What size dowel, how deep it goes in, how much sticks out, and how many per foot
← Back to the calculatorThe buried end must be at least the development length — 16 in for #4 Grade 60 in 4,000 psi concrete — and the exposed end laps the next pour’s bars: 21 in for a Class B lap. That is a 37 in cut, and six of them from a 20 ft stick.
A dowel is not a lap splice and not an anchor. A lap joins two bars end to end so the pair carries tension through one pour; an anchor develops a bar into concrete to hold a load. A dowel does both jobs across a construction joint — the buried end develops in the first pour, the exposed end laps into the second — and if either end is short, the joint becomes the plane the load was supposed to cross.
The embedment figures here come from the same ACI 318-19 §25.4.2.2 simplified table the development length page carries: Grade 60 bar, 4,000 psi normalweight concrete, no epoxy. Change the strength, the grade or the coating and the embedment moves; the method does not.
A dowel is a short length of bar embedded in one pour and left projecting into the next. It does three jobs at once: it transfers load across the construction joint, it locates the second pour, and it ties the two pours into one structural element. You meet them wherever a wall meets a footing, a slab meets a wall, or two pavement sections meet.
Two things it is not:
That distinction is where dowel details go wrong: the buried end gets sized like an anchor (too short), or the exposed end gets sized by eye (too short to lap). Each end has its own rule, and the cut length is the sum.
The buried end has to develop the bar’s yield force in the first pour. That is exactly what development length measures, so the embedment is the development length — not a round number and not “half the wall thickness.” ACI 318-19 §25.4.2.2 gives the simplified value for a straight deformed bar in tension, and §25.4.2.1 sets a floor of 12 in for any bar.
| Bar | db (in) | ld (in) | ld | ld (mm) |
|---|---|---|---|---|
| #3 | 0.375 | 12* | 1 ft 0 in | 305 |
| #4 | 0.500 | 16 | 1 ft 4 in | 406 |
| #5 | 0.625 | 19 | 1 ft 7 in | 483 |
| #6 | 0.750 | 23 | 1 ft 11 in | 584 |
| #7 | 0.875 | 34 | 2 ft 10 in | 864 |
| #8 | 1.000 | 38 | 3 ft 2 in | 965 |
| #9 | 1.128 | 43 | 3 ft 7 in | 1,092 |
| #10 | 1.270 | 49 | 4 ft 1 in | 1,245 |
| #11 | 1.410 | 54 | 4 ft 6 in | 1,372 |
Swipe the table sideways for more columns →
What the cover does not do is change these numbers. The expression behind them takes fy, f'c, the bar diameter and four factors, and cover is not one of them, so a #8 needs 38 in whether the chair is 1.5 in or 3 in. Cover decides whether the bar is in the right place, not how long the embedment is. The development length page works the arithmetic symbol by symbol — it is the same arithmetic that decides a dowel’s embedment.
The exposed end laps the bars of the next pour. So its length is a lap length: 1.0 × ld for a Class A splice, 1.3 × ld for Class B. Class B is the default whenever nobody has verified that the steel is at least twice what analysis requires and that no more than half the bars are spliced at one section — which, in practice, is most of the time.
| Bar | ld (in) | Class A lap | Class B lap (default) |
|---|---|---|---|
| #3 | 12 | 12 in | 16 in |
| #4 | 16 | 16 in | 21 in |
| #5 | 19 | 19 in | 25 in |
| #6 | 23 | 23 in | 30 in |
| #8 | 38 | 38 in | 50 in |
Swipe the table sideways for more columns →
This is where the “6 in of dowel” detail comes from and why it fails: 6 in is neither a #4 embedment (16 in) nor a #4 lap (21 in). If the drawing genuinely calls a short projection, it is not a lap — it is a different connection, and the wall’s bars must be anchored some other way. The drawing wins, but it is worth knowing which of the two you are building.
Count a dowel line the way you count a mat: floor(clear length ÷ spacing) + 1. The +1 is the dowel at the far end, and it is the one people forget — a 20 ft wall at 16 in on centre is 16 dowels, not 15, because 240 ÷ 16 = 15 spaces and 16 bars close them.
| Spacing | Dowels per foot | Dowels in a 20 ft wall |
|---|---|---|
| 12 in o.c. | 1.00 | 21 |
| 16 in o.c. | 0.75 | 16 |
| 18 in o.c. | 0.67 | 14 |
| 24 in o.c. | 0.50 | 11 |
Swipe the table sideways for more columns →
The spacing is the drawing’s call, but the count rule is the same at any spacing — and it is the same rule the spacing calculator uses for a mat, which is why a dowel takeoff and a mat takeoff never disagree when they are done the same way.
Cut length = embedment + projection. For the #4 example that is 16 + 21 = 37 in. Then a 20 ft stick (240 in) yields floor(240 ÷ cut) pieces — whole pieces only, and the remainder is scrap.
| Cut length | Pieces per 20 ft stick | Offcut |
|---|---|---|
| 16 in | 15 | 0 in |
| 20 in | 12 | 0 in |
| 24 in | 10 | 0 in |
| 30 in | 8 | 0 in |
| 37 in | 6 | 18 in |
| 48 in | 5 | 0 in |
Swipe the table sideways for more columns →
Cuts that divide 240 evenly waste nothing; 37 in leaves 18 in per stick. That is the same takeoff logic as any other bar — the offcut is a remainder, not a percentage — and the estimator page works it on a full job.
“Dowel” covers two different products, and mixing them up is a real failure mode:
Put a deformed rebar dowel into a pavement joint and you have built a tie bar by accident: the joint cannot move, so the movement shows up as a random crack somewhere else. If the drawing says “dowel” near a joint that is meant to open, check whether it wants smooth. The driveway spacing page covers where those joints go; the dowel inside them is a different bar.
Epoxy coating protects the bar, but it also reduces bond, so the embedment has to grow. The coating factor is 1.2 when cover is at least 3db and clear spacing at least 6db, and 1.5 otherwise — the same ψe the development length formula uses:
| Bar | ld uncoated | ld epoxy (1.2) | ld epoxy (1.5) |
|---|---|---|---|
| #4 | 16 in | 19 in | 23 in |
| #5 | 19 in | 23 in | 29 in |
| #6 | 23 in | 28 in | 35 in |
Swipe the table sideways for more columns →
So an epoxy-coated #4 dowel in a tight detail is a 23 in embedment, and with the 21 in lap that is a 44 in cut — not the 37 in of the uncoated detail. The coating does not change the lap rule; it changes the length the lap is computed from.
Everything above is the cast-in case, where the concrete is poured around the bar. Drilling rebar into existing concrete changes one thing only: the buried end develops through an adhesive ring instead of through cast-in bond. The arithmetic that follows from it is fixed and small — the bit is 1/8 in larger than the bar, the hole is 1/2 in deeper than the bonded length, the adhesive is the annulus times the depth, and the count is still floor(clear ÷ spacing) + 1.
How to install dowel bars into a cured pour, in the order that decides the result: mark, drill, clean, inject, insert. Adhesive that goes into a dusty hole is wasted adhesive, and a bar twisted into a gelled cartridge never comes back out. Four tools do the whole job — a hammer drill, a nylon brush, a blow-out pump and the cartridge gun. Keep a wrap of tape on the bit at the drilled depth: a hole that is 15 in instead of 16.5 in has 9 percent less bonded area and nothing about it looks wrong from the surface.
That gap is not clearance for the bar to slide into — it is the annulus the adhesive has to fill, and it is the number every cartridge table is written around. Tighter and the bar scrapes the adhesive off the hole wall on the way in; wider and you spend adhesive filling a gap that carries no load.
| Bar | Bar dia (in) | Drill bit | Bonded length (in) | Drill to (in) |
|---|---|---|---|---|
| #3 | 0.375 | 1/2 | 12 | 12.5 |
| #4 | 0.500 | 5/8 | 16 | 16.5 |
| #5 | 0.625 | 3/4 | 19 | 19.5 |
| #6 | 0.750 | 7/8 | 23 | 23.5 |
| #7 | 0.875 | 1 | 34 | 34.5 |
| #8 | 1.000 | 1-1/8 | 38 | 38.5 |
Two details matter more than they look. The hole floor does not need to be flat, but it must not be cone-shaped from a worn bit, because a cone leaves the bottom of the bar sitting in air. And the hole must be blown out before it is measured: a 16.5 in hole holding an inch of dust is a 15.5 in hole.
Volume is the annulus between hole wall and bar, times the drilled depth. It is worth doing once per bar size, because the result is not intuitive: a #4 hole is small but there are a lot of them, and the count falls fast as the bar grows — the annulus climbs with the square of the diameter while the cartridge does not.
| Bar | Adhesive per hole (in³) | Holes per 10 oz | Holes per 20 oz | Holes per 600 ml |
|---|---|---|---|---|
| #3 | 1.07 | 16.8 | 33.6 | 34.1 |
| #4 | 1.82 | 9.9 | 19.8 | 20.1 |
| #5 | 2.63 | 6.9 | 13.7 | 13.9 |
| #6 | 3.75 | 4.8 | 9.6 | 9.8 |
| #7 | 6.35 | 2.8 | 5.7 | 5.8 |
| #8 | 8.03 | 2.2 | 4.5 | 4.6 |
Scale it once to see where the money goes: a 100-hole #4 job is 182 in³, about ten 10 oz cartridges. The same 100 holes in #6 is 375 in³, about twenty-one.
Concrete dust is fine, it coats the wall of the hole, and adhesive bonds to dust instead of to concrete. That is the failure nobody sees: the bar looks set, the adhesive looks solid, and the dowel comes out at a fraction of its rated capacity. Every adhesive system's evaluation report spends more words on cleaning than on mixing, for that reason.
The sequence that works is short: nylon brush sized to the hole, in and out with a twist, blow from the bottom with oil-free air, then brush and blow again. Two rounds, not one. If the hole will sit overnight, cap it. What does not work: rinsing with water, blowing with a compressor that carries oil, or trusting the drill's own spiral to clear the hole — each leaves a film on the wall, and the film is what the adhesive grabs.
The bit has to match the anchor, not the drill you happen to own: a carbide-tipped masonry bit in the nominal size from the table — 5/8 in for #4 — run in a hammer drill, not a rotary-only drill and not a wood bit. The hole does not need to be pretty; it needs to be round, full depth and the right diameter.
Two substitutions are common and both cost capacity. A diamond core bit cuts a cleaner hole but leaves a smoother wall, and the adhesive tables are written around the rougher hammer-drilled surface; if you core, the bonded length becomes a question for the system's evaluation report, not for this page. And an SDS bit that has drilled a hundred holes is no longer 5/8 in — worn bits drill undersize, which is the one direction you cannot fix afterwards.
Hitting steel is the other half of the bit question. A hammer drill that stops cutting and starts ringing has found existing reinforcement. Stop, back out, and shift the hole along the joint; do not cut the bar and do not drill past it. A hole that is half in concrete and half in steel has no adhesive ring on the steel side, and that side carries nothing.
Horizontal and overhead holes add one more rule: a thixotropic, non-sag adhesive, injected from the back of the hole with a nozzle long enough to reach it, and a stop collar or tape wrap on the bar to hold it at the bonded depth while it sets. A sagging adhesive leaves the top of the hole empty, which is the side that carries the tension.
A 20 ft wall on a footing, #4 dowels at 16 in on centre, Grade 60 bar in 4,000 psi concrete, uncoated, Class B lap into the wall:
Three sticks deliver 18 pieces for 16 dowels, so the offcut is two 37 in pieces — the same remainder arithmetic as the cut-length table. At 12 in on centre the wall takes 21 dowels, which is 3.5 sticks → 4 sticks, and the same 37 in cut.
A short bar cast into one pour and left standing for the next. The buried end develops in the first pour, the exposed end laps into the second, and together they carry the load across the construction joint. It is not a lap splice (both ends of a lap are bonded into the same pour) and not an anchor (an anchor holds a load in one pour).
At least the development length. For Grade 60 bar in 4,000 psi concrete the simplified ACI 318-19 values are 16 in for #4, 19 in for #5, 23 in for #6 and 38 in for #8, with 12 in as the code minimum for any bar. Epoxy coating raises those by 20% to 50%. The embedded end has to reach yield before the joint can slip, so a dowel set 6 in into a footing is not a dowel — it is a pin.
Cut length = embedment + projection. For a #4 dowel in 4,000 psi concrete with a Class B lap into the wall, that is 16 in + 21 in = 37 in. The projection is not a guess: if it laps the next pour’s bars it must be a lap length, and Class B (1.3 × ld) is the default when nobody has checked the splice class.
Whatever the drawing calls for, and it has to match the bars it laps with. A #4 dowel laps #4 wall bars; matching sizes keeps the lap a lap. The size then sets the embedment through the development length — #4 needs 16 in, #6 needs 23 in at 4,000 psi — so the smaller bar is not automatically the cheaper detail once you count the cut length.
One divided by the spacing in feet: at 12 in on centre it is 1.00 per foot, at 16 in it is 0.75, at 18 in 0.67 and at 24 in 0.50. Count a joint the same way a mat is counted — floor(clear length ÷ spacing) + 1 — so a 20 ft wall at 16 in o.c. takes 16 dowels, not 15.
Not as a pavement dowel. Contraction and expansion joints need smooth dowel bars, usually with one end coated or sleeved as a bond breaker, so the joint can open and close. A deformed rebar dowel bonds to both slabs, locks the joint, and the crack appears somewhere else. Deformed rebar dowels belong in structural joints — footing to wall, slab to wall — where the two pours are meant to act as one.
Only where the detail calls for corrosion protection — exterior walls, marine or deicing exposure, thin cover. Coating is not free structurally: the epoxy factor is 1.2 when cover and spacing are generous and 1.5 otherwise, so a #4 dowel’s embedment goes from 16 in to 19 in or 23 in. Coating changes the embedment, not the cut length logic.
One 20 ft stick is 240 in, and the count is floor(240 ÷ cut length). At a 37 in cut that is 6 dowels with 18 in left over; at 20 in it is 12 with nothing left; at 30 in it is 8. The remainder is why dowel orders are counted in whole sticks — the offcut is not a percentage.
5/8 in. The rule is 1/8 in over the bar: #4 is 1/2 in, so the bit is 5/8 in. #3 goes in 1/2 in, #5 in 3/4 in, #6 in 7/8 in, #7 in 1 in and #8 in 1-1/8 in. The gap is the adhesive, so it is not clearance you can spend.
The bonded length plus 1/2 in. For #4 Grade 60 in 4,000 psi concrete the bonded length is 16 in, so the hole is 16.5 in. The spare half inch stops the bar bottoming out on dust or a rough hole floor, which would leave it short of the depth you designed.
Nine #4 holes. A 5/8 in hole around a 1/2 in bar leaves 0.110 in² of adhesive, times 16.5 in of depth is 1.82 in³ per hole, and a 10 oz cartridge is 18.05 in³. #3 gives about 17 holes, #5 about 7, #6 about 5, #8 about 2.