Mortise and Tenon vs Dowels vs Pocket Screws: Why Every Joint-Strength Test Disagrees

By August the hills here in Vermont go heavy and green, the shop holds its heat into the evening, and I have learned to cut joinery early, while the air is still cool enough to think in. This week the bench held chair rails, and the mail held a question from a reader building their first chair: which joint? They had done their homework, which was exactly the trouble. They had found a magazine test in which a mortise-and-tenon carried 1,444 pounds, and a hobbyist's rig in which the same joint gave up at 172. They had found pocket screws beating a Festool Domino in one test and losing to it soundly in another. Which number, they asked, was true?

All of them. Every one of those tests broke real wood and reported honest pounds. What changes is the rig, and once you learn to read a rig, the disagreement stops being noise and becomes the most instructive thing in the whole literature.

The Missing Lever Arm

Nearly every hobby test reports pounds at failure, and pounds alone mean almost nothing. When a sitter tips back, what the chair's rear joint feels is not force but bending moment: force multiplied by distance from the joint. Move the same push twice as far out along the leg and you have doubled the load without adding a pound.

Matthias Wandel, the engineer behind woodgears.ca, was unusually plain about his geometry. He applied a hydraulic jack 20 centimeters from the post, read the force off a bathroom scale, and averaged 172 pounds for spruce mortise-and-tenons, 135 for dowels, 110 to 115 for pocket holes. The one peer-reviewed study in this story, from Kasal, Eckelman and colleagues at Muğla and Purdue, broke 360 L-shaped mortise-and-tenon joints for BioResources in 2015 and skipped pounds entirely, reporting moment instead: 70 to 279 newton-meters depending on species, glue, and tenon size.

Here is arithmetic I did at my own desk, offered as a back-of-envelope and not a published finding. Wandel's 172 pounds is about 765 newtons; applied at 0.2 meters, that is roughly 153 newton-meters. The university's Scotch pine tenons averaged 140 in compression, and its pine-with-PVA specimens spanned 70 to 183. Wandel broke spruce, a comparable softwood, glued with yellow PVA. The bathroom scale and the university press agree almost exactly once they speak the same units. The eightfold spread against Fine Woodworking's 1,444 pounds is mostly geometry and rig, not sloppiness.

Three Ways to Break a Corner

Three different tests hide under the phrase joint strength: shear, which slides one member past the other; pull-apart, which drags the joint straight from its socket; and racking, the lever-action twisting a chair actually endures. Wood magazine's testing in 2006 and 2007, some of it run at Iowa State University, makes the point all by itself. The same mortise-and-tenon carried roughly 1,000 pounds in shear and somewhere between 2,500 and 4,700 in pull-apart, the wood itself failing before the joint let go. Direction alone moved the answer several times over.

The academic data sharpens the picture. When Kasal's team loaded identical corners diagonally, once closing the corner and once opening it, capacity shifted only seven percent, but the failure changed its nature: the mortise wall breaking in tension across the grain in one case, shearing along the grain in the other.

This is how two famous numbers can both be true. Rockler's pocket-hole page cites unnamed independent testing in which a pocket-screw joint held 707 pounds in shear against 453 for a mortise-and-tenon. Wandel's pocket holes averaged 99. His was a bending test at a 15-centimeter arm; theirs, whatever its provenance, was shear. A chair's back leg lives in bending, which is to say the shear figure flatters pocket screws for precisely the duty a chair never asks of them.

A Tenon Is a Range, Not a Joint

Fine Woodworking's eighteen-joint shootout (issue #203, early 2009, by Douglas Moore and Thomas McKenna; the article sits behind a paywall now, so its numbers survive mainly in forum records that largely agree with one another) refutes its own headline internally. A quarter-inch tenon failed at 717 pounds; a traditional mortise-and-tenon at 1,444; a half lap topped the field at about 1,600. Dowelmax dowels, at 759, beat the quarter-inch tenon and lost by nearly half to the traditional one. Whether dowels beat mortise-and-tenon depends entirely on which row you quote.

Kasal's 360 specimens put figures to the geometry. Tenon length mattered more than width: lengthening from 30 to 40 millimeters added 43 percent capacity under tension, while widening from 40 to 50 added about 12. Size even flipped the failure mode. The 30-millimeter tenons let go at the glue line; the 40s and 50s split the post or broke the tenon itself, at which point the rig is measuring the tree rather than the joinery. Wandel met the same wall from the other side: his negotiated dowel-versus-tenon rematch in red oak ended close to a tie because both joints had grown strong enough that the wood broke first.

What the Glue Line Knows

The Forest Products Laboratory's Wood Handbook states the ground truth plainly: an excellent bond is one where the wood breaks well away from the glue, and a sound long-grain joint is at least as strong as the wood around it. End grain is the exception. It is many times more porous than side grain, it drinks adhesive away from the joint, and that overpenetration is a primary reason strong butt joints are so hard to make. Every pocket-screw joint is a butt joint; the screws carry it. Wandel showed how little the glue helps: adding it raised his pocket-hole average from 99 pounds to 111.

On a true tenon, glue choice moves the needle only modestly. Wandel's spruce tenons averaged 222 pounds with yellow glue, 202 with Titebond III, and 236 with epoxy, in small samples; Kasal's polyurethane joints ran about 16 percent stronger than PVA. The Handbook adds a caution no break test can show: PVA yields under continued stress. A chair racks a little every day of its working life, so a one-shot test flatters a PVA joint's decades.

Species sets the ceiling. The Handbook has bonded assemblies gaining strength with wood density up to roughly 700 to 800 kilograms per cubic meter; below that, the wood limits the assembly, not the bond. Hence Wandel's maple joints running nearly double his spruce, and Kasal's Turkish beech beating Scotch pine by a third in compression.

Stiff Before Strong

Ultimate load is not even the number your chair cares most about. Wandel watched pocket-hole joints deform visibly from about 50 pounds onward before their sudden end, and an end-grain screw with washers reached 210 pounds in spruce, beating his 172-pound tenons, while flexing far too much for furniture. Peak pounds and usable stiffness are different rankings. Failure character differs too: readers on the Festool Owners Group forum observed that a broken Domino joint still hangs together, where a failed pocket-screw joint pops and leaves a bare, unglued butt. The peer-reviewed data offers one real comfort. None of Kasal's 360 specimens fell below 70 percent of its group average, which is why the authors recommend designing to about 70 percent of predicted capacity. Well-made joints are consistent; a one-sample video test samples that consistency exactly once.

Who Paid for the Broken Board

After Wandel's first test favored the tenon, a doweling-jig maker asked him to remove its name, then agreed to a rematch on negotiated specifications. Dowels resisted first splitting better, 121 pounds to 103; the tenon took more force to fully separate, 138 to 123. Same rig, same day, two honest numbers, opposite headlines. The manufacturer insisted only the first be published, Wandel refused, and the collaboration died. That afternoon holds the whole industry problem in miniature.

So read provenance the way you read grain. Dowelmax's strength-test page leans on a UK magazine test run on apparatus Dowelmax itself shipped over, and offers no failure loads at all; Wikipedia records the company's own shear test at 890 for multiple dowels against 540 for a doubled Domino. Rockler's 707-pound claim names no laboratory, no species, no geometry. Honest dowel-favorable results do exist: the Working by Hand blog records a 2010 Canadian Woodworking oak T-joint test in which three dowels held 650 pounds against a tenon's 500, and a UK Furniture & Cabinetmaking ram test in which pocket holes, at 280 pounds, lost to the Domino's 400, the exact reverse of Fine Woodworking's 698-to-597 verdict. And beneath all of it, at least one content site now publishes workshop "tests" claiming biscuit strengths several times anything a documented rig has ever produced, with no methodology attached. The noise floor is rising.

Ask five questions of any test before you let it pick your joinery. How far from the joint was the force applied? In which direction: shear, pull-apart, or racking? How large was the tenon or dowel relative to the stock, and did the wood fail or the joint? What species and what glue? Who commissioned the number, and is it peak load or first movement? For a chair, the honest configuration is racking at a long lever arm in your own species, and there every independent rig in this story, from the bathroom scale to the university press, puts a generously sized tenon on top, with dowels close behind when they are sized to the stock, and pocket screws' shear-test glory somewhere it cannot help you. Cut the tenon long before you cut it wide, choose the densest wood the design will bear, and let the marketing departments argue with the lever arm. The quiet snick of a chisel has outlasted louder claims than theirs.