A jersey leans because its yarn still turns

A single-jersey garment that has been washed a few times has its side seam somewhere other than its side. The wales lean, the seam follows them, and the shirt turns slowly on the body of whoever is wearing it. The trade calls it spirality, it is the commonest complaint about single jersey, and the standard remedies are all about the yarn: use a lower twist, use a plied yarn, steam it, or knit it with two yarns of opposite twist.

Every one of those is a remedy aimed at the yarn’s residual torque, and every one of them works. But there is a fabric that has none of them applied to it, is knitted from the same yarn on the same machine, and does not spiral at all: a 1×1 rib.

So the yarn is not the whole story, and the part that is missing is a count.

The claim #

A loop knitted on the front bed and one knitted on the back are mirror images, so their torques oppose, and a fabric that knits equally on both beds has no net torque whatever the yarn’s twist is.

That gives a count. Take a structure, count the knitted loops each of its wales makes, sort them by which bed they are on, and the imbalance is the difference over the total. Single jersey is one. The ribs and the interlock are zero.

The count has to be made per fabric and not per structure, and that qualification is the whole of the interesting part.

Where the torque comes from #

A yarn leaves the spinning frame with twist in it and with a torque it has not been allowed to release. That torque is what makes a length of singles yarn snarl on itself when the ends are brought together, and it is a standing quantity: the yarn is in equilibrium only because its two ends are held.

Knit a loop from it and the torque acts on the loop. It tries to rotate the plane the loop lies in, and the accumulated effect over a wale is a lean — the wale is no longer perpendicular to the courses, and the whole fabric skews.

How much lean per unit of torque is a measurement, and this collection carries it in the form the trade uses: a lean in degrees per unit of twist factor above the balanced value. That relation is fitted rather than derived, and it is marked as fitted wherever it is used. A twist factor of 4.0 against a balanced 2.4 gives 14.4 degrees.

Why the ribs are exempt #

A rib knits alternate wales on opposite beds. The two beds face each other, so a loop drawn on the front and a loop drawn on the back are mirror images through the plane of the fabric — the yarn goes round the needle the other way.

A mirror image reverses a handedness, so it reverses the sign of a torque. The torque a front loop applies to the fabric and the torque a back loop applies are equal and opposite, and a structure with equal numbers of each nets to zero.

That is a statement about the structure and not about the yarn, so it holds at any twist factor. It is why a rib knitted from a hard-twisted singles yarn, which would produce a badly spiralling jersey, produces a rib that hangs straight.

A half-cardigan is the intermediate case and confirms the arithmetic rather than illustrating it: it knits four loops on the front and two on the back in its repeat, giving an imbalance of a third and a predicted lean of a third of the jersey’s.

The case that makes the count non-trivial #

An interlock and a tube both knit equally on the two beds. Counted over the whole structure they have the same imbalance, which is zero, and they behave completely differently: an interlock hangs straight and a tubular fabric twists.

The count is wrong because the two are two fabrics each, and a two-component structure is two fabrics that happen to have been made on the same machine at the same time. The imbalance that decides whether anything leans belongs to a fabric.

Split them into components and the two separate cleanly.

An interlock’s two components each contain one front wale and one back wale. Each component straddles the beds; each is balanced; neither leans.

A tube’s two components are the front bed’s wales and the back bed’s wales. Each component is wholly on one bed; each has an imbalance of one; each leans, and they lean in opposite directions — which is exactly the observed behaviour of a tubular fabric, whose two faces skew against each other and whose seam wanders.

And what separates them is which beds each component draws its wales from, which is a quantity this collection’s integrity criterion computes and which nothing about the point-paper representation shows. The criterion was built to answer whether a draft describes one cloth or several. Here it is deciding whether a fabric spirals, which is not a question anybody built it for.

Why the count is a permission and not a prediction #

It is worth being precise about what the count does and does not settle, because the two are easy to run together.

What it settles is whether a structure can lean at all. A balanced structure’s torques cancel exactly, at every twist, so the answer is no and the answer does not depend on anything measured. That is a permission, in the sense this collection uses the word: a statement about what is possible rather than about what happens.

What it does not settle is how far an unbalanced one leans. That needs the relation between residual torque and lean, which is fitted, and the residual torque itself, which depends on the yarn’s twist, its fibre, whether it has been steamed and how long ago it was spun. Two jerseys from two yarns lean by different amounts and the count says nothing about which.

The distinction matters because the two halves have different reliabilities. The permission is exact and structural; the magnitude is a measurement with a range around it. A reader who takes the 14.4 degrees at a twist factor of 4.0 as a prediction has taken the weaker half for the stronger.

It also says where the useful engineering is. A garment that must not spiral is not a garment made from a carefully chosen yarn — that only reduces the effect. It is a garment made in a structure whose count is zero, and the choice of structure is free of every uncertainty the yarn carries.

A test the account has to pass #

An account of spirality that leans on the yarn’s residual torque has an obvious prediction attached, and it is worth writing down because it is the thing that would falsify this rung.

A yarn with no residual torque should produce no spirality in any structure. A two-fold yarn twisted the opposite way to its singles is very nearly torque-balanced, and a steamed yarn has had its torque set out. Single jersey knitted from either should hang straight, and it does — which is why plying and steaming are the standard remedies and why they work in a way that changing the machine does not.

The converse test is the one this rung adds. A structure with a zero count should hang straight from any yarn, including a hard-twisted singles that would spiral badly as a jersey. That is also observed, and it is the observation the count explains.

The two together pin the mechanism down between them: the yarn supplies the torque and the structure decides whether the torques cancel, and removing either removes the effect. An account that had only the first half would predict that a rib spirals, and an account that had only the second would predict that a plied jersey spirals. Neither does.

What was counted, and how #

The loops are counted from each structure’s own two grids, one per bed, and the counting is of knitted loops rather than of needles — a tuck is not a loop and a miss is nothing.

The components come from the connectivity of the wales, which is the same computation that decides whether a structure hangs together.

Four assertions carry the argument. Single jersey must be wholly unbalanced. The ribs must be balanced exactly, at zero and not at a small number. The interlock must have two components with a worst imbalance of zero, and the tube two components with a worst imbalance of one. And the criterion that separates the last two must be the beds each component draws from, which is asserted directly.

The lean itself is the measured relation and is marked as such wherever it appears. Everything this rung claims is a statement about which structures lean, and the degrees are the trade’s number rather than this collection’s.

The table the count produces #

The count is a permission and the essay says so, and it is worth setting out for the structures a knitter chooses between — because the permission alone sorts them, and the sorting is the design decision.

Take each structure’s repeat, count the knitted loops on each bed per component, and the imbalance is the difference over the total.

Three things fall out that a rule about balanced structures does not give.

Every even rib is exempt and every uneven one is not. A 2×2 rib is as free of spirality as a 1×1, and a 1×2 is a third of a jersey — which is a real intermediate rather than a small effect, and it is the structure a knitter reaches for when a rib’s appearance is wanted with less width contraction. The lean comes with it, and nobody warns about that because the rule everybody carries is “ribs do not spiral”.

The half-cardigan and the 1×2 rib land on the same figure by different routes. One is unbalanced in its knits and the other in its beds, and the count does not distinguish them because it is a count of knitted loops. That is a coincidence of these two structures rather than a general fact, and it is the kind of agreement that makes a count worth trusting: two quite different repeats reduced to one number that predicts the same thing.

And the two zeroes are not the same zero. The ribs’ is a cancellation within a component; the interlock’s is a cancellation within each of two components. Both give a fabric that hangs straight, and only the second survives the fabric being cut — because an interlock cut across its wales is still two balanced components, while a rib cut along a wale line leaves a strip that is wholly on one bed and leans like a jersey.

That last is a practical consequence nobody states. A narrow strip cut from a rib can spiral even though the fabric it came from does not, if the cut isolates wales of one bed — which is exactly what a cut-and-sew rib trim is, and exactly the component where spirality is complained about in finished garments.

Where the model stops #

The lean per unit of twist is fitted. Nine degrees per unit of twist factor above balance is a working figure and not a derivation; deriving it would need a loop’s torsional compliance and the moment a residual torque applies to it, and this collection has neither. Every number in degrees here inherits that.

The balanced twist factor is quoted. A twist factor of 2.4 is taken as the point at which a cotton singles yarn has no residual torque, which is a convention with a range around it rather than a constant.

Tucks are counted as not-loops. A half-cardigan’s back bed makes two knitted loops and two tucks in its repeat, and the tucks are given no torque at all. A tuck does wrap the yarn round a needle and presumably carries some share, so the third-of-a-jersey figure for a half-cardigan is a lower bound rather than a prediction.

And the count says which structures lean, not how much. A structure with an imbalance of a third is predicted to lean a third as far as one with an imbalance of one, and that proportionality is an assumption about torques adding linearly which nothing here tests.

What spirality has in common with the rest of this field #

Everything else in this collection’s account of what a fabric fails to give back has been about friction and geometry. Spirality is the one item that is genuinely a stored load, and putting it beside the others is instructive.

A woven cloth’s excess is put in by the loom and comes out over five washes. A knit’s distance from its fully relaxed state is put in by the knitting and comes out over three named states. Both are positions: the fabric is somewhere it would not choose to be, friction is holding it there, and agitation lets it move.

A yarn’s residual torque is not a position. It is a load the yarn has been carrying since it was spun, and it does not go away when the fabric relaxes — it is what drives the relaxation, in the direction of the lean. So a jersey does not spiral less as it settles; it spirals more, because settling is what lets the torque act.

That is the sign that separates the two mechanisms and it is checkable without any instrument. A shrinkage gets smaller wash by wash. A spirality gets larger, up to whatever the structure permits, and then stops. Anybody who has washed a T-shirt several times has the data.

The generalisation #

A property that presents as belonging to a material can be cancelled by an arrangement, and the cancellation is exact rather than approximate when it comes from a symmetry. The torque in a knitting yarn is real and is unchanged by putting it into a rib; what the rib does is arrange the loops so that the torques sum to zero, which works at any magnitude and needs no tuning.

That is a much stronger kind of fix than reducing the cause. Lowering a yarn’s twist reduces spirality proportionally and never to zero; knitting a rib removes it entirely. Wherever a symmetric arrangement is available, it beats a reduction, and the standing example in engineering is a differential pair against a well-matched single-ended amplifier.

The second lesson is about the level a count is taken at. A count over the wrong object gives the same answer for two things that behave oppositely, and no amount of care about the counting fixes it. Here the wrong object is the structure and the right one is the fabric, and the difference is invisible until something computes which is which.

Who found it, and when #

Spirality in single jersey is as old as the fabric and its dependence on yarn twist is thoroughly documented; the remedies — plied yarns, steam setting, two-feeder alternation of S and Z — are standard practice. That ribs and interlocks do not spiral is equally well known and is usually explained by saying that the structure is balanced, which is right and is the sentence this rung puts a count behind.

What appears to be this collection’s own is the observation that the count must be taken per component rather than per structure, and that an interlock and a tube are the pair that shows why. The integrity criterion that separates them was built four fields ago for an entirely different question, and this is the second time it has answered one it was not designed for.

Where the ladder goes next #

The torque this rung counts is a memory of spinning, held in the yarn and released when the fabric relaxes — so it belongs beside the other things a knit releases when it is allowed to, which is the path its three named states lie on.

Sideways, the criterion that separated the interlock from the tube is the same one that asks whether a double jersey hangs together, and the yarn’s twist has its own ladder in the helix that makes a bundle behave like a cylinder.