Woven on its side, a tapestry's slits carry its weight
Worth reading first: A tapestry's slit is the chord of one end · A cloth does not mind a hole · Balance, and what an unbalanced cloth does.
A tapestry’s slit is the chord of one end worked out how long a slit tapestry is cut along its outlines. Where an outline runs straight up the warp it is cut for its whole length. Where a curve turns vertical it is cut for the chord whose sagitta is one end spacing, between and whatever the weaver does. Everywhere else the cuts are short and harmless.
All of that assumed the picture was woven the way it is seen, with its vertical along the warp. The workshops that made the great European tapestries usually did not. From the late Middle Ages on, most were woven on their side: the cartoon was turned a quarter turn behind the warp, the weaver built the picture from one side edge to the other, and when the finished piece was hung the warp ran horizontally across the wall.
The reasons usually given are several. A loom’s width sets one dimension of what it can weave and its warp’s length the other, so turning a wide picture lets it be woven on a narrower loom. The weft’s fine ribs then run vertically on the wall, which suits a picture full of standing figures. And the figures’ long verticals become horizontals on the loom, where a colour change between passes leaves no slit at all. It is the third reason that is computable, and it turns out to be an exchange with two sides.
Turning exchanges the two axes and nothing else
A quarter turn does one thing to a cartoon: what ran up the picture now runs across the loom, and what ran across now runs up. The slit rule does not change. A slit is still a run of passes in one gap between ends, it is still long where an outline runs along the warp and short where it crosses it. What changes is which of the picture’s outlines run along the warp.
So the exchange can be read straight off the picture. Every line that is vertical in the picture is cut when woven upright and costs nothing when turned. Every line that is horizontal in the picture costs nothing when upright and is cut when turned. A curve’s two vertical tangents become its top and bottom, and its top and bottom become vertical tangents.
In an ordinary striped cloth the same arithmetic costs nothing, because a stripe is a partition of the warp and every pick crosses every band; it is only a discontinuous weft that turns an edge along the warp into a cut. For straight lines that is the whole story. A column’s edge, 120 cm long, is a 120 cm slit upright and nothing at all turned; a floor line 90 cm long is nothing upright and a 90 cm slit turned. For curves there is more, because a curve’s vertical tangent and its horizontal one are not equally curved.
A tall shape is flattest where it turns vertical
An ellipse of half-width and half-height has its vertical tangents at the ends of its horizontal axis, where its radius of curvature is . A tall ellipse, larger than , is therefore flat where it turns vertical — it is nearly straight along its sides — and its slits there are long. Turned, its vertical tangents move to the ends of its long axis, where the radius is , and it is sharply curved.
The slit goes as the square root of the radius, so the ratio between the two weavings is . Turning a shape shortens the slits at its verticals by its aspect to the power three halves, and lengthens them by the same factor for a shape wider than it is tall.
That is a strong effect. A robe drawn as an ellipse 28 cm wide and 92 cm tall, the one in the figure above, has an aspect of about 3.3, and its side slits fall from a mean of 8.5 cm to 1.4 cm when the cartoon is turned: a factor of six. A head, at 14 by 18 cm, is nearly round and gains little. The shapes that a picture of people is made of are tall, and every one of them is cut less when the picture is woven on its side. That is the cartoon-level reason the workshops’ practice made sense for the pictures they wove.
But every horizontal is cut instead
The exchange has a price, and it is the picture’s horizontals. A floor line, a horizon, the top of a wall, the hem of a tablecloth, the lintel over a door — every one is a single uninterrupted change of colour, which woven upright is a boundary between passes and leaves nothing, and woven on its side is a boundary along the warp and leaves one slit as long as itself.
That is the floor line in the figure above. Woven upright it costs nothing; turned, it is a single cut 90 cm long, the whole width of the hanging, and it is by far the longest slit the turned tapestry has. The columns’ tops, 12 cm each, are the next longest. The robe’s sides, which were the curve problem when upright, have shrunk to under 3 cm between them.
So turning a cartoon does not reduce its slits so much as move them from its verticals to its horizontals, with the curves favouring whichever axis is shorter. Whether it helps depends on which the picture has more of.
A landscape loses both ways
A picture of standing figures between columns is the case the practice was made for. A landscape is the opposite, and the same calculation says so.
The horizon is the reason. Upright, it costs nothing; turned, it is a cut the whole width of the hanging. The hills are wide, low ellipses, so the aspect rule works against turning them too: their flat crowns become vertical tangents on the loom, and leave slits of five centimetres each where, upright, their steep sides left under two.
For a landscape, then, turning is a loss on every count. The cut grows sixteen-fold, from 11 cm to 176 cm, and the longest slit from under two centimetres to a metre and a half. A workshop that wove everything on its side would have been right for its figure subjects and wrong for its verdures, the leafy landscape tapestries that were a large part of the trade.
Upright, nothing hangs across a slit
So far this is about length, and length is only half the story. A slit that runs along the direction a hanging is pulled is a very different object from one that runs across it.
A tapestry hangs from its top edge, and each horizontal line through it carries the weight of all the cloth below that line. Woven upright, the warp hangs vertically and so does every slit. The weight runs along the slit, in the same direction as the warp ends either side of it, and nothing has to cross the cut: the two ends that bound it each carry their own share straight down. A vertical slit is a place where the cloth is not joined sideways, and almost no load in a hanging is sideways. This is the same reason a cloth does not mind a hole the way a plate does: each thread carries its own load, and a cut beside a thread that is not asked to share does not change what it carries.
Woven on its side, the warp hangs horizontally and so does every slit. Now the weight runs across it. The cloth below a horizontal slit hangs from the cloth above it, and along the slit’s whole length there is no thread to hang from: the weight that would have crossed it has to be carried round its two ends. A long horizontal slit high in a hanging is a hinge with the whole of the tapestry below it pulling it open.
The weight that has to go round
That can be made into a number without a model of how cloth stretches. Give each slit a load measure: its length times the depth of cloth hanging below it. That is the area of tapestry whose weight would have crossed the slit and must now go round it. It is a ranking rather than a stress — two slits with the same measure divert the same weight, and how the cloth around them copes is a question about the cloth.
For the figure, turning cuts the total slit by more than three quarters and puts what is left under load: 5,999 square centimetres of cloth, the weight of nearly half the hanging’s area, now hangs across a cut instead of beside one. Most of it is the columns’ tops — 24 cm of cut with 142 cm of cloth below — and only a third is the floor line, which is nearly four times longer but sits low, with 22 cm below it. Where a horizontal sits in the picture matters as much as how long it is, and a short cut at the top of a turned tapestry carries more than a long one near the bottom.
For the landscape, turning gives sixteen times as much slit and puts all of it under load, with the horizon alone accounting for three quarters of the measure.
What a finer warp does to a straight line
The chord of one end found that a finer warp shortens a curve’s slits only by the square root of the sett ratio. For straight lines it does not shorten them at all.
A vertical straight edge sits in one gap for its whole length at any sett, so its slit is its length; a horizontal straight edge woven on its side is the same. Doubling the warp takes 35 cm off the figure’s upright slits, all of it from the curves of the robe and the head, and leaves the columns’ 480 cm untouched. It takes a centimetre and a half off the turned weaving’s 120.
That makes the comparison sharper. Against a picture’s straight lines, the only remedies are the orientation and the join; the sett has nothing to offer. A cartoon of columns, doorways and floor lines is cut by exactly its straight lines along whichever axis the warp runs, and the designer’s choice of axis is the whole of the decision.
How much there is to sew
A slit is sewn once the tapestry is off the loom, and a weaver leaves the short ones. So the real bill is the total length of slits longer than whatever the weaver is prepared to leave open, and it depends on that threshold.
The curves are flat where a cartoon’s long straight edges hold them up: the figure’s upright curve never falls below 480 cm, which is its columns, and its turned curve never below 90 cm, its floor line. Above about ten centimetres nothing changes at all until the threshold passes the straight edges themselves. The bill is set by the straight lines, and the curves decide only its small change.
Read with the load measure beside it, the choice for the figure is not obvious. Upright, the weaver sews four and a half times as much slit, but every centimetre of it runs down the hanging and carries nothing across. Turned, there is less to sew, but the seams that close it are what hold the cloth below them. A seam that carries weight is a seam that can fail, and a failed seam in a turned tapestry is a cut that gapes; in an upright one it is a cut that hangs closed.
The weft carries everything anyway
There is a larger consequence of hanging a tapestry on its side, and it is not about slits. In a turned tapestry the weft runs vertically on the wall, so the weft carries the whole weight of the piece, slit or no slit. That is a reversal of almost every other woven cloth, where the warp is the strong system and the weft fills in: a tapestry’s warp is the strong, plain, hidden thread, and its weft is the soft, dyed, decorative one, and the hanging puts its weight on the weaker.
Conservators attribute a great deal of the damage to old tapestries to exactly that — the weft, especially in its most heavily dyed colours, creeping and breaking under centuries of its own weight, until the warp shows through. That is not computed here; it needs a model of creep in dyed wool, which is a different calculation from any here. But it belongs beside the slit result, because both say the same thing: turning a tapestry makes its decorative system structural, at its slits and everywhere else.
Why the practice persisted
Given all of this, why weave on the side at all? Partly because the subjects suited it. The workshops’ figure subjects — courts, processions, histories — are full of tall shapes and verticals, and for them turning cuts the slit by large factors, as the figure cartoon shows. Partly because of the loom: a wide hanging woven on its side is woven on a warp as long as it is wide and a loom as wide as it is tall. And partly because the weft’s ribs, turned vertical, catch the light along a picture’s standing forms.
None of those reasons needs the load to be ignored. A turned tapestry’s slits were sewn, and sewn well; the stitches that close them were made to carry. A sewn slit is a seam with no allowance at all, and a loaded seam in a loose weft-faced cloth fails the way a seam slips before it breaks describes — the weft pulling out beside the stitches rather than anything snapping — which is the failure conservators most often find along old sewn slits. The calculation here says what those stitches had to do, which the practice took on without counting.
Whether the cut can be told from the tear
It is worth asking what a loaded slit is like, compared with a tear, which in a cloth asks fewer threads than a pull because it takes them one at a time. A tear in a woven cloth stops where the grip is: it runs across threads until friction holds the next one and it breaks instead of sliding. A slit is a cut already made, with no thread across it at all, so there is nothing to grip along its length and the question is only what happens at its two ends.
At each end the weight diverted round the slit has to pass through the first few passes where the weft crosses again. In a continuum that would be a stress concentration, rising with the square root of the cut’s length. In a cloth whose threads carry their own loads it is not obvious that it rises at all, and that is the measurement this calculation stops short of: how much of the diverted weight the first crossing weft takes, and how fast the share falls off with the passes above it.
The model, and the numbers in it
Each cartoon is a list of shapes in centimetres painted in order over a ground, sampled at the centre of every end and every pass, at eight ends and 32 passes a centimetre unless a figure says sixteen. A gap is open on a pass when the two ends either side carry different colours; a slit is a maximal run of open passes in one gap. Turning is a quarter turn of the cartoon, after which the same rule is applied and the slits are drawn back in the picture’s own orientation. The load measure is each slit’s length times the height of the picture below it, summed over slits longer than a centimetre. The ellipse comparison uses the exact curvature at each vertical tangent and the mean slit over every placement of the cartoon against the warp. The two cartoons are constructions to make the comparison, not copies of any tapestry; the one-centimetre threshold is a yardstick, not a trade figure.
What the drawings cannot show
How far a loaded slit opens. The load measure ranks the slits by the weight they divert. Turning it into a gape needs the cloth’s stiffness across the warp and along it, which in a weft-faced cloth are very different — the same imbalance that makes an unbalanced cloth wrinkle one way and not the other — and a model of the cloth round a cut, which nothing here has.
The seams. A sewn slit is a joined slit, and everything above is the bill before the sewing. What the seam can carry depends on the thread and the stitch, and is not modelled.
The weft’s creep. The weight a turned tapestry puts on its weft everywhere is named and not computed.
Joins other than the slit. Dovetailed and interlocked joins remove slits at the loom. A workshop that interlocked its horizontals when weaving on the side would pay a different price, in a ridge on the back rather than a cut, and that is not counted here.
Who found it and when
That European tapestries were usually woven with the cartoon turned, so that the warp runs horizontally when hung, is standard in the literature on the medium, as is the observation that this turns the picture’s verticals into horizontals on the loom. The damage that hanging puts on the weft is a commonplace of tapestry conservation. What is added here is the exchange made exact: the aspect to the power three halves for a shape’s verticals; the cut that every horizontal becomes; the load measure that separates an upright slit, which nothing crosses, from a turned one, which the cloth below hangs across; the two cartoons that come out on opposite sides; and the observation that a finer warp does nothing to straight lines, so that orientation and join are the only remedies for them.
Still open: what a dovetail trades
Both weavings here are slit tapestry. The weaver’s other answer is the join: two wefts sharing one end, turning round it in alternate groups of passes, so that no slit is longer than a group. That bounds every cut at the loom whatever the orientation — the floor line and the columns alike — at the price of an edge that zig-zags by one end with a period of two groups.
So a dovetail converts a slit’s length into an edge’s roughness, and the exchange rate is the group size. A group of four passes at 32 passes a centimetre cuts nothing longer than an eighth of a centimetre and serrates the edge every quarter. The question the slits leave is whether there is a group size at which the serration is finer than an eye at the hanging’s viewing distance can see — which would make the dovetail a free join for long straight edges — and how that size compares with the one-centimetre slit a weaver will leave open.
What links here
Computed from the collection rather than written here: the essays that point at this one.
Reads more easily once this is understood
Essays that name this one as worth reading first.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- A damask's edge floats further than its figure — both name figure and ground, weft-faced
- A woven outline is a staircase — both name figure and ground, sett
- Turn the cloth and the shine changes hands — both name anisotropy, weft-faced
- Wicking is slower along a crimped thread — both name anisotropy, sett
Named objects
A flat tag is an object no other essay names yet.
AnisotropyFigure and groundSettSlitStress concentrationTapestryWeft-faced