Knits and other structures

A mitre rule is a measurement of the loop

A mitred square is cast on as an L and decreased at its corner, and the trade's rule — a double decrease every other row — makes it square in garter stitch. Drawn at the shape of its loops, a mitred square is a kite whose corner is twice the arctangent of the decrease rate over the loop's aspect, so the rule is square at an aspect of exactly two and nowhere else. In plain knit, whose loops are 1.25 to 1.29 times as wide as they are tall, the same rule knits a kite of 115 and 65 degrees. Four decreasing rows in five square it dry; seven in nine come within two degrees in every relaxed state. And the pick-up ratio is the same number again: three stitches for four rows sits three to six per cent tight.

Worth reading first: A fashioned edge has a quantised angle · A knitted disc is flat at one shape of loop · A heel turns a right angle because of the loop.

A mitred square is one of the first shaped things a knitter makes and one of the most satisfying. Cast on an odd number of stitches, mark the middle one, and work a double decrease there on every other row. The cast-on, which went on as a straight line, folds at the middle stitch into an L; the decreases stack into a diagonal from the corner of the L to the centre; and when the last stitch is fastened off the piece is a square, with a crease across it like a mitred picture frame. Join enough of them along their edges and a blanket builds itself one square at a time.

The rule is always given for garter stitch. That is not an accident of tradition, and it is not a matter of taste. The rule makes a square in garter and nothing else in particular, because it is a statement about garter’s loop — and the same rule worked in plain knit makes a kite with corners twenty-five degrees out of true.

Mitred squares drawn at the shape of their loops. Mitred squares of 12 stitches an arm, drawn stitch by stitch at the fabric's own aspect, the two arms shaded apart and the mitre dashed. every other row, decreasing every other row at a loop aspect of 1.279 (fully-relaxed): a kite with 114.8° at the cast-on corner, 65.2° at the centre and right angles at the arms' ends, its selvedges 1.564 times its cast-on arms; 4 rows in every 5, decreasing 4 rows in every 5 at a loop aspect of 1.250 (dry-relaxed): a kite with 90.0° at the cast-on corner, 90.0° at the centre and right angles at the arms' ends, its selvedges 1.000 times its cast-on arms. A mitre is square only where its decrease rate in rows equals the loop's aspect. What the drawing cannot show is the double decrease itself, which pulls three loops into one along the mitre.
Fig. 1 Two mitred squares of twelve stitches an arm, drawn stitch by stitch at the shape of plain knit’s loops, the two arms shaded apart and the mitre dashed. Decreased every other row, fully relaxed, the piece is a kite of 114.8 and 65.2 degrees whose selvedges are 1.56 times its cast-on arms. Decreased on four rows in every five and dry relaxed, it is square.

The two arms are two triangles

Follow one arm of the L. It starts as nn stitches along the cast-on, and the corner decrease takes one of them away every mm rows, so after nmnm rows it is gone. Its rows are stacked one on another, each a stitch shorter at the mitre end than the one before it every mm rows. Drawn at the size of its stitches, the arm is a right triangle: one leg along the cast-on, nn wales long, the other up the selvedge, nmnm courses long, and the mitre as its hypotenuse.

The second arm is the same triangle reflected in the mitre, because it is knitted row for row alongside the first, either side of one decrease. The piece is the two triangles joined on their long side.

That is enough to say what shape it is. A wale is krk_r times as wide as a course is tall, where krk_r is the loop’s aspect — Munden’s ratio of courses to wales per unit length, which does not depend on the yarn or on how big the loop is. So the arm’s selvedge is nm/krnm/k_r wale widths long against a cast-on of nn, and the mitre makes an angle with the cast-on of

θ=arctan⁡mkr.\theta = \arctan\frac{m}{k_r}.

A kite, and when it is a square

Two right triangles joined on their hypotenuse make a kite. Its corners at the ends of the arms are right angles, since each is the right angle of a triangle. The corner at the start of the mitre is 2θ2\theta, and the corner at the centre, where the last stitch is fastened off, is what is left:

cast-on corner=2arctan⁡mkr,centre=180∘−2arctan⁡mkr.\text{cast-on corner} = 2\arctan\frac{m}{k_r}, \qquad \text{centre} = 180^\circ - 2\arctan\frac{m}{k_r}.

The selvedges are m/krm/k_r times the length of the cast-on arms. The piece is a square exactly when m=krm = k_r — when the number of rows per decrease equals the number of courses that fit in one wale’s width — and at no other rate. Decrease faster and the cast-on corner closes up; decrease slower and it opens out, and the selvedges grow longer than the arms.

Where each decrease schedule squares a mitre. The corner a mitred square makes at its cast-on, against the loop's aspect, for decreases every row (square at an aspect of 1.000), 4 rows in every 5 (square at an aspect of 1.250), 7 rows in every 9 (square at an aspect of 1.286), 3 rows in every 4 (square at an aspect of 1.333), 2 rows in every 3 (square at an aspect of 1.500), every other row (square at an aspect of 2.000). Each curve crosses ninety degrees where the aspect equals its rate. The shaded band is plain knit's aspect between its dry-relaxed value, 1.250, and its wet-relaxed value, 1.293; the trade's rule of a decrease every other row crosses ninety at two, far outside it.
Fig. 2 The corner a mitred square makes at its cast-on, against the loop’s aspect, for six decrease schedules. Each crosses a right angle where the aspect equals its rate in rows a decrease. The shaded band is plain knit between its dry-relaxed and wet-relaxed aspects, 1.250 and 1.293. The trade’s rule of a decrease every other row crosses a right angle at an aspect of two, far outside the band.

The rule states garter’s aspect

The trade’s rule decreases every other row, so m=2m = 2, and it is square only where kr=2k_r = 2: where two courses fit in the width of one wale. That is precisely the knitter’s description of garter stitch. Garter is knitted on both faces in alternate rows, so its courses nest into ridges, and the long-standing rule of thumb is that it knits about twice as many rows as stitches to the inch, one ridge of two rows standing as tall as one stitch is wide. A garter ridge looks square for that reason, and a mitred square in garter comes out square for the same one.

So the mitre rule is not a recipe that happens to work in garter. It is a measurement of garter’s loop, written as an instruction, and it has been checked every time a mitred square came out square. It is worth saying what this does and does not establish. Munden’s constants, the ones every number here rests on, are for plain knit; nobody has published the equivalent pair for garter, and composite structures do not inherit their constants from their parts. The figure of two is the trade’s, and the evidence for it is the squares themselves.

In plain knit the rule makes a kite

Work the same rule in plain knit and the arithmetic is no longer kind. Plain knit’s loop is far less flattened than garter’s: in the three relaxed states Munden measured, a wale is 1.250, 1.279 and 1.293 times a course. With m=2m = 2 the cast-on corner comes out at 116.0 degrees dry relaxed, 114.8 fully relaxed and 114.3 wet relaxed, and the centre at 64 to 66 degrees. The selvedges run 1.55 to 1.60 times the length of the arms.

The first figure draws it. Nothing about the piece looks wrong while it is being knitted — every decrease is where the pattern says, the mitre is straight — and it comes off the needles as a lozenge half as long again as it is wide. The straight mitre is the reason it is not obvious: a knitter expects a skewed square to show a crooked line somewhere, and this one has none. The error is in the corners, all at once, and it is the whole of the difference between kr=2k_r = 2 and kr=1.28k_r = 1.28.

Decreasing on every row overshoots the other way: the cast-on corner closes to 76 degrees and the centre opens to 104, with selvedges about four fifths of the arms.

Mitred squares drawn at the shape of their loops. Mitred squares of 12 stitches an arm, drawn stitch by stitch at the fabric's own aspect, the two arms shaded apart and the mitre dashed. every row, decreasing every row at a loop aspect of 1.279 (fully-relaxed): a kite with 76.0° at the cast-on corner, 104.0° at the centre and right angles at the arms' ends, its selvedges 0.782 times its cast-on arms; 3 rows in every 4, decreasing 3 rows in every 4 at a loop aspect of 1.279 (fully-relaxed): a kite with 92.4° at the cast-on corner, 87.6° at the centre and right angles at the arms' ends, its selvedges 1.042 times its cast-on arms; 7 rows in every 9, decreasing 7 rows in every 9 at a loop aspect of 1.279 (fully-relaxed): a kite with 90.3° at the cast-on corner, 89.7° at the centre and right angles at the arms' ends, its selvedges 1.005 times its cast-on arms. A mitre is square only where its decrease rate in rows equals the loop's aspect. What the drawing cannot show is the double decrease itself, which pulls three loops into one along the mitre.
Fig. 3 Three schedules in fully relaxed plain knit. Decreasing every row: a kite of 76.0 and 104.0 degrees, selvedges 0.78 of the arms. Decreasing on three rows in four: 92.4 degrees, a little open. Decreasing on seven rows in nine: 90.3 degrees, as near square as the cells can show.

A square needs a rate between one row and two

A square in plain knit needs a decrease every 1.25 to 1.29 rows, and a knitter can only decrease on whole rows. The way to a fractional rate is the one a fashioned edge uses: decrease on qq rows out of every pp, which averages a decrease every p/qp/q rows. The available rates are fractions, and they are quantised in the same way the angles of a fashioned edge are — a short schedule gives a coarse choice, and fine control costs a long one to count.

Among every schedule of up to ten rows, only two land inside plain knit’s band at all: four decreasing rows in every five, a rate of exactly 1.25, and seven in every nine, 1.286. The first squares the mitre in the dry-relaxed state and in no other, because dry-relaxed plain knit is the one state in which Munden’s two constants stand in the ratio of five to four. Washed, the same square has corners of 88.1 and 91.9 degrees.

And the wash moves the corner

A square that is square in one state is a little out of square in another, since the loop’s aspect moves when the fabric is wetted and relaxed — a knit relaxes for as long as it is allowed, and its loops change shape as it does. Across plain knit’s three relaxed states the aspect moves by 3.4 per cent, and the cast-on corner of any mitre moves with it, by about two degrees.

The decrease schedules that come near a square in plain knit. Every schedule of q decreasing rows in every p, up to 10 rows, whose mitre comes within 6 degrees of square in plain knit, with its corner at the cast-on in each relaxed state: 6 rows in every 7, 86.1°, 84.1°, 84.7° dry, wet and fully relaxed; 5 rows in every 6, 87.7°, 85.7°, 86.3° dry, wet and fully relaxed; 4 rows in every 5, 90.0°, 88.1°, 88.7° dry, wet and fully relaxed; 7 rows in every 9, 91.6°, 89.7°, 90.3° dry, wet and fully relaxed; 3 rows in every 4, 93.7°, 91.8°, 92.4° dry, wet and fully relaxed. The nearest across all three is 7 rows in every 9, within 1.6 degrees; only four rows in every five is exactly square, and only when dry.
Fig. 4 Every countable schedule of up to ten rows whose mitre comes within six degrees of square in plain knit, with its cast-on corner dry relaxed, fully relaxed and wet relaxed. Four rows in every five is exactly square dry and 88.1 degrees wet; seven rows in every nine is 91.6 dry, 90.3 fully relaxed and 89.7 wet, within 1.6 degrees in all three. No schedule is square in every state.

So no schedule is square in every state, and the question becomes which is nearest across all three. Seven decreasing rows in every nine — a rate of 1.286 — sits inside the band, and its corners are 91.6, 90.3 and 89.7 degrees, never more than 1.6 out. Four in five is better dry and worse wet, reaching 1.9 out. Three in four, a schedule a knitter might reach for as the simplest fraction near the target, is open by 1.8 to 3.7 degrees. A designer who knows whether the piece will be washed can choose; one who does not should take seven in nine and accept a corner that is out by less than the eye can see across a ten-centimetre square.

What garter does when it is washed is not known here. Its aspect of two is the trade’s for garter as knitted, and whether a garter mitre stays square through a wash depends on whether garter’s ridges relax in proportion, which is exactly the number nobody has measured.

A pick-up ratio is the same number

The second rule of modular knitting is about joining. A new square is started not by casting on but by picking up stitches along the edge of a square already made, and the trade gives a ratio for that too. In garter it is one stitch for every ridge — one for every two rows. In plain knit it is the famous one: pick up three stitches for every four rows.

These are not two more rules. A picked-up arm of nn stitches hangs from a selvedge of nmnm rows, and it lies flat against it only if nn wales are as long as nmnm courses — the condition m=krm = k_r again. So a pick-up ratio of qq stitches for pp rows claims an aspect of p/qp/q, and the fit of the picked-up fabric to the edge it hangs from is

fit=kr qp,\text{fit} = \frac{k_r\, q}{p},

above one if the new fabric is wider than its edge, below one if it is narrower.

How well a picked-up edge fits the selvedge it is knitted from. A square knitted onto another's selvedge, picking up stitches at five rates, and the width of the new fabric as a share of the edge it hangs from in plain knit's three relaxed states: one stitch a ridge, garter's rule, 0.625 to 0.646 to 0.640; three for every four rows, 0.938 to 0.970 to 0.959; seven for every nine rows, 0.972 to 1.005 to 0.995; four for every five rows, 1.000 to 1.034 to 1.023; one for every row, 1.250 to 1.293 to 1.279. A fit of one needs a pick-up rate equal to the loop's aspect, which is the same condition that squares a mitre. Three stitches for four rows sits 3 to 6 per cent short in every state.
Fig. 5 A square picked up along another’s selvedge at five rates: the width of the new fabric as a share of the edge it hangs from, across plain knit’s three relaxed states. Garter’s rule of a stitch a ridge, worked in plain knit, is 35 to 38 per cent short. Three for every four rows is 3 to 6 per cent short in every state. Seven for every nine straddles a perfect fit, four for every five runs up to 3 per cent full, and a stitch a row 25 to 29 per cent full.

The three-for-four rule claims an aspect of 4/3, 1.333, which is above plain knit’s whole relaxed band. In every state Munden measured, a picked-up edge at three for four is three to six per cent narrower than the selvedge it hangs from, so the seam is drawn in very slightly. Seven for nine fits to within half a per cent fully relaxed, and four for five to nothing at all dry.

Two readings of that gap are possible and the arithmetic cannot choose between them. One is that the rule is simply rounded to the simplest fraction near the answer, and three for four is the nearest a knitter can count without a pencil. The other is that the rule is deliberately a little tight, since a picked-up band that pulls in slightly lies flat and one that is slightly full ripples, and a knitter would rather have the first. Either way, the rule is a measurement of the loop with a rounding or a bias in it, and the bias is now a number.

A blanket of the wrong mitres does not lie flat

The two rules together say what goes wrong in a modular blanket knitted in plain knit by garter’s instructions. Each square is a kite whose selvedges are about 1.56 times its arms. The next square is picked up along one of those selvedges at a stitch every two rows — garter’s ratio — so its arm is nn stitches hanging from a selvedge of 2n2n courses. In plain knit those nn wales are only 0.64 of the selvedge’s length, and the seam has to gather the selvedge in to meet them.

The trouble is in the edges more than in the corners. A kite still has two right angles, and the other two add to 180 degrees, so a corner that is too open on one square is matched by one too closed on the next. But every join asks a fabric to be about two thirds the length of the edge it is joined to, and no arrangement of the squares removes that. That is why a modular blanket in plain knit puckers in a way a knitter tends to blame on tension, and why patterns for one are so reliably written in garter.

A mitre is a protractor for the loop

The relation can be run backwards. If a mitred square’s cast-on corner is measured, the aspect of the fabric it was knitted in follows at once:

kr=mtan⁡(corner/2).k_r = \frac{m}{\tan(\text{corner}/2)}.

So a mitred square is an instrument as well as an object. Knit one at a known schedule, lay it flat, and the corner says how many courses fit in a wale’s width — in whatever stitch, yarn and hand the square was knitted in, which is exactly what Munden’s constants cannot say for any structure but the plain loop they were measured on.

It is a coarse instrument, and it is worth knowing how coarse. Near an aspect of two, with a decrease every other row, the corner moves by about 0.57 degrees for each per cent the aspect moves. A protractor read to a degree on a ten-centimetre square therefore pins garter’s aspect to within about two per cent, which is no better than counting ridges and stitches on a swatch and a good deal less convenient. Its value is not precision. It is that the reading is of the shape the knitter actually cares about, so a mitre that comes out square has measured garter’s aspect as two to the accuracy that matters for squares, and one that comes out at 95 degrees has measured it as about 1.83.

That is also the quickest way to test the two readings of the three-for-four rule. A plain-knit mitre decreased on three rows in every four, knitted by the same hand that picks up three for four, should come out open by two to four degrees if the hand’s loops are Munden’s, and square if the rule was written for loops a little flatter than his.

Garter has a second reason to be the stitch for squares

The aspect is not the only reason the trade’s squares are garter. Plain knit curls — its side edges roll one way and its top and bottom edges the other — and a square of it, even a true one, will not lie flat at its edges without blocking. Garter does not curl, because its alternate courses face opposite ways and their bending cancels. A pattern writer choosing a stitch for a piece that has to lie flat and meet its neighbours edge to edge would choose garter for that alone. It is a coincidence of the stitch that its loop is also the one the simplest decrease schedule fits, and the coincidence is why the rule has never been written for anything else.

Why the loop’s aspect keeps turning up

This is the fourth shape in which the same constant has turned a count of stitches and rows into an angle. A fashioned edge stands at an arctangent of a fraction times krk_r. A heel turns through two thirds of π\pi divided by it. A knitted disc lies flat at one value of it. And a mitre is square when its decrease rate equals it.

The reason is that every one of these shapes is built by counting in two directions, wales one way and courses the other, and the plane does not care about counts; it cares about lengths. The aspect is the exchange rate between the two counts, so any shape that depends on a count of rows against a count of stitches depends on it. What makes the mitre unusual among them is that the trade wrote the exchange rate into its instructions twice — once as a decrease rate and once as a pick-up ratio — so the instructions can be read as measurements, and checked against Munden’s.

The model, and the numbers in it

Every angle here comes from one model: a knitted fabric as cells one wale wide and one course tall, with the ratio of the two given by Munden’s constants for relaxed plain knit — 5.0 and 4.0 dry, 5.3 and 4.1 wet, 5.5 and 4.3 fully relaxed, from his measurements of the late 1950s. A mitred square is drawn as two triangles of such cells, the second reflected in the mitre. Garter’s aspect of two is the trade’s rule of thumb, used only where the essay says so. The schedules counted are every qq in pp with pp up to ten, reduced so that each rate is counted once.

What the drawing cannot show

The decrease itself. A double decrease pulls three loops into one and leaves a ridge of crowded stitches along the mitre, and the cells either side of it are distorted in a way no cell model describes. The drawings place the cells as if the mitre were a fold, which is the right picture away from it and the wrong one on it.

Hand-knitted loops. Munden’s constants come from machine-knitted fabric relaxed under controlled conditions. A hand knitter’s loop depends on how the yarn is held and how tightly the stitches are worked, and its aspect may sit a few per cent off the band, which is the same size as every effect computed here. That is one reason the three-for-four rule could be right for the knitters who wrote it and still disagree with the band.

Garter’s own constants, which would turn the claim that the trade’s rule measures an aspect of two into a check of it.

Who worked out which part

Mitred squares are old enough that nobody is credited with them, and the modular method of building whole garments from squares picked up from one another was set out at length by the Danish designer Vivian Høxbro, whose work on what she called domino knitting reached English readers in 2002. The constants are Munden’s, from 1959 and the papers that followed, and the loop’s aspect as an exchange rate between counts and angles is the thread of the essays on fashioned edges, tubes, heels, discs and balls with stacked increases. What is added here is the mitre as a kite with corners 2arctan⁡(m/kr)2\arctan(m/k_r); the trade’s mitre and pick-up rules read as statements of an aspect; the plain-knit kite of 115 and 65 degrees; the schedules, of which four in five squares a dry mitre and seven in nine comes nearest across every state; and the three-to-six per cent by which three-for-four draws a plain-knit seam in.

Still open: whether a garter mitre survives a wash

The prediction with a number in it is about garter. If a garter mitre is square as knitted, its aspect as knitted is two, and the question is what a wash does to it. Plain knit’s aspect moves 3.4 per cent between its relaxed states and moves a mitre’s corner by about two degrees; if garter’s moved as much, a garter square knitted perfectly would come out of the wash a degree or two out, and a blanket of them would gain or lose a few millimetres along every join.

A garter swatch measured dry and after a full relaxation — wales and ridges per ten centimetres, and nothing else — would give garter’s own constants and settle it. It would also say whether the curl that garter avoids and the square it makes are the same fact about its ridges, or two facts that happen to coincide at an aspect of two.

Shares its objects with

Essays naming at least two of the same things, that neither author linked.

Named objects

A flat tag is an object no other essay names yet.

CourseLoop aspectMunden constantsRelaxationSelvedgeShapingWale