Posted by R. John Howe on 05-09-2005 11:13 AM:

Harold Keshishian at TM, Saf 4

(Note: Even if you've looked at this post before it may be worth your while to do so again. I was disappointed in the photos I was able to take at the TM of this piece and knew I had some betters ones somewhere. This morning I found them and Steve has helped me replace and add to the originals.)

Dear folks –

The fourth piece that Harold Keshishian presented in his TM rug morning on saffs offered a number of interesting features.

First, it was part of a larger fragment that Harold owns.

If you look at the photo above, you will see that the piece on the board is the section that is in the lower right hand corner of the photo Harold is holding. Harold owns this larger piece as well. The larger piece measures 15 feet by 25 feet. It in turn came from an even larger complete piece with several rows of multiple niche designs. The original piece was 16 feet by 35 feet.

The colors, on the strip that Harold brought, are remarkable.

The drawing is superb.

Here are some additional shots.

Here is one more look at one corner.

Harold said this piece was likely made in northeast Persia, north of Lahore about 1600. (He said that one well-known rug expert once told him that it was made 10 miles north of Lahore. Harold smiled and said we no longer attempt this kind of precision.)

Harold indicated that another noteworthy feature of this rug is that it is tied with a “jufti” knot. He demonstrated what he meant with his fingers and a handkerchief.

He first demonstrated how an ordinary symmetrical knot is tied. It goes around two warps from the front and comes up between them.

Harold said that the jufti knot is different from an ordinary symmetrical knot in two ways.

First, he said, it is tied around four rather than two warps.

Second, it does not come up between any of the four warps it encircles but rather comes up (after circling) on the sides. He made a particular point of indicating that such a jufti pile knot does not come up between any of the encircle warps.

I had heard previously of the “jufti” knot, of course, but had not heard this description of how it might be tied. In fact, I was sure I had seen drawings of juft knots in which the pile threads DID come up between the encircled warps. So Harold’s indication sent me to my books.

I consulted both Cecil Edwards, whom I remember complained loudly about the jufti knot in his book “The Persian Carpet.” And I looked to see what Marla Mallett has to say in her book “Woven Structures.

Both Marla and Edwards show versions of the jufti knot in which the pile cords do come up between the warps encircled. Marla indicates that the defining characteristic of the “jufti” knot is that it is one tied over more than two warps. She indicates (and this makes sense once one thinks about it) that there are a number of ways in which “jufti” knots can be and are tied. She offers an asymmetric version that is tied over six warps and around another ordinary asymmetric knot.

I think Harold’s point may be that some versions of the “jufti” knot (perhaps that in his fragment ) are tied with a jufti knot that is constructed so that the pile thread comes up on the outside (both sides) of the encircled warps, rather than coming up between any of them. I was glad to add this item of information to my modest bank of structural knowledge.

There is one last aspect of this nice piece that deserves mention. This larger fragment that Harold owns was purchased in London and came from the lavish home in Switzerland of Paulette Goddard, the movie star. Ms. Goddard, some of us will remember, was once married to Charles Chaplin (1936-1942). So this is a rug with an interesting provenance.


R. John Howe

Posted by R. John Howe on 05-10-2005 10:07 AM:

Dear folks -

The use of the "jufti" knot in what is clearly a piece done with considerable care and artistry attracts my attention.

As many of you will know the "jufti" knot is widely denigrated in the literature as a way of weaving a rug more quickly, but at a reduction in quality. It is seen as a sign that time and cost have entered the equation to the detriment of the art and the product.

Listen to Cecil Edwards hold forth on the jufti knot in his book "The Persian Carpet." He is writing in the earlier 1950s at the end of a long career as the representative. in Iran, of an English rug making firm.

"...I shall have a good deal to say in the course of this survey about this method of weaving. For it is a fraudulent device: a device whereby the weaver ties one knot in place of two, and so doubles the output to the detriment of quality. In a closely woven fabric the fraud is difficult to detect before the weft is passed; and almost impossible to detect afterwards. Only the feel of the carpet betrays it. Because, of course, a fabric in which a large proportion of the knots are tied on four strings of warp instead of two must lack density, and withouth density a carpet will not give good service.

"The type of knot is known in Persia as the "jufti" or "juft ilmeh," i.e. the double knot. Twenty-five years ago it was little used except in northern Khurasan. But is has spread with alarming rapidity --- particularly in the last decade..."

Edwards goes on angrily about the use and seeming spread of the jufti knot in several places in his respected book, but in the section on Khurasan, he says briefly, without elaborating, that the use of the jufti knot in that location was not as the result of the intent to reduce quality, but was rather the traditional way in which asymmetric knots (only, he says, those using symmetric knots in Khurasan did not use the jufti version) were tied there.

But Edwards is writing about practices in the early 20th century and while he saw and comments on some antique Persian rugs he encountered in northern Khurasan, he does not mention the type(s) of knots used in them.

The fact that Edwards seems to acknowledge that the use of a jufti version of the asymmetric knot was part of the tradition of northern Khurasan, made me wonder whether the locals had noted (in say about 1600 or so) that their jufti knot might have some advantages in weaving very large carpets (as some of these safs seem to have been). Might someone not have noticed that use of the jufti knot would reduce the weight of very large carpets sharply and make them easier to move and to adjust on occasion?

One other point here. Notice that Edwards claims that the jufti knots in northern Khurasan were always versions of the asymmetric knot. But Harold has reported that jufti knots are (at least sometimes, perhaps in his fragment) tied in way that would seem closer to, but not identical with a symmetric usage.

Interesting things.


R. John Howe

Posted by R. John Howe on 05-11-2005 09:26 AM:

Dear folks -

I am posting in this thread again only to pull it up to the top so that you will be tempted to look at the images in it again.

I was disappointed in my TM photos of this piece and looked around and found better ones which I have now inserted. Please take a look. I think this is a remarkable fragment.


R. John Howe

Posted by Filiberto Boncompagni on 05-11-2005 10:08 AM:

Yes, John, you are right, the fragment is beautiful. Wonderful colors too.
If I understood well, Mr. Keshishian owes also the rest of the carpet. Did he explain what happened to it? I mean, why it was cut?
Another matter… Isn’t Lahore in what nowadays is called Pakistan?


Posted by R. John Howe on 05-11-2005 10:51 AM:

Filiberto -

Yes, the strip that Harold brought to the TM is part of a larger fragment that he owns as well. These in turn come from a slightly larger complete rug.

I have not heard Harold say why this piece is fragmented, but as you know this often happens to old pieces. Sometimes the owners have had them cut to "fit" into some place in a given room. In other instances, dealers cut them up because they think they can get more from several smaller fragments than from the complete piece.

Lahore is in what is now Pakistan and in fact on its far east side.

Of course the political boundaries of this area of the world have changed dramatically from time to time. I have looked for a map of this area in about 1600 but have not found one.


R. John Howe

Posted by David R.E. Hunt on 05-12-2005 12:32 AM:

No Map But...

Hi John

I found this "Timeline Of Art History" from the Metropolitan Museum Of Art containing the followimg quote.

"Akbar was also the first great Mughal patron of the arts. Of his various building projects, the most ambitious was the new capital city of Fatehpur Sikri, near Agra. Built mostly between 1571 and 1585, when Akbar adopted Lahore as his principal residence, the palace buildings at Fatehpur Sikri reflect a synthesis of Timurid traditions of Iran and Central Asia with indigenous traditions of Hindu and Muslim India.

Although he is said to have been illiterate, Akbar assembled a royal atelier, first at Fatehpur Sikri, then at Lahore, from which he commissioned numerous illustrated manuscripts that incorporate Persian, Indian, and even European elements. In fact, the artists who worked for Akbar, the first great Mughal patron of the arts of the book, included Persians as well as Indian Muslims and Hindus. This collaborative process helped to foster the development of a specifically Mughal style, which was initiated under Akbar and is demonstrated by pages from diverse late-sixteenth-century manuscripts. This style of painting was further developed and refined during the reigns of Jahangir and Shah Jahan during the seventeenth century."

Be sure not to miss the links to the right of the "Timeline Of Art History" page.


Posted by Filiberto Boncompagni on 05-12-2005 04:57 AM:

Map of Mughal Empire in India, 17th century, showing also Lahore:



Posted by R. John Howe on 05-15-2005 09:40 AM:

Depictions of Knots

Dear folks -

As I've indicated above, Harold's description of how a jufti knot is tied reminded me of how various the depictions are of even the two basic knots in rug books.

So last night I pulled a few off the shelf and scanned some of the variations. I thought it might be useful to see that what we often now see as obvious and settled has not always been so.

First, a review of names. Most will know that the "symmetric" knot is also referred to in the literature sometimes as the "Turkish" knot or the "Ghiordes" knot. Similarly, the "asymmetric" knot is often referred to as the "Persian" knot or the "Sehna" knot (this latter despite the fact that the asymmetric knot is not the one used by Senneh weavers.

I am going to give you some depictions of these two knots arranged chronologically in terms of the books from which they were taken.

The first is from Mumford's old "chesnut" published in 1900.

Mumford's depiction of the asymmetric (Sehna) knot is quite similar to others we shall see but notice how odd his drawing of the symmetric (Ghiordes) seems at first glance. I do not think it is at bottom really different in structure from what we see more frequently, but for some reason he thought it necessary to show the collar sharply angled and to de-emphasize the similarity in the way the knot moves around the warps on both sides. Odd.

Mumford does though, right away, indicate that the asymmetric knot is tied either open to the right or open to the left.

Then I looked to see what Hawley had to say in 1937.

Hawley is important in this respect because he is one of the first of the traditional writers to pay particular attention to structure and has a table at the end of each section of his book in which he summarizes structural differences of that group in a conveniently comparative way.

Hawley is insisting on a degree of complexity in the making of the "symmetric" (Ghiordes) knot. His first drawing looks like those most usually seen nowadays but his other two accentuate that when the pile threads come up between the two warps around which they are tied they can do so either right over left or left over right. This distinction has not been treated seriously as we come forward in time and I am not sure that its omisson is not appropriate. It's not clear to me why Mumford felt it important to depict these three versions of the symmetric knot.

This same year a lady name Holt put out a book full of "fine writing" something to which some of the older rug writers tended but one that is sometimes treated with some respect since she seems to have see, for example, that there was Chinese influence in many middle eastern designs.

I find Holt's knot depictions confusing and I think her depiction of the asymmetric knot is incorrect.

The first source of potential confusion is that two of her three knot drawing show the knot tied continuously between warps without being cut. I don't think this is how knots are tied, but it also works to make the depiction of their construction puzzling. Holt does show a "cut" symmetric knot and it appears to have a basic structure that we would recognize but notice that Holt too, seems to think the crossing of the pile threads one over the other as it comes up between the warps in a symmetric knot is important to notice.

I think Holt's depiction of an asymmetric (Sinnah) knot is simply incorrect. First, only a continuous version is given but more importantly notice that the knot's drawing seems to suggest that while it goes around one warp it does not interlace a second. Holt's drawing seems to suggest that asymmetric knots are not tied around two warps each but pushes her depiction of them toward something closer to a "Spanish" knot tied on a single warp.

Next I looked at Edwards, who was writing in 1953.

Edwards' depiction of the symmetric knot is much like Mumford's first one and those that we see most frequently in the literature today.

Notice, though, that Edwards draws his knots with the pile ends pointing down. Different authors take different positions on this. I think Edwards, who was the agent of a rug producer was indicating that this is how the knot looks to a weaver as it is being tied. The pile cord is taken from the front over the two warps, passed around their outside to the back, come up between them and then are pulled DOWN. He is honoring the weaver's perspective in his drawings.

Edwards also give us cross-sectional drawings of the construction of the symmetric and asymmetric knots. That is how they look if view from the ends of the warps. Again the pile ends point down as they would as the weaver ties them.

Now we come to the German author Hubel, who wrote a nice survey in 1964. Here are his depictions of the asymmetric and symmetric knots.

Hubel, too, has pile pointing down. He also retains three varieties of symmetric knot. Two of them seem to suggest that the knot is pulled either to the right or to the left.
Jacobsen, the dealer in Syracuse, NY, who sold lots of oriental rugs by mail order, wrote a couple of books that sold well. Here are his 1962 knot drawings.

Jacobsen also has pile pointing down, includes cross-sectional views from warp ends and give us drawings of the jufti knot in which the pile cords do come up between the warps. I will show you the Edwards and Mallett versions of the jufti knot at the end of this sequence.

Now we come to the knot drawings that Eiland has been using in his comprehensive guides. These drawings are taken from the most recent one he did with his son in 1998 but he has used these drawings since 1973.

Notice that Eiland has dropped the perspective of the weaver in favor of that of the rug owner or customer. Rugs, as most of us experience them have the pile pointing up, and Eiland makes this adjustment of tradition.

He does not seem to think that the varieties of symmetric knot are important enough for separate depiction, but he retains separate drawings of the two asymmetric knots.

In 1997 in his Lexicon, Stone gave us these drawings:

Stone is retaining the perspective of the rug owner rather than the rug weaver and adds a touch of perspective, giving his drawing a sort of three-quarter view that addresses both the "top" and "end of warp" perspectives offered by some others nicely.

In 1998, Marla Mallet published the first edition of here "Woven Structures" book that contains her own drawings of the basic knots.

Since Marla has often argued in her writings for the importance of retaining the perspective of the weaver in our analysis of rug structures it should not be surprising to discover that the cut ends in her knot drawings point down again.

Notice also that her depictions are more "relaxed" in some sense than are many of the others. She also show some crossing of the knot cords in the symmetric knot as they come up between the warps but does not seem to think that the fact that this could be done left over right or right over left and the fact that the symmetric knot "collar" can be slanted to the right or to the left require separate treatment.

Well, that is my little tour of various depictions of the two basic pile knots in the literature. It is interesting to me to see how even basic things are often seen differently.

One last thing, the drawings of the "jufti" knot's construction that triggered this post.

You have already seen Jacobsen's above.

Here is that provided by Edwards:

Notice that Edwards indicates that the jufti knot is tied differently in Khurasan and Kerman.

Here is the jufti knot drawing provided by Stone:

Here is what Marla provides early:

And here is her drawing of a curious version of the jufti knot tied over six warps and around an ordinary asymmetric knot.

This (together with the two versions shown by Edwards) seems to me to demonstrate clearly that there are a variety of ways in which the jufti knot can be tied.

Comments and corrections are invited.


R. John Howe

Posted by Filiberto Boncompagni on 05-16-2005 08:09 AM:

Hi John,

Thank you for the scans.

I agree with your comments on Mrs. Holt’s depiction and I feel also that Hubel’s sketch of symmetric knots is somehow implausible.

I mean, when the weaver beats down the warps, “T. II” and “T.III”, should end like “T.I”, shouldn’t they?

Another observation: I find the renderings adopting the “weaver’s perspective” more logical.



Posted by R. John Howe on 05-16-2005 10:00 AM:

Hi Filiberto -

Thanks for your comments.

Yes, the cant of the symmetric knots shown in some of these drawings (you focus on Hubel's) would not be very obvious once the wefts were inserted and beaten down.

But remember what these drawing are usually trying to depict, not what a given knot looks light after it has been tied and beaten, but rather how the knot is constructed. Notice the very real looseness that Marla adopts in her drawings.

I can testify that you can detect that some symmetric knots have been pulled right or left when they were tied just by running your hand from a lower part of a rug toward the higher one at various angles searching for that which is roughest.

I have watched some DOBAG weavers using a symmetric knot and was impressed with finding that, although they seemed to pull the knots down straight as they tied them, they were in fact pulling them quite sharply to the right.

For me the main purpose of this little exercise was to examine what Harold's comment about the jufti knot seemed to suggest: that there are in fact a variety of ways in which many of the knots used to weave rugs can be tied. Some of the portrayals are in error, but some real variations are being noted sometimes.

At first, I was prepared to be skeptical of Harold's indication about the construction of the jufti, because I knew I had seen drawings of them in which the pile threads did come up between the warps.

But after examining the varieties of depictions of even the two most basic knots, as well as some of jufti knots (we have drawings of at least three distinctive versions of the latter), I think there is room for correctness in Harold's description.

He's been looking at rugs and knots for a very long time and is unlikely to have made this claim without some concrete experience.


R. John Howe

Posted by Tim Adam on 05-16-2005 11:33 AM:

Hi John and Filiberto,

I am wondering whether the images T.II and T.III don't just depict symmetric knots on depressed warps?


Posted by R. John Howe on 05-16-2005 11:51 AM:

Hi Tim -

Good observation. That may well be the intent since the warps shown are staggered in the latter two drawings but not in the first.


R. John Howe

Posted by Filiberto Boncompagni on 05-16-2005 12:16 PM:

Right, Tim.
And the warp depression could help to retain the slant of T.II and T.III.
My apologies to Mr. Hubel.