Domestic Hardwoods: A Builder's Guide to North American Species
Janka hardness, specific gravity, stiffness and shrinkage for thirteen North American species, the three places the hardness ranking will mislead you, and the arithmetic that tells you what a tabletop is actually going to move.
Almost every hardwood guide on the internet ranks species by one number, and it is the wrong number. Janka hardness measures how far a steel ball sinks into a board. It does not measure stiffness, it does not measure bending strength, and it has nothing at all to say about the thing that actually wrecks furniture, which is seasonal movement. This guide puts all five properties side by side, out of the USDA Forest Products Laboratory Wood Handbook, and shows you the three cases where following the hardness ranking gets you a worse result than ignoring it.
I have been putting this one off for a long time, because the topic is enormous and I did not want to write another chart with a hardness column and some adjectives. Most of us pick wood the way we pick paint. Walk in, look at the rack, grab something that seems decent and fits the budget. For a garage shelf that works fine. For furniture that has to live in somebody's house for twenty years, the species is a set of engineering decisions you are making whether you know it or not.
I found that out the expensive way. Early on I built a dining table out of red oak. Weeks of work. My wife set a glass of ice water down at dinner and inside forty minutes there was a ring in the surface that no amount of refinishing ever fully hid. A film finish would have prevented it, and I know that now, but at the time I did not know that red oak has open vessels running the length of the board with nothing plugging them, and that white oak, sitting on the same rack for about twenty percent more, does not. That is not a taste difference. It is an anatomical one, and it is measurable.
So this is the version of that article I wish somebody had handed me. Practical decisions first, then the material science, and the science is not decoration. Every number below traces to a specific table in a specific document, and where the common wisdom disagrees with the data, I have said so and shown the numbers.
The five numbers that actually decide a species
A species is not one property. It is at least five, they are measured independently, and they do not move together. Getting good at wood selection is mostly a matter of knowing which one governs the thing you are building, because for any given project four of the five do not matter much and one of them decides everything.
Ranges shown span the thirteen species tabulated in this article. Nothing here is a design value; these are clear-wood material properties.
Specific gravity is the root property, the oven-dry mass of the wood divided by the mass of an equal volume of water. Nearly everything else correlates with it, which is why it is the value structural codes actually key off. Hardness is resistance to denting. Modulus of elasticity is stiffness, meaning how much a board deflects under load, and it is what governs a shelf. Modulus of rupture is bending strength, meaning the stress at which the board breaks, which is a different question from how far it bends first. Shrinkage is dimensional change with moisture, and it is the property that opens joints, cups panels and cracks tabletops.
Here is the part that gets skipped. Those properties are measured on small, clear, straight-grained specimens in a laboratory. The board on the rack has knots, grain deviation and a real moisture content. The Wood Handbook values are excellent for comparing species against each other, which is what this article does with them. They are not design values, and the difference matters enough that I gave it its own section further down.
How the Janka hardness test actually works
The test is beautifully simple, and its simplicity is exactly why people over-read it. A steel ball is pressed into the face of a board until it is embedded to half its diameter. The force required, in pounds-force, is the Janka number. One ball, one direction, one measurement.
The ball diameter is 0.444 inches, and that number is not arbitrary. At half embedment the circle of contact is the ball's full diameter, so the projected area works out to
So the test was designed to be an honest pressure measurement, not an arbitrary index. What gets published as a species Janka value is side hardness, load applied perpendicular to the grain, and the Wood Handbook figure is the average of readings taken on the radial and tangential faces. End hardness, pressing into the end grain, runs higher and is almost never what a chart is quoting.
The ball diameter is not arbitrary. At half embedment the projected contact area is 0.1548 in², which is one square centimetre, so the reading is a pressure. Side hardness is the average of readings on the radial and tangential faces.
Now the useful question. Across the thirteen species in this article, what does the Janka number actually predict? I ran the correlations against the other Wood Handbook properties for the same thirteen species:
| Relationship | r2 | What that means |
|---|---|---|
| Janka vs specific gravity | 0.88 | Hardness is essentially a density measurement in disguise |
| Janka vs bending strength | 0.70 | Related, but a third of the variation is something else |
| Janka vs stiffness | 0.64 | Weak. Hardness is a poor proxy for how far a board bends |
| Stiffness vs bending strength | 0.87 | These two do travel together |
Read that top row again. Hardness is density. If you know the specific gravity of a domestic hardwood you can predict its Janka number to within about seven percent, and the Janka column adds almost no information that the density column did not already give you. What it does not give you is stiffness, and stiffness is what governs the single most common hardwood failure I get asked about, which is a shelf that sags.
The species property table
This is the reference I actually wanted to exist. Thirteen species, seven properties, all at 12% moisture content, all from one table so the numbers are mutually consistent. Scroll it sideways on a phone.
| Species | Specific gravity | Janka lbf | Stiffness MOE million lbf/in2 | Bending MOR lbf/in2 | Crushing lbf/in2 | Shear lbf/in2 |
|---|---|---|---|---|---|---|
| Pignut hickory Carya glabra | 0.75 | 2,140 | 2.26 | 20,100 | 1,980 | 2,150 |
| Shagbark hickory Carya ovata | 0.72 | 1,880 | 2.16 | 20,200 | 1,760 | 2,430 |
| Pecan hickory Carya illinoinensis | 0.66 | 1,820 | 1.73 | 13,700 | 1,720 | 2,080 |
| Hard maple Acer saccharum | 0.63 | 1,450 | 1.83 | 15,800 | 1,470 | 2,330 |
| White oak Quercus alba | 0.68 | 1,360 | 1.78 | 15,200 | 1,070 | 2,000 |
| White ash Fraxinus americana | 0.60 | 1,320 | 1.74 | 15,400 | 1,160 | 1,910 |
| American beech Fagus grandifolia | 0.64 | 1,300 | 1.72 | 14,900 | 1,010 | 2,010 |
| Red oak, northern Quercus rubra | 0.63 | 1,290 | 1.82 | 14,300 | 1,010 | 1,780 |
| Yellow birch Betula alleghaniensis | 0.62 | 1,260 | 2.01 | 16,600 | 970 | 1,880 |
| Black walnut Juglans nigra | 0.55 | 1,010 | 1.68 | 14,600 | 1,010 | 1,370 |
| Black cherry Prunus serotina | 0.50 | 950 | 1.49 | 12,300 | 690 | 1,700 |
| Soft maple, red Acer rubrum | 0.54 | 950 | 1.64 | 13,400 | 1,000 | 1,850 |
| Yellow-poplar Liriodendron tulipifera | 0.42 | 540 | 1.58 | 10,100 | 500 | 1,190 |
Three places the hardness ranking lies to you
Sort that table by Janka and you get one order. Sort it by stiffness and you get a different one. Sort it by stability and you get a third. Here are the three specific disagreements that cost people real money.
Four species are highlighted because their rank moves furthest between columns. Yellow-poplar is last in hardness and mid-table in stiffness. Pecan hickory is third in hardness and eighth in stiffness. Cherry is last in stiffness and first in stability. Sorting a chart by Janka tells you almost nothing about the other two columns.
1. The softest wood on the list makes a stiffer shelf than cherry
Yellow-poplar is 540 Janka. Black cherry is 950, which is 76% harder. Every chart on the internet will tell you to build the shelf out of cherry. Run the deflection and the ranking inverts, because a shelf does not fail by denting, it fails by sagging, and sagging is governed by modulus of elasticity, where poplar is 1.58 million and cherry is 1.49 million.
The same inversion runs the other way too. Yellow birch is 1,260 Janka, ninth of the thirteen, and 2.01 million stiffness, third of the thirteen. Pecan hickory is the mirror image: third hardest at 1,820 Janka, but eighth in stiffness at 1.73 million, below white oak and white ash. Buying pecan because the hardness chart put it near the top gets you a floor that resists denting beautifully and a shelf that is unremarkable.
2. Walnut is a stronger beam than red oak
Black walnut is 1,010 Janka against northern red oak's 1,290, so walnut reads as 22% softer and gets treated as the delicate one. Its modulus of rupture is 14,600 lbf/in2 against red oak's 14,300. Walnut is the stronger of the two in bending. It will dent more easily and it will break later, and those are simply different properties.
Walnut does have a real weakness, and it is not the one people warn you about. Its shear strength parallel to grain is 1,370 lbf/in2, the lowest of any of the furniture species in the table and 41% below hard maple. Shear parallel to grain is what resists a tenon cheek tearing out, a dovetail pin shearing off at its base, or a screw splitting away a strip of end grain. So the correct caution about walnut is not "go easy, it is soft". It is "give the joinery more glue surface than you would in maple, and do not skimp on tenon length".
3. Cherry dents because of a property nobody quotes
Cherry's compression perpendicular to grain is 690 lbf/in2. Hard maple's is 1,470, more than twice as much. That is the property that governs crushing, and it explains two things at once: why a cherry tabletop picks up marks from a dropped bowl, and why cherry needs cauls under clamps when maple shrugs them off. Yellow-poplar is worse again at 500, which is exactly why poplar makes a stiff shelf and a terrible workbench top. Stiff and dent-resistant are unrelated properties and poplar splits the difference in the most confusing possible way.
The rule I use now: pick the property before you pick the species. Shelf or rail that must not sag, use stiffness. Surface that takes impact, use hardness. Wide panel or tabletop, use tangential shrinkage. Joinery-heavy piece in a soft species, check shear. Only one of those is the Janka column.
Why wood moves, and how much
Every cracked tabletop, every drawer that sticks in August, every panel that rattles in January, traces to one mechanism. Wood is hygroscopic. It exchanges moisture with the air around it forever, it swells when it takes moisture up and shrinks when it gives it off, and no finish stops that. Finishes slow the rate. Nothing prevents the exchange.
Below roughly 30% moisture content, the fibre saturation point, that exchange changes the dimensions of the board. Above it, the extra water sits in the cell cavities and the board does not move at all. So all the movement you care about happens in the range your shop actually lives in.
Tangential and radial shrinkage values are white oak, Wood Handbook Table 4-3. Movement figures are computed for a 36 in width over 5.8 points of moisture content. Longitudinal movement is real but around 0.1% to 0.2% green to ovendry, which is why nobody designs for it.
Wood does not move equally in every direction, and the three directions are not close. Along the grain, movement is so small it can be ignored for furniture. Radially, from the pith outward, a species might shrink 4 to 7% going all the way from green to oven dry. Tangentially, following the growth ring around, it shrinks roughly twice that. Sugar maple is 4.8% radial and 9.9% tangential. American beech is 5.5% and 11.9%.
The arithmetic, in a form you can actually use
Published shrinkage values are green to oven dry, which is a range no piece of furniture ever experiences. The Wood Handbook gives the relation for converting them to real conditions, and it is a straight line: shrinkage is proportional to moisture content below fibre saturation.
Differentiate that and you get the number that is genuinely useful in a shop, which no chart seems to publish: each single point of moisture content is worth S0 / 30 percent of the board's width. For flatsawn stock use the tangential value, for quartersawn use the radial one. White oak flatsawn is 10.5 / 30 = 0.35% of its width per point of moisture content. Quartersawn, it is 5.6 / 30 = 0.187%.
How many points does a shop actually swing?
You need the moisture swing to finish the calculation, and this is where most advice goes vague. The Wood Handbook tabulates equilibrium moisture content directly against temperature and relative humidity. At 70 degrees Fahrenheit, which is close enough for any heated shop:
Wood Handbook Table 4-2. EMC barely changes with temperature across normal shop conditions, so this one curve covers roughly 30 to 80 °F. The curve is shallow through the middle and steep above 80% relative humidity, which is why an uncontrolled damp shop is so much worse than a slightly damp one.
| Relative humidity | Wood EMC | Typical when |
|---|---|---|
| 20% | 4.5% | Hard winter, forced-air heat, no humidifier |
| 30% | 6.2% | Normal heated interior in winter |
| 40% | 7.7% | Spring and autumn |
| 50% | 9.2% | Conditioned space, year round |
| 65% | 12.0% | Unconditioned shop in summer |
| 80% | 16.0% | Humid summer, no dehumidifier |
So a piece that lives in a heated house through winter and an unconditioned one in summer swings from about 6.2% to about 12.0%, which is 5.8 points of moisture content. That is the number I use for everything below. If your house is conditioned year round, halve it. If your shop is a barn, increase it.
What that does to a 36 inch tabletop
Three quarters of an inch is not a rounding error. It is why a tabletop gets fastened with buttons or slotted cleats and never glued to its base, why a frame and panel door floats its panel instead of trapping it, and why a breadboard end is pinned in slots. Those are not traditions. They are 0.73 inches of unavoidable annual travel, designed around.
Drawn in section. Which detail you choose is a question of load and appearance, not of how much movement you expect: all four accommodate the full seasonal travel if the slots are cut long enough. Size the slot from the movement figure for your species and width, not from habit.
Every one of those details does the same job in a different place: it holds the part where it belongs in one direction while letting it travel in the other. Once you have the movement number for your species and width, choosing between them is straightforward, and getting it wrong is what splits a top down a glue line in the second winter.
| Species | Tangential green to ovendry | Radial green to ovendry | T/R ratio | Flatsawn inches | Quartersawn inches |
|---|---|---|---|---|---|
| American beech | 11.9% | 5.5% | 2.16 | 0.83 | 0.38 |
| Pignut hickory | 11.5% | 7.2% | 1.60 | 0.80 | 0.50 |
| White oak | 10.5% | 5.6% | 1.88 | 0.73 | 0.39 |
| Shagbark hickory | 10.5% | 7.0% | 1.50 | 0.73 | 0.49 |
| Hard maple | 9.9% | 4.8% | 2.06 | 0.69 | 0.33 |
| Yellow birch | 9.5% | 7.3% | 1.30 | 0.66 | 0.51 |
| Pecan hickory | 8.9% | 4.9% | 1.82 | 0.62 | 0.34 |
| Red oak, northern | 8.6% | 4.0% | 2.15 | 0.60 | 0.28 |
| Soft maple, red | 8.2% | 4.0% | 2.05 | 0.57 | 0.28 |
| Yellow-poplar | 8.2% | 4.6% | 1.78 | 0.57 | 0.32 |
| Black walnut | 7.8% | 5.5% | 1.42 | 0.54 | 0.38 |
| White ash | 7.8% | 4.9% | 1.59 | 0.54 | 0.34 |
| Black cherry | 7.1% | 3.7% | 1.92 | 0.49 | 0.26 |
Two things that table corrects
White oak is not the stable one. It is the third most active species on this list flatsawn, moving 0.73 inches where cherry moves 0.49. White oak has an excellent reputation for stability and it is half deserved: its T/R ratio of 1.88 is better than red oak's 2.15, so it distorts less. But total movement and distortion are two different questions, and on total movement white oak is near the top of the list. I had this wrong in the previous version of this article and the Wood Handbook numbers are what corrected me.
Black cherry is the most dimensionally stable species here. Not walnut. Walnut wins the T/R ratio at 1.42, which is why it stays flat and resists cupping, and that is a real advantage for a wide top. But cherry moves the least in absolute terms, 0.49 inches against walnut's 0.54, and considerably less than either oak. Cherry has quietly been the stability champion the whole time while walnut collected the credit.
T/R ratio versus tangential magnitude, and why you need both: the ratio predicts distortion, meaning cup, crook and twist, because it measures how unequally the board moves in its two cross-grain directions. The tangential number predicts total width change, which is what your joinery has to absorb. A species can be good at one and poor at the other. Yellow birch has the best ratio in the table at 1.30 and still moves 0.66 inches.
Grain, pores and why finishing goes wrong
Everything in the finishing section of a normal hardwood guide is a consequence of cell structure, and once you can see the structure the advice stops being a list of rules to memorise. Hardwoods are classified by how the vessels, the pores you can see on end grain, are distributed through a growth ring.
Porosity classes as defined in Wood Handbook Chapter 3, which describes ring-porous vessel diameter dropping by an order of magnitude or more across the growth ring. The pore sizes here are schematic and drawn to show the pattern, not to scale.
Ring-porous species put a band of very large vessels down at the start of each season's growth and then switch abruptly to dense fibrous latewood. The Wood Handbook describes the vessel diameter dropping "often by an order of magnitude or more" across that transition. Red oak, white oak, ash and hickory are all ring-porous. That earlywood band is why oak looks like oak, and it is why a film finish sprayed straight onto unfilled oak still feels like corduroy after four coats. You are not feeling a finish defect, you are feeling vessels that the finish sank into. Grain filler exists for exactly this.
Diffuse-porous species spread small vessels evenly through the ring with no sharp earlywood to latewood boundary. Maple, birch, aspen and yellow-poplar are the Wood Handbook's examples. Nothing to fill, so these go straight to topcoat. The trade-off arrives with pigmented stain: the density variation is gradual rather than banded, which means stain penetration varies smoothly and unpredictably across the surface. That is blotching. It is a diffuse-porous problem, and it is why hard maple and cherry blotch while oak, which is far more open, does not.
Semi-ring-porous falls between. Black walnut is the Wood Handbook's named temperate example, and it explains walnut's finishing behaviour precisely: enough pore structure to take a light fill if you want a mirror surface, not enough to demand it.
| Species | Pore structure | Grain filler | Pigmented stain | What I actually do |
|---|---|---|---|---|
| Red oak | Ring-porous | Required for a smooth film finish | Takes it evenly | Fill, then stain, then topcoat |
| White oak | Ring-porous, heartwood vessels plugged with tyloses | Required, though the tyloses reduce depth | Takes it evenly | Clear finish on quartersawn, fill if I want gloss |
| White ash | Ring-porous | Required for a smooth film finish | Takes it well | Fill, or lean into the open texture with oil |
| Hickory | Ring-porous | Required | Uneven, heartwood and sapwood differ sharply | Clear, and design around the colour variation |
| Black walnut | Semi-ring-porous | Optional, light fill for gloss | Rarely needed | Clear with a UV inhibitor |
| Hard maple | Diffuse-porous | Not needed | Blotches badly | Clear only, or dye if colour is unavoidable |
| Black cherry | Diffuse-porous | Not needed | Blotches | Clear, and let UV do the colouring |
| Yellow birch | Diffuse-porous | Not needed | Blotches | Dye and a washcoat, or clear |
| Yellow-poplar | Diffuse-porous | Not needed | Blotches, and the green streaks show through | Prime and paint |
Telling the species apart at the rack
Two of these confusions cost real money, and both have a definitive answer.
Red oak versus white oak
The Wood Handbook is blunt about the mechanism. White oak heartwood pores "are usually plugged with tyloses, which tend to make the wood impenetrable to liquids", which is why white oak makes tight cooperage and red oak does not. Red oak "generally lacks tyloses in the pores". Tyloses are balloon-like growths that push into the vessel from the surrounding cells and seal it shut.
The Wood Handbook states that white oak heartwood pores are usually plugged with tyloses, which tend to make the wood impenetrable to liquids, and that red oak generally lacks them. The blow test follows directly from that. It is a shop test rather than a standard, and it is worth running on your own boards: the handbook notes that many chestnut oak pores lack tyloses despite chestnut oak sitting in the white oak group.
That single anatomical fact generates every practical difference between the two woods. It is why white oak heartwood is rated Resistant for decay and red oak is not in that group at all. It is why white oak goes outdoors and on boats. It is why my ice water left a ring in a red oak tabletop and would not have in white oak. And it gives you a shop test: cut a short offcut, put one end in soapy water and blow through the other. Red oak passes air and you get bubbles. White oak heartwood generally will not.
The exception worth knowing: "white oak" is a group of species, not one tree, and the Wood Handbook notes that many chestnut oak heartwood pores lack tyloses even though chestnut oak sits in the white oak group. If watertightness is the reason you are buying white oak, the blow test on your actual boards beats the label on the rack.
Hard maple versus soft maple
These are sold within a few dollars of each other and are genuinely different materials. Hard maple is sugar maple at 1,450 Janka and 0.63 specific gravity. Soft maple, most often red maple, is 950 Janka and 0.54. That is a 34% drop in hardness. In stiffness the gap is much smaller, 1.83 against 1.64, only 10%, which is the useful part: for a shelf or a rail, soft maple gives you most of hard maple's performance for less money and far less tool wear. For a workbench top or a cutting board, where compression perpendicular to grain governs, hard maple's 1,470 against soft maple's 1,000 is the number that matters and the substitution is a bad one.
What you can and cannot calculate with these numbers
This is the section I care most about, because it is where the internet's hardwood content and actual engineering practice diverge completely.
Everything above comes from tests on small, clear, straight-grained specimens. Those are material properties. They are not design values. A design value has a grade-based reduction for knots and grain deviation baked into it, plus a safety factor, and it is what a code-compliant calculation has to use. For domestic hardwoods, published design values exist for a surprisingly short list.
| Species group | Bending Fb | Shear Fv | Compression ⊥ Fc⊥ | Modulus E | Specific gravity G |
|---|---|---|---|---|---|
| Beech-Birch-Hickory | 1,450 | 195 | 715 | 1,700,000 | 0.71 |
| Northern Red Oak | 1,400 | 220 | 885 | 1,400,000 | 0.68 |
| Red Maple | 1,300 | 210 | 615 | 1,700,000 | 0.58 |
| White Oak | 1,200 | 220 | 800 | 1,100,000 | 0.73 |
| Red Oak | 1,150 | 170 | 820 | 1,400,000 | 0.67 |
| Mixed Oak | 1,150 | 170 | 800 | 1,100,000 | 0.68 |
| Mixed Maple | 1,000 | 195 | 620 | 1,300,000 | 0.55 |
| Yellow Poplar | 1,000 | 145 | 420 | 1,500,000 | 0.43 |
| Cottonwood | 875 | 125 | 320 | 1,200,000 | 0.41 |
| Aspen | 875 | 120 | 265 | 1,100,000 | 0.39 |
Ten groups. That is all of them. Black cherry, black walnut and white ash do not appear in Table 4A at all, at any grade. Neither does any species outside those ten. So if you are sizing a structural member out of walnut, there is no published reference design value to size it with, and no amount of Wood Handbook clear-wood data substitutes, because clear-wood values carry no reduction for the knot that is actually in your board.
For furniture this rarely bites, because furniture members are governed by stiffness and serviceability rather than by strength, and a deflection calculation with a clear-wood modulus and an honest safety margin is a reasonable engineering approach. For anything holding up a structure, it matters a great deal. Our own engineering methodology page sets out how the analysis behind our plan reviews handles it.
Creep is the factor people forget. A shelf does not stop at its calculated deflection. Under sustained load, wood keeps deforming. The NDS applies a factor of 1.5 to the long-term component of deflection for seasoned lumber in dry service. So the 36 inch cherry shelf that calculates to 0.032 inches settles at roughly 0.047 inches over years. Design for the number after creep, not before.
Span beats species, every single time
Here is the comparison that ends most shelf arguments. Same 30 lb load, same 3/4 inch stock, deflection after creep, against the L/360 line commonly used for serviceability.
Load held constant at 30 lb total, so deflection goes with the cube of the span. The three species curves sit within a few thousandths of each other while the span axis moves them by a factor of eight. Tabulated modulus raised 10% per Wood Handbook Table 5-3 note c, then multiplied by the NDS creep factor of 1.5 for seasoned lumber in dry service.
| Clear span | L/360 limit | Yellow-poplar | Black cherry | Hard maple |
|---|---|---|---|---|
| 30 in | 0.083 in | 0.026 in | 0.028 in | 0.022 in |
| 36 in | 0.100 in | 0.045 in | 0.047 in | 0.039 in |
| 48 in | 0.133 in | 0.106 in | 0.113 in | 0.092 in |
| 60 in | 0.167 in | 0.207 in | 0.220 in | 0.179 in |
Going from cherry to hard maple at a 48 inch span buys you 19% less sag. Shortening the span from 48 inches to 36 inches buys you 58% less sag, in whatever species you already own. One extra support is worth more than any species upgrade on the shelf. If you want to run your own numbers, our shelf sag calculator does this arithmetic for a specific shelf.
The species, one at a time
Now the part everybody actually came for, with the data threaded through it.
Red oak, the default
Red oak is what "hardwood" means in America. It is at every yard, it is the cheapest of the serious furniture species, and at 1,290 Janka and 1.82 million stiffness it is genuinely capable. Its stiffness is fifth of the thirteen here, ahead of white oak, which surprises people.
Two things will bite you. The open ring-porous vessels have to be filled before a film finish or the surface will telegraph every pore. And red oak carries high tannin with no tyloses to keep moisture out of those vessels, which is the combination that put a ring in my table. The same tannins react with iron to form iron tannate, a permanent blue-black dye, so steel wool residue followed by a water-based finish gives you stains that do not sand out. I have done that too.
On movement, red oak flatsawn moves 0.60 inches across 36 inches, which is mid-pack, but its T/R ratio of 2.15 is the second worst in the table. It moves a moderate amount and it moves unevenly, which is precisely the recipe for a cupped flatsawn panel.
White oak, the one that handles water
Tyloses. That is the whole story and everything else follows from it. Plugged vessels give white oak heartwood a Resistant decay rating, watertightness for cooperage, and genuine outdoor capability. At 1,360 Janka it is slightly harder than red oak, and at 2,000 lbf/in2 shear it is meaningfully stronger in the direction joinery cares about.
Quartersawn white oak shows the medullary rays as broad flecks across the face, which is the Craftsman and Mission look, and the Wood Handbook notes that conspicuous rays are the identifying mark of quartersawn oak generally. What people underrate is that quartersawing white oak is not just a look: it takes the movement from 0.73 inches to 0.39 across a 36 inch top. If you are building a wide top in white oak, quartersawn stock is an engineering decision that happens to be beautiful.
Expect to pay 15% to 30% over red oak, more for quartersawn.
Hard maple, the wear surface
Hard maple earns its place on two numbers: 1,450 Janka, and compression perpendicular to grain of 1,470 lbf/in2, the second highest of the furniture species here. That combination is why every commercial butcher block, bowling lane and gym floor is maple. It resists both denting and crushing, which are different failures.
It is also 2,330 lbf/in2 in shear, the highest in the table, so maple joinery is forgiving. The costs are real though. Maple is hard on cutting edges, and figured maple will tear out without mercy unless your irons are genuinely sharp. And as a diffuse-porous species it blotches with pigmented stain badly enough that I stopped trying. Dye if you must colour it, otherwise clear finish and let it amber on its own.
Black cherry, the quiet engineering winner
Cherry is my favourite and I have already admitted the bias, but the data backs it more than I expected. It is the most dimensionally stable species in this article, moving 0.49 inches where white oak moves 0.73. It machines beautifully, a sharp plane leaves a surface ready for finish, and the UV darkening from pale pink to deep reddish brown over six to twelve months is a genuine photochemical change, not a stain.
The honest weaknesses are both in the table. At 1.49 million, cherry has the lowest stiffness here, so cherry shelves sag more than anything else on the list. And compression perpendicular of 690 lbf/in2 is the lowest of the furniture species, so it dents. For casework, boxes, doors and desks those two barely matter. For a shelf spanning four feet or a workbench top, they matter a lot.
One thing to warn a client about: leave a book on a cherry top for three months and you will lift it to find a pale rectangle. It evens out, but it takes months.
Black walnut, the one with a hidden weak axis
Walnut is the prestige species and the numbers mostly justify it. Modulus of rupture 14,600 lbf/in2, higher than red oak. T/R ratio 1.42, the second best here, so wide walnut tops stay flat. Semi-ring-porous, so it finishes without the fight oak puts up. Heartwood rated Resistant for decay, the same group as white oak, which almost nobody knows.
The weak axis is shear, 1,370 lbf/in2, lowest of the furniture species and 41% below maple. Give tenons more length and dovetails more shoulder in walnut than you would in maple, and treat short-grain anywhere in the design as a real risk rather than a theoretical one.
Two practical notes. Nearly every walnut board carries pale sapwood, so buy for yield and plan the layout. And walnut fades under UV while cherry darkens, so a top that starts as dark chocolate drifts to medium brown unless the finish carries a UV inhibitor.
The supporting cast
Hickory is not one wood. True hickories, shagbark at 1,880 Janka and pignut at 2,140, are the hardest and stiffest domestic species you will meet. Pecan hickory, at 1,820 Janka, is nearly as hard but noticeably less stiff at 1.73 million and much lower in bending strength, 13,700 against shagbark's 20,200. Charts that print "hickory, 1,820" are quoting pecan and implying the true hickories. All of them chew tool edges and tear out unpredictably.
White ash works like a friendlier oak: 1,320 Janka, 15,400 bending, and a T/R ratio of 1.59 that is better than either oak. It is the traditional choice for tool handles and steam bending. Emerald ash borer has devastated eastern populations and supply is genuinely uncertain, so buy it when you see it and do not design a product line around it.
Yellow-poplar is a hardwood botanically and soft in practice at 540 Janka, but 1.58 million stiffness puts it ahead of cherry. It is the correct answer for paint-grade work, drawer boxes, jigs and secondary structure, and it is a better shelf than its reputation.
Soft maple at 950 Janka is not a hard maple substitute for wear surfaces, but it takes stain more evenly and costs less. Yellow birch is the sleeper: 2.01 million stiffness, third here, at 1,260 Janka and a modest price. American beech is strong and cheap and moves more than anything else in this article at 0.83 inches, which is why it ended up in tool handles and factory furniture rather than in wide tops.
Buying: board feet, thickness and what cheap lumber really costs
Hardwood is sold by the board foot, which is 144 cubic inches: a piece 12 inches square and 1 inch thick. Thickness comes in quarters of an inch of rough sawn stock, so 4/4 is nominally one inch, 8/4 is two. Surfacing takes its cut, and 4/4 surfaced two sides typically lands at 13/16. Buy 4/4 for a 3/4 inch part and you have almost nothing left for flattening a board that arrived with a bow.
Grade is the other axis, and it is a yield specification rather than a quality judgement. FAS gives you long clear cuttings, Common grades give shorter ones for less money, and if your parts are short the lower grade is often the cheaper route to the same finished piece. Our guide to FAS, Selects and Common grades covers how the yields work and how to decide.
| Species | Approx. $/BF | Relative to red oak | Availability |
|---|---|---|---|
| Yellow-poplar | $3 to $5 | 0.6× | Everywhere |
| Red oak | $5 to $8 | 1.0× | Everywhere |
| Soft maple | $5 to $8 | 1.0× | Good |
| White ash | $5 to $9 | 1.1× | Declining, borer pressure |
| Hickory | $6 to $10 | 1.2× | Good |
| Hard maple | $7 to $11 | 1.4× | Good |
| White oak | $8 to $13 | 1.6× | Good, rising |
| Black cherry | $8 to $14 | 1.7× | Good |
| Black walnut | $12 to $20+ | 2.5× | Moderate, wide stock at a premium |
The thing I would tell my younger self: cheap lumber is priced per board foot and paid for in shop hours. Warped stock, hidden checks, wild grain that tears whatever direction you come at it. By the time you add the wasted material and the extra milling, the saving is gone. If the piece is seen and used daily, buy the species that suits it. If it is painted or hidden, that is where poplar and soft maple make their money.
Matching species to project
The decision is always the same shape: work out which property governs, then pick from the column that ranks well on it.
Stiffness from Wood Handbook Table 5-3b, movement computed for a 36 in flatsawn width over 5.8 points of moisture content. The stiff-and-stable corner is nearly empty, which is the honest summary of hardwood selection: you are almost always trading one against the other, and quartersawing is the only move that improves the vertical axis without changing species.
The five mistakes I still see most often
Ranking on Janka. It correlates with density at r2 = 0.88 and with stiffness at 0.64. If your failure mode is sagging, hardness is close to irrelevant.
Treating kiln-dried as a service condition. Six to eight percent is where the board left the kiln. In an unconditioned summer shop it is heading for twelve. Sticker your stock and give it at least a week, and check it with a meter if the piece matters.
Gluing a wide top to its base. Three quarters of an inch of annual travel does not negotiate. Buttons, slotted cleats or figure-eight fasteners, always.
Pigmented stain on a diffuse-porous species. Maple, cherry and birch blotch for structural reasons that no amount of careful application fixes. Dye and a washcoat, or clear.
Red oak anywhere damp. Open vessels plus high tannin. It is not a durability grade issue, it is an anatomy issue, and white oak is sitting on the next rack over.
Frequently asked questions
Pignut hickory at 2,140 lbf Janka, followed by shagbark hickory at 1,880 and pecan hickory at 1,820. Among the furniture species you will actually find at a yard, hard maple leads at 1,450 and white oak follows at 1,360. Black locust and osage-orange are harder still but are not commercially sold as furniture lumber in most regions.
No, and for most projects it is not even in the top two. Across thirteen domestic hardwoods, Janka correlates with specific gravity at r-squared 0.88, meaning it is largely a density measurement, but only 0.64 with stiffness. Stiffness governs shelf sag, shrinkage governs tabletop cracking, and compression perpendicular to grain governs denting on a work surface. Choose the property that matches your failure mode first, then read the ranking for that property.
For flatsawn white oak, about 0.73 inches across the year in a house that swings from roughly 6% to 12% wood moisture content. Cherry moves 0.49 inches over the same swing, walnut 0.54, red oak 0.60 and hard maple 0.69. Quartersawn stock roughly halves those figures. The calculation is width times tangential shrinkage divided by 30, times the change in moisture content in percentage points.
Black cherry, by total movement. It shrinks 7.1% tangentially green to ovendry, the least of the common furniture species, so it changes width less than walnut, either oak or maple. Black walnut wins a different contest: its tangential to radial ratio of 1.42 is the best of the five, which means it distorts and cups less even though it changes width slightly more. If you are fighting width change use cherry, if you are fighting cupping use walnut.
Tyloses. White oak heartwood pores are usually plugged with these balloon-like cell growths, which makes the wood essentially impenetrable to liquid, gives its heartwood a Resistant decay rating and makes it suitable for tight cooperage and outdoor use. Red oak generally lacks them, so its vessels run open along the board. White oak is also slightly harder at 1,360 lbf against 1,290, and stronger in shear at 2,000 lbf per square inch against 1,780. Red oak is a little stiffer and noticeably cheaper.
I would not. Red oak has open vessels with no tyloses to block liquid, so water wicks deep into the board, and it does not appear in the Forest Products Laboratory's list of decay-resistant heartwoods. White oak is the right choice: its heartwood is rated Resistant and the plugged pores are why it has been used in boats and barrels for centuries. If watertightness is the reason you are buying, test your actual boards, because many chestnut oak pores lack tyloses despite chestnut oak sitting in the white oak group.
It is tangential shrinkage divided by radial shrinkage, and it predicts distortion rather than total movement. A ratio near 1.0 means the board shrinks nearly equally in both cross-grain directions, so it stays flat. A ratio above 2.0 means it moves roughly twice as much one way as the other, which is what makes flatsawn boards cup. Walnut is 1.42 and yellow birch 1.30, both excellent. Northern red oak is 2.15 and American beech 2.16, both poor. Read it alongside the tangential number, because they answer different questions.
Because it is diffuse-porous. Its vessels are small and spread evenly through the growth ring with no abrupt earlywood to latewood boundary, so density varies smoothly and unpredictably across the surface and pigmented stain sinks in unevenly. Ring-porous woods like oak have a banded structure that absorbs far more consistently, which is why oak stains well and maple does not. Use a dye rather than a pigmented stain, seal with a washcoat first, or finish clear and let the wood amber naturally.
Of the common species, hickory and yellow birch, at 2.16 and 2.01 million lbf per square inch modulus of elasticity, then hard maple at 1.83. The surprise is that yellow-poplar at 1.58 beats black cherry at 1.49 despite being far softer, so a poplar shelf sags about 6% less than a cherry one of the same size. Span matters more than any of it: deflection goes with the cube of the span at constant load, so shortening a 48 inch bay to 36 inches removes 58% of the sag while the best species swap available removes 19%.
Only ten domestic hardwood groups appear in the 2018 NDS Supplement Table 4A: Aspen, Beech-Birch-Hickory, Cottonwood, Mixed Maple, Mixed Oak, Northern Red Oak, Red Maple, Red Oak, White Oak and Yellow Poplar. Cherry, walnut and ash have no published reference design values at any grade. The Wood Handbook strength figures are not a substitute, because they come from small clear specimens and carry no reduction for the knots and grain deviation in a real board.
A week in the space where you will work it, stickered so air reaches every face, is my minimum. The reason is in the equilibrium moisture content table: kiln-dried lumber leaves at 6% to 8%, but at 70 degrees and 65% relative humidity the same board is heading for 12%. Milling to final dimension before it settles means the piece keeps moving after assembly, which is how joints open and glue-ups bow. If the piece matters, check with a moisture meter and aim to be within a point or two of where it will live.
Hard maple is sugar maple, 1,450 lbf Janka and 0.63 specific gravity. Soft maple is usually red maple, 950 lbf and 0.54, so 34% softer. In stiffness the gap is only about 10%, 1.83 against 1.64 million, so for shelves and rails soft maple gives you most of the performance for less money and much less tool wear. For work surfaces the property that matters is compression perpendicular to grain, where hard maple is 1,470 against soft maple's 1,000, and the substitution is a poor one.
No, though it will show its history. Walnut is 1,010 lbf Janka so it dents more readily than oak or maple, but its bending strength of 14,600 lbf per square inch actually exceeds northern red oak's 14,300, and its tangential to radial ratio of 1.42 makes it one of the flattest-staying species for a wide top. The property to respect is shear parallel to grain, 1,370 lbf per square inch, the lowest of the furniture hardwoods, so give tenons extra length and avoid short grain in the base.
On ring-porous species, yes, if you want a smooth film finish. Oak, ash and hickory lay down a band of very large earlywood vessels, and the Wood Handbook describes vessel diameter dropping by an order of magnitude or more across the growth ring in these woods. Finish sinks into those vessels, so an unfilled oak surface still feels textured after four coats. Diffuse-porous woods like maple, cherry and birch have nothing to fill and go straight to topcoat. Skipping filler on oak is a legitimate choice if you want the open texture, but it should be a decision rather than an omission.
Sources
- USDA Forest Products Laboratory, Wood Handbook: Wood as an Engineering Material, General Technical Report FPL-GTR-282 (2021). Table 5-3b for strength properties, specific gravity and side hardness at 12% moisture content; Table 4-3 for radial, tangential and volumetric shrinkage; Table 4-2 for equilibrium moisture content against temperature and relative humidity; Equation 4-9 for the shrinkage to moisture content relation; Table 14-1 for heartwood decay resistance; Chapter 3 for porosity classes; Chapter 2 for the tyloses and oak identification passages.
- American Wood Council, National Design Specification for Wood Construction, 2018 Supplement, Table 4A. Reference design values for visually graded dimension lumber, and the complete list of domestic hardwood species groups that have them.
- American Wood Council, National Design Specification for Wood Construction (2018), section 3.5.2. The creep factor of 1.5 applied to long-term deflection for seasoned lumber in dry service.
Every property value on this page was read out of those documents with pdftotext and is checked on every deploy by a script that re-extracts the tables and asserts the article's data still matches, so a transcription error cannot survive a build. Every arithmetic result, including the movement figures, the deflections and the correlations, is computed rather than quoted. Pricing is my own regional observation and is the only thing here that is not sourced. Species properties vary: the Wood Handbook puts the coefficient of variation on shrinkage at about 15% and on ovendry density at about 10%, so treat every figure as a species average rather than a promise about your board.
Related reading on The Plan Stack: lumber grades covers how FAS, Selects and Common actually differ and how to buy for yield, the pilot hole chart sizes screw holes by species density, which is the same specific gravity column used above, and the shelf sag calculator runs the deflection arithmetic from this article for a shelf you are actually building.