Most people learn joinery by technique. Here is how you cut a mortise. Here is how you cut a dovetail. The mechanics come first, and the reasoning — if it arrives at all — comes years later, after you've made pieces that held and pieces that failed and spent some time figuring out why.
Invert that. Every joint in furniture construction is answering a structural question. Know the question before you pick up a chisel.
What the joint is solving
There are four things a joint has to answer.
What loads will it carry, and which direction do they run? Tension (the joint being pulled apart), compression (pushed together), racking (lateral force — the kind that turns a square frame into a parallelogram), torsion (twisting), or some combination. A chair leg-to-seat joint sees tension and racking together. A shelf-to-case joint sees mostly compression and racking. The loads determine the joint family.
What does wood movement do to this joint over time? Wood expands and contracts across the grain. A well-fitted mortise-and-tenon between a chair apron and leg is stable because both members move in roughly parallel directions. A rigidly glued breadboard end fights the panel it's supposed to hold flat. It loses, every time, and the panel cracks. The joint has to accommodate movement or it fights it. Fighting always loses.
How much long-grain glue surface does the joint create? End grain is a lousy glue surface — it absorbs adhesive before the glue can set up and doesn't hold well in tension. Long grain is excellent. A mortise-and-tenon works because the geometry creates long-grain contact on six surfaces. A butt joint at a corner doesn't work for the same reason.
What does assembly actually require? Can you get clamps on it? Can you close all the joints before the glue sets? Decent joinery sometimes gets cut for pieces where the assembly sequence is physically impossible. The joints are fine. The thinking isn't.
Four families
Hundreds of named joints. Four structural families.
Butt joints: surface meets surface, no mechanical interlock. Glue and fasteners carry everything. Not wrong. Just limited. Appropriate where loads are low or assembly access is constrained.
Dado, rabbet, and groove joints: one member seats into a recess cut in the other. Mechanical resistance in one axis, glue handles the rest. The defining move of case construction.
Mortise-and-tenon: projecting tenon into a mortise. Resists racking and tension mechanically; glue handles compression and seating. The workhorse. It has been the workhorse for a few thousand years because the geometry is honest about the loads.
Dovetails: the tenon geometry resists tension without glue. The interlock is mechanical. This is why it appears at case corners and drawer fronts, exactly where tension loads are highest. The fact that it's visually distinctive is beside the point.
The diagnostic frame
A failed joint is information.
A gap at the tenon shoulder means the shoulder was never tight, or the tenon dried and shrank after cutting. A split along the grain next to a joint means cross-grain was rigidly glued and the wood moved when the joint didn't. A frame that racks means the joinery doesn't resist lateral load — undersized tenons, insufficient shoulder bearing, or the wrong joint family for the application. A drawer that swells shut every July isn't a failure. It's a predictable outcome of not accounting for solid wood movement in the drawer fit.
The failure tells you exactly what the joint wasn't doing. That's worth knowing, even after the fact.
The rule underneath all of it: know what the joint is solving before you start laying out.