Main Menu - Misc. - Clothing/Textiles - Medieval Wales - Names - Other Medieval - Publications - Harpy Publications

This page last modified May 31, 2004

A Very Brief Illustrative Discussion of the Llangors Embroidery

Thrown together in a hurry by: Heather Rose Jones

Included here is the text of a posting I made to the SCA_Welsh mailing list, followed by a few redrawings of illustrations from the original article, plus a picture of my own first experiments with this technique.

I cannot emphasize enough that anyone interested in this piece should buy the original publication! (Available through Oxbow/David Brown Books <>.)

Go directly to the pictures.

The Article

Granger-Taylor, Hero and Frances Pritchard. "A Fine-Quality Insular Embroidery from Llan-gors Crannog, near Brecon," in Pattern and Purpose in Insular Art, edited by Mark Redknapp et al. Oxford: Oxbow Books, 2001, 91-99

The article is a combination of some fascinating and useful details and some maddening and tantalizing omissions. Here's a detailed summary of some of the information, organized by topic.


The history of the settlement where it was found dates the item fairly narrowly to the couple of decades around 900 CE. Geographically, we're talking about a region in the south-eastern part of Wales with reasonably close connections with some parts of England as well as with Wales. The site was a royal capital, and the amount of work that went into the embroidery probably indicates that this was a very high-status garment, not something everyday.


The ground fabric is a tabby-weave linen with a thread count around 25 threads per cm. Just as a comparison, the type of linen generally found in the stores today labelled "handkerchief linen" (the finest linen you can ordinarily get in yardage) seems to run around 15 threads per cm, while some drafting linen that I happen to have lying around (which is impossibly fine) runs around 30 threads per cm. So if you visualize the normal weights of linen that you might be used to finding in fabric stores, this is probably twice as fine as those.

Embroidery materials

The embroidery is done in silk thread. There are two types, both reeled silk rather than spun (i.e. with extremely little twist): one plied from two strands (used for the main designs) and the other unplied and more glossy (used for the backgrounds). It's quite probable that multiple colors were involved as well, but because the textile was burnt before being waterlogged all visual color information is lost. (I don't know whether there's any potential for chemical analysis.) The size of the thread can be judged by the technique (see next).

Embroidery technique

The embroidered areas are solidly covered (although there are also extensive amounts of unembroidered textile in the find). The basic stitch is a sort of "counted stem stitch" or one might think of it as an "elongated needlepoint". Each stitch goes over three ground threads and up one thread from its starting point, then comes back under one thread in each direction so that when it emerges on the right side again, it is in the same row as the previous emergence, but two threads advanced. There is a row of embroidery for every thread of the fabric (so you can figure that the embroidery threads are about the same size as the ground threads, or maybe a little bigger). Successive rows are offset by one thread, so the face of the embroidery gives a sort of twill effect. Or rather, given that the background and the pattern are worked at right angles to each other, the effect is very much like a damask pattern weave. (One very likely theory is that the technique is designed to imitate a damask-type pattern weave in needlework.) Not only are the ground and figure at right angles in each motif, but the direction may be alternated in successive motifs, so that sometimes the threads of the main figures follow the warp and sometimes the weft.

Embroidery motifs

The motifs are stylistically reminiscent of contemporary imported silks from the Near East, or possibly even the Far East. There are two narrow band-motifs -- one with small "lions" in facing pairs, one with simple "trefoils" emerging from alternate sides of the band and separated by diagonal lines. These are described as being around 12-14 mm wide -- that's about half an inch.

Then there is a much larger band pattern with hexagonal frames formed by stylized vegetation (with the occasional leaf, flower, or fruit) enclosing stylized birds of various types. This pattern comes in pairs that are mirror-images of each other.

It appears that there may be yet a different set of patterns in the triangular gusset (see below), but these appear to be worn enough that they are hard to make out.

In addition, the narrow bands are edged by a row of stitches that cross diagonally over two threads in both directions, then return back to the original row but advanced one stitch. There are also some sewn reinforcements at the point of the triangular gusset, but it looks to me from the photo as if these are made from applied braid like that covering the seams (see below) rather than being purely embroidered.


Narrow braids, believed to be flat fingerloop braids, are applied over the line of the seams on the right side of the fabric. These are presumably sewn down in the process of sewing the seam, but the details are not given.

Other sewing

The description mentions a hem around 80 mm deep (that would be very roughly around 3"), but doesn't describe the sewing technique, other than mentioning that there's another braid sewn along the bottom edge of the hem.

The construction of the underlying object

Based on lining up elements in the embroidery and seams, a small amount of cut/construction information can be retrieved, but evidently much of the item survives as isolated pieces of undecorated linen whose relative arrangement can't be retrieved. The embroidered portion reconstructs into a narrow triangular gusset -- the point is around a 30 degree angle, and the part that survives appears to be around 30 cm long. This is attached at either side to rectangular panels. And here's where I want to know more about the reasoning that went into some of the suggestions and conclusions of the article, because some of their ideas aren't the ones that seem most obvious to me. The gusset is solidly embroidered, and the pairs of broad bands parallel the edges of the gusset on the rectangular panels. On one side, there is enough surviving textile to show a pair of broad bands (this would be roughly 15 cm wide), then an expanse of unembroidered linen somewhat wider (??maybe 20 cm??), then the beginnings of another wide band. On the other side, there's fragments of the first of what is presumably another double-band. The authors, noting that these two partial bands on the edges of the reconstructed portion _could_ be lined up to form another double band, continue on to hypothesize that the piece might be a sleeve with a gusset, and then argue against this hypothesis on the grounds that, if you're going to have embroidered and unembroidered portions of a garment, it doesn't make sense for the underarm to be heavily embroidered. And then they consider the possibility that "the gore ran downwards from the top of the shoulder" (which I'm having trouble visualizing), but then object that this would have key portions of the motifs oriented upside down (ignoring the matter that the double-bands with the birds will necessarily have either some or all of the birds oriented non-vertically).

This is where I'd like to know what the underlying reasoning process was, because the type of gore that they say "are often found under the sleeves of garments, at the armpit" are not, in my experience, necessarily characteristic of early medieval garments (to the extent that we have examples of these at all), at least in Europe. Furthermore, underarm gussets are not typically this acute in shape or this long. (The only relevant example of underarm gussets I can think of is the 11th century Danish linen shirt from Viborg, where the gussets are square.)

They seem to give much shorter consideration to the possibility of side-seam gussets (e.g. beginning around the waist of the garment) or some similar configuration (e.g., center front and back gussets). But, in my opinion, this is the much more obvious likelihood. I.e., that what we have in the one gusset with associated material on the sides is, at best, roughly 50% of the circumference of the garment, and represents some portion of a tunic skirt. If this is correct, and the gusset starts around the waistline, and my measurement estimates (based on the scaled photos) are reasonable, then we're looking at a skirt circumference of around 140 cm around the hips (and flaring somewhat more widely below that), which is quite reasonable for adult dimensions. This interpretation would then have the broad bands of embroidery running vertically along the edges of the front and back panels of the garment (sort of like clavii, but at the edges of the panel rather than leaving blank space at the edges).

Gripe time: the description fails to indicate where the hem that they describe is in relation to the set of embroidered pieces whose relationship can be reasonably determined. It may be that the hem is in one of the unembroidered pieces whose relationship to them can't be determined, but it would be really nice to know. Second gripe: they describe a "belt loop, 45 mm long, made from a short piece of silk tape or ribbon". Why do they describe this as a "belt loop" when belt loops are (with one 1st century exception) unheard-of in medieval garments? Was this piece of tape attached to the fabric? At both ends? If so, what was it attached to? If what they found was a 45 mm long piece of silk tape and they have no actual evidence how it functioned on the garment, then I wish they would have refrained from labelling it a "belt loop", to the confoundment of less knowledgeable researchers in the future who take the description at face value. Ok, end of major griping.

The final piece of fascinating information on construction is that the embroidered portions (or possibly all?) of the garment is lined, with plain fabric of the same type as the outer fabric and cut in the same way. This is almost as radical a piece of information as the embroidery itself.

Ok, so what sort of practical information does this give us about early medieval Welsh clothing?

The Pictures

Some day I may try to get permission to include scans of the original photos here, but in the mean time -- buy the book for yourself! (Even if you're only interested in the one article!)

First we have a rough outline of a part of the original textile. The edge of an embroidered band lies in the bottom part of the image, with a row of applied braid at its upper edge, and then a section of plain (umembroidered) textile in the upper half of the image. The embroidered portion has a row of small "lions" along the upper edge, and then part of a set of vine-frames below that. Although it's purely coincidental, this has ended up being only slightly larger than the original size.



This next image is a much smaller-scale view of the set of embroidered pieces whose relationship to each other can be reasonably guessed at. In the upper part, you can see the gusset running horizontally (large end at the left, point at the right). The ruler at the bottom of the image has one centimeter blocks, so you can see that this is reduced about 8-10 times from life-size.


Here's a schematic diagram of the embroidery technqiue. This diagram doesn't (to my knowledge) reproduce any actual specific motif -- it's just using square pattern blocks against a background to illustrate what's going on. The intent is that the small internal squares represent the "main pattern" and are the same color, but with the embroidery running different directions on alternating blocks, while the background is also a single color, but similarly alternating directions in successive blocks. At the lower edge there's a row of a wider stitch (over two rows rather than one) that seems to edge the narrow bands.


So no sooner have I read through the article than I must experiment with this embroidery technique myself. Here are my first and second attempts -- the lower one is the first, the upper one is the second. A detailed discussion will follow. These should display about life-size.


Test #1 (Lower) "trefoil" design

The motif is about 25 threads wide (it should be a little narrower, but there are a few faults in the pattern) so that's about 40 threads per inch, or roughly 17 per cm, compared with the original find's 25 threads per cm. The ground fabric is a piece of linen I use for testing or demonstrating embroidery techniques, so I didn't necessarily choose it as being close to the original thread count. The thread I used is identical for both the pattern and ground (unlike the original) -- it's Kreinik's "Ping-Ling" silk, used in two strands (of a 6-ply thread). Ping-Ling is a glossy reeled silk, but with a noticeable twist, which is why my stitches have a bit more definition than those in the book's sample above. (For the next experiment, I'll use an untwisted silk floss.) Alternating the directions of the stitches (which is most visible in the yellow pattern motifs below) does create an interesting visual effect, but not one that can be seen in a flat-on scan. It's most visible when the embroidery is viewed obliquely from the side, and especially when it is somewhat in motion.

This sample -- which comes out to roughly one square inch of embroidery -- took me a total of maybe 3-4 hours to do. (I wasn't keeping track.) Naturally, being in practice, and working on a larger piece continuously, would speed up the rate of work somewhat, but still this gives you a notion of the enormous amount of time the original garment must have taken. The technique is also -- if you'll excuse my language -- a bit of a bitch to do. It would be easier if the regions being done in a single color+stitch were larger -- as they are in the vine-and-bird sections. What takes time is not the stitching itself so much as turning corners, counting out the pattern edges, and especially working in small spaces between existing worked sections. I'm of two minds whether this works better as an actual counted-thread pattern (i.e., you figure out how many stitches to do in each row based on a thread count) or whether it would work to draw the pattern on the fabric and work that way. Perhaps both -- counted-thread for the small bands, and a drawn pattern for the larger sections.

Working the technique over single threads (rather than pairs of threads, which is more common in counted work) has some major disadvantages when you turn around at the end of a row of work. It might be interesting to try a larger-scale version of the technique worked over pairs of threads, with correspondingly larger motifs. On the other hand, there's no arguing with the fact that the original is worked on single threads.

Contemplating using this technique on an actual garment: the use of both embroidered areas and plain-fabric areas on the original garment conveniently licenses an embroiderer to choose how much area you want to cover. Starting with something along the lines of narrow bands along the bottom hem and sleeve hems might be something less daunting to try than the full original (whatever it may have been).

Test #2 (Upper) "lion" design

My second test piece is shown above alongside the first test. This time I used Kreinik "Soie Platte" which is an essentially untwisted silk floss -- a much better match for the thread used in the original embroidery. I had been planning to do a larger scale test piece, working over pairs of threads rather than single threads, but a single strand of Soie Platte was about the same thickness as the double-strand of Ping-Ling that I'd used previously, so I let my materials dictate the scale. Although it's hard to see in this scan, the difference in floss type made an enormous difference in effect. The surface of the embroidery is much smoother and glossier, and the difference in highlights between the horizontal and vertical sections of the work is much more pronounced (but, again, doesn't show up well in the scan, as it is best seen at an angle). Interestingly, although the piece took about the same time to work as my first test, I felt much less frustrated in working on it. Part of this may have been that the Soie Platte was less prone to kink up and knot, but part of it was simply better understanding the mechanics of what the stitch was doing.

Despite the description of each stitch going over three threads, the essential basis of this stitch is a two-thread unit. (Remember that the stitch goes over three, but then back one thread each time.) This tripped me up on the first test, because I kept expecting there to be a three-row repeat for the stitch offset (to get the "twill" effect), but it's only a two-row repeat. So if the pattern is conceived of as built up out of blocks of 2x2 threads (i.e., two rows with each row focusing on two-thread units) then it procedes naturally. You're doing two essentially identical rows in each pattern unit (although you can smooth out some corners by doing a half-stitch in one row, as I've done on the lions' faces, once you get the hang of the basic stitch). You work one row across from left to right, flip the work around 180 degrees, work the next row back (but still from left to right).

And here's the trick to getting the "twill" offset to come out perfect. Start each row with a 2-thread stitch -- i.e., from your starting point, put the needle through two threads to the right, and up one, from where the thread emerges, then bring it back up through the fabric one thread back and one down. (You should be able to do this in a single motion -- i.e., catching a single "cross" with the needle, under and up again.) From there, continue with the three-thread stitch, i.e. over three threads and up one, then back and down one. This has the effect of advancing the work two threads with each stitch, including the first stitch, making your natural 2x2 "pattern block". When you get to the end of a row, do your "back one and down one" finish the same as for any stitch, but then plunge the needle down through the same hole as the previous stitch -- i.e., the last stitch of the row is an "over one, up one" stitch. (This helps add a little bulk to the end of the row to avoid things looking ragged.)

Then, bring the needle back up in the same vertical column that you went down in, but in the next row to start the new stitches. (Whether the next row of stitches is above or below the current one depends on which way you're working, but you'll only be going under a single thread vertically in either case.) When you've finished the second row, to complete a pattern row, you'll be coming up to start the next pattern row wherever the pattern dictates, but remember to only go under a single thread in a vertical direction. (I may try to add illustrations for this at some future point -- learning embroidery techniques from verbal descriptions alone is not ideal.)

Small discontinuities in the pattern can be dealt with by leaving a float on the back and contining working the same horizontal thread, but stopping and starting the work as described above. When I did the lions, I did this for everything except the large areas of the body, but it probably would have been just as easy to work straight across in all cases. For both lions, I did the (red) background first and then filled in the pattern color. I don't know whether this would make any difference. For a larger pattern, I suspect that doing the main motifs first might be safer, and then filling in the background afterward.

In both test pieces, the resulting fabric is slightly stiff, but still quite flexible. Working this embroidery on a relatively large area would tend to cause the fabric to lie flat (which would show off the work nicely). So a relatively large embroidered area on a tunic would tend to stand out flat, with the non-embroidered areas on the garment being more likely to fall into folds.

I could wish that the surviving material included something that was clearly part of a neck opening. The embroidery all seems to be relatively "straight and square", but necklines in northern Europe at this period are generally shown as being rounded. So it would be interesting to see how the embroidery behaved around a curved neckline (if at all).

I'm now starting to get excited about trying this stuff out on an actual garment-sized project, but first I'm going to do some color research on surviving eastern silks from this general era to get a better notion of what sort of palette to aim for. Then I can spend five years or so embroidering.

This site belongs to Heather Rose Jones. Contact me regarding anything beyond personal, individual use of this material.

Unless otherwise noted, all contents are copyright by Heather Rose Jones, all rights reserved.