The Glue Is Rarely What Breaks in a Good Joint. The Wood Is.

If a mortise and tenon is glued and has a long enough tenon, the published tests I read show it failing in the wood around the glue line, not in the glue. That changes what an extra pin or screw can do. A pin cannot reinforce a glue line that is not the weak part. It can only take wood away from the part that is. I set out to show that more fasteners add little strength. The tests support the first half of that claim well. They support the second half only by inference, and I say below where the data run out.

The rule and the old book

The shop rule is simple: pin it, dowel it or screw it, and the joint gets stronger. I wanted to know where the rule came from, so I went to William Fairham's Woodwork Joints, an early twentieth century British manual. Fairham does not say the pin adds strength. He gives a reason, and it is a clamping reason. Drawbore pinning, he writes, is "employed when it is not convenient to obtain the necessary pressure by using a cramp" [1]. The offset hole means that "if a hardwood pin be driven into the joint it will draw the shoulders of the tenon to a close joint and effectually secure the parts" [1]. For indoor work he says joints "may be united with glue" [1].

So the old book treats the pin as a clamp that stays in the wood. Somewhere between Fairham and the modern shop, "a pin can replace the clamp" became "a pin adds strength." I like old books for exactly this. They give reasons, and the reason here is narrower than the rule.

Question

In a glued solid-wood frame joint, where does the failure happen: in the glue line or in the wood? If the wood fails first, what does an added fastener change?

Data and where it came from

I used four kinds of evidence.

1. A controlled lab study. Kasal, Eckelman, Haviarova, Erdil and Yalcin (2015) tested L-shaped mortise and tenon joints in Turkish beech (Fagus orientalis, mean density 0.60 g/cm³) and Scotch pine (Pinus sylvestris, 0.45 g/cm³) at 12 ± 0.2% moisture content. They used PVA and one-part polyurethane glue, with tenon widths and lengths of 30, 40 and 50 mm, under compression and tension loading [2].
2. The Forest Products Laboratory adhesive chapter. The Wood Handbook (Frihart and Hunt, 2010) defines a good bond by wood failure [3].
3. Two more lab studies in beech. A frame reliability study compared mortise and tenon joints with dowel and staple joints in Oriental beech glued with PVA [4]. Hajdarevic and co-authors tested standard and double mortise and tenon joints in European beech at 12.1% moisture content, with 20 mm and 30 mm tenons, nine replicates per configuration [5].
4. Bench tests by Matthias Wandel (woodgears.ca). These are not peer reviewed. They do give failure loads and photos of the broken surfaces in spruce and maple. They give no moisture content [6][7][8].

I add one dovetail study. Wang and co-authors (2020) tested glued box-frame dovetails in poplar modified with melamine-formaldehyde, using PVAc [9]. The abstract and summary I read give no moisture content.

Method

I collected each failure load with its failure type, species and moisture content where the source gives them. I did one conversion myself, without the Lab, so that the bench tests and the beech study use the same units. Wandel applied his loads 20 cm from the post [6], so moment equals force times 0.2 m:

$$M = F \times 0.4536 \times 9.81 \times 0.2$$

with $F$ in pounds and $M$ in newton-metres.

Result

Failure type: mostly wood, with one clear exception

The Wood Handbook sets the standard. An excellent bond is one where the wood breaks away from the adhesive joint, so the bond is as strong as the solid wood [3]. Industry reports bond quality as percent wood failure for this reason. If the wood tears, nobody has to argue about whether the bond formed [3].

Wandel's glue test fits that standard. He glued spruce cross-grain joints (5.3 × 3.8 cm) with six adhesives. All six failed in the wood, between 100 and 150 lb, and "with chunks of wood torn out of the post and/or rail" [7]. In his mortise and tenon and dowel tests, the parts separated at the glue line. But wood fibres stayed stuck to the mating surface: "Bits of the post adhered to the dowels, so even though the joint failed along the glue line, it doesn't suggest that the glue itself failed" [6]. In the spruce frame test, the mortise and tenon also "failed right at the glue line," with fibres moved from mortise to tenon or the other way [8].

Kasal's beech and pine joints give the exception, and it delights me because it is so clean. Joints with 30 mm tenons failed by glue-line fracture. Joints with 40 and 50 mm tenons failed by splitting of the post or fracture of the tenon [10]. Same species, same 12% moisture, same glue: tenon length alone moved the failure from the glue to the wood. The post was 60 × 21 mm in section [10].

The Wang dovetails failed in the wood too. In one loading direction, "failure of the dovetail joint always occurred at the root of the tenon" [9]. That is the narrow neck of the tail, where the section is smallest. The glue did not let go first.

Failure load: the joints compared

Source	Species, moisture	Joint	Failure load	Failure type
Wandel [6]	Spruce, MC not given	Mortise and tenon, 6 tests	140 to 190 lb, mean 173 lb (about 154 N·m)	At glue line, wood fibres on both faces
Wandel [6]	Spruce, MC not given	3 dowels, 3 tests	120 to 145 lb, mean 135 lb (about 120 N·m)	Glue line, fibres on dowels
Wandel [6]	Spruce, MC not given	Single screw into end grain	85 lb (about 76 N·m)	Rail crushed into post, screw heads pulled in
Wandel [6]	Maple, MC not given	Mortise and tenon / 3 dowels	300, 270 lb / 230, 245 lb	Not separated in summary
Reliability study [4]	Oriental beech, 20 °C and 65% RH	Mortise and tenon	204 N·m mean	Not reported in what I read
Reliability study [4]	Oriental beech, 20 °C and 65% RH	Dowel	154 N·m mean	Not reported in what I read
Kasal et al. [2][10]	Beech and Scotch pine, 12%	Mortise and tenon, 30 mm tenon	Not extracted	Glue-line fracture
Kasal et al. [2][10]	Beech and Scotch pine, 12%	Mortise and tenon, 40 and 50 mm tenon	Not extracted	Post split or tenon fracture

My arithmetic for the spruce means: (170 + 175 + 140 + 180 + 185 + 190) / 6 = 173.3 lb, and (120 + 140 + 145) / 3 = 135 lb. Wandel reports an average of 172 lb [6], so his set may differ slightly from the six values in his table.

One ratio surprised me. The dowel joint reached 135 / 173.3 = 0.78 of the mortise and tenon in spruce. In beech, the dowel joint reached 154 / 204 = 0.75 [4]. Two independent setups, two species, one hobby rig and one lab, and they give the same ratio to within 0.03. I would not lean hard on two numbers. Still, it suggests that geometry sets the ranking, not the glue.

The screw is the weak member of the table. In spruce a single end-grain screw failed at 85 lb, about half the mortise and tenon. Wandel describes the failure: "the rail would push itself into the post, and the screw heads would pull into the wood" [6]. Washers raised it to 210 lb, but the joint opened too far for furniture use [6].

What about pins?

Here I have to be honest about the gap. I did not find a peer-reviewed furniture study that tested the same glued mortise and tenon with and without a pin, at a stated moisture content. Fine Woodworking tested 18 frame joints in cherry in 2009. Its editors picked the pinned mortise and tenon to finish first, but a half-lap won [11]. I could not open the article, so I do not report its pinned result.

What I can do is reason from the failure types above. This is inference, not a test result. If a long tenon fails by splitting the post or breaking the tenon [10], a pin hole goes through exactly those two parts. It removes cross-section from the tenon. It also adds a hole to a mortise wall that already fails by splitting. A pin also adds a second load path after the glue line lets go, and in a short tenon (where the glue line does fail first [10]) that second path may matter. So my expectation is this. A pin adds little or reduces peak strength in a long, well-glued tenon. It adds some residual strength in a short one. And it adds what Fairham said it adds: clamping pressure where no cramp fits [1].

Sensitivity: which assumption moves the result most

Tenon length moves it most. In Kasal's study, the change from a 30 mm tenon to 40 mm changed the failure type from glue line to wood [10]. Tenon length also had a larger effect on moment capacity than tenon width [2]. So "the glue is never the weak link" is false for short tenons. The rule I can defend is narrower: if the tenon is long enough, the glue line stops being the weak part.

Species and density come second. Beech joints had 33% higher moment capacity than Scotch pine in compression and 16% higher in tension [2], at the same 12% moisture content. The Wood Handbook notes that wood failure becomes harder to get as density rises, and drops faster above about 0.7 to 0.8 g/cm³ [3]. Wandel's maple test fits that pattern. In hard maple cross-grain joints, carpenter's glue failed in the glue line at a mean of 141 lb with no fibre tear-out, while three other glues failed in the wood at 159 to 179 lb [7]. In spruce, all six glues tore wood [7]. Dense wood can be stronger than the bond, so in dense species the glue line can become the weak part again.

Adhesive type comes third. Polyurethane joints were about 16% stronger than PVA in Kasal's beech and pine, in both loading directions [2]. The Wood Handbook adds a caution: flexible adhesives such as polyurethane can give higher bond strength but lower wood failure, because the adhesive deforms and takes stress off the wood [3]. So "wood failure" and "strong joint" are not the same measurement. A high wood failure percentage says the bond formed. It does not say the joint is the strongest one you could make.

Fit matters less than the rule says, in one rig. Wandel found no detectable difference between loose, snug and overly tight spruce tenons, provided glue filled the joint [6]. That is one bench rig with no moisture data, so I hold it loosely.

Moisture is the input I trust least. The lab studies sit at 12% [2][5]. None of the sources I read tested the same joint after seasonal cycles. That is the gap I care about most, because cross-grain movement in the mortise cheeks attacks the glue line slowly, over years, and a one-day test at 12% cannot see it.

Where this leaves my position

I held, at 0.6 confidence, that a well-fitted glued joint in solid wood fails in the wood before the glue line in most tests, so extra fasteners add little. After this reading I keep 0.6, with a narrower scope. The wood-failure half holds for tenons of 40 mm or more in medium-density species at 12% moisture [10][3]. It fails for 30 mm tenons [10], and it weakens in dense maple [7]. The "fasteners add little" half is still my inference, not a measured result. I lean toward tradition when a test is missing, and I know it, so I mark it as inference here.

One test would change my mind: the same glued tenon, with and without a drawbored pin, in one species at a stated moisture content, cycled through humidity before the load test, with failure load and failure type for each. If the pinned joints keep more strength after cycling, the pin earns its place for reasons Fairham never claimed.

The rule ignores what the grain does. The tenon cheeks are long grain against the long-grain walls of the mortise, and that is where the glue holds best. The mortise wall splits along its grain, and every pin hole is a new place for a split to start.