Movement joints in tiling

Why does a tiler stop a floor short of the wall and hide the gap rather than tiling right up to it?

A tiled floor that stops short of the wall, with the gap filled by something soft and hidden by the skirting, looks like a corner cut — as though the tiler dodged a fiddly cut and covered the evidence.

But the same detail turns up in expensive work, in specifications, and in the failures of floors that lacked it. So the interesting question is what a rigid, well-bonded floor needs a soft edge for, when nothing about it is ever going to move.

Begin with the assumption in that last clause. Is a tiled floor one material? It is at least four — tiles, adhesive bed, screed and the structure under it — each changing size for different reasons and on different timescales.

Cementitious screed shrinks as it dries, and that shrinkage is large, slow and one-way, still running months after the floor looks finished. Fired ceramic does the opposite over years, taking up moisture and expanding slightly, irreversibly. Everything moves with temperature, and a concrete structure creeps under load long after it is built. So the tiled surface and its base are two sheets of unlike material, glued together, each with its own programme of dimensional change.

Now ask what becomes of the difference between them. In the middle of a floor, nothing dramatic — neighbouring tiles push on each other equally from both sides. At the edges it must go somewhere, and if the edge is a rigid wall, there is nowhere.

What accumulates there? Take the component that computes cleanly, thermal movement, at a coefficient of about seven millionths per kelvin for ceramic (a recalled typical figure — published values run roughly five to nine, so this is an order-of-magnitude sum, not a specification). Ten metres of unbroken run through a twenty-degree swing gives 7 × 10⁻⁶ × 10 000 mm × 20 = 1.4 mm.

Its smallness is the lesson rather than an objection. A millimetre and a half is nothing if there is somewhere to put it, and non-negotiable if there is not, because the alternative to moving is force. And movement is proportional to length, so the demand at the edge grows with the unbroken field — which is why the answer is a spacing rule rather than a one-off allowance.

Then the failure. A tile field is very strong in compression along its plane, so it does not crush; but it is a thin, stiff sheet held down by a bond weak in peel, and a restrained sheet in compression leaves the plane long before it crushes. The tiles arch up in a ridge — tenting — often with a bang, sometimes weeks after the last temperature change, so the compression released must have been building.

A contrast is worth drawing, because a restrained slab in the cold is the standard story of cracking in roads and floors: something shrinks, cannot contract, goes into tension and cracks at intervals. A tiled floor commonly fails the other way — the screed shrinks while the tile field does not, so the field ends up slightly too long for its base and is loaded in compression, and compression in a bonded thin layer buckles rather than cracks. Same accumulated difference, opposite sign, different-looking failure.

Two practical details follow, and both are where the craft is lost. The perimeter is the priority, because there the restraint is total. And the joint must run right through — grout, adhesive, down to the base — or it is a decorative line transmitting load as before. A soft joint is not an absence of tiling; it is a component whose job is to carry no force.

Is any of this trade inertia? A little. National standards give spacings that differ from one another and are conservative rules of thumb rather than a calculation for the floor in front of you; I will not quote one, since the value depends on which code you work to. And plenty of small domestic floors survive with no joint at all — exactly what the mechanism predicts for a short run in a stable room, and why the detail is easy to dismiss until someone tiles a long corridor.

The model is not "materials expand, so leave a gap". It is that a bonded assembly of dissimilar materials accumulates a difference along its length, that the difference is absorbed either as displacement or as force, and that the force route fails at a strain far too small to notice.

Two refinements. Temperature is the component that can be calculated, but usually the smallest one; screed shrinkage and the irreversible moisture expansion of ceramic are larger and, being one-way, do not helpfully reverse each winter, so anyone reassured by the 1.4 mm above has taken the easiest term for the total. And an underfloor-heated floor is a genuinely harder case rather than a warmer one, because the heat is applied to the assembly directly and cycles it deliberately.

The test that would decide it is what the failures scale with. If accumulation is the mechanism, tenting should track unbroken run length and imposed swing — long corridors, conservatories, heated floors — and be largely indifferent to tile quality or adhesive brand. Find small enclosed bathrooms tenting as often as long runs, at equal materials and exposure, and the accumulation story is wrong: the cause would be bond failure or a defective substrate, and the joint would be treating a symptom.

• Why roads and floors crack at regular spacings when the same accumulated movement puts them in tension instead of compression.
• How decoupling membranes take a different route to the same problem, by letting the tile field slide over its base.

Every term the collection defines is gathered in the glossary.