Sources and further reading

This is the largest disagreement on the site, and averaging does not settle it. Roubo says the bench must be no higher than the top of your thighs, ordinarily two and a half pieds, 812 mm; higher, and it robs you of your strength. Nicholson gives no body landmark: benches vary "to accommodate the height of the workman", and the medium is about 2 ft 8 in, 813 mm.

Schwarz stands with his arms relaxed and measures from the floor to where the little finger joins the hand. That gives him 34 in at 6 ft 3 in, and he stays inside a historical range of 28 to 36 in. Maguire measures to the underside of the palm with the hand flexed level and shop shoes on, then adds a couple of inches for mixed work and an inch or two more for delicate work or a bad back. Sellers gives no landmark at all. He works at 38 in at about 5 ft 10 in, tells you to start tall, try it and cut the legs down, and he rejects knuckle rules outright.

The difference comes from how each man works. Roubo and Schwarz plane on the top against a stop and want their weight over the plane. Sellers holds most work in the vise and blames low benches for sore backs and necks. For close work Schwarz does not raise the bench; he sits down, or raises the work in a second vise.

For someone about 1800 mm tall the two ends are roughly 810 mm and 965 mm, a gap of about 150 mm. The first figure is derived from a single ratio, Schwarz's own bench over his own stature, and he calls himself an outlier.

On the sheet Each sheet starts at 0.45 of your height, marked as an estimate, derived. It asks you to measure yourself, and it records whose rule you used.

One-fifth, one-third and one-half are all in print. Fairham gives one-third of the stock "in practically all cases where a single tenon is used". Nicholson, the oldest English text read here, gives "about one fifth of the thickness of stuff" for carpentry framing that lies unsupported, and argues that one-third splits the mortise cheek there. Schwarz reports one-third from "traditional texts" he does not name, and allows one-half, "an old rule", only when everything is as thick as a bench leg.

Much of the difference is scale and loading. Where one-half would leave thin mortise walls, Schwarz cuts a bareface tenon: half the thickness of the stretcher, with one shoulder on the outside and none inside.

Nicholson reduces the tenon at a frame corner by "about one third" of the rail's breadth, which leaves tenon and mortise two-thirds of it, and adds that the mortise "must never be cut out to the extremity, but always at least one fourth of the whole breadth farther in". Ellis, as Schwarz reports him, makes the tenon half the width of the rail, and divides it into two or more tenons where its width would pass six times its thickness. Ellis's own page on this was not read for the site.

On the sheet Where the end rails of the Nicholson bench are tenoned, the tenon is drawn two-thirds of the rail's width, Nicholson's figure, and the joint note says that Ellis is reported to prefer half.

Schwarz disagrees with himself across fifteen years. In 2005 he gives 1/4 in (6 mm) pegs for cabinet work and 3/8 in (10 mm) for benches, doors and windows; Fairham bores general joinery with a 3/8 in twist bit. By 2020 Schwarz pegs his own bench with 5/8 in (16 mm), with 3/4 in (19 mm) as his second choice.

His reason for caring: "A too-skinny peg will split when driven in. A too-thick peg will wreck the tenon, usually popping the end off."

On the sheet The Roubo sheet starts at 5/8 in and offers 3/4 and 3/8 in. The Moravian sheet starts at 3/8 in for the middle rail and offers 5/8 in. Myers gives no size, so both figures are general bench practice. The Nicholson sheet draws a 10 mm peg where the end rails are drawbored.

Schwarz in 2005 glued the peg: "It cannot hurt." Schwarz in 2020 waxes it instead: "Glue makes a peg swell up. Wax helps a peg do its job, which is to slide into place."

On the sheet The Roubo sheet says waxed and not glued. That is the later of his two statements, taken because the sheet starts from his 2020 bench; it is not a rule the sources agree on.

The rule most often repeated today, 1 in 6 for softwood and 1 in 8 for hardwood, is not in the older texts. Fairham says "one in eight is considered correct" and that the slope "should on no account be less than one in six", and makes no difference between hard and soft wood. Schwarz's survey of the older books finds about 10 degrees for all joints in Ellis and Hasluck, 7.5 degrees for exposed work and 10 for heavier work in Audel, 12 degrees for coarse work in Hayward, and 15 in Hoard and Marlow. Those five were read only through his survey. He concludes that "the advice on softwoods and hardwoods seems to become more common with modern writing".

What the older books do agree on is a range of roughly 7 to 15 degrees, with the steeper slopes for coarse or heavy work. Roubo gives no slope for his leg dovetail.

On the sheet The lapped dovetails of the Nicholson end rails and the tray dovetails of the Moravian bench are drawn at 1 in 6, the steepest Fairham allows, bench work being coarse work. No source gives a slope for either bench; the figure is our choice from a general joinery rule.

Three sizes are on record, and the choice cannot be undone once the top is bored. Roubo's holes are 14 to 16 lignes, 31.6 to 36.1 mm, for shanks of 12 to 15 lignes. Schwarz bores 1 in (25.4 mm), which he calls "a more traditional size", and writes that "bigger holdfasts that are closely fit to their holes work better". The Gramercy holdfast is made for a 3/4 in (19 mm) hole, the modern commercial size.

On the sheet The Roubo sheet starts at Schwarz's 1 in, shows Roubo's 14 to 16 lignes as the band and allows anything from 19 to 36.1 mm. The Nicholson sheet starts at 3/4 in, as the knockdown bench is bored, and offers 1 in. The Moravian sheet starts at Myers' 3/4 in and offers 1 in.

Roubo wrote in the units of the pied du roi: one pouce is 27.07 mm, about 1.066 in. Schwarz reads the figures as inches ("5 in to 6 in thick and 20 in to 22 in wide"), so his Roubo is about 6.6 percent smaller than Roubo's.

Schwarz dates the book to about 1774. The title page of the part that holds plate 11 reads M. DCC. LXIX: 1769.

On the sheet The bands on the rules are Roubo's figures converted at 27.07 mm to the pouce. The defaults are Schwarz's, in his inches.

In the bench description the top is "vingt à vingt-deux pouces" wide, 541 to 596 mm. A few pages earlier, laying out the shop, he says benches are ordinarily eighteen to twenty pouces wide, 487 to 541 mm. Both figures are his.

On the sheet The band on the width rule is the figure from the bench description.

The English translation of 2008 differs from the French in four places. It gives the top as 20 to 25 wide, where the French has twenty to twenty-two. It gives the stretchers as "4 inches wide or less", where the French says "au moins", at least. It makes the pad of the holdfast 10 inches high, where the French says six pouces. And it gives the tool rack a width of "6 to 8 inches", where the French gives no width: the six to eight lignes are how far the rack stands off the bench.

On the sheet The French is followed in all four.

Roubo brings each leg up through the top as a sliding dovetail and a through tenon, wedged from above. Schwarz built it that way in 2010 and has since changed his mind: he writes that the through joint "offers few functional advantages compared to a blind mortise and tenon", that its end grain stands proud as the top shrinks and must be planed flush again and again in the first years, and he prefers a blind drawbored tenon. His book is not consistent on the point. A caption in the same chapter says the annoyance is worth it.

Roubo also disagrees with the custom he records. The custom left no shoulder at the back of the leg, and plate 11 draws it that way; he recommends a rear shoulder so that the top does not sink on its legs with age. Schwarz notes that the dimensions in the text "don't really jibe with the drawings".

No source states a rule for dividing the leg between dovetail, gap and tenon. Three builds are documented, and Myers says the thicknesses "are really up to you".

On the sheet Both joints are offered. Outside the three documented layouts the proportions are marked derived.

The text says only that the stretchers are assembled into the legs. The plate draws two peg ends at each visible end of the front stretcher. Whether they were drawbored is not stated.

One phrase, "about one pied from the end of the bench", sits between the description of the back row of holdfast holes and the planing stop, and can be read with either. The 2008 translation attaches it to the holes. The plate draws the stop in the left overhang, hard against the leg. The text does not say which leg the front row of holes is spaced from; the translation supplies "right front".

The text gives no figure for the overhang of the top at the ends. The plate shows one at each end, but it is a perspective view with no scale, so no number can be taken from it.

Schwarz dates the Mechanic's Companion to 1812. The Lost Art Press transcription of the bench passage gives 1811. A later edition was the one read for this site.

Ninety years apart they describe much the same bench, 10 to 12 ft long, and differ on the rest. Nicholson makes it about 2 ft 6 in wide with a medium height of about 2 ft 8 in. Ellis makes it 27 in wide with a "common height" of 28 in.

Both are traditional. Nicholson's plate reads as plumb, though that is a judgement from a small perspective engraving, and the knockdown bench of 2014 is plumb. Ellis sets the legs "on the rake, the better to withstand the thrust of the planing", with the feet kicked out toward the ends, and gives no angle.

Schwarz raked the legs of his 2006 bench 20 degrees, "a feature I found on a number of vintage benches", and later wrote that the angled legs did not make the bench more stable in use. The 20 degrees is his own figure.

The drawing of the knockdown bench gives a top of two 1-1/2 in layers, 3 in in all. The Anarchist's Workbench lists the same top at 2-1/4 in. This is not resolved.

The drawing of the 2006 bench makes it 35 in high and its top 1-1/4 in thick. The book says 34 in, the legs having been trimmed and later restored, and about 2-3/8 in after a second layer was added.

The planing stop of the knockdown bench is 2-1/4 in from the front edge on the drawing and "3 in or so" in the article. Its hardware is counted as fifteen, sixteen, or fourteen and a T-nut, depending on where you read.

Nicholson writes of "side boards" in the plural, so a back apron is implied, but his plate does not show it. Ellis makes both sides the same, 9 by 1 in. The suggestion that Nicholson's back board is narrower is one modern reader's guess; neither the plate nor the text says so.

Nicholson has "two rails dovetailed into the legs" and does not say what kind of dovetail; the lapped form is the usual reading. Ellis uses "mortise and tenon joints, glued and wedged". Maguire's own chapter summaries say only "lap joints"; a builder who followed his videos reports half-lap dovetails throughout, which is secondhand.

Schwarz has done it two ways. His 2006 bench has drawbored mortise and tenon joints. The knockdown bench has the rails let into notches, glued and screwed, and he regretted it: "I wish I had joined the lower stretchers and legs with a full mortise-and-tenon joint instead of notches and screws."

On the sheet All four are offered: the notch, the lapped dovetail, the wedged tenon and the drawbored tenon. The sheet starts with the notch because it starts from the knockdown bench. No source gives the proportions of the dovetail or the tenons for this bench; they are general joinery rules, and the sheet says so.

Ellis has them "ploughed, tongued, and glued, and cleaned off flush on the top side". Schwarz glues them edge to edge. Maguire leaves them unglued with gaps between and nails them down, because nails "will flex and bend allowing the boards to expand and contract". Nicholson does not say.

On the sheet The order of work follows the fastening you choose: with nails the planks go down loose, Maguire's way; with bolts or screws they are glued edge to edge, and the note adds that Ellis ploughs and tongues them as well.

Myers, who measured it, writes that "the exact age of the original is not known for sure or who made it". A retailer dates it to the late 1700s; the museum has published no such attribution that we could read. Schwarz's remark about a bench "likely dating back to the 18th century" is about another bench in the collection, the oldest, a massive 8 ft trestle bench.

Myers gives 16 degrees twice in his 2012 article and again in 2016. The label on his own drawing says 15, and its linework is drawn at 16.3. No measurement of the original's angle is published.

On the sheet The sheet starts at 16 degrees and offers 15. That is our reconciliation; no source gives it. 16 is his repeated statement, and with a 12 in overhang it reproduces the shoulder distance labelled on the drawing to within about 2 mm.

The labels on Myers' one-sheet drawing and its linework do not agree. The stretcher shoulders are drawn 54.8 in apart and labelled 4 ft 8-1/8 in. The tenon is drawn 1-1/2 in thick in the end view and labelled 1-1/8 in in the detail. The chain of heights in the end view is drawn to one set of figures and labelled with another.

On the sheet Where they conflict the labels and the article are followed. Linework is used only for what is not labelled, and the sheet says where.

The overhang of the top is "about 12 1/2 inches" in 2012 and "only 12 inches" in 2019; the drawing scales to 12-3/4 in. The vise backer is 2 in thick in 2012, "a 3 in. board" in 2014 and "2 3/4 in. or so" in 2019. The dowels that locate the top are 1/2 in in 2012 and 5/8 in by 2014.

Schwarz: "If it can get loose, it will get loose." His have worked loose slowly with use and the seasons and need the keys driven again, though he grants that friends report theirs never do. Myers reports flawless service over four years of knocking his bench down and setting it up.

Myers puts the joint's bad name down to too little wood beyond the key. In his pull tests no tenon failed: the face of the leg crushed and then the key broke, white oak keys at about 4000 lbf, white pine at about 700.