there’s an elephant in the room. a sticky yellow one. i know you’ve seen it. the video about endgrain and glue. and i have lost count of the number of times someone has mentioned it to me and asked if this has changed the way i think about furniture and glue and removed any of my disdain for butt-joints and the people who think they’re acceptable in fine-woodworking. and honestly the answer is no. while some of you may have been shocked by the results of the “myth-busting” about endgrain gluing, i wasn’t. i already knew those things. and none of those are the reason i think you should avoid gluing endgrain. by the way, i don’t think the video was bad or that the test was inaccurate. i think it was seriously overhyped and sensationalized but a content creator trying to make a living can be excused for trying to go viral, especially if they’re successful about it. my only problem is that what it’s managed to do is confuse a lot of beginners and probably lead to some seriously-structurally-unsound furniture being built that would otherwise not have been attempted. if it’s grabbed some people’s attention, though, and attracted them to the craft, perhaps it’s all good in the end. but let’s take a look at some of the things that are relevant to glue that i can address in a more thoughtful form than is possible in a few minutes of video — again, not a criticism of the creator, just a natural limitation of modern attention-spans.
- mechanical joints and glue joints
- glue strength in general (how to pick a glue)
- how endgrain performs when glued
- wood-movement and glue joints
- water (and why it’s relevant to endgrain, glue, life, the universe and everything)
a trip down memory lane
first, though, let’s talk a little about where we came from. no, not the internet. glue. pre-humans and early homo-sapiens actually pioneered the use of glue before the use of anything we’d think of as functional language. they took tree-sap and used it to — you’re going to laugh at this — stick two pieces of wood together. i know, right? some things never change. almost a hundred-thousand years ago, stone-age laborers began making more complex glues by mixing tree-resin (not far from what we now call epoxy, practically-speaking) with iron oxide (rust — they’re literally making thickened-epoxy like what you use to build a boat) to adhere tool components like handles to axes. of course, the axe-head was made of stone but if you’ve ever knocked a wedge into an axe handle through the eye with some glue on it, you’re following in a tradition that’s probably older than you could possibly have imagined.
in more recent history, from the dawn of written language in ancient china and mesopotamia, spreading through india and egypt, clay and wooden daily-use objects were frequently repaired (and still survive at least as fragments) using bitumen — what we now call simple cement — and this was used for everything from fixing a broken bowl to attaching jewels to statues of great rulers in their eternal resting places. hot glue appeared a little later — you’ve probably heard of “pitch”, which is really just organic tar extracted from a tree using heat and applied to a surface. it’s surprisingly strong and we’re really only up to about 3500bce. by the time of the new kingdom (about 1500bce), egyptians were turning to more lively glue sources, turning animals into stickiness and producing what we now call “hide-glue”, which isn’t really any stronger than the thickened resins but is significantly easier to work with. they even used it for wood lamination — seriously, there are glue joints on king tut’s tomb and they’re done with hide-glue.
the greeks (and later the thieving romans) developed techniques we now think of as veneering and marquetry using egg-based glues and grain adhesives while the mongols conquered most of the known world using weapons mixing horn, bone, wood and metals adhered with hide-glue. it didn’t really matter which protein — they all worked. the point was to take a complex organic strand and heat it so it would liquify then polymerize as a solid resin bond on two sides of a surface. by the nineteenth century, grain-based glues were being used for things like stamps and hot rubber was being used to make something far closer to what we now think of as contact cement and urethane glues (yes, the nineteenth century — no gorillas we harmed in the production of this adhesion). in the twentieth century, though, major advances happened mostly because of military need for simple glues that worked without heat, mixing powder or preserving starches (hide glue and grain glue). rubber was expensive and plastic was very complicated to make — it still is, reflected in the price of a bottle of structural epoxy.
so they turned to water-based glues. just after the first-world-war ended, pva was born (in germany — where else?) to provide adhesion in a premixed, long-lasting glue that simply worked. it was the first time you could just pour something on, no preparation required, let it dry and have a serious hold. the world had shifted and most people didn’t notice. like many things developed in germany at the time, it didn’t work. but it was a great idea. over the next couple of decades, it was refined in japan and turned into a usable product (anyone out there in the japanese woodworking community who doesn’t want to use pva glue because it’s “not traditionally japanese” needs to stop talking and sit down because it was, practically-speaking, turned into a practical and marketable product there for the first time). despite being destroyed literally for fun in a war that should never have happened and was plagued by mistakes, kyoto imperial university persevered and, a few years after the occupation began, managed to get “kuralon” on the market — universities don’t usually get loans to make industrial products but desperate times… by the seventies, oil shortages and trade crises had shifted industrial demand from petroleum-based adhesives to something that could be produced very cheaply as a water-based alternatives. the pva-glue revolution had finally arrived.
now we’re here at a point where “woodworking” is synonymous with yellow glue. it feels like it’s been around forever but it’s realistically only been on the scene during my lifetime. and the controversies have been raging from the beginning — is it strong enough?
well, i don’t think it matters. i mean, it is. but it’s not the right question. glue test after glue test and what have we discovered? it’s strong as all fuck. but what does that really mean in-practice?
mechanical joints and glue joints
i think the largest issue with the debate about glue is what people are trying to do with it. they’re using the glue to stick two pieces of wood together. sometimes this is a good idea but it’s mostly just a shitty premise to start from. you don’t want to stick two pieces of wood together with glue. you want to stick two pieces of wood together with mechanical joinery and the glue is backup. only backup. it’s nice that it’s strong. but it shouldn’t be the cause of the strength in your furniture, just a way to hold everything tight because the mechanical joints aren’t always perfect. think of glue as wearing your seatbelt. you’re already going to try to drive to avoid the accident. you’re not going to speed. you’re staying in your lane. you’re driving on safe roads and paying attention. the seatbelt only has to work if everything else in your life goes completely to shit. that’s glue. it’s the shit-avoidance backup plan. you don’t put on a blindfold and get on the highway and say “no, i’m ok. i’ve got my seatbelt on”. relying on glue is a little like that.
where should you use glue?
glue as a single-source adhesive between wood has a very specific purpose. it holds boards together where there is a significant amount of non-moving surface-to-surface contact. note i haven’t said what direction the grain is moving here. it doesn’t matter. let’s say that again for the people in the back who want to have a fight about this. it. doesn’t. matter. the point is that it has to be wood that isn’t going to move differently and the surface area has to be proportionally-massive. so let’s think of some examples where this is relevant.
if you have to make a tabletop, it’s unlikely you’re going to find a meter-wide board. three-meters long is probably not an issue but unless you’ve got trees deserving places in scripture they’re not going to be that kind of width. yes, you might be a slab that’s approaching that kind of width but most people don’t make their dining-tables in the style of thor’s drinking-house or a viking party tent. they use lumber from the yard and that’s probably not going to be much beyond about 40-50cm wide. and that’s a narrow tabletop. so you joint the edges and glue them together.
no, you don’t need biscuits or dowels or (gasp) dominos or metal fasteners or anything. glue will work just fine for this. pva glue. as long as you squeeze it like you’re a dog and it’s the most exciting leg in your life. why? because it’s a huge amount of surface-to-surface contact and the wood (this is important) isn’t moving on one board relative to the other. it’s not expanding and contracting. it’s not doing anything. the two pieces could have actually lived that way inside the tree and the tree wouldn’t have cracked. so this is totally ok. the same goes for case-sides and other edge-laminated components.
what’s the other place we usually see this? face-laminations. whether you’re making plywood (you should definitely make plywood cause it’s all the awesomes) or bent-laminations or simply trying to make a thicker board, the glue surface is huge. it’s the whole surface. all of it. again, this isn’t because it’s face-to-face. it’s because it’s so much surface the glue isn’t going anywhere. starting to see a pattern?
i can hear the question coming, though. wood movement happens (mostly) across the grain. if you rotate your plywood laminations 90-degrees, you have crossgrain gluing, which is completely against the main guideline i just mentioned. and you’re absolutely right. but there’s a reason this works for plywood and doesn’t work for other types of lamination — proportional strength to potential absorption. if you have a thin piece of wood (let’s say 3mm or even less), there’s a very small amount of potential water it can take in. will it try to expand with moisture? yes. but that wood isn’t going to take in much and the expansion will be fairly small, even on a wide panel. it’s still going to try to move. but the sheer quantity of glue compared to the tiny amount of pressure that water can exert, though, will mean the wood simply can’t move. the next piece, which isn’t going to expand in that direction, is glue solidly to it and keeps it from expanding at all. the first piece locks the second down in return from moving on the other linear axis. the key here is the amount of surface area compared to the amount of potential expansion pressure. try this with 12mm-thick boards and you’ll rip the glue apart. but thin veneers just don’t have the strength to break the glue bond and this is where the power of plywood and its stability come from. yes, you could in theory do this with thicker panels if you used strong-enough adhesives. but you’d have to pretty-much encase them in liquid plastic to get that strength and it’s unfeasible. not to mention industrial players have no interest in making this possible because nobody in commercial manufacturing needs that kind of technology — plywood already works so why fix a problem nobody’s got by producing something extremely complex that nobody needs?
let’s review. if you’re going to use glue as a way to join pieces of wood together without functional joinery, it’s either edge-to-edge or face-to-face. this isn’t because endgrain-to-endgrain gluing doesn’t work. it’s because it’s irrelevant. yes, it’ll work. but you don’t need it to. it’s a proportional thing. don’t believe me? let’s look at some math.
the silly thing people try to do is adhere two long boards together using glue on the endgrain without a mechanical joint. let’s compare the three situations using common dimensional lumber. i’m assuming these pieces are all well-prepared, jointed and smoothed on all faces. none of that is relevant to the equation in practice, despite what you might have seen in a video. yes, they’ll make it work better. but they won’t change the comparison.
i’m going to give two examples, one large and the other small. let’s imagine attaching two identical boards together in three different ways — two 50x250mm boards (in america, this is a 2×10) and two 25x75mm boards (1x3s). i’ll treat the first as 1.2m (about 4’) and the second as 600mm (about 2’). i’m going to use the named dimensions but you could do the same calculations with nominal yields or planed dimensions and get much the same results. they’d just be harder numbers to talk about so this is easier to understand as a concept.
50mm x 250mm x 1.2m (2” x 10” x 4’)
the total volume of these boards is 30 000 000 cubic millimeters. their total surface area is 1 490 000 square millimeters. these are big numbers but you’ll see why it’s important in a minute. just think of it as thirty-million and one-and-a-half-million for simplicity’s sake.
gluing these face-to-face will give you a total glue surface area of 600 000 square millimeters (the large face of each board). gluing them edge-to-edge will give you 120 000 square millimeters (the long edge of each board). end-to-end, that surface shrinks to 25 000.
surface-area to volume looks like this… (and yes i know i’m comparing dissimilar units but the ratio doesn’t have to be a measurement unit because our comparison is unit-neutral as they’re the same.)
- face-to-face = 2% (two percent)
- edge-to-edge = .4% (four tenths of a percent)
- end-to-end = .08% (eight hundredths of a percent)
what does this mean? the glue bond, all other things being equal (which they’re not) between edges is five times better and that on the faces is twenty-five times better than end-to-end.
let’s look at it when we think in terms of surfaces, which is less significant but demonstrates a similar pattern…
- face-to-face = 40% (forty percent)
- edge-to-edge = 8% (eight percent)
- end-to-end = 1.7% (one and seven-tenths of a percent)
again, the comparison looks like about five times better for edge and twenty-five times better for face gluing. i know what you’re thinking, though. why does this matter?
well, let’s think in terms of torsional strength. if you’re going to look at the forces acting on the boards after they’re glued together, you need to think of them working like opposing levers. if you have two boards glued face-to-face, you have a very small amount of leverage — limited by the thickness of the boards. edge-to-edge, you can apply more levering force but it’s still quite limited — limited by the width of the boards. for end-to-end joints, though, you can go all the way to the extreme points on the full lengths of each board and apply far more pressure. it will break. think of it like a children’s see-saw. the glue joint is the fulcrum in the middle. if you put your board on there along the glue line, the face glue-up sticks out only 50mm on each side. the edge glue-up sticks out 250mm on each side. the end glue-up sticks out 1.2m on each side and even those children pushing down with minimal pressure will make those boards crack at the seam — unless they’re really weak boards and they’re going to crack anyway, which is certainly possible but we’re assuming for the purposes of this example you’re using lumber strong enough for the application and just trying to make sure the glue will hold them together.
25mm x 75mm x 600mm (1” x 3” x 2’)
you might think that was an unreasonable example. you don’t want to glue two pieces of construction lumber together. you want something much smaller. like the size of a typical small table stretcher — 25x75x600, for example. that’s about the size you’d use to support a tabletop on a side-table. let’s do the same calculations and see how it works out.
- total volume = 1 125 000 cubic millimeters
- total surface area = 123 750 square millimeters
- face-to-face glue surface = 90 000 square millimeters
- edge-to-edge glue surface = 30 000 square millimeters
- end-to-end glue surface = 1 875 square millimeters
(these calculations were done per board rather than total, by the way, differently from the ones above to avoid anyone telling me one way shows different results — the two methods have the same proportional difference. i love math.)
- face-to-face (volume) = 8% (eight percent)
- edge-to-edge (volume) = 2.7% (two and seven-tenths of a percent)
- end-to-end (volume) = .17% (seventeen-hundredths of a percent)
- face-to-face (surface) = 73% (seventy-three percent)
- edge-to-edge (surface) = 24% (twenty-four percent)
- end-to-end (surface) = 1.5% (one-and-a-half percent)
in comparison, things look very familiar but more extreme — in the volume calculations, edge is sixteen times as high as end with face being three times as high again. in surface calculations, it’s also a factor of sixteen between end and edge and three between edge and face. so the result when using smaller boards is even more significant — going smaller doesn’t mitigate the force-balance problem. it actually exacerbates it. but you already knew this. glue two popsicle-sticks together end-to-end and you might as well have attached them with gift-wrapping tape for all the strength it has but glue their faces together, even with white-glue, and you’ll have a hard time ever getting them to shift. it’s not the strength of the glue joint. it’s the strength of the sheer quantity of surface-to-surface contact the glue is allowing compared to the leverage you can put on it to get them apart.
where shouldn’t you use glue?
realistically, everywhere else.
what do i mean by that? as you’ve seen, there’s simply no strength end-gluing boards together. it’s not because end-grain doesn’t glue nicely. it’s because of how lumber is cut. if it was a huge end-grain surface and the board was only a few millimeters long, the glue would work just fine. let’s take an example of where that could happen — it doesn’t because it’s a silly example but if you wanted you could definitely do it.
take a thick board (perhaps a 150x150mm or 6×6” beam) and slice 3mm (1/8”) veneers from it — all endgrain. now laminate them together to form … oh. now you see the issue here. you’re just gluing the beam back together and the result is actually a series of very strong glue joints, end-to-end. but why would you bother? we don’t avoid endgrain glue-ups in situations where it is functional because there aren’t situations where it can work. it’s because those situations never show up in our actual construction methods. you just wouldn’t cut the board that way because it’s silly and wasteful.
so let’s say you want to join boards together. you could use a butt-joint (it’s not called that for this particular reason — butt is just the old word for “end” — but i like to think of it as the joint used when your thinking occurs in that particular piece of the anatomy instead of the head) but what are the other options?
- mortise-and-tenon
- wedged mortise-and-tenon
- drawbored or pinned mortise-and-tenon (are you seeing a pattern here?)
- dovetails (or even pinned dovetails)
- dowels or floating-tenons (the deified domino or the manual version — it’s the same joint)
- pinned rabbets
there are lots of others but these are the most common. by the way, i’m a japanese-style woodworker. these joints have slightly different shapes and vastly-different names in japanese but they’re practically the same joints. if you think japanese joinery is complex, that’s simply because you haven’t noticed pretty-much every joint is either a tenon going in a mortise or a dovetail-shape between pins, sometimes both at the same time. the tendency to make things appear complex is a great way of keeping people outside the craft and making everyone impressed and willing to pay high prices but it’s unrealistic. they’re the same joints cut with very similar tools using slightly-different methods but well within the same ballpark, if not the same starting lineup.
a mortise-and-tenon joint looks like a butt-joint with two significant differences. and these are important for what we’re thinking about today. in terms of glue surface, it has a almost the same total glue surface. imagine taking a 50×100 and gluing its end to the face of another. the butt-joint has 10 000 square-millimeters of glue surface. cut a through-tenon (not even pinning or drawboring or wedging) that’s 20x75mm and you’ve realistically not changed your total gluing area in a measurable way. but what have you changed? try levering on the end of your board? now you’re not pressing against the glue at all. you’re pressing against the inside of the mortise and its shoulders. the glue doesn’t have to hold it torsionally — only from pulling the two pieces straight apart, which isn’t a force that’s ever going to happen in the normal use of a table-stretcher. the glue is just there to keep everything from looking messy and sliding around — you could honestly use children’s white glue and it would be perfectly-sufficient for the purpose. so all the debates about glue strength in mortise-and-tenon joinery and people paying for premium glues and obsessing over clamping pressure? meaningless. the joint is strong because it’s mechanically-strong. the glue is the back. the seatbelt.
what about dovetails on the sides of a box? let’s think about a box made with the same boards — 50x100s. on each corner, you’re looking at 10 000 square millimeters of glue surface if you butt-joint them. great. sounds like a lot. when you dovetail them, though — let’s say you have three tails —, you end up with six glue surfaces. i’m not going to take you through all the angled-surface calculations but it is approximately 30 000 square millimeters of glue surface (well, if it was finger-joints, that’s what it would be and it’s actually a little more because of the angles but let’s not worry too much about that cause it’s already three-times as large). but — and this is a big but(t) — that’s not the part that matters in this joint. having the dovetail means the overwhelming majority of the torsional forces, like in the mortise-and-tenon, are already being absorbed by the walls of the joint, not the glue. so not only do you have three-times the glue-surface (and then some), the glue isn’t even the thing doing the holding. if you have accurate dovetails, you don’t even need clamps. or, for that matter, glue, in many instances.
a pinned rabbet, by comparison, has somewhat less advantage and this is, practically-speaking, why you don’t see these as commonly in furniture construction. dowels and floating-tenons, however, work the same way integral-mortise-and-tenons work — not shifting the quantity of glue surface in a way that really matters but removing the glue from the equation by turning the joint into a mechanical one where torsional forces are absorbed by the physics, not the chemistry — the glue.
glue strength in general (how to pick a glue)
now that we’ve dealt with the issue of joinery and glue, there are, of course, a few other things we should probably briefly touch on for the sake of being complete. these could easily be complete articles on their own (and likely will be if anyone’s interested in reading them) but i’ll treat them briefly here because they’re relevant and i haven’t discussed them much in the past. there has been some discussion on forums and in magazines of these issues but it’s actually surprisingly thinly-spread (yes, a glue pun) and they can be rather sticky topics.
picking a glue generally comes down to three things — cost, strength and speed.
there are various options for glue in the modern world.
- pva
- epoxy
- urethane
- hide
- grain
- ca
i’m not going to get into the weeds about how each of these works chemically (unless you really want to ask — in which case, i’ll do my best to explain it) but here’s the simple version. pva glue is water-based glue. it’s easy to clean, quick to use, cheap to make and stable at room-temperature. it’s the most common thing most of us use.
epoxy is liquid plastic. it’s extremely expensive, very strong, cumbersome to work with because it’s almost always a binary chemical when you buy it (two-part, sometimes more). it gives you a more flexible open-time and it’s not water-based so it won’t swell the wood or raise the grain. short-time epoxies (1-minute, 5-minute, 30-minute) are realistically temporary bonds when it comes to wood. they’ll hold it quickly but they’re brittle and don’t have the structural strength. that being said, if your joint is a mortise-and-tenon, this is probably sufficient and you can definitely use 5m epoxy for that if you prefer, though you’ll spend a fortune on it compared to pva.
urethane glue (popularized by gorilla) is another plastic but it works rather differently. it’s moisture-activated (no, you don’t have to add water to it — ambient moisture is enough — though some people do and there’s nothing wrong with that approach). it’s messy, difficult to work with, painfully-expensive and smells awful (yes, so does epoxy, unfortunately). will it work? absolutely. should you use it? not unless you really want a headache trying to clean it up. it has only one real advantage over pva (and many disadvantages) — being non-water-based, the wood won’t swell. so if your joint is too tight, you can force it together more easily with urethane adhesives. it might, however, be better not to make your joint too tight. you are, after all, a woodworker. you have the tools.
hide glue is realistically boiled animals. i’m a hardcore animal-rights activist. the ethical issues with hide glue aside, though, it’s not useful. it takes many times longer to set than pva. you can buy it premixed but it mostly arrives in powder that you have to mix — and heat. compared to pva, which just comes from the bottle ready to use. it’s also water-sensitive in a way pva isn’t. get the joint wet at all and it’s got the holding strength of dish-soap. people talk about it being reversible as if that’s a benefit. either you want your furniture to stay together or you don’t. i want mine to stay together. forever. you can reverse pva with heat. but you won’t. nobody wants to repair your desk thirty years after it was made. they want a new desk. reversible isn’t a benefit. it’s a curse. step away from the hide glue.
grain glue (rice, wheat-starch, corn, etc) is very traditional. there’s nothing wrong with traditional when it compares well with modern alternatives. there are certainly applications where this can be useful — like adhering paper in traditional bookbinding where the hold doesn’t need to be very strong. there’s no application in the modern woodworking world where anyone should be thinking about these glues, though. anything grain glue can do, pva and epoxy can do better. it’s also water-sensitive like hide-glue, extremely messy, time-consuming to prepare and a disaster to clean up when you’re finished.
the other glue on the list is cyanoacrylate (ca) glue. this is superglue, crazyglue, quickbond glue, etc. it comes in various forms but you can think of it as being much the same as 1m epoxy. it’s not quite based in the same plastic chemistry but it’s surprisingly close — really just an organic version of petroleum-resin that’s air-activated. i know that’s a bit of a mess from a chemical perspective but that’s not the point. it works like a quick epoxy and the result is the same — it’s messy and brittle. you can hold things together for a few minutes this way but it has realistically no longterm or structural strength. it will, however, in many cases hold up well to water and heat. the fact that it’s easy to just crack off, though, means this property is rarely useful in woodworking. use it to hold parts together while you find better ways to make the bond permanent. don’t count on it. it won’t save you.
so the moral of this little story is simple. use pva glue. for just about everything. if you need something stronger — or something weather-resistant, for example — or something that fills gaps better than water-based glue, structural epoxy is your friend. it is, however, much slower to use both in mixing and curing and i would suggest avoiding it unless you have to. it also smells bad (because of all the plastics) compared to pva, which is far less toxic.
if you’re curious, i use titebond 1 and 3 and, when necessary, totalboat epoxy. yes, i use other things from time to time — ca glue for making jigs, for example, and quick-set epoxy to do minor repairs — but those are the staple of my adhesives. whether you buy titebond or another manufacturer’s pva, though, it should be fine. there’s a gorilla-branded version and even elmer’s makes one. just pick what’s cheap, really. and do joinery that doesn’t depend on the glue for extreme strength. you won’t be disappointed in the joinery-focused method. rely on glue and you’ll always be let down.
how endgrain performs when glued
endgrain performs badly when glued for a very simple reason — it’s the same reason wood gains and loses moisture through the endgrain. think of a board as a collection of very small straws. the straws terminate at the endgrain. so when you apply glue to the end it absorbs into the board rather than sitting on the surface waiting for adhesion. yes, you can certainly prepare the surface by sealing it — like with a first-application of glue, often called “sizing the joint”. this mitigates the issue. but it’s totally irrelevant. if your joint is structurally-sound and mechanical, this glue joint is fine the way it is. if your joint is shaped like your butt, it doesn’t matter how much glue you get on there and you can seal the endgrain with epoxy — and, for that matter, use epoxy as the adhesive for its extra strength. it won’t matter. for the reasons we’ve already discussed. it can’t resist the forces applied to it unless the joint has its own internal, mechanical strength.
so it’s the wrong question. how does endgrain perform when glued? who gives a shit? if you’re relying on endgrain joints, that’s the least of your design and construction problems.
wood-movement and glue joints
the other reason, though it’s a secondary one, is about wood movement. wood moves many times more across its width than along its length and the endgrain doesn’t really move at all (it can swell but only with a huge shift in moisture and this is unrealistic in a natural environment unless you’re literally dipping it in a pot of water).
imagine that same butt-joint where the end of one board is joined to the face of another with nothing but glue. if we’re talking about a stretcher going into a table-leg, the stretcher will expand and contract along the vertical axis and the leg will shift along one or both horizontal axes — there are times this isn’t going to be the case but it’s generally the way it works out — the grain runs along the stretcher but up and down the leg. it’s just a function of how boards are cut from trees. and the same applies (though slightly less aggressively) if the wood is rift or quarter. the expansion is less but the axis of expansion is still perpendicular between the two parts.
that means the endgrain is shifting relative to the face or edge of the leg. every year. multiple times. often multiple times a day. these shifts are tiny unless you have a seriously-huge table with bulky stretchers and legs. but they’re significant over time and there’s no glue we use in woodworking (there are flexible adhesives that can absorb this much movement but they have other disadvantages, mostly either strength or density) that can take that relentless shifting. so, even if the bond is strong when it’s first applied, over time it will fail. of course, it’s not strong at first, push on the far end of the stretcher and you’ll lever it off the leg in no time. but, assuming you don’t actively try to tear your table apart (though i assume you don’t have kids if you think this is a realistic thing to expect), it’s still going to fail simply because of the shifts of humidity in nature. it’s weak in both ways — structurally and temporally. that, of course, isn’t to dispute the findings of recent on-camera experiments. the glue is perfectly-strong from a numeric perspective. it just can’t resist leverage at a distance (your stretcher is proportionally-long) and it can’t hold up to the movement over time the wood will experience.
if you’re using glue to glue endgrain in a miter situation, the problem is even more significant but you might not have thought about it. a miter is a bad piece of geometry for glue. let’s imagine the moisture is at a midpoint when you glue the corner together at 45-degrees. if the wood expands and contracts 2% along its width, the angle will shift quite significantly from 45 in both directions.
let’s do a concrete example. a 100mm-wide board that expands and contracts 2% (flat oak or cherry don’t move quite this much but it’s a simple calculation and you can substitute your species of choice). let’s imagine you glue the miter joint together at exactly the humidity midpoint of the year and the two extremes will be +/-2% from there.
when the frame is assembled, the hypoteneuse (the glue-surface line) of your miter-triangle is 141.4mm long. when summer comes, the board will now be 102mm wide, making the glue-joint 144.2mm. in the middle of winter, the board will be 98mm wide and the glue-line 138.6mm. the result will be that, if the joint is perfectly-tight in the fall and spring, in the summer you’ll see a gap on the outside and the winter will bring a gap on the inside. this constant shifting isn’t just ugly. the movement will tear the glue apart. there are solutions to this, of course. use a different joint or add a spline, either visible or hidden. but a plain miter at this size isn’t going to hold together for long. if your miter is much, much smaller, you can get closer and closer to the glue being able to withstand the shifts — a 3mm-thick mitered case side is likely to hold together relatively well but a 200mm-wide frame board will pull apart in no time unless it’s secured with something other than that glue-joint. if you’re curious, by the way, a metal fastener like a stable on the back of the frame will do little to keep it together. you can use a spline or turn the joint into a mitered bridal-joint (yes, that kind of bridal — there’s a reason we talk about it having male and female parts that … slide together so nicely) and that will stop it from being able to shift through expansion and contraction.
water (and why it’s relevant to endgrain, glue, life, the universe and everything)
water is the source of life.
in terms of glue, though, most of our glues are either water-based or water-sensitive. the only thing that’s not likely to shift from the application of water even once it’s dry is epoxy. pva will weaken with water and protein glues like hide or grain will simply fall apart when they get wet (which people constantly talk about as an advantage — i’m not sure who wants their furniture to fall apart but i’m not one of them). but this is, yet again, asking the wrong question.
the right question might be why you’re taking your side-table in the bathtub with you. but that’s not even the right question.
the real question to ask is where the moisture is going. and that’s quite relevant to our issue here. boards expand and contract across the growth-rings (grain). we know this. we’ve had it driven into our heads since we took shop class as preteens. but the moisture, we often forget, enters and exits the boards almost completely through the endgrain. if the endgrain is covered with glue … what happens? well, one of two things can result. either the water won’t get in or come out, which is unlikely. or the water will literally push the glue out of the way.
if it’s strong epoxy, the first will happen at first and the chemical bonds will eventually break down over time. remember, not 100% of the moisture is getting in and out through the endgrain so if the board absorbs moisture through its face and edge in small quantities then tries to lose it, it’s still pushing on the endgrain from the inside and working its way through the microscopic bonds of the epoxy — the outside moisture doesn’t have much effect but what’s inside certainly will over years.
if it’s water-permeable glue like pva or protein or urethane, however, this process will happen very quickly. it will literally dissolve some of the glue every time there’s a moisture shift. and that can happen from morning to night and back again every day of the year, with larger swings happening between seasons.
final thoughts
theories are great. demonstrations are amazing. and videos are a stunning way to teach things so people engage and remember. we are a culture of useless myths and traditions in many ways and these need to be examined in the clear light of day and most must be eliminated. i think it’s a wonderful thing to do. but we have to ask the right questions. we can collect and share data about anything. but how does that data relate to what we do in our everyday lives?
ask yourself this question when you’re designing something — is this the best way to attach these pieces together? does it matter if the glue-bond is strong or is there something else that’s going to make this fail?
after weeks of seeing debates and, sadly, hate and emotion come into this topic — who the actual fuck is that invested in glue? certainly not the companies that make it! — i think it’s time we all calmed down and started to think more realistically about the situation. yes, endgrain can be glued. yes, it will hold together almost as well as other grain structures if we use it in the same situations. but we’re woodworkers. we’re building furniture. and the places we use endgrain aren’t the same as those where we use edges and faces of boards just because that’s how trees grow and are milled into lumber.
so let’s be peaceful, practical and gentle in our responses. let’s think instead of allowing our emotions to run away with us. and let’s stop getting stuck (yes, stuck, like glue) in our old ways of thinking and use things like logic and math to give us real answers. and our personal experiences. don’t believe this technical explanation? take two long boards and glue them together, end-grain to end-grain. then stand on one and try to pry the other off it. i know what will happen. so do you. but you’re welcome to try — be careful you don’t hit yourself in the face when they fly apart, though, as i’m not going to be responsible for your broken nose.
this has been a rather lengthy exploration (and i didn’t even go into the details of the different types of glue and their applications in woodworking!) but i hope it’s brought some answers to all the people who have seemed to have myriad millions of questions about this topic of late. i wonder if the questions were always there in the background but stifled from embarrassment — or if they were just spontaneous thoughts. either way, thanks for taking a trip down a very adhesive road with me today. be at peace in your shop and don’t forget to take a moment to thank the tree that gave you happiness today.