Why pavers sink

A dip in a paver patio is a report about a layer you cannot see any more. The useful thing is that the report is readable: where the low spot is, and what shape it takes, narrows it down to one layer, and the layer decides whether the repair is an afternoon or a weekend.

The wrong move is available first and looks like the right one — lift the low paver, throw sand under it, tap it down. It is even printed on the bag. Whether that works depends entirely on which layer moved, and most of the time it is not the one you just added sand to.

Put a number on “sunken” before doing anything

Belgard publishes the tolerances an installed pavement is supposed to be accepted at, and they are the same measurements you can take on a finished patio years later:

What	Figure
Surface flatness	±3/8 in (10 mm) over 10 ft (3 m), non-cumulative
Lippage between adjacent units	maximum 1/8 in (3 mm)
Joint width	1/16 in to 3/16 in (2 to 5 mm)
Joint fill depth below the chamfer	maximum 1/4 in
Slope for drainage	minimum 2%

So the first tool is a 10 ft straightedge, not a shovel. ICPI’s construction bulletin uses the same ±3/8 in figure for final surface elevations and adds that pavers are deliberately laid 1/8 to 1/4 in high “to compensate for possible minor settling,” which is normal in every flexible pavement. A patio that has come down by a quarter of an inch has done what it was built to do.

Then look at the shape of it. Belgard’s maintenance guide separates two distresses that get the same name: depressions are areas “that have settled within the base or subgrade” and are “typically not load related,” turning up near edge restraints, utility structures and penetrations, and at utility crossings; ruts are depressions in wheel paths from repeated traffic. A dip along a border is an edge problem. A rectangular dip in the middle of a field is a trench nobody compacted. A general waviness is the bedding layer or the base.

The bag and the technical bulletin disagree by a factor of two

Here is the same pavement described by two documents you can buy or download in an afternoon: the data sheet for a bagged patio paver sand, and ICPI’s bulletin on bedding sand. Both were read in July 2026.

	QUIKRETE Patio Paver Sands data sheet (© 2018)	ICPI Tech Spec 17 (rev. Feb 2018)
Bedding depth	“Spread QUIKRETE Patio Paver Sand to a depth of 1 to 2 inches (25 mm to 50 mm)”	screed loose to 1 in–1 1/2 in, which compacts to 3/4 in–1 1/4 in; nominal 1 in with a construction tolerance of ±3/8 in
Moisture	“Wet the sand with a fine spray and screed”	never “bone dry” or saturated; 6% to 8% shown to be optimal; squeeze a handful and it should hold together without shedding water
Seating the pavers	“tap and level the pavers in place with a rubber mallet”	pavers seated with a minimum 5,000 lbf (22 kN) plate compactor
Base below	“Place 2 to 4 inches (50 to 100 mm) of QUIKRETE All Purpose Gravel (No. 1151) or crushed rock”	ICPI Tech Spec 2: patios need a minimum 4 in after compaction over well-drained soils, and 2 to 4 in more in cold climates on continually wet or weak soils
Gradation	“properly graded aggregates” — no standard named on this sheet	ASTM C33 or CSA A23.1 FA1 concrete-sand grading, with a maximum of 1% passing the No. 200 (0.075 mm) sieve

Both documents are honest about what they are. One is a consumer bag instruction; the other is a technical bulletin whose stated scope is vehicular applications, written for specifiers. The sand under a patio does not know which of the two it was bought under. Two rows of that table are where sunken patios come from.

The bedding depth. Take the top of the bag’s range and you have laid twice the nominal thickness, on a base that may be half the recommended depth. The bulletin’s own figures show what that layer then does: 1 to 1 1/2 in loose compacts to 3/4 to 1 1/4 in. That is roughly a quarter of its depth lost to consolidation alone. Read across to the ±3/8 in tolerance and the arithmetic is uncomfortable — a 2 in loose bed can consume the entire allowable surface deviation on its own, before any weather, traffic or subgrade movement has contributed anything. (That inference is mine; the thicknesses either side of it are printed in the two documents.)

The fines. ICPI’s table caps material passing the No. 200 sieve at 1%. For the Canadian grading it says outright that this is a reduction from the 3% that standard allows; for ASTM C33 it describes its 1% as an “additional limitation” without printing C33’s own figure, so only the ICPI side of that comparison is quoted here. What the extra fines do is described in one sentence worth memorising: as the layer approaches higher moisture, “smaller particle sizes (fines) become suspended in water, forming slurry that lubricates the entire bedding layer.”

That is the mechanism behind a patio that is fine in August and rocking in March.

Where the ASTM claim actually sits on the shelf

If you want to check what you are about to spread, the useful discovery is that the standard reference is not on the product named after the job. QUIKRETE’s Sand & Gravel Products data sheet, revised November 2020, says of All-Purpose Sand (No. 1152) that it “Meets ASTM C33, Standard Specification for Concrete Aggregates.” The Patio Paver Sands sheet, covering base sand No. 1150-49 and jointing sand No. 1150-59, describes them as “properly graded aggregates for the installation of patio pavers” and names no standard at all. Both are the same manufacturer’s documents, both linked below, and it takes a minute to confirm.

The point is not that one bag is bad. It is that “paver sand” is a marketing category and ASTM C33 is a gradation, and only one of those is written down.

The material with a published prediction of this exact failure

There is one bedding material whose failure mode is stated in advance. ICPI:

“Limestone screenings and stone dust are not recommended for bedding sand. In addition to being unevenly graded and having excessive material passing the No. 200 (0.075 mm) sieve, screenings and stone dust will break down over time from wetting and abrasion… Depressions will eventually appear in the pavement surface with limestone screenings or stone dust.”

Screenings are cheap, they screed beautifully, and in some regions they are what the yard hands you when you ask for bedding material. If that is what is under your patio, the diagnosis is finished.

The field test on delivered sand needs no laboratory: hard enough particles “will not break under the pressure of a Swiss Army pocket knife.” Specifying rather than guessing, the primary criteria are Micro-Deval degradation of no more than 8% (CSA A23.2-23A or ASTM D7428) and constant-head permeability of at least 2.83 in/hr (ASTM D2434).

Sand that left, by one of three exits

Bedding sand does not compress much. It goes somewhere.

Downwards. Dense-graded base aggregate carrying 5% to 12% fines is what stops the sand migrating into the base surface. A clean, open-graded stone base under a sand-set patio, with nothing between them, is an open door.

Sideways. ICPI’s edge restraint bulletin asks for a 12 in wide strip of geotextile along the base, turned up the sides of the restraint, wherever sand could escape from beneath the pavers or through the joints of the restraint — and is equally clear about the two places fabric does not belong: not across the whole base surface, and never on top of the bedding sand.

Nowhere, which is worse. Over an old concrete or asphalt slab the water that comes through the joints has no way out, and the bedding layer stays saturated. The published fix is drainage through the base: 2 in diameter weep holes around the perimeter on 10 ft centres and at the low points, filled with washed pea gravel and covered with geotextile so the sand cannot follow the water out. A paver surface laid over an existing slab that dips in the wet season is usually this, and no amount of re-sanding the joints touches it.

Joints matter here too, and not only cosmetically. Full joints at 1/16 to 3/16 in are what makes neighbouring units share a load; joints that have emptied leave each paver working alone above a bed that is now free to move under it.

The edge, which is where it usually starts

Edge restraints exist to “prevent spreading of the pavers from horizontal forces,” and ICPI’s bulletin contains two prohibitions that describe most failed borders:

“Soil backfill is never a suitable edge restraint and edge restraints should never be installed on top of the bedding sand.”

The third figure is the one nobody checks: the base has to extend past the spikes by at least its own thickness — a 6 in base should run at least 6 in beyond them — which “contributes stability to the restraint especially in soils subject to heaving.” A border course that drops while the field stays flat is not a bedding problem, and sand under it will not hold.

The fix that lasts, and the shortcut that is printed on the bag

The shortcut, verbatim from the QUIKRETE sheet’s jointing section: “If a paver has settled lower than the surrounding surface, carefully remove the paver and add Patio Paver Jointing Sand to the sand base. Replace the paver and tap with a rubber mallet or similar tool until the paver is level with the surrounding pavers.” That restores the surface, and if the cause was one unit seated badly on day one, it is the correct repair.

Belgard describes the same physical act with an acceptance test attached: after compaction the replaced unit must be flush to plus 1/8 in with its neighbours, and if it is not, “it should be removed, and the thickness of the bedding layer should be modified.” Same operation, but with a number that has to pass.

Where the base, the subgrade or a trench moved, the repair is reinstatement, and ICPI’s procedure is worth following in order:

1. Take up pavers a few feet wider than the failed area on each side, so you are working outward from undisturbed bedding sand rather than guessing at a level.
2. Leave an undisturbed strip of sand 6 to 12 in wide against the standing pavers. It is your screed guide and the support for temporary edging.
3. Brace the opening with temporary plastic or metal edge restraints spiked to the base, so the pavers you did not lift stay where they are.
4. Scrape sand off the sides and bottoms of every lifted unit. Sand left clinging to a paver stops it seating.
5. Bedding sand that has mixed with soil, base or debris “should not be reused” — replace it with clean sand.
6. Rebuild the base in lifts: 2 to 4 in per ICPI, and Belgard specifies 4 in with the reasoning spelled out, that the compactor which fits in a small excavation is a small one. Where the hole reached the subgrade, compact the soil before any base goes back.
7. Compact to at least 98% standard Proctor density on the soil subgrade and 98% modified Proctor on the base.
8. Re-screed the bed to 1 in, relay in the original pattern, then at least two passes with a minimum 5,000 lbf plate compactor, overlapping onto the undisturbed pavers.
9. Sweep in joint sand and compact again “until the joints can no longer accept sand.”
10. Finish it high: after compaction the reinstated area should be no higher than 1/8 in at the edges and 3/16 in at the centre. Level on the day means low by the spring.

For the materials, bedding sand and joint sand are separate purchases in separate quantities — the paver sand calculator works the bed at 1 in and the joints from the paving area, and the paver calculator covers replacements and base tonnage if the repair reaches that far.

If units have to be swapped out, ICPI’s advice is to scatter the replacements through the existing pattern rather than dropping them in as a block, because weathering will have moved the colour of everything already on the ground.

The short version

A sunken paver is worth ten minutes of diagnosis, because this is the one hard surface you can open and close with the same material. Straightedge first, then work out which document your patio was built to. If the answer is the bag, the bed is probably twice as thick as it should be over a base that is half as deep, and the sand you were about to add will settle exactly like the sand already under there.