<---previous article


Rugs of East Turkestan Khotan,
Yarkand, or Kashgar?



by George O'Bannon

Originally appeared in Oriental Rug Review.



The appearance of a number of rugs from East Turkestan in recent auctions and the record setting prices achieved for many of them prompted me to take a closer look at the literature on these rugs. At the same time, I knew of a large group of these rugs in the inventory of Timur Shah of Afghanistan's Nomadic Rugs in Atlanta. I conducted a structural analysis of several and sent samples of the colors of seven to Paul Mushak for dye analysis. The following article and Dr. Mushak's accompanying dye study report the results of this survey.

The current classification of these rugs as Khotan, Yarkand, and Kashgar is based primarily on structural characteristics and secondarily on designs. In the last 10 years, a broad spectrum of these rugs and information on their structure have been published in a multitude of books and auction catalogs. The recent sale of the Meyer-Muller Collection which contained some excellent examples of East Turkestan rugs saw record prices. Two of these lots illustrate not only excellent types but attribution problems. Illustrations 8, Yarkand, and 9, Khotan, are from Christie's sale of September 11, 1990.

The differences between the two rugs are striking, and the structure is different. A case for origins in two different weaving centers or areas seems reasonable. (Illustration 8 sold for $143,000 and 9 for $52,800, including buyer's premium.) Yet a closer look at published examples and the pertaining literature leads to confusion.

I am not the first to raise questions about the classification of these rugs. Donald Wilber raised similar issues in an article in Oriental Rug Review in Vol. V, No. 8, November, 1985. It generated responses from Murray Eiland, Ulrich Schürmann, Charles Grant Ellis, and Wilber again in subsequent issues. Light was let in, but nothing was resolved. The following is another gleaning of the same ground but, I hope, including some new questions, some suggested new avenues for study, and a restatement of Eiland's remark that "we should not accept any of this uncritically," that the attribution of East Turkestan rugs will be handled more cautiously.




Illustration 8. Yarkand carpet, Christie's Lot 128, 9/11/90


Typically, Oriental rugs are divided into five major areas of production: Turkey, the Caucacus, Iran, Central Asia, and China. This division is still used by most writers but, in recent years, some writers have sub-divided the Chinese group into three areas: China, Tibet, and East Turkestan. There are good and valid reasons for this subdivision, based primarily on technical features. The rugs now classified as Chinese have fairly homogeneous characteristics and, using various differences, classifications are being employed to divide them into stylistic and age groupings. Tibetan rugs are woven with the cut senneh loop technique peculiar to them thus justifying a separate category. The East Turkestan rugs differ significantly in patterns, colors, and technical features from the Chinese and Tibetan rugs and form a cohesive group.

It has become common practice to further divide the East Turkestan group into three subdivisions: Khotan, Yarkand, and Kashgar. Motivations for this study included a desire to determine how these classifications came about, what the differences are, and on what factual basis the distinctions were constructed.
The literature on East Turkestan rugs is relatively sparse. The publications in which they are discussed in some detail or where a number of pieces are illustrated are:

Bennett, I., Rugs and Carpets of the World, A. & W. Publishers, New York, 1977

Bidder, H., Carpets from Eastern Turkestan, Zwemmer, London, 1964

Dimand, M. S. and Mailey, J., Oriental Rugs in the Metropolitan Museum of Art, Metropolitan Museum of Art, New York, 1973
Eiland, M., Chinese and Exotic Rugs, New York Graphic Society, Boston, 1979

Eskenazi, J., L'Arte del Tappeto Orientale, G. Mondadori, Milan, 1983

Herrmann, E., Seltene Orientteppiche and Asiatische Teppich und Textilkunst, annual catalogs, Munich

Larsson, L., Carpets from China, Xinjiang and Tibet, Bamboo Publ., London, 1988

Lorentz, H. A., A View of Chinese Rugs, Routledge and Kegan Paul, London, 1972

Natschlager, H., and Volker, A., Knupfteppiche aus China und Ost-Turkestan, Residenz Verlag, Vienna, 1986

Rippon Boswell, Antique Chinese Carpets, Masterpieces from the Te-Chun Wang Collection, London, 1978

Schürmann, U., Central Asian Rugs, Allen and Unwin, London, 1969
Spuhler, F., Konig, H., and Volkmann, M., Old Eastern Carpets, Callwey, Munich, 1978

Straka, J., and Mackie, L., The Oriental Rug Collection of Jerome and Mary Jane Straka, The Textile Museum, Washington, D.C., 1978

Volkmann, M., Old Eastern Carpets, Callwey, Munich, 1985
Wilber, D. N., "East Turkestan Rugs: A Puzzlement," Oriental Rug Review, Vol. V, Nos. 8, 9, and 10.

Bidder's Carpets from Eastern Turkestan is the only book solely concerned with the rugs from this region of the world. He provides the most thorough analysis of the history and development of rug weaving in this region, the various cultural influences upon it, and the ethnic make-up of the region's population. All subsequent writers give due reference to his pioneering effort to identify the unique characteristics of these rugs.



Illustration 9. Khotan Carpet, Christie's Lot 114, 9/11/90


East Turkestan is the Chinese province of Sinkiang. The population is predominantly ethnic Turk; they speak a Turkic language and are Muslims. Like Tibet, it is remote, surrounded by mountain chains, and at its center is the Tarim Basin, a vast desert area. On leaving China proper, the famed Silk Road divided and passed on the southern and northern edges of the Tarim Basin. Along these routes oasis towns developed. On the southern route were Khotan and Yarkand. Kashgar was at the western end where the two routes rejoined and then proceeded to Tashkent and Samarkand in West Turkestan and India to the south.
Bidder illustrated almost the entire range of rug designs produced in this region. He also provided a description of their technical features:

Warp: Always cotton
Weft: Mostly wool, brown shades, three shoots; less frequently dyed red and blue; and most rarely cotton wefts dyed red and yellow
Knot: Persian, left opening
Edge: Overcast, wool, dyed one color
Ends: Kilim of 9-18 weft shoots, cotton
Fringes: One end uncut loops, one cut loops

He describes the five major design groups to be found in these rugs as:
1. Vase-Pomegranate Pattern
2. Medallion Design
3. Coffered Gul Pattern
4. Old Turkish Motifs and Borders, Hellenistic, Buddhist and Chinese Influences
5. Muslim All-over Patterns and Ornamentation



.
1. Prayer Rug
Prayer rugs of any vintage are rare from East Turkestan and dated or inscribed pieces even rarer. This rug is interesting for several reasons: the unpatterned mihrab; the outline of the mihrab, which has a strong Indian feeling; the patterns in the spandrels, which are typical East Turkestan motifs; and the borders, the outer two being East Turkestan and the inner one a Turkoman pattern.
Size:2'10"x4'7"
Beg. End: Top, l" kilim, 1" cut fringe
Fin. End: Fringed out 1/4" into pile.
Colors: CAMEL, orange-red, dark blue, maroon, tan, light blue, yellow, green
Knot: Persian left; 9h. x 7v. =63 p.s.i.
Warp: cotton, white, Z6 and 8, S, severely depressed
Weft: Cotton, faded purple Z8, S, two shoots.
Edge: Overcast blue wool
Comment: Dated 1331 (1912)


He also discusses the development of the prayer carpet and concludes with a chapter on "the so-called Chinese carpets" from Kansu, Suiyan, Ninghsia and Paotou. These latter are areas or towns within China proper. I should add that the subtitle of Bidder's book is "Known as Khotan, Samarkand and Kansu Carpets."

The subtitle is important because Bidder points out that Samarkand was the term used most frequently in the West to describe these rugs. One purpose of the book was to correct this error and substitute East Turkestan or Khotan for Samarkand. Throughout the text he refers to Khotan rugs. He mentions other areas where rugs were woven such as Yarkand, Kashgar, and other lesser towns, but he does not describe any distinct differences between the rugs of these areas.

Regarding the vase-pomegranate pattern he says it "occurs frequently as a pattern on Khotan carpets." Of the five carpets illustrating this pattern, two are attributed to Kashgar, two to Yarkand, and one to Khotan. The Khotan rug and one of the Kashgar rugs have identical one-directional designs and border patterns! He does not explain why one is Kashgar and the other Khotan. Except for condition and the absence of one dark blue pomegranate on the Khotan rug, they are a pair. (This rug is published as Plate 104 in Volkmann.) Did Bidder notice this? What was his reason for making these distinctions? He doesn't say.

My interpretation of this is that in describing East Turkestan rugs he labels all of them Khotan. The main structural features were the most common throughout the region, but exceptions were noted. These exceptions are never assigned to specific towns or areas. By calling some rugs Khotan, some Yarkand, and some Kashgar, he is either reporting his conclusions or dealer attributions given to him.

About the medallion design he says it is "the most common in Khotan." Of the 12 illustrated, only two are attributed to Yarkand and the rest to Khotan. Of the coffered gul rugs, which Bidder considered to be a Turkoman design, two are attributed to Kashgar, the rest to Khotan.

Of the other designs illustrated, only one is attributed to Kashgar, the rest to Khotan. All silk rugs illustrated are attributed to Khotan. No technical analyses of any specific rugs are provided. Dating of these carpets ranges from 16th to mid-19th century.

In reading Bidder today it is clear to me that he used Khotan as the generic name for these rugs. He provides no basis for a division into more discreet groups. If one compares border patterns of rugs attributed to Kashgar, Yarkand, and Khotan, no one pattern consistently applies to a specific attribution. Although Bidder lived in China, he does not say whether he traveled to East Turkestan and/or bought rugs of the type at source. In his preface, he thanks two dealers from whom he bought and learned about these rugs. (The collection of one of these Te Chung Wang is listed above.) Does one assume that his specific attributions are based on their information? In looking at his illustrations today, all are pieces one would attribute to the pre-synthetic era or not later than mid 19th century. Undoubtedly he was seeing, collecting, and writing about antique pieces. In view of his definition of these rugs as primarily cotton warped and wool wefted with three shoots, was that the type that his dealers primarily showed and sold to him? We will probably never know the answers to these questions but they should be borne in mind.

Schürmann's book was the next major influence in identifying these rugs. Thirty-three pieces are assigned to East Turkestan and, except for one piece, are attributed to Kashgar, Yarkand, and Khotan. No explanation is provided for these attributions, but several of the pieces are accompanied by structural analyses. The Khotan pieces uniformly have wool wefts with three shoots between each row of knots. The Yarkand pieces have cotton wefts, usually two shoots, sometimes dyed blue, and with warp depression. One piece has a third wool weft. Only two Kashgar pieces have structural data, but one has cotton wefts with no color specification or number of shoots and the other two shoots of undyed cotton. Excluding one fragment, they are dated 17th to 19th centuries.




5. Lattice Design Rug This lattice design has apparently been used for centuries in East Turkestan and is one of the most frequently seen types. If one did not recognize the synthetic colors of this rug, the inner Turkoman border is another way of placing it with the end of the 19th/early 20th century weavings
Size: 4'1" x 6'3"
Beg. End: 1/2" kilim, 2" uncut fringe
Fin. End: Eight shoots weft, 2" cut fringe
Colors: RED, blue, light blue, yellow, light green, gold, rust, brown, ivory
Knot: Persian, left; 9 h. x 9 v. = 81p.s.i.
Warp: Z4, S, cotton, white, severely depressed
Weft: Z2, S, cotton, gray, three shoots
Edge: Overcast red wool


Schürmann's book was much more influential on those interested in Oriental rugs in the 1970s than Bidder's. He was an important and well known dealer. His book on Caucasian rugs was enormously successful in making that group of rugs understandable. Thus his presentation of Central Asian rugs -- both West and East Turkestan -- was well received and avidly read. As the first writer to dwell on structural analysis and introduce it as a basis for rug study, Schürmann's descriptions of East Turkestan rugs were pored over by those interested in organizing rugs into rational groupings.

It appears that Eiland was strongly influenced by Schürmann's structural information. Characteristically, his book is written in a clear, well-organized, rational style. He covers much of the information provided in Bidder and provides an update on weaving from this area into the 20th century. He provides information on dating these rugs based on the presence of hand- or machinespun cotton warps and wefts. The use and colors of synthetic dyes are presented. Eiland describes the structure of the three rugs types as follows:


"Most probably from Khotan"
Warp: Cotton
Weft: Wool, usually three shoots; later rugs, e.g. 2nd half l9th century, may be cotton, often machine-spun; many show mix of cotton and wool

"Oldest Yarkand is more likely to be"
Warp: Cotton
Weft: Cotton, two shoots rather than three, at times dyed blue and yellow

"Kashgar rug has a much more tenuous identity"
Warp: Cotton
Weft: Cotton, undyed, two shoots

"Present day East Turkestan rug...standardized"
Warp: Cotton
Weft: Cotton, two shoots, dyed (lavender, violet-purple) or undyed
He provides a variety of illustrations of these rugs which include pre- and post-synthetic examples with technical analyses of each. The attributions given conform to the above characteristics.

In discussing designs Eiland makes some adjustments to Bidder and some attributions which identified certain rugs with specific areas.

1 . Medallions
Most common design and "most of the medallion rugs are of a weave that I identify with Khotan."


2. Coffered Gul
Leans to the Turkoman origin and most "appear to be from Khotan."

3. Vase-Pomegranate
"More frequently associated with Yarkand, although some pieces show the three wool wefts characteristic of Khotan"

4. Designs from other areas
Discusses Iran, China, and India as design sources. Identifies some designs as Moghul India inspired. "Turkoman derived rugs seem to be from Yarkand."

5. Pictorial rug
Are a late development and characteristics conform to the standard weave of late l9th and 20th century. No specific town attribution.
Eiland knows when he is on unsure ground and each of the above attributions is carefully hedged. Regrettably, these attributions have been hardened into fact by collectors and dealers. This is fostered by a desire to categorize rugs and to do so on a rational basis. The basis is structure.

Since Bidder was closest to the source, let's return to him and consider some additional information in his book. About raw materials for carpet weaving he says that the wool came primarily from the Kirghiz and that the autumn shearing was priced 25% higher than that of the spring. He also asserts that the Kirghiz sold the wool already spun as yarn.

As for cotton it "is grown extensively within the Khotan oasis, and has a high reputation for its excellent quality ... the larger part of the annual harvest is woven in Khotan itself, however, and is exported widely as finished products ... especially as 'kham' or 'matta', a sturdy hand woven cotton fabric." If cotton were so plentiful and wool could vary so much in price, would a Khotan weaver use expensive wool for wefts, when cotton was so readily available, probably cheaper, and locally spun into a thread suitable for wefts? Is it perhaps the reverse that cotton warps and wefts were use in Khotan? Is it possible that the wool wefted pieces more appropriately belong to Yarkand or Kashgar because of their proximity to Kirghiz, Kazakh, and Turkoman sheep raisers of Kashmir and West Turkestan where wool may have been cheaper?

If the coffered gul is related to Turkoman ornamentation, why did it survive at the oasis city, Khotan, furthest removed from the Turkomans? If it is a Turkoman design, why in the late l9th century did the rugs with decidedly Turkoman guls evolve in Yarkand instead of Khotan? Should we be concerned about such inconsistencies?





3. Kirghiz Rosette Rug Few of these rugs have ever been published for they are considered by dealers in both Central Asia and the West to be among the least desirable. Several of these rugs are illustrated in Tsareva, Rugs and Carpets from Central Asia. No.137 has the same field design. Note the weaver's adjustment of the field pattern after finishing the first row of rosettes. Could this be the Kirghiz version of the Mina Khani pattern?
Size: 4'4" x 9'8"
Beg End: 1" kilim, 10" braided fringe
Fin. End: 1/2" kilim, l0" braided, fringe
Colors: RED (abrashed), blue, dark blue, light red, brown, yellow
Knot: Persian, left; 6h. x 7v. = 42 p.s.i.
Warp: Z2,S, wool, tan, some white
Weft: Z2,S, wool, brown, one shoot
Edge: Selvedge, three cord brown wool


Illustration 8, the most common three medallion design, was published by Schürmann, plate 79, as Yarkand. Bidder and Ei1and assert that the medallion design is primarily Khotan. Why then Yarkand? The answer seems to be the use of three blue cotton wefts instead of wool. Schürmann says that Yarkand designs are stronger and more striking, and that is certainly true of this rug. In Spuhler/ Konig/Volkmann Plate 103 is virtually identical to this piece in color, design, and structure. Eskenazi's Plate 293 is identical except yellow is substituted for green in the main border. Specific structure is not given but the introduction to his Yarkand section specifies cotton wefts. But Volkmann, Plate 107, shows a similar piece with three wefts consisting of two wool and one cotton. Eskenazi's Plate 298, called Khotan, is a vase-pomegranate rug with the same distinctive trefoil border and ostensibly with wool wefts. What is the controlling factor, design or structure?

Bidder reports that carpet weaving was done by men. He provides details on the looms and costs of production. For example, "In the 60s and 70s of the l9th century there were in the Khotan oasis alone 200 recognized masters of carpet weaving (aksakal) and they were in charge of the work. They were responsible for the purchase and dyeing of the wool, and for engaging the weavers according to the work on hand. "What is described here is factory production. Scales of payment are given, e.g. "each worker was paid 7-8 francs per square meter." Although drawn patterns were not used, the designs were passed on in an oral tradition. With this type of organization there is no reason to suppose that designs would remain restricted to one town. These professionals probably moved back and forth depending upon employment and took patterns with them. They undoubtedly shared and passed on designs that were popular or sold well. They could teach a whole group of weavers a design they had not known before. Nothing in Bidder's description should lead one to think that patterns could, would or should remain localized.

As Mushak's dye analysis shows, designs and weave characteristics may be the wrong approach to decipher the areas of production for East Turkestan rugs. The red dye in the antique rug in this sample is madder from India. Bidder has good information on the dyes used in East Turkestan rugs. Using accounts of British agents who opened East Turkestan up to trade and political influence as early as 1867, he tells how vegetal dyes moved in from West Turkestan and India, when synthetics began to appear and that by 1906 the art of dyeing with vegetal dyes was virtually forgotten. In discussing a particular red shade found in Khotan rugs, he presciently reports that "the red dyeing process...was not carried out there by Tibetans, but by the members of an Indian, probably even Kashmiri, guild who held a monopoly on it. Whether the strikingly warm, deep glowing red of older Khotan carpets comes from this dye can now be ascertained only by chemical analysis, not from the traditions of the Khotan dyers." Also about this red , he says,"the indigo had come from India via Bokhara, and the madder via Badakshan to Kokand." Cochineal is also mentioned as a widely used red dye. Mushak's analysis confirms the Indian origin of madder in at least one rug. If more were sampled, what percentage of East Turkestan rugs have this dye? More needs to be done along these lines.

In most rug weaving areas, yellows usually came from local sources. Bidder mentions several in East Turkestan. Since yellow is one of the major colors in East Turkestan rugs, analysis of the specific types of yellow dyes coupled with information on the botanic sources available around Kashgar, Yarkand, and Khotan could help in identification. Yellow dyes from India might indirectly tell something about the movement of dyes through trade routes. More attention to these kinds of facts are likely to provide more precise answers than designs.




7. Antique Lattice Rug Rugs of this type are usually attributed to circa 1800. Schürmann's Plate 89 has the same palette but different design and is dated 18th century. Certainly silk pieces with similar dyes are so dated. The softness of the reds and yellows are undoubtedly due to its great age.

Size: 3'8"x6'4"
Beg. End: None
Fin. End: None
Colors: LIGHT BLUE, red, brown, blue, pale yellow, white
Knot: Persian, left; 8h. x 7v. =56 p.s.i.
Warp: Z4,S, cotton, white
Weft: Z1, wool, white and brown, three shoots
Edge: Overcast red wool



Mention should be made of silk and metallic rugs, as these have attracted much attention recently. Bidder is not clear on this type of rug although a few are illustrated. Schürmann includes six attributed to Kashgar and two to Yarkand. In discussing these rugs, Eiland quotes a Chinese source who attributes silk rugs to Kashgar and a British source who mentions Yarkand and Kashgar. He says, "there are no silk and metal carpets in designs that can be specifically identified with Khotan.'' Today in the auction and trade catalogs, most silk East Turkestan rugs are attributed to Kashgar.

The alacrity with which the rug world has seized this scant information to create specific rug categories is frightening. I have only touched on a few points which I feel are relevant here. Those interested in other aspects in dispute should read the chapter on these rugs in Bennett's book as well as the Wilber material.

The group of rugs selected for this survey included one rug believed to be antique, two were Kirghiz rugs -- probably from West Turkestan -- and four were thought to be late l9th and early 20th century. All of the latter were believed to have synthetic dyes solely or combined with some vegetal dyes. The Kirghiz rugs are distinctly different from the East Turkestan pieces but were included as a contrast and to publish examples of these little known rugs. The results of the dye analysis is are in Mushak's accompanying article.

Structural analysis data on an additional eight, turn-of-the-century East Turkestan rugs were added to this group. The combined data was fairly consistent with some deviations.

1. The design types included were:

Single Medallion, 1 (Illustration 4); Double Medallion, 2; Vase-Pomegranate, 2; Lattice, 2 (Illustrations 5 and 7); Persian All-over, 2; Gerus, 1; Persian Medallion, 1; Prayer Rug, 1 (Illustration 1); Donkey Bag, 1 (Illustration 6); Kirghiz Lattice, 1 (Illustration 2); Kirghiz Rosette, 1 (Illustration 3).

2. All were woven with a left opening Persian knot except one which opened right.

3 . Knot density ranged from 36 to 81 p.s.i. The most typical count was 8 horizontal by 7 vertical with 56 p.s.i.

4. All had cotton warps except for the Kirghiz rugs, which were wool.

5. Warps were about evenly divided between Z2, Z3, Z4 ply (8) and Z6, Z7, Z8 ply (7). Eiland believes cotton plies of five and under are handspun and over five are machine-spun. Machine-spun thread would certainly have entered the area by 1870 but he believes probably earlier. This provides a clue to dating.

6. Wefts in all of the newer rugs were cotton, except one which has one wool weft and two cotton. All but two rugs had three shoots of weft between each row of knots. One rug with two shoots was the prayer rug which is one of the youngest rugs in the group (dated 1912).

7. Weft ply in the rugs ranged from Z2 to Z8 about equally divided between the low and high plies. With a few exceptions (Illustration 4) , low warp ply was matched by low weft ply and vice versa.

8. Several rugs -- Illustrations 1 and 5 -- had synthetic dyed wefts typical of late l9th and 20th century products, most commonly a faded purple to blue-gray.

9. All the East Turkestan rugs had overcast, one-color, wool edges except one which has a selvedge (Illustration 4). The two Kirghiz rugs have selvedges.

10. Where present, kilims on each end were about one inch wide and fringes were looped on one end and cut on the other.

The structural differences between the antique rug and the turn-of-the-century pieces are immediately obvious. Distinguishing the subtle changes in materials and dyes in the course of the second half century require much closer analysis. But none of this information provides a clue as to the oasis town of production.

Noting the inconsistencies in this sample, I undertook a survey of the data on 70 rugs published in recent years to see how consistent they were. The Hermann publications and the collection catalogs of Volkmann, Straka, To-Chun Wang, and Spuhler/Konig/Volkmann provided the data. All of the rugs were attributed to the l9th century or earlier.

This is not a scientific sample because the rugs represent what is desired by dealers and collectors and not a random sample of all East Turkestan weaving. For example, the percentage of silk rugs is very high in comparison to the total production of the region. Of the 70 pieces, 14 have silk pile, or 20% of the sample. Of the Khotan rugs, three are the pattern Bidder called Pu-10, a tie-dye pattern from Tibet. It is one of the rarer patterns and, by comparison, there were only four coffered gul rugs, a common pattern, in the Khotan group. However, since collectors and dealers tend to buy rugs on the basis of pattern and color, the data on structure may be considered as representative of what was being produced. Of the 70 rugs four were labeled East Turkestan; 15 were Kashgar; 10 were Yarkand; and 41 were Khotan. To focus on the normal and unusual features, I will discuss them by these four groupings.




6. Donkey Bag The two panels are from a double donkey bag. The pattern seems to be a pictorial of a vase with two flowers and leaves.
Size: 1'8"x3'2"
Beg. End: 2" kilim with 3/4" slits
Fin. End: Same
Colors: ORANGE, ivory, yellow, brown, pink (faded), green (faded)
Knot: Persian, left; 7 h. x 7 v. = 49 p.s.i.
Warp: Z3, S, cotton, white
Weft: Z4,S, cotton, white, three shoots
Edge: Woven strips, sewn on


The Kashgar Group
These showed the broadest diversity in every aspect. Of the 15 pieces, 10 were silk. There were 12 different design types. They ranged from traditional East Turkestan patterns to Chinese, Persian, and Mughal.

The wool pile rugs averaged 733 knots psdm (range 644-8l2), the silk rugs 257l (range 1596-5400). The highest knot counts were those woven on silk warps and wefts (6). Three had cotton warps and wefts, and one had cotton warp and wool wefts, three shoots, a Khotan structure! All had three weft shoots but one, which had two.

Warp and weft colors covered the entire spectrum: red, yellow, brown, white, blue, orange, and pink. The ply of warps and wefts were all under Z5 except one which was recorded as Z3-6, a silk weft.

The field colors were red (6), yellow (4), blue (4) and pink (1).

The Yarkand Group
This was also a diverse group. Of the 10 pieces, six were silk. There were six different design types. They were limited to East Turkestan designs (Vase-Pomegranate, three Medallion, Saphs) and Chinese (Pictorial, Lotus, Rice). There were two saphs and one prayer rug, the only prayer rug in the entire survey.

The wool pile rugs averaged 1183 knots psdm (range 792-1764); the silk rugs 1288 (range 980-1764). The silk rugs were woven on cotton warps and wefts with one exception where silk was used with cotton in the wefts. One wool pile rug had two weft shoots, all the others had three. Another wool pile rug had Zl brown wool wefts, the supposed Khotan identifier.

Warps were all undyed. Weft colors include blue, brown, yellow, red, and violet. The ply of wefts were mostly Z with none over Z5. The warps were evenly split between those above and below Z5. The field colors were red (4), yellow (3), blue (1), varied (2. Saphs).

The Khotan Group

These rugs represented 60% of the sample and contained the least variety. None of the pieces had any silk whatsoever. There were only nine design types. With three exceptions, they were typical East Turkestan designs: Three Medallion (12); Vase-Pomegranate (7); One, Two or Three-Medallion variant (6); Coffered Gul (5 ); Kejebe Medallion (2); Five Flower (1); and a saph. The three exceptions were Tiger stripes (2), Pu-lo (4) and a Zili Sultan (1). The last rug was a most unusual one and was labeled Khotan or Kashgar.

These rugs averaged 672 knots psdm (range 380-1020) and are obviously the most coarsely knotted of the groups. One rug had an asymmetric knot open to the right and one a symmetric knot (Straka, pl. 57). Five rugs had only two weft shoots between each row of knots. Two had four weft shoots. Six had some cotton wefts in addition to the typical natural white and brown wool wefts. Twenty rugs had Zl1wefts. The rest had Z2, Z3 or Z4 wefts.

Warps were all undyed. In six the warps were Z6, Z7 or Z8. The field colors were red (20), yellow (5), blue (5), salmon/pink (5), white (4), and varied (2).

The East Turkestan Group
Obviously a problematic group for identification, these four pieces exhibit many odd features. Three are in the Straka Collection and, were it not a Textile Museum publication, one might be tempted to question the data.

One is a silk rug, Straka pl. 58, on white cotton warps and two shoots of wefts. It has a Persian knot open right. The pattern is unique. A lattice of flowers and stars encloses alternately Tibetan dorges and an endless knot. At 42 knots p.s.i., it is quite coarse for a silk rug.

Another, Straka pl. 56, is a typical one-medallion design woven with a symmetric knot and a coarse 30 knots p.s.i. Cotton warps and brown wool wefts are what one would expect. The selvedge is unusual. The color palette is limited and the early 19th century date seems early by current dating standards.

Straka pl. 55, a lattice design, and the fourth rug are straight Khotans by the current definitions.

What this larger sampling shows is a fairly consistent usage of certain structural features for these groups. The majority of silk rugs are called Kashgar or secondarily Yarkand. The most coarsely woven rugs are Khotans; Yarkands are on average more finely woven (because they have two thinner wefts and not three which require less space); and the finest are the Kashgars, even in the non-silk rugs. Kashgar is a catchall category for rugs whose features do not fit the specified ones for Yarkand and Khotan, as witnessed by the number of structural exceptions and variety of designs.

But the main point is that, in this much larger sample of older rugs, exceptions to internal consistency of these current definitions of Kashgar, Yarkand, and Khotan are obvious. One could counter this by saying that if the same person categorized all of the pieces, that these inconsistencies would not exist. But how would one handle the few pieces with symmetric knots, asymmetric knots open right, or wool wefts with only two shoots between each row of knots? Is that what the East Turkestan category is for?

The current terminology has been built on Bidder's information which is sound as far as he took it. The accretions of specificity which have been added since have not been based on additional hard information but by wishful thinking to organize marketing and collectors' needs for categories.

This area of rug weaving needs some scholarly attention. As the Chinese excavate in Sinkiang and Mongolia, they are unearthing more and more fragments of early rugs. Rug scholars have paid scant attention to these materials or those of Stein, LeCoq, Pelliot, and Otani who excavated there at the turn of the century. A great Buddhist civilization established itself there from the third century B. C. onwards. Its records and arts have scarcely been examined for references to weaving or influence upon the development of rug styles. Given the amount of material available, certainly a chronology of the development not only of rugs but other textiles should be possible. Even for recent centuries there should be materials in Chinese and other Asian languages which could provide answers to some of these questions. With that kind of data, more specific concepts may emerge of what Khotan, Yarkand, Kashgar, and other types of rugs are. Until then such terms will remain nothing more than "brand names."