Of all the major Bronze Age civilizations, only the Inca of
South America appeared to lack a written language, an
exception embarrassing to anthropologists who habitually
include writing as a defining attribute of a vibrant,
complex culture deserving to be ranked a civilization.

The Inca left ample evidence of the other attributes:
monumental architecture, technology, urbanization and
political and social structures to mobilize people and
resources. Mesopotamia, Egypt, China and the Maya of Mexico
and Central America had all these and writing too.

The only possible Incan example of encoding and recording
information could have been cryptic knotted strings known
as khipu.

The knots are unlike anything sailors or Eagle Scouts tie.
In the conventional view of scholars, most khipu (or quipu,
in the Hispanic spelling) were arranged as knotted strings
hanging from horizontal cords in such a way as to represent
numbers for bookkeeping and census purposes. The khipu were
presumably textile abacuses, hardly written documents.

But a more searching analysis of some 450 of the 600
surviving khipu has called into question this
interpretation. Although they were probably mainly
accounting tools, a growing number of researchers now think
that some khipu were nonnumerical and may have been an
early form of writing.

A reading of the knotted string devices, if deciphered,
could perhaps reveal narratives of the Inca Empire, the
most extensive in America in its glory days before the
Spanish conquest in 1532.

If khipu is indeed the medium of a writing system, Dr. Gary
Urton of Harvard says, this is entirely different from any
of the known ancient scripts, beginning with the cuneiform
of Mesopotamia more than 5,000 years ago. The khipu did not
record information in graphic signs for words, but rather a
kind of three-dimensional binary code similar to the
language of today's computers.

Dr. Urton, an anthropologist and a MacArthur fellow,
suggests that the Inca manipulated strings and knots to
convey certain meanings. By an accumulation of binary
choices, khipu makers encoded and stored information in a
shared system of record keeping that could be read
throughout the Inca domain.

In his book "Signs of the Inka Khipu," being published next
month by the University of Texas Press, Dr. Urton said he
had for the first time identified the constituent khipu
elements. The knots appeared to be arranged in coded
sequences analogous, he said, to "the process of writing
binary number (1/0) coded programs for computers."

When someone types e-mail messages, they exist inside the
computer in the form of eight-digit sequences of 1's and
0's. The binary coded message is sent to another computer,
which translates it back into the more familiar script
typed by the sender. The Inca information, Dr. Urton said,
appeared to be coded in seven-bit sequences.

Each sequence could have been a name, an identity or an
activity. With the possible variations afforded by string
colors and weaves, Dr. Urton estimated, the khipu makers
could have had at their command more than 1,500 separate
units of information. By comparison, the Sumerians worked
with fewer than 1,500 cuneiform signs, and Egyptian
hieroglyphs numbered under 800.

Dr. Urton concedes that his interpretation of a khipu
writing system may be hard to prove. No narrative khipu has
been deciphered. Spanish conquerors, who suspected the
knotted strings might contain accounts of Inca history and
religion, destroyed those they came across as idolatrous
objects. The few existing descriptions of the khipu by
explorers and missionaries lack enough detail for an
understanding of the way the Inca made and "read" them.

Other Inca scholars generally agree that the khipu may have
served as more than accounting devices or memory aids, and
may have been a medium for recording historical
information. But they reserved judgment on Dr. Urton's
binary code hypothesis.

"Most serious scholars of khipu today believe that they
were more than mnemonic devices, and probably much more,"
said Dr. Galen Brokaw, a specialist in ancient Andean texts
at the State University of New York at Buffalo. He was
quoted in an article about the khipu in the June 13 issue
of the journal Science.

Dr. Patricia J. Lyon of the Institute of Andean Studies in
Berkeley, Calif., was unmoved from her position that the
khipu were mnemonic devices, personalized visual and
tactile cues for the recall of the information retained in
the memory of the maker. If that was the case, the khipu
would not be a form of writing because they would have been
understood only by their makers, or someone familiar with
the same memorized accounts or narrative.

"People feel this great need to pump up the Inca by
indicating that the khipu were writing," Dr. Lyon said.

Dr. Urton said in an interview that others would soon be
able to test his theory and possibly find other patterns
and clues in the khipu he studied. A detailed khipu
database, financed by the National Science Foundation and
prepared with the help of Dr. Carrie Brezine, a Harvard
mathematician and weaver, is expected to be ready this fall
and will eventually be available online.

Experts in the culture of early Peru think it
understandable that textiles would have been the chosen
medium for writing. The Sumerians and Babylonians wrote on
clay, the Egyptians on stone and papyrus. The Inca may have
used cloth, though, to store and communicate knowledge
because to them cloth was a widely used marker of status,
wealth and political authority.

Dr. Heather Lechtman, an archaeologist at the Massachusetts
Institute of Technology who specializes in early Andean
technology, said that "fibers were the heart of Andean
technologies of all kinds, even long before the Inca, and
so it doesn't surprise me that people would have thought of
using khipu perhaps for some sort of writing system."

Early Spanish colonists gave conflicting accounts of the
practice. A drawing of a khipu maker in an Inca storehouse
seemed to reflect the view that the knotted strings
involved record keeping. A Jesuit chronicler said the khipu
were like ledgers or notebooks that overseers and
accountants used "to remember what had been received and
consumed."

Another account tells of Spanish travelers who came upon an
old Indian man who tried to hide the khipu he was carrying.
Under questioning, the Indian claimed the khipu recorded
the activities of the conquerors, "both the good and the
evil." The Spanish burned the khipu and punished the
Indian.

Not until the 1920's did scholars seem to reach a consensus
on what the khipu were. From studies of a collection of
knotted strings at the American Museum of Natural History
in New York City, L. Leland Locke, a science historian,
concluded that they did not represent a conventional scheme
of writing but signs recording columns of numbers. Khipu
makers must have been bookkeeping bureaucrats.

This remained the prevailing opinion until the last two
decades. Husband and wife researchers, Dr. Robert Ascher, a
retired Cornell archaeologist, and Dr. Marcia Ascher, a
mathematician at Ithaca College, reopened debate by
pointing out that khipu seemed to use numbers as both
numbers and labels. They estimated that about 20 percent of
existing khipu were "clearly nonnumerical" and could have
been examples of an early form of writing.

Dr. Urton has carried the idea further. A creator of khipu,
he posits, made a series of choices involving the type and
color of string and each knot. Each choice contributed to
creating a binary signature. A certain string configuration
could represent signs that stood for a value, object or
event, much as graphic signs do in familiar forms of
writing.

Emboldened by this insight, Dr. Urton said in his book that
the Inca "may well have been recording full
subject-object-verb notations in the khipu."

Dr. Urton based his research primarily on khipu specimens
at museums in the United States, Germany and Peru. A
discovery in 1997 in northern Peru, at a burial site of the
Chachapoya culture, yielded 32 khipu with exceptionally
elaborate and varied types of string patterns. Strands
hanging from the horizontal cord had their own secondary
and tertiary pendants.

These complex pendant attachments, he wrote, "must have
been an important mode of binary coding in the khipu."

A close examination of Dr. Urton's new database of khipu
elements by other scholars, including linguists and
pattern-recognition experts, may win wider support for the
writing hypothesis.

"It's much too early to say anything about how this will
all come out," said Dr. Lechtman of M.I.T.

More definitive would be the discovery of an Inca "Rosetta
stone." It was such a trilingual inscription that finally
enabled scholars to decipher Egyptian hieroglyphics.

A colonial governor had khipu makers "read" some strings
and scribes record the accounts in Spanish. This could have
been a start toward decipherment, if only the khipu had
been preserved.

A prospective Rosetta stone was announced in 1996 by an
Italian amateur historian, who claimed to have found a
translation into Spanish of a song encoded in a khipu. But
other researchers have not been allowed to examine the
material, and Dr. Urton said that many questions had been
raised about its authenticity.

Dr. Urton holds out more hope of making a breakthrough
discovery in the Chachapoya material. Most of the khipu
there appear to be from the early colonial period. For that
reason their encoded messages are more likely to have been
transcribed in Spanish documents as the sought-after
Rosetta stone of Inca writing. If, that is, the Inca wrote
with strings and knots.