A joint works loose in its second dry winter and you'd swear you did nothing wrong. The shoulders looked tight at glue-up. The fit was snug going together. Then a January comes when the shop's been heated for weeks, you go to move the piece, and there's a little play in a joint that had none. You blame the glue, or the weather, or yourself. The decision that actually did it was made a year earlier, in about two seconds, and you don't remember making it. You decided which wood got the tenon and which got the mortise.
When both pieces are the same species, this never comes up. They shrink and swell together and the joint just rides the seasons. The trouble starts when you mix woods for looks, an oak leg with a walnut rail, a maple frame with an ash stretcher, and you cut the joint the way the design happens to fall rather than the way the wood wants it.
Here's the thing the design doesn't tell you. Different woods move different amounts across the grain. In a mortise and tenon, what matters isn't how much either one moves on its own, it's the difference between them. If the wood wrapped around the outside, the mortise, moves more than the wood inside it, the tenon, then when the air dries out the mortise closes down harder than the tenon shrinks, and the joint pinches tighter. Wood handles that fine. It's just squeezing across the grain, which is the direction wood is strongest in. But flip it. Put the bigger mover on the inside as the tenon, and in dry weather that tenon shrinks away from a mortise that barely moves. The fit goes slack, the glue line gets pulled in tension, and it happens in deep winter, the driest, worst time to ask a glue joint to hold.
So the rule, and it runs backwards from how most people pair wood: the lower-movement species should be the tenon, the higher-movement species the mortise. Cherry, walnut, and ash move relatively little. They make good tenons. White oak and hard maple move more. They should be the mortised piece. Put a walnut rail's tenon into a white oak leg and the joint tightens every winter for the life of the piece. Reverse it, oak tenon into a walnut mortise, and that's the one that goes slack in a decade. Same two boards, opposite outcome, decided entirely by which one you mortised.
A few riders worth carrying to the bench. Leave a millimeter or two of clearance at the bottom of the mortise, not for movement but so the shoulder seats clean and there's somewhere for squeezed-out glue to go. In a dry climate, lean to the larger gap. If you're turning round tenons for a chair, the rule matters more, not less, because a round tenon goes slightly oval as it dries, and a low-movement wood like walnut keeps that distortion small enough not to matter.
And when the design won't let you orient it the safe way, when the part that has to be the tenon is your bigger mover and there's no arguing with it, that's not a shrug, it's a signal to change the glue. A joint living in the loosening orientation puts its glue line under slow winter tension, and plain yellow PVA creeps under sustained load. It lets go a little at a time. Hide glue doesn't creep, and it re-tacks a touch in humid summers, a quiet self-repair the yellow bottle can't offer. Forced into the risky orientation, reach for hide glue or a cross-linking PVA instead.
None of this is the difference between a joint that holds and a joint that explodes. Under normal household humidity, the mismatched joint doesn't crack itself apart, it just loosens slowly over years. So the real stakes are tight-for-a-generation versus needs-re-gluing-in-a-decade. That's a difference nobody notices on the day you build it and everybody notices on the day it fails. If you're building work meant to outlive you, it's worth the two seconds at glue-up to put the wood that moves less on the inside.