Floating panels
A Socratic walk-through of floating panels — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a joiner leave the panel of a door sitting loose in its frame instead of gluing it fast?
A frame-and-panel door is built with care: the rails and stiles are jointed, glued, clamped and squared. Then the panel that fills the middle — the largest piece of wood in the assembly — is deliberately left loose, sitting in a groove with a gap at its edges and nothing holding it but gravity and its own thickness.
To anyone assembling it, this looks like an unfinished job. The panel rattles. Every instinct says to run glue into the groove and make the door one solid thing. Doing so is the classic way to destroy it, and the reason is not about strength.
Reasoning it through
REASONING #Start with what wood does after it becomes furniture. Wood is hygroscopic: it exchanges moisture with the air until its moisture content matches the humidity. Humidity changes with the seasons — damp in one part of the year, dry when heating is on — so a wooden object never stops moving, swelling and shrinking indefinitely.
Now the crucial fact, and it is the one that makes the whole craft practice make sense: that movement is wildly unequal in different directions. Along the grain, wood barely moves at all — a length of timber is essentially the same length wet or dry. Across the grain it moves substantially, and a wide board can change its width by a noticeable fraction of an inch between a damp summer and a dry winter. The difference between the two directions is roughly two orders of magnitude, and it comes from the structure: the cell walls swell in thickness far more readily than the long fibres stretch.
So a panel in a frame is an assembly of parts that want to move by very different amounts. The stiles either side of the panel run vertically, so their length is stable; the panel's width runs horizontally across its grain, so its width is not. The frame is dimensionally stable in exactly the direction the panel is not.
Follow what happens if you glue them together. In a dry spell the panel tries to shrink across its width and cannot, because it is fixed to a frame that is not shrinking. The panel is now in tension, and wood is weak in tension across the grain. It splits. In a damp spell the reverse: the panel tries to expand, cannot, goes into compression, and either crushes its own fibres — so that it is permanently narrower when it dries again, and splits later anyway — or pushes the frame apart at the joints.
The loose panel avoids all of this by a very simple move: it lets the panel change size without asking anything else to change with it. The groove is cut deeper than the panel needs, so there is room at the edges for expansion; the panel is centred so it has room in both directions; nothing is glued except, at most, a spot in the middle of the top and bottom edges so the panel stays centred and expands symmetrically.
Two craft details now become obvious rather than arbitrary. The panel is finished before assembly, so shrinking does not reveal an unfinished stripe at its edges. And small rubber balls or springs are sometimes set in the groove — not to hold the panel but to stop it rattling while still letting it move.
The analogy
THE ANALOGY #Think of a long steel bridge sitting on its abutments. It is not bolted rigidly to the ground at both ends. One end sits on rollers or a sliding bearing, and the deck has expansion joints — gaps that open and close with the temperature.
Those gaps look like a defect to anyone who wants a bridge to be solid, and they are the opposite. A bridge fixed rigidly at both ends would not resist thermal expansion; it would generate enormous forces trying to, and something would buckle or crack. The joint does not weaken the bridge. It removes the load entirely by declining to resist a movement that was going to happen anyway.
A bridge expands and contracts along its length and by a predictable amount from temperature alone, whereas a panel's movement is across the grain, depends on humidity that lags the weather by weeks, and varies with the species and even the way the board was sawn — so the joiner's allowance is a judgement about a range rather than a calculation from a coefficient.
Clarifying the model
THE MODEL #The panel is not left loose because glue is weak. Modern adhesives are commonly stronger than the wood itself, which is the point people reach for when they argue the joint should be glued. That strength is exactly the problem: a glue line stronger than the wood guarantees that when the assembly is stressed, the failure happens in the timber rather than in the joint. Making the bond stronger moves the crack; it does not prevent it.
Which way the board was cut matters, and it is why some panels move less. Quartersawn timber, where the growth rings run roughly perpendicular to the face, moves appreciably less in width than flatsawn timber from the same tree, and it also stays flatter. That is a large part of why quartersawn stock is prized for panels and doors, and it means the allowance a joiner leaves is not a fixed number but depends on the material in hand.
Sheet goods change the answer. Plywood and MDF are engineered to be dimensionally stable in the plane of the sheet, so a panel made from them can be glued in without splitting. Floating remains correct for solid wood, and applying solid-wood rules to a sheet panel is merely unnecessary while the reverse is destructive — a good reason the conservative habit persists.
The movement is slow and lagging, which hides the cause. Wood responds to the average humidity over weeks, so a door made in a damp workshop and installed in a dry heated room may take a season to show the problem. The delay is why failure gets blamed on the timber or the glue rather than on a decision made at assembly.
The falsification test. If the mechanism is restrained cross-grain movement, then a glued panel should split across its width along the grain, in the dry season, and a panel of the same wood cut so that its stable direction runs across the frame should not. If glued panels split randomly in direction and season, or split just as often when floated, the account would be wrong and something else — adhesive failure, or stress from the frame joints — would be doing it.
A picture of it
THE PICTURE #How to readThe line is the change in the panel's width, across the grain, over a year; the flat bars are the change in the frame's height along the grain over the same year, which is why they sit on zero. Read the vertical gap between them as the amount of movement that has to go somewhere. If the panel is free, the gap is absorbed in the groove and nothing happens; if it is glued, that same gap becomes force, and the wood fails in tension at the dry end of the cycle. The millimetre values illustrate the scale of the asymmetry rather than measuring any particular panel.
What became clearer
WHAT CLEARED #The loose panel is not an unfinished joint or a tolerance for sloppy work. It is a designed clearance, and it exists because wood moves a great deal across the grain and almost not at all along it — so a panel and the frame around it are inevitably trying to be different sizes at different times of year. Gluing them together does not prevent that; it converts a harmless change of dimension into a stress the wood cannot survive. The joiner is not failing to fix the panel in place. They are declining to fight a movement that will win.