A mosaic chart is a stack of nested rows
Worth reading first: A float limit forbids almost every chart and costs almost nothing to break · What a tuck costs · Knit, tuck and miss.
A float limit forbids almost every chart wrote stranded knitting in stitch notation as two courses a row, one colour knitting where the chart says and missing everywhere else, the other its complement. Every stitch is then one knit and one miss, a colour’s float is a run of the other colour along the row, and the knitter’s float limit is a rule about runs. Because floats run only along rows, the rows were independent, and the count of allowed charts was one row’s share raised to the number of rows.
There is a second way to put two colours into a knitted fabric, and it is the opposite arrangement. Work each row with one yarn only. At every stitch the row’s yarn either knits or passes behind, and a stitch it passes is slipped from one needle to the other unworked, so it stays the loop the row below made — in the colour the row below made it. Two rows of blue, two of light, two of blue: wherever a light stitch is slipped through a blue stripe, light shows in the blue, and the reverse. That is slip-stitch colourwork, and its most systematic form is Barbara Walker’s mosaic knitting, where the colours alternate every two rows and no stitch is slipped for more than one stripe.
In the notation that is just as exact. A slip is a miss: the miss already in the alphabet, with the old loop held on the needle — not a tuck, which takes the new yarn into the old head, but a stitch the row simply passes by. What changes is what the chart is allowed to say, and the rule turns out to bind columns where the stranded rule bound rows.
A column changes colour only on a row of the new colour
Every row is worked in one colour, and every cell in it is either knitted in that colour or slipped, showing whatever is below. So a cell can differ from the cell beneath it only by being knitted, and a knitted cell is the row’s own colour. A column can change colour only on a row worked in the colour it changes to. A blue row can turn light columns blue; it cannot turn a blue column light. A light row does the reverse.
Read a row as the set of its blue cells and that is a statement about two sets. On a blue row the blue cells can only grow — every blue cell below is still blue, slipped or knitted over — so the row contains the row below. On a light row they can only shrink, so the row lies inside the row below. A chart can be knitted one colour a row exactly when every pair of consecutive rows nests, one inside the other, and the direction of the nesting says which colour the upper row has to be worked in. Where a row is the same as the row below, either colour will do.
That is a complete criterion, and it can be checked on any chart by eye: no row may turn one column blue and another light.
The diamonds fail at their corners
The diamond chart that served as the stranded essay’s example passes most of the test and fails it twice.
As the diamonds widen, each row’s blue contains the last, and as they narrow, each lies inside it; those six steps can each be knitted in one colour. The first and last steps cannot. The bottom row has blue at the corners between diamonds, the next row has blue at the diamonds’ points, and the step asks the same row to turn the corners light and the points blue. Two colours on one row is precisely what one yarn cannot do, and no choice of stripe order rescues it.
Walker’s form of the craft asks for more than nesting. Its colours alternate, so the direction is fixed in advance — up on a blue row, down on a light one — and the diamonds’ three widening steps in a row break that too. The mosaic version drawn above is what that rule makes of the target: a row of the wrong colour keeps the colour below, so cells the chart wanted but the row’s yarn could not supply are simply absent. Sixteen of the 192 cells above the first row move, and they are exactly the places where a column was asked to change on a row worked in the other colour. The rest survives, which is why mosaic patterns are drawn as outlines a stitch wide with a staircase on each flank: a line that steps up and to the side every other row is the shape alternation can always deliver.
Walker’s alternation makes the columns independent
Fix the alternation and the nesting has a fixed direction at every step, up on blue rows and down on light ones. Now look at one column alone. On a blue row it may turn blue or stay as it was; on a light row it may turn light or stay; and nothing about its neighbours enters, because the condition is per cell. The columns are independent, and a chart is any choice of allowed column for each column.
A column of h rows, with the starting colour fixed, is a sequence of colours that may rise only on the rows of one parity and fall only on the others, and those are counted by the Fibonacci numbers: 2, 3, 5, 8, 13 for one to five rows, F(h + 2) in general. Equivalently, a column’s slipped cells — the cells not in their row’s colour — never sit two rows running. A cell slipped through a blue stripe is light; the stripe above it is light, the cell cannot turn blue there, and so it is in that row’s colour whatever is done to it. Strings with no two adjacent marks are what the Fibonacci numbers count.
So the number of charts Walker’s rule allows, w stitches by h rows, starting in either colour, is 2F(h + 2)^w − 2^w, the subtraction removing the 2^w charts of constant columns that are allowed with either starting colour and would otherwise be counted twice. It agrees with a count of every chart at three by four, four by five and five by three.
The contrast with stranded knitting is now exact. Stranded factorises by rows and mosaic by columns. At 24 stitches and twenty rows the stranded five-stitch limit allowed one chart in about 3,800. Walker’s rule allows one in about 10^42 — a vastly stricter rule over the whole catalogue, made of a column rule that admits F(22) = 17,711 of a twenty-row column’s 1,048,576 colourings and is applied to each of 24 columns independently. A woven designer’s float limit couples every thread to every other and can only be counted by search; the two knitted rules each factorise, one by rows and one by columns. The pattern-book consequence is the same one the stranded count reached: a rule that excludes almost every chart excludes almost nothing a designer working inside it would draw, and mosaic designers draw inside it from the first line.
The golden ratio from the other direction
Per cell, a long column under Walker’s rule grows by the golden ratio φ ≈ 1.618 every row, since the Fibonacci numbers do. That is log₂φ = 0.694 bits of choice in every cell.
The stranded essay found the same number for a different rule. A row whose floats are at most two stitches long is a string of runs of one and two, and those are counted by the Fibonacci numbers too, so a two-stitch limit leaves each stitch log₂φ bits. Walker’s mosaic leaves every cell exactly what a two-stitch float limit leaves a stranded row, and the reason is not knitting: both are strings in which some mark may never appear twice in a row. In stranded knitting the forbidden pair is three stitches of one colour, read along the row; in mosaic it is two slips, read up the column.
The coincidence is exact and it is also misleading if read as equivalence. A stranded chart under a two-stitch limit has its freedom along the rows and none forbidden up the columns; a mosaic chart has its freedom up the columns and, before any float limit, none forbidden along the rows. The entropy is the same and the charts are nothing alike.
What a free order of colours buys
Walker’s alternation is a discipline, not a necessity. The one-yarn rule only requires that rows nest, and a knitter is free to work two blue stripes running, or three, whenever the chart’s next step is up again. The cost is in the fabric: a light stitch slipped through two blue stripes is held across four knitted rows, stretched and pulled in, which is why Walker’s form forbids it.
Counted by a transfer over the rows — a row may go to any row that contains it or lies inside it — the free order allows more charts than the strict one, by a factor that peaks at 7.22 at four stitches and falls steadily after, to 2.67 at fourteen. The reason it falls is the same reason the columns were independent under alternation. A repeated colour is a choice made for the whole row, and in a wide row it has to suit every column at once; the wider the repeat, the more likely some column wants the other direction, and the free order collapses back towards the alternating one. At the widths mosaic patterns are actually drawn, the free order is a factor of a few on a count of 10^40 and more, and the price of it is the elongated stitch Walker’s rule was written to avoid.
The float behind a slipped stitch
A slipped stitch is passed with the working yarn held behind it, so a run of slips along a row leaves a float on the back, exactly as a run of the other colour does in stranded knitting. Mosaic knitters therefore keep a float rule as well, and it is the stranded rule with the colours turned into slips: on a blue row the floats are runs of light cells, on a light row runs of blue cells.
With both rules in force the chart no longer factorises either way: the slips are fixed by the columns and the float limit reads them along the rows. Counted by a transfer over rows at twelve stitches and twenty rows, a limit of three slips keeps 29 per cent of Walker’s charts and a limit of five keeps 89. A limit of one, which would forbid any two adjacent slips in a row while Walker’s rule already forbids them up a column, keeps about one chart in 143 million.
That makes the two crafts’ floats directly comparable, and both lie on the side of the fabric that shows the loops’ heads. A stranded row puts a float behind every stitch and its limit binds the lengths of runs of colour; a mosaic row puts a float only behind its slipped stitches, and at most every other stripe of any column is slipped. So a mosaic fabric carries less yarn across its back than a stranded one — less of the held-off second layer that makes a stranded fabric warm — and the next section is the same statement in yarn.
A mosaic fabric’s yarn depends on its chart
What a tuck costs priced a miss at one wale spacing of float against a knitted loop of wale spacings — Munden’s constant, about four: replacing a knit with a miss removes of a loop. A slipped stitch is a miss, so a slip-stitch fabric saves exactly that for every stitch it slips.
Here the two crafts part completely. A stranded fabric, as the stranded essay showed, spends a quarter more yarn than plain jersey whatever its chart, because every stitch carries one loop and one float. A mosaic fabric spends less than plain jersey, and how much less depends on the chart: between nothing, for a chart that slips nowhere, and 37.5 per cent dry-relaxed at the most a Walker chart can slip, which is half its stitches, since no column may slip two stripes running. Drawn at random from Walker’s charts at twenty rows, a chart slips 28.4 per cent of its stitches — the long-column share tends to , 27.6 per cent — and spends 21.3 per cent less yarn per stitch than plain.
The figure is per stitch, and that matters more here than in the stranded case. A slipped stitch is a loop stretched through a whole stripe more than its neighbours, and it pulls the fabric in along the wales; a mosaic fabric has more stitches in a square metre than a plain one at the same loop, so its weight per square metre does not fall in step with its yarn per stitch. A knit’s weight nearly names its yarn for plain jersey because its loops are all one kind. A mosaic fabric’s loops are two kinds in a proportion its chart sets, and nothing here measures how much the elongated kind pulls in.
How the charts were counted
The fabric is written as one course a chart row — really two knitted rows, right side and wrong side, worked alike — in which every stitch is a knit in the row’s colour or a slip showing the colour below. Rows are read as sets of blue cells. The first row stands on a foundation and is unconstrained.
Walker’s count is 2F(h + 2)^w − 2^w in exact integer arithmetic. The free-order count is a transfer over the 2^w possible rows, moving from a row to any row that contains it or lies inside it, computed with a subset sum and a superset sum at each step. The float-limited count is the same transfer with Walker’s directions and each row’s slipped cells — its cells not in its colour — required to have no run longer than the limit, read round the repeat. The nearest Walker chart to a target is built row by row for each starting colour, keeping a target cell wherever the row’s yarn can supply it and the colour below otherwise, and the start that moves fewer cells is kept. The yarn uses Munden’s for the three relaxed states, as the stranded accounting did; the slip share of a random Walker chart is the share of marked cells over all F(h + 2) column strings, by a transfer on the last cell.
Required of it, not shown: that Walker’s closed form and the free-order transfer both agree with a count of every chart at three small sizes; that every Walker chart nests; that the mosaic version of the diamonds alternates its nesting at every row; that a long column’s slip share approach ; and that the free-order count refuse a width it cannot carry.
What the counting leaves out
Charts are counted, not designs, exactly as in the stranded case, and the shares describe the catalogue rather than any pattern book. The two readings that do not depend on what is drawn are the per-cell freedom and the yarn arithmetic.
The elongation is not modelled. Every slipped stitch stands a stripe taller than its neighbours and draws the fabric in; garter-stitch mosaic and stocking-stitch mosaic behave differently there, and a free colour order that slips a stitch through two stripes draws it in further. All of it changes dimensions and none of it changes a count.
Tension is absent again. A float behind a run of slips is laid at the width the fabric has on the needles, just as a stranded float is, and a knitted float shortens its fabric; how tightly a knitter lays it is the difference between a flat mosaic and a puckered one, and nothing here prices it.
Who knew which part
Slip-stitch colourwork is old knitting practice. Barbara G. Walker systematised its strictest form in Mosaic Knitting (1976), with charts in which every row is worked in one colour, the colours alternate every two rows, and a stitch is slipped through one stripe at most — the rule counted here. The Fibonacci count of strings with no two adjacent marks is classical, and the transfer matrices are the same method as the stranded count and the run-length-limited codes behind it.
What is added here is the criterion that a chart can be knitted one colour a row exactly when its consecutive rows nest; the count of Walker’s charts, 2F(h + 2)^w − 2^w, from the columns being independent; the golden-ratio freedom per cell and its coincidence with a two-stitch stranded limit; the factor a free order of colours buys, peaking at four stitches; the share of Walker’s charts a back-float limit keeps; and a slip-stitch fabric’s yarn, which unlike a stranded fabric’s depends on the chart.
Still open: how far a slipped stitch pulls the fabric in
Every count here is a count of charts, and every yarn figure is per stitch. What would turn the second into a weight is the one number the notation cannot supply: how much shorter a course of stitches becomes, along the wales, for every stitch in it that is slipped through a stripe.
It has a measurement that isolates it. Two swatches of one yarn on one needle, one plain and one in a mosaic chart whose slip share is known exactly — a checkerboard of single slips, a quarter of its stitches — relaxed alike, measured in courses per centimetre. Their ratio, divided by the slip share, is the pull-in per slip, and with it the per-stitch saving above becomes a weight per square metre. The prediction is a direction: the mosaic swatch must have more courses per centimetre than the plain one. If the excess is large, a mosaic fabric weighs more than plain jersey despite its yarn per stitch; if it is small, the saving survives into the weight, and slip-stitch colourwork is the one way to put two colours into a knit that makes it lighter.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- A second bed changes what a float is — both name float, miss, stitch notation
- A cloth's derivation class is its census of small patches — both name census, float
- A damask's edge floats further than its figure — both name census, float
- A fashioned edge has a quantised angle — both name loop length, munden constants
- A jersey has no float for a wire to catch — both name census, float
- A jersey's drape does not know its loop — both name loop length, munden constants
Named objects
A flat tag is an object no other essay names yet.
CensusFloatLoop lengthMissMunden constantsStitch notationTwo-colour symmetry