Insights — 19 November 2026
Expansion joints: why caulk is not a joint system, and what a moving joint needs
Walk a parkade that is more than a few years old and you will usually find the same repair in the same place: an expansion joint that has been caulked, and then caulked again over the old caulking. The sealant is split down the middle, or peeled cleanly away from one edge, and on the level below a rust-brown stain runs down the wall under the joint line. Each round of caulking was cheap, each one failed, and the water never stopped moving.
The mistake is rarely the sealant itself. It is treating a moving joint as a gap to be filled rather than a system to be built. A bead of caulk is one component of that system — and on its own, without the edges, the geometry, the movement measurement and the waterproofing tie-in around it, it is a component set up to fail. This article covers what a joint in a concrete structure is actually doing, why caulking alone cannot survive it, how to read a failed joint, and what rebuilding one properly involves.
The movement
What the joint is doing all day
Concrete moves. It shrinks as it dries in its first years, it expands and contracts with temperature every day and every season, and a suspended slab also deflects slightly under live load — every vehicle that crosses it. A structure cannot stop that movement, so it is designed to concentrate it: the expansion joint is the one place in the building where the slab is supposed to separate, so that it does not crack somewhere it should not. A joint that moves is not a defect; it is the structure doing what the engineer intended.
That means the joint never rests. It is widest on the coldest night of the year, narrowest on the hottest afternoon, and on a trafficked suspended slab it also flexes a little with every axle. Whatever material sits in that gap has to stretch, compress and recover through all of it, for years, while wet, cold, hot, salted and driven over. Very few materials can, and none can when installed in the wrong shape.
In a parkade the joint is also a waterline. Rain and snowmelt on an exposed deck carry de-icing salt, and an open or failed joint delivers that chloride-laden water directly onto the slab edge, the beam and the reinforcing steel below — the exact mechanism behind most of the corrosion repair this region pays for, and one of the reasons parkade membranes fail years before they should. A leaking joint is the start of a concrete repair bill several times the cost of the joint work that would have prevented it.
The failure
Why the bead of caulk fails
A word on terms: caulking is field-applied elastomeric sealant, and sealant is a legitimate part of joint work. The reference standard, ASTM C920, classifies sealants by the movement they are rated to take, and the manufacturers whose products we install — Tremco, Sika, MAPEI and BASF among them — publish those ratings and the installation geometry each product needs on its datasheet. Installed inside its rated movement, in the right shape, on sound prepared edges, a sealant joint works. The failures come from using the bead as the whole answer.
The first way it goes wrong is geometry. A sealant stretches like a band, and its shape decides how much strain it feels: a bead that is too deep for its width cannot elongate, so it tears down the middle the first winter the joint opens. Worse, sealant gunned straight into an open joint bonds to the bottom as well as both sides. Held at three faces, the material cannot stretch at all — three-sided adhesion is one of the most common defects found in failed joints, and it is prevented by a foam backer rod that costs almost nothing.
The second is the edges. Sealant is only ever as good as what it grips, and the arrises of an old parkade joint are carbonated, soaked in oil and salt, and often cracked or spalling from years of wheel impact. New sealant tooled over failing edges reproduces the failure on schedule. The third is movement: if the joint moves more than the installed sealant's rated class, it fails by fatigue no matter how well it was applied — and in most re-caulking, nobody has measured the movement at all. The fourth is traffic. In a drive aisle the joint takes wheel loads directly, and a wide, unsupported bead gets ploughed out by tyres and snow-clearing equipment. Wide, trafficked joints are not a caulking problem; they need a system designed to carry wheels across the gap.
Side by side
Reading a failed joint
| What you see | What it means | What the fix involves |
|---|---|---|
| Sealant split down the centre | Cohesive failure: the joint moves more than the sealant can take, or the bead is too deep for its width | Measure the movement, re-size the joint, match the movement class, rebuild the geometry over a backer rod |
| Sealant peeled cleanly off one edge | Adhesive failure: the bond face is contaminated, carbonated or breaking up | Cut the edges back to sound concrete and prime per the datasheet before any new sealant |
| Crumbling concrete along the joint edge | Nosing failure: traffic is pounding an unsupported arris | Reconstruct the nosings in repair mortar before the joint is sealed — sealant cannot fix concrete |
| Stains or drips on the level below the joint line | The joint is passing water: the waterproofing tie-in has failed or never existed | Detail the joint and the deck membrane together, not as two separate trades |
| Sealant proud of the deck or ploughed out in drive lanes | The bead is carrying wheel loads it was never designed for | An engineered trafficked joint system across the gap, not a bigger bead |
The symptom names are generic; the rated movement, geometry and preparation each product needs are on its manufacturer's datasheet, which is where they should be checked.
The build
What rebuilding a moving joint involves
A joint that answers the table above is built in a sequence, and the sequence is most of the value. Done in order, the same work also phases cleanly — a joint is a line, not an area, so sections of it can be rebuilt while the deck around them stays in use.
- Measure the movement, not just the gap
The width of the joint on the day you look at it is one data point, taken at one temperature. What matters is the range: how much the joint opens and closes through the year, and whether traffic flexes it. That measurement — not habit, not whatever is in the van — decides whether a field-applied sealant is even eligible, and in what class.
- Cut out and rebuild the edges
Old sealant, old backer and contaminated bond faces come out, and the joint faces are cut back to sound, clean concrete. Spalled or broken nosings are reconstructed in repair mortar so that the new joint has solid, square arrises to grip and traffic has something other than the sealant to run on. The edges are the foundation of the joint; this is the step that re-caulking skips.
- Set the geometry
A backer rod sized to the joint is set to the depth the sealant's datasheet calls for, giving the bead its proper width-to-depth proportion and limiting adhesion to the two faces that should carry it. Primers go on where the manufacturer requires them. None of this is visible once the joint is tooled — which is exactly why it is where corners get cut.
- Choose the closure by movement and traffic
Smaller joints inside a sealant's rated class are sealed with the matched product. Wide joints, joints with large movement and joints crossed by traffic get an engineered joint system instead — among them the BASF Jeene structural sealing joint, a pre-compressed gland bonded into the prepared joint, which we install where a drive-aisle joint has already defeated repeated caulking.
- Tie the joint into the waterproofing
On a coated deck, the traffic membrane and the joint are detailed together, each terminated onto the other the way the membrane manufacturer draws it. A joint sealed in isolation leaks at the interface between the two systems — which is why the drips below a "re-caulked" joint so often continue as before.
Joint work is assessed, not priced blind: the movement, the condition of the edges and the traffic over the joint decide the system, and none of the three can be known from a photo. Our expansion joints page covers the service itself, and our concrete repair page covers what happens around the joint when the water has already been through it. If joints are on your maintenance list — or on your depreciation report — the useful first step is having them measured and walked, level by level.
Questions we get asked
Expansion joint FAQs
How long should a sealed expansion joint last?
There is no honest single number, because the life of a joint is decided before the sealant goes in. A joint built inside the sealant's rated movement class, over sound rebuilt edges, with the geometry the manufacturer's datasheet asks for, gives years of service through thousands of movement cycles; a bead gunned into an unprepared gap can fail in its first winter. The useful question is not how long the product lasts but whether the joint was measured, prepared and detailed as a system — and whether anyone looks at it again afterwards. Joints should be walked as part of routine parkade maintenance: a small failure caught early is a repair; the same failure left through two winters becomes concrete repair below the joint.
The joint was re-caulked last year and it is leaking again. Why?
Because re-caulking replaces the one component that probably did not cause the failure. If the original sealant tore down the middle, the joint moves more than that class of sealant can take, and the new bead inherits the same movement. If it peeled off an edge, the bond faces are contaminated or deteriorating, and new sealant is being asked to grip the same failed surface. If water is arriving from behind the joint, the waterproofing tie-in has failed and no amount of sealant at the surface will stop it. Repeating the caulking without re-measuring the movement, rebuilding the edges and checking the membrane detail repeats the result.
What is a backer rod, and why does it matter?
A backer rod is a compressible foam rod pressed into the joint before sealing. It does two jobs. It sets the depth of the sealant relative to its width, which controls how much strain the sealant feels when the joint moves — a bead that is too deep for its width cannot stretch and tears. And it stops the sealant bonding to the bottom of the joint. A sealant stuck to both sides and the base is held at three faces, so when the joint opens the material cannot elongate and fails almost immediately. The rod is a few cents of foam that decides whether the metres of sealant above it can do their job, which is why its absence is one of the first things to look for in a failed joint.
Can expansion joints be replaced while the parkade stays open?
Usually, yes — joint work phases well, because a joint is a line, not an area. The work is sequenced in sections, with the stalls and lanes either side of each section closed and traffic routed around them, and the rest of the parkade in use. How quickly each section can take traffic again depends on the products installed, the substrate and the temperature, so the schedule is set at assessment against the manufacturers' datasheets rather than promised in advance. The sequencing is part of the design: a joint that crosses a drive aisle is planned around access hours, and the waterproofing tie-in either side of it is programmed in the same closure so the joint is not opened twice.
Keep reading
Related reading
The service itself: joint assessment, nosing reconstruction, sealant and engineered joint systems on occupied properties.
Parkade waterproofingThe membrane the joint has to work with — traffic coatings for suspended slabs and exposed decks.
Why parkade membranes fail earlySix failure mechanisms, most of them at details and terminations — joints included.