteners of any kind, just wood fitted to wood and joints so precise that I genuinely could not work a single piece of paper into the seam. I tried. I stood there with a thin strip of paper trying to find a gap, and there wasn't one. When I asked the lead carpenter through a translator how those joints had survived 800 years of seasonal movement and humidity cycles and earthquakes, he thought about it quietly for a moment, and then he said, the joint gets stronger when it is loaded. That was his entire answer. And I spent the next three days trying to fully understand what he meant by it. Today, I'm going to walk you through the three joints at the heart of this tradition. What they are, how they're cut, what makes them behave the way they do under load, and what any of us can take back to our own shops from what these craftsmen figured out over more than a thousand years of building in earthquake country. And stay with me to the end, because I have a layout technique I picked up from watching that restoration crew that changed how I approach every single drawboard joint I've cut since I got home. Japanese temple joinery is not decorative craft. It is structural engineering solved entirely in wood, developed over more than a thousand years by working tradesmen who needed their buildings to survive in one of the most seismically active places on earth. These are not joints designed to look impressive in a museum. They are joints that have been holding up real buildings through real earthquakes for centuries without a single nail. That track record is the only argument they need. Japan sits on the Pacific Ring of Fire, and the history of major seismic events there stretches back as far as written records go. The carpenters building large timber structures from the 7th and 8th centuries onward understood a problem that took structural engineering in the western world considerably longer to fully appreciate. A rigid structure fights lateral movement until it reaches its limit, and then it fails, often catastrophically. A structure built to flex to absorb that lateral energy and return to its original form survives the same event without losing its integrity. And for the Miyadaiku, which translates as shrine and temple carpenters, that engineered flexibility had to be achieved entirely in wood because no adhesive and no simple metal fastener can handle the kind of cyclical, multi-directional dynamic load a major earthquake produces repeatedly over the span of centuries. Some of the structures built using these methods are still standing today. Horyu-ji Temple in Nara, widely considered the oldest surviving wooden structure in the world, has been standing in some form since the late 7th century. Toshodai-ji Temple, also in Nara, dates to the 8th century. These are not preserved ruins. These are buildings that have been in continuous use, maintained using the same joinery principles they were built with through over a thousand years of seismic activity. The apprenticeship for a Miyadaiku traditionally lasted between 10 and 15 years before a craftsman was trusted to cut primary structural joints without direct supervision. That duration reflects the reality of what these joints demand. The tolerances are unforgiving, and the understanding of how each joint behaves under different load combinations takes years of accumulated physical experience to develop. When I was at that restoration site in Kyoto, the lead carpenter had been working his trade for over 40 years. I watched him fit a joint by feel as much as by eye, drawing his fingertips slowly along the mating faces, reading the wood through his hands, making adjustments so fine that I genuinely could not see what was changing between passes. But each reassembly was tighter than the one before it. The principle running underneath all of this is what translates roughly as working with the nature of wood rather than against it. Wood is not a static material after it leaves the mill. It swells in humid conditions, shrinks in dry ones, compresses under sustained load, relaxes when that load is removed. The Miyadaiku do not design to prevent that movement. They design joints that use it. Joints where the natural mechanical response of wood to load makes the connection tighter rather than looser. The Japanese term for this behavior is draw tightening, and it is the central organizing principle behind every joint I'm going to walk you through today. The first joint is the kanawasugi, which you can sound out as ka-na-wa, t-s-u-g. This is a double-wedge scarf joint, and its purpose is to join two timbers end-to-end to create a single, longer structural member. If you've ever needed to extend a beam, you know the fundamental challenge. A scarf joint has to transfer load across a connection that is inherently weaker than continuous timber. The conventional approach most of us reach for is a lapped scarf secured with a bolt, and the bolt carries most of the work. It holds, at least for a while. But over years of cyclical loading, the wood fibers around the bolt hole compress slightly. The bolt loses its clamping force incrementally, and the joint begins to work under dynamic load. It's a slow deterioration, but in structural work, it's the kind of thing that compounds over decades. The kanawasugi uses no fasteners at all. The two mating faces are cut with a series of interlocking steps and precisely angled shoulders, and two hardwood wedges are driven in from opposite sides of the joint after assembly. The geometry is designed so that downward load on the beam drives those wedges deeper into their seats. The weight the beam is carrying is what tightens the connection. The heavier the load, the tighter the joint becomes. That is draw tightening in its most direct and elegant expression. To cut this joint accurately, you begin with a marking gauge set to exactly one third of the beam's total thickness. That proportion is not a starting suggestion. The middle third of the joint is what carries the wedges, and the outer thirds form the locking shoulders. And if those proportions drift during layout, the geometry stops functioning as intended. You mark out both halves with the gauge and work the cuts with a ryoba, sounded out as ri-o-chi-ba, which is a Japanese double-edged pull saw with rip teeth on one side and crosscut teeth on the other. The crosscut side typically runs between 17 and 22 teeth per inch and cuts on the draw stroke, keeping the blade in tension throughout the cut. The kerf is roughly half the width of a standard western handsaw, and the surface quality off the blade is clean enough that very little chisel work is needed to bring the shoulders to final fit. When I compare the surface left by a sharp ryoba to what I get from my western tenon saw on the same cut, the ryoba wins every time in terms of surface quality. The wedges themselves are traditionally made from Japanese white oak, called shirakashi, tapered at a ratio of 1 in 12, meaning one unit of rise for every 12 units of length. That shallow taper is deliberate. A steep wedge will pop loose under vibration. A shallow wedge at 1 in 12 stays, seated under load, and gets driven tighter as the load increases. When I first cut this joint in American white oak, which is a very close substitute for shirakashi in terms of hardness and grain density, my first attempt was genuinely poor. I had gaps in the shoulder fits and the wedges were seating unevenly. My second attempt was better, but still not right. My third attempt was something I was proud of, and the moment I applied load to that joint and felt it tighten rather than deflect, I understood at a physical level what that carpenter in Kyoto had been trying to tell me. The second joint is the shachi-sen, pronounced sha-chi-sen, and it is a drawboard peg joint. That takes a technique many of us are already comfortable with and advances it in a direction most of us have never considered. Standard drawboring is familiar to a lot of timber framers. You drill a hole through the mortise wall, offset it slightly from the matching hole through the tenon, and drive a round peg through both. The offset forces the tenon tight against the shoulder as the peg tries to straighten itself through the misaligned holes. It works well, and western timber framers have relied on it for centuries. I've used it myself in a lot of projects. The shachi-sen version uses a peg that is tapered along its length and octagonal in cross-section, driven through a hole that is shaped to bear against four of the eight faces of the peg, while leaving the remaining four faces with deliberate clearance. What this geometry produces is a peg that wedges itself into compression from four directions simultaneously as it is driven home. It doesn't just pull the joint tight the way a round drawboard peg does. It locks itself into a compressive state that resists being worked loose under dynamic load, and the draw-tightening principle applies here just as it does in the kanawatsugi. Under lateral racking load, which is precisely the force an earthquake produces, the peg is driven harder into its seat rather than being worked free. The load that would destroy a conventional fastened connection makes this one stronger. The wood species for these pegs carries more importance than people usually expect. Traditional practice calls for Japanese white oak because of its hardness, its tight and consistent grain, and its resistance to compression across the grain. Domestic white oak is a sound substitute for North American shops. What you need to avoid is any species soft enough to crush under the bearing load the peg is carrying. A peg that compresses in place defeats the entire design because the joint depends on the peg developing and maintaining compressive force. Soft wood simply cannot do that. When I tried cutting the octagonal tapered peg in my own shop, the part that demanded the most patience was getting a consistent taper along the full length of the peg. I started with a draw knife to rough the shape down close, then used a shoulder plane to clean and flatten each of the eight faces. It took me four attempts before the taper was consistent enough to seat correctly through the full depth of the hole. When that fourth peg drove home and the joint pulled up tight in a way I had never felt from a round peg, I understood why this design has persisted through a thousand years of structural use. If you're finding this as valuable as I did when I first started digging into it, go ahead and subscribe to Dale Oakfield's Workshop so you don't miss what's coming. I put a lot of time into these and I'd love to have you along for the next one. All right, let's get to the third joint. The third joint is the ariotoshi, pronounced a-ri-e-o-to-shi, which translates roughly as a dropped dovetail. A sliding dovetail is something many woodworkers have cut before, typically at a router table or with a dado stack. The dovetail key slides into a matching socket and resists withdrawal along one axis. It's useful for fixed shelving, attaching cross members to case sides, and similar applications where you need mechanical resistance to pulling the joint apart. The ariotoshi takes that basic mechanical concept and modifies it to solve a specific problem that comes up constantly in large timber framing. When you're working on a structure where multiple beams are coming together, you often cannot slide a joint in from the end of a member because other timbers are already in position and the clearance simply does not exist. The ariotoshi is designed to be assembled by dropping the dovetail key straight down into its socket from above, rather than sliding it in laterally from an open end. The socket has a mechanical stop machined into its base and the key seats against that stop when it reaches full depth. Once seated, the joint resists withdrawal upward, resists lateral racking, and resists rotation, all without a single fastener and all from a vertical assembly motion that works even when three other beams are already in place around it. The dovetail angle used in this joint is one in six, meaning the face of the dovetail flares outward one unit for every six units of depth. That is a shallower angle than the one in eight you commonly encounter in furniture dovetail work and the reasoning behind it is structural rather than aesthetic. A steep dovetail angle in a timber housing generates significant splitting force across the grain of the surrounding wood when the joint is loaded laterally. The one in six angle provides enough mechanical resistance to prevent withdrawal under load while keeping the splitting forces within what the wood can carry without fracturing at the housing. It is not an arbitrary proportion. It is a proportion arrived at through centuries of observing what fails and what holds. I cut this joint in red oak and tested it against a simple half lap carrying an equivalent racking load. The Ariyotoshi held firm at a load level where the half lap was deflecting visibly and beginning to show stress at the glue line. What struck me was that the resistance in the Ariyotoshi increased slightly the longer I held the load on it which is the draw tightening signature. The geometry was responding to the load exactly the way it was designed to respond. Now let me share some patterns I've noticed over the years. Habits that most of us carry into the shop without ever stopping to examine them. I want to be clear that I am not pointing fingers at anyone here because I did every single one of these myself for longer than I care to admit. The first pattern is treating glue as the primary structural element of a joint. Most of us reach for glue automatically because it is fast it works and we were taught that a well-glued joint is stronger than the wood itself. For furniture that gets normal household use that is often true but glue fails under cyclical stress applied repeatedly over decades. It fails as wood moves through its seasonal cycle year after year for a century. The Miyadaiku consider the need for glue a sign that the joint design has not completed its job. The geometry is meant to carry the load. If it cannot do that without adhesive assistance the geometry needs to be reconsidered. The second pattern is using metal fasteners to cover a loose fit. A joint comes out slightly sloppy and a screw goes in to solve the problem. I understand the impulse completely because I've done it but what I didn't fully reckon with is that a metal fastener introduces a rigid point into a system that is designed to flex. Under repeated dynamic loading the wood moves and the metal does not and over enough cycles the wood compresses around the fastener the hole elongates and the connection that felt completely solid for the first several years begins to work loose. These Japanese joints have no such conflict because every element is wood and every element flexes together as a system. The third pattern is cutting joints to a loose fit on the assumption that glue will compensate. The draw tightening principle is entirely dependent on full bearing contact at the specific faces within the joint where the geometry is doing its work. A gap means the load is being transferred through a smaller contact area than the design intends which concentrates stress and accelerates wear at those faces. The precise fit is not perfectionism it is load bearing engineering. The fourth pattern is assuming that more fasteners means a stronger connection. I used to think that adding a second peg or a third bolt to a joint I wasn't confident about was making it more robust. What I understand now is that adding fasteners to a poorly fitted joint does not fix the joint it just distributes the failure differently. The connection is only as strong as the contact between the wood faces and no number of fasteners changes that fundamental reality. The fifth pattern is thinking about wood as a static material once it has been dimensioned and cut. We mill it, we build with it and we expect it to stay exactly where we put it indefinitely. The Miyadaiku think about cut wood as a material that is still responding to its environment and will continue responding for as long as it exists. They design every joint to accommodate that ongoing movement rather than resist it. That shift in thinking from stopping wood movement to designing for wood movement changed how I approach every structural project I have taken on since that week in Kyoto. Now for the bonus tip I mentioned at the very beginning. This one comes directly from watching the layout sequence at the restoration site and I have not found it described in any woodworking book I own western or Japanese. When those carpenters laid out the peg hole for a shachisen joint they did not mark the peg hole location at the same time they marked the mortise. What they did instead was assemble the joint completely dry tap it carefully to hand tight and then mark through the outside face of the mortise wall using a brad awl driving it gently but deliberately through the mortise wall and into the surface of the tenon below. That mark reflects the exact position of the hole given the actual fitted position of that specific tenon inside that specific mortise. Not a measurement transferred from a drawing not a calculated dimension based on nominal sizes. The actual position in the actual wood after cutting and fitting. They then disassembled the joint located the brad awl impression on the tenon face and shifted that mark precisely 1 16th of an inch toward the shoulder of the tenon before drilling. That offset is the drawbore. But because the starting position of that offset was taken from the real fitted condition of the joint rather than from a theoretical measurement that may have drifted during cutting the drawbore is working from reality. I started using this approach about two years ago. Before it my drawbore joints were competent but inconsistent. Some pulled up perfectly tight on the first drive of the peg. Others came up almost right close enough to live with but not what I was after. After switching to this layout method the consistency changed in a way I noticed on the very first joint I tried it on. The joint pulled up square and firm on the first peg drive and every joint I have cut this way since has done the same. It is a small adjustment in the layout sequence but the results are not small. It is the difference between a joint that performs reliably and one that performs most of the time. If any of this shifted how you think about what a joint is doing under load I would genuinely love to hear about it down in the comments. Tell me which of these methods you've tried. What worked the way you expected? What surprised you? Or what gave you trouble? And if you want to keep going deeper into woodworking traditions from people who have been refining these solutions for far longer than most of us have been alive hit subscribe and ring that bell so you don't miss the next one. I appreciate you giving me your time today. It means a lot. I'll see you in the next one.