Unbroken seal secrets

The Indus script is the most significant undeciphered script in archaeology, since it was the writing system of one of the great civilisations of the ancient world, that of the Indus Valley, dating to the third millennium BCE.

Archaeologists have identified more than 1,000 Indus sites covering an area of some 800,000 square kilometres, encompassing all of what is now Pakistan and parts of western India and Afghanistan. While the extent of the Indus traditions is impressive, most archaeologists think that there was never a state-level government in the Indus Valley. The people did, however, develop an extensive system of trade, both between Indus Valley sites and with the adjacent cultures in the Persian Gulf, Mesopotamia and the Indian subcontinent. In order to facilitate trade they also developed a system of weights and measures and a unique system of writing that survives on nearly 4,000 artefacts - the majority sealstones - containing 7,199 signs.

The script first came to the public eye in 1924 with the publication of Sir John Marshall's article in The Illustrated London News announcing the new discoveries at Mohenjo-daro and Harappa, the two chief Indus cities.

Excavations continue to this day at Harappa by a joint interdisciplinary team from Harvard University and the University of Wisconsin led by Richard Meadow and Mark Kenoyer. Thanks to their efforts we are coming to a detailed understanding of the complexities of Indus culture. Since December 1994, when Robert Knox and John Chadwick reviewed Deciphering the Indus Script by Asko Parpola in The THES , the Kenoyer and Meadow team have published at least six major works (some more than 100 pages long) summarising new finds, including several hundred new inscriptions with examples of the earliest Indus writing apparently dating to 3300-3500 BCE. At the same time, the publication of the two volumes of the Corpus of Indus Seals and Inscriptions , edited by Parpola and collaborators, has revolutionised the study of the script by making available photographs of most of the Indus texts. Together with R. S. Bisht's discoveries at Dholavira (in Kutch) in the past decade, all this new information is greatly improving our chances of a systematic decipherment of the Indus script.

Gregory Possehl, the author of Indus Age: The Writing System , is a well-known Indus archaeologist at the University of Pennsylvania who has been studying the subject since the mid-1960s, initially under the direction of the late Walter Fairservis Jr, an archaeologist who wrote many books on the Indus civilisation and published his own controversial decipherment, first in Scientific American . Possehl is thus not a trained epigrapher, and this is perhaps a weakness of his book but also a strength, because it allows him to be more objective about the decipherments proposed by others.

Since 1924 there have been 100 or more published attempts to decipher the Indus script, none demonstrably successful. Most have claimed to decipher the majority or even all of the signs and texts. They have generated a great many books, articles and reviews, in which eminent academics have discussed the various hypotheses about the origin and nature of the Indus people, their language and their writing. Nevertheless, there is little the experts have agreed on. About the only characteristic of the script that enjoys nearly universal acceptance is that the texts are most often read from right to left. It is the lack of long texts, of a bilingual text (like the Rosetta Stone) and of a known root language that has made progress on the Indus script slow. Many of the early attempts at decipherment supposed that because the Indus script had no known proto-script, the shapes of its signs must have been borrowed - from Mesopotamia or Egypt or even (absurd as it now seems) Easter Island. Most decipherments continue to rely on circular arguments, in which in order to begin a decipherment the researcher assumes that some language or other is the language of the script; popular choices have been Sanskrit, Brahmi, Dravidian, Sumerian and Austro-Asiatic (not forgetting a putative lost language). While none of these can be demonstrated to be the language of the Indus script, Dravidian and Indo-European dominate the literature.

In an 85-page chapter, Possehl reviews some 35 of these decipherments from around the world in varying degrees of detail. He does not endorse any specific decipherment and disavows any personal interest in attempting a decipherment, but he does offer certain views on the script. For example, he states that "Since the inscriptions are short and are probably names and/or titles and epitaphs, they lack the structure of a text with words contributing to the cast of a sentence and sentences becoming paragraphs, each with its own linguistic structure formed according to grammatical rules." This is clearly a continuation of the view of Fairservis (Possehl's former mentor) and others: that the seals were some form of personal identification; and that it was the seals and not the seal impressions that were read, because the seals themselves are not heavily worn and so could not have been used often to make impressions. In support of the latter argument, there is a dearth of tags: unbaked clay sealings bearing seal impressions.

I disagree with these views for several reasons. Seals are typically less than five centimetres square and are carved intaglio (in the negative), ie the signs in the seal are the reverse of those in the seal impression. To read the seal must have been very difficult without close inspection and some degree of practice - which begs the question, why carve in the negative if the impressions would seldom be read? While Possehl lists 48 tags, I count 117 tags in the photographic corpus (for example L-211). This is still not very many given the total number of seals, but there may be good reasons. The main one is the caustic nature of the Indus Valley environment, which has destroyed almost all unfired clay. By far the largest number (96) of surviving tags is from a building, described by the excavator as a warehouse, in a port settlement (Lothal) that was much smaller than Mohenjo-daro, Harappa or Dholavira. They survived because the building accidentally burned down in antiquity.

I therefore postulate a different use for the seals: as markers of merchandise, rather than personal "signatures". Using the seals, the tags and some of the longer texts, it can be demonstrated that longer texts can be created by recombining short texts (see accompanying photos). The order of recombination can be worked out by a careful analysis of the long inscriptions, including the tags. The seals contain all of the necessary elements of syntax to form long messages, as seen on tags such as K-89 from the northern site of Kalibangan, which contains five different seal impressions, although many tags consist of just one seal impression.

It seems possible, given the almost universal presence of perforated bosses on the back of seals and/or other perforations, that seals were strung together on cords or wires and used as required by the particular nature of the shipment to create impressions on tags - a rudimentary form of printing. We can expect that some seals contain elements of syntax such as names, places, contents of loads, modes of transport, agreed values or quantities, protective incantations and other mercantile information. Further, some seals will contain one element while others will have several elements. The wear on a particular seal would therefore be much reduced because it would not have been constantly used. This would have been a very efficient system for a merchant who was trading in one commodity with several people in different locations.

Analysing the decipherments as a whole, Possehl calls for some agreement on the details of an Indus sign list. He summarises the current unsatisfactory situation: "just about anything is fair game. Everyone is guessing about the set of primary signs and making claims about sign variation that may or may not be true. From an epistemological point of view these claims are nothing but assertions and 99.9 per cent is probably useless speculation." He points out that only Iravatham Mahadevan, the leading Indian student of the script, has attempted to determine sign variation using the statistical and positional behaviour of signs. Most sign lists, including Mahadevan's and Parpola's, contain 250 to 425 signs. My own count of signs is 584. Whoever is right, these are too many signs for the Indus script to be an alphabet or a syllabic script like Cretan Linear B; it is probably logosyllabic, ie a mixture of logographs (word signs) and syllables, like Egyptian hieroglyphs and Mayan glyphs.

Unfortunately, there are frequent printing errors in this book. Most are relatively minor typos, but others are more serious, for instance a repeated sentence on page 18, inverted graphics on page 107, and a misprinted Indus sign on page 134. The misprinting of the Dholavira sign on page 51 is the most serious example, because this sign is unique in the corpus for its context and size. When it was found in 1990, Possehl tells us that it was "lying face down on the floor of one of the western chambers of the north gate to the 'Castle' high mound".

The letters measure about 37 by 25 cm and are made of white steatite originally inset into a large wooden plank; they probably served as an entrance sign board for the city. Possehl omits one of the ten signs in his transcription, the seventh sign (reading from the right) in the transcription above. This will seem on casual inspection a relatively minor error, but given that the particular sign sequence occurs elsewhere, the misprint could be important. I posit that some part of the sign sequence from Dholavira should be the site's locative (place name). But which part? Signs that are locatives may reasonably be expected to occur elsewhere in the corpus of inscriptions. Five consecutive signs from the Dholavira inscription are repeated on three artefacts from Mohenjo-daro with the inscriptions below. The top and middle ones are found on bronze implements, while the bottom inscription is found on a tag. It seems logical that the five-sign sequence in common is the locative of ancient Dholavira.

In addition to the locative, of the two texts on the bronze implements both contain the "spiked" sign, but only the bronze celt (an ancient kind of chisel) has the additional "hatted" sign. From this distribution I suggest that the spiked sign is the logograph for bronze, and the hatted sign (more tentatively) the logograph for celt. With these identifications the tag from Mohenjo-daro can be read - see above. There are still several unknowns: the direction of movement of the bronze; the quantity signified by the last two signs (on the left); and the sound values of the signs. (Note that this reading does not require a knowledge of the language of the script, or sound values at all.)

In his concluding chapter, Possehl makes several insightful statements. First, he calls for the study of the Indus script in its physical context, that is, chiefly a thorough examination of the kind of object that has been inscribed and of its geographical position when found. Second, he recommends further study of the script within the context of the Indus culture as a whole. Third, he suggests "a more methodical, deliberate approach to the decipherment programme, with a kind of team spirit". Collaboration was required for the decipherment of Mayan hieroglyphic writing, as Michael Coe showed in his 1992 book, Breaking the Maya Code . Michael Ventris, who deciphered Linear B, strongly believed in collaboration too. Once Mayanist scholars began to cooperate, publish notes and short papers, and meet regularly to debate readings and ideas, progress in deciphering the Mayan script accelerated at a phenomenal rate. This lesson should not be lost on those Indus scholars still working seriously on the script.

There are relatively few examples of books on the Indus script that examine the subject without offering a complete and definitive decipherment, often without any discussion of methodology. Possehl succeeds in his goal of taking stock of the current state of the decipherment without making a case for a decipherment of his own. The way in which he reviews the major decipherments makes this book at once useful to the professional and accessible to the neophyte. Treating the decipherments chronologically gives an understanding of the way in which our views of both Indus culture and the script have changed over the past 80 years. Possehl also notes the advent of new technology - computers - and its limited effect on the methodology of would-be decipherers. The book's bibliography is enormous, with almost 1,000 entries; it is potentially a very useful research tool, and a significant contribution to the value of this book.

Indus Age: The Writing System is a necessary addition to the library of any serious Indus scholar - as necessary for students of the undeciphered Indus script as Coe's Breaking the Maya Code is for students of the now-deciphered Mayan writing.

Bryan Wells is a graduate of the University of Calgary, Canada, where he has been working with the Mayanist David Kelley on the decipherment of the Indus script before beginning a PhD at Harvard University.