Posted by Steve Price on 01-02-2005 10:10 PM:

Databases would be useful if we could trust them

Hi People

Attributing dates is, in principle, based on more or less the same paradigm as attributing geographic or tribal origin. We examine a large number of rugs of different (but known) origins and assemble databases of properties associated with each time or place (or tribal group). This is usually not done in formal terms; the databases are in the mind.

This should be a highly accurate and reliable approach, but there is one serious catch to it: the origins of the rugs in the database are largely presumed, not actually known with real certainty.

One of the paradoxes in this is that the cells in the database with the largest numbers of truly known examples are the most recent ones, and these are the least interesting to collectors. The further back we try to go in time, the smaller the number of rugs with documentable ages becomes and, hence, the less certain the criteria for their identification become.

Let me illustrate with a simple example: the Pazyryk rug. It is the only known piece of it's age. If we came across another rug made by the same people at about the same time, would we be able to identify it as such? No, because the one example gives us no information about the range of variation those people produced at that time.

This kind of problem, it seems to me, is very widespread but not generally acknowledged. The number of extant rugs of documentable age that are more than 150 or 200 years old is very small, too small for us to actually know the extent of variability in the original population of, say, 16th century rugs from around Erivan (or almost any other venue). Yet, now and then I see collectors try to decide the quarter century in which a piece thought to be of considerable age was woven. I believe this to be completely futile. Does anyone see it differently? If so, why?


Steve Price

Posted by Richard Farber on 01-03-2005 02:07 AM:


i agree that the attempt to put dates within quarter centuries is most often based on insufficient information an often more a question of wish than fact.

i do think that within a body of objects to be dated a feel for earliest, earlier, mid, late very late can be sometimes be judged. and if there is some documentation as to the century the stuff was made in than some guesswork might well be attempted.

as to your Pazyryk rug statement . . . here i think that there might be a good chance that a second rug from the period would be quickly recognized irregardless of the possible variations that would accord within a population making such an object . . . same thing with an animal carpet from the 500 or 600 years ago . . . or an embroidery from spain from the 15th cent. i believe that there woud be enough clues. . . and if i didnt beleive this i would be questioning some of the basic ideas of art and music historians.


richard farber

Posted by Steve Price on 01-03-2005 08:14 AM:

Hi Richard

I agree that many of us feel pretty comfortable in judging the relative ages of a few pieces, but that isn't the same thing as actually knowing that we're right or placing even approximately correct dates on each.

Questioning some of the basic ideas of art and music historians isn't a bad thing to do. I'm no art historian, but I can tell you that the history of science is filled with incorrrect notions that remained in place until questioned and I have no doubt that contemporary science has its share as well.

Back to the Pazyryk. It's actual origin (who wove it or where it was woven) isn't known, and the only reason we are confident about its great age is that it was found in a long-frozen grave that could be dated with reasonable certainty. If it had been found by a picker and showed up in a dealer's inventory, it would probably be attributed to the early or mid-19th century and there would be general agreement that it was made in some particular village or by some particular tribal people (if it attracted widespread attention for one reason or another). That is, without knowing the provenance, most experts would err by a few millennia in dating it.

The basic clues in rugs are the size, shape, knot type and density, materials, dyes, motifs, layouts and designs. Combinations of these form the basis for attribution of time and place of a weaving. They often give ambiguous conclusions even when we have hundreds (or thousands) of pieces from the presumed origins with which to compare them. That is partly because weavers are innovative and sometimes make mistakes, all of which leads to variability in the fruits of their labors. Reliable attribution requires knowing the range of variability. This, in turn, requires having a significant number of samples and, most important, documentation of the actual time and place of weaving of those samples.


Steve Price

Posted by R. John Howe on 01-03-2005 09:17 PM:

Steve -

A couple of years ago, I got interested in exploring whether there might be some basis for suggesting that a given Turkmen engsi had actually been made for use as a cover for the door opening in a yurt.

I gradually collected data on over 630 Turkmen engsis.

Estimated age was, of course, one variable that seemed an important basis for sorting these pieces out (it seemed likely that the ones estimated to be older should be expected to exhibit other features that might make it more likely that they had been made to serve as door covers.

There were/are severe methodological flaws in the age estimates I had from these pieces. Many were from rug books but a large majority were from auction catalogs for the last 25-30 years. There are, of course, serious biases and great gaps of knowledge reflected in these estimates. I ultimately adopted three categories:

Before 1850
1850 to 1900
After 1900

Now these are larger periods but the foolishness of this exercise was very apparent to me as I worked.

I did not confirm (even slightly) the basic hypothesis I was testing (that is whether size, especially width of about 120 cm might be more frequent on pieces estimated to be older). The hypothesis may well have been simply and rightly disconfirmed, but I was/am uneasy about whether I was able to sort the rugs in these three groups, even approximately, by their true ages.

I notice some dealers now say only "19th century." They have moved to hundred year intervals. Even that seems daring to me, but is probably less so than I think.


R. John Howe

Posted by Patrick Weiler on 01-03-2005 11:04 PM:

The Pazyryk Score


If a composer were to write a score in the style of old classical music, would we be able to determine the age of the composition?
Science can inspect the paper and ink, but not the score.
We could compare the style with known composers, but if the ink was the same composition as old ink and if the paper were old, it would be nearly impossible to know the age of the composition. Even today compositions are being discovered and attributed to old composers. Even painting attributions confuse the experts. Fakes in museums are common.
Rug dyes and wool can be inspected. New copies of old weavings have been made with old wool from jajims. A case could be made for experts being quite good at determining the relative age of a weaving, but absolute conclusions within a range of a couple of decades is a fools errand. Unless there is an actual date woven in, even a half century is only plausible.
This is why many collectors, museums and auction houses rely on "experts" to vet their acquisitions. But a concerted fake can be nearly impossible to discern.
The answer? Only buy really cheap garbage that nobody else wants. At least you know it's real.

Patrick Weiler

Posted by Steve Price on 01-03-2005 11:51 PM:

Hi Patrick

One very important difference between music manuscripts and old rugs is that we usually have a significant number of manuscripts in which the time, place and composer are documented. Thus, there are concrete standards against which to compare a questionable one.

Not so for old rugs. Despite the apparent confidence with which rugs are attributed to this or that quarter century (or half century, or any other interval), there are rather few specimens from most weaving regions whose ages can be documented. This, in my opinion, is the crux of the problem. And this would be true even if there were no fakes with which to contend. In the absence of a decent sample of specimens of known (not guessed, known) ages from a particular weaving area, we cannot construct a set of criteria in which characteristics that allow conclusions about the age of any specimen to be specified, except in very broad terms.

The biggest exception to this dilemma, unfortunately, is modern production. Someone familiar with the subject can date modern rugs with pretty high accuracy. To collectors, these are the least interesting rugs.

Rug databases for attributional use are well illustrated by Peter Stone's excellent work, Rugs of the Caucasus. He tabulates a large number of characteristics of Caucasian rugs, divided by specific geographic origin. The problem: it's presumed, not demonstrated, that every rug in the database originated in the geographic region to which it is assigned (the rugs and their attributions come from published sources). I suspect that most actually are correctly assigned, and if this is the case, the errors introduced to the database simply become low level noise. I am much less confident that a database using published estimates of age would have as small a percentage of incorrect age attributions.


Steve Price