The Indus script: a challenging puzzle
The historical context of the Indus script: expanding horizons The first example of the Indus script was published more than a century ago. It was a very typical Harappan stone seal, bearing on its face the carved image of a 'unicorn' bull and a line of unknown writing. A drawing of the seal appeared in an archaeological report on the ruin mounds at Harappa in the Punjab (Cunningham 1875). But it was not until the 1920s that the Indus civilization was really discovered. Extensive excavations were carried out at Harappa (Vats 1940) and at Mohenjo-daro in Sind, 600 km southwards on the Indus (Marshall 1931; Mackay 1938). These great metropolitan centres represented a previously unknown culture that had been forgotten already in antiquity. The uppermost levels at Harappa and Mohenjo-daro belong to the Mauryan and Kusana periods, when the actual history of India begins, but they are separated from the Harappan strata by a hiatus of some 1500-2000 years. The number of 'Harappan' settlements so far located is around one thousand. They cover a huge area that extends from Sutkagen-dor on the Makran coast in Pakistan near the Iranian border to Alamgirpur near Delhi and from Shortugai in Afghanistan to Daimabad near Bombay (for mutually supplementing site lists see Jansen 1979: 298-306; Mughal 1981; Joshi et al. 1984). The temporal dimension, too, has expanded enormously. The excavations in progress at Mehrgarh on the border of South Baluchistan and Sind have unearthed a settlement whose 8 periods span c. 7000-2000 B.C. This now provides a chronological backbone for the development of the various 'Early Harappan' neolithic cultures and the 'Mature Harappan' Indus civilization (Jarrige and Meadow 1980; Jarrige 1984). We are also gaining an increasingly clearer idea of the gradual replacement of the Indus civilization by other cultures in the 'Late Harappan' period, roughly the first half of the 2nd millennium B.C. (Allchin & Allchin 1982: 229 ff.) There is direct evidence for maritime trade between the Indus civilization and Western Asia from the time of Sargon the Great (24th century B.C.) to the Ur III and Isin-Larsa dynasties. This consists, mainly, of seals with Indus script excavated in Mesopotamia and the Gulf (Gadd 1932; Brunswig et al. 1983), and of Gulf seals found at the Harappan seaport of Lothal in Gujarat (Rao 1963) as well as in the now submerged town off Bet Dwarka island in Saurashtra (The Statesman, Delhi, April 17, 1985). Moreover* thecuneiform texts of that time speak of sea trade with the foreign countries of Dilmun (the Gulf/Bahrain), Makkan (Oman and the Makran coast) and, most distant, Meluhha (Pettinato 1972). Meluhha is now almost unanimously identified with the Indo-Iranian borderlands and the Indus. Several tablets refer to a colony of acculturated Meluhhan traders in Lagash (S. Parpola et al. 1977). The extent of the inscribed material A primary task in the study of the Indus script is the collection of all extant material, making it available in a critical edition, and preparing various kinds of indexes, statistics and concordances. Working tools of this kind have been produced in recent years with constant improvement. (Koskenniemi et al. 1973; Mahadevan 1977; Koskenniemi and Parpola 1979-82. The inscription numbers quoted below are those of the last mentioned publications.) However, there is still much to be done in this field. An important current project is a comprehensive photographic corpus of the Harappan seals and inscriptions, being prepared in collaboration by the Archaeological Survey of India, the Department of Archaeology and Museums, Government of Pakistan, and the University of Helsinki. The writer would be most grateful if readers would kindly notify him of any relevant material outside the few obvious collections. At present, about 4200 Indus seals and inscriptions coming from 60 different sites are known. 2090 of these come from Mohenjo-daro, 1490 from Harappa, 240 from Lothal, 140 from Kalibangan, 83 from Chanhujo-daro, and 44 from 15 different sites in the Near East. However, of these 4200 objects 250 are uninscribed (mainly seals with geometrical motifs like the swastika which became common towards the end of the Indus civilization) and about 400 are broken or otherwise illegible. About 510 different texts are variously repeated so that they occur altogether 1860 times: some 1350 may thus be excluded as duplicates (Koskenniemi and Parpola 1980). Often, but not always, the types of objects bearing such identical texts and their pictorial motifs (if any) are also identical. There are, then, about 2200 different and more or less legible inscriptions. All these texts are short. Their average length is just 5 signs. The longest text, on a three-sided 'amulet' (2757, 1573), consists of 26 signs. The longest inscription on any single side occurs on a seal (1400) with 17 signs in three lines. The Indus seals Approximately 60 per cent of the Indus texts are carved on stamp seals made of steatite. About 10 per cent of seals are rectangular and contain nothing but text, while a few display some rare form. The normal type, however, is square, with one line of text at the top of the face and beneath it a pictorial motif (Plate 1). The iconography of the seals (and amulets) is standardized; among the typical animal figures, which are naturalistic and often feed from a trough, by far the most common is the 'unicorn' bull facing an unidentified cult object (Plate 1). Occasional anthropomorphic deities (Plate 2) and other rare scenes give precious insights into the Harappan religion.
A considerable number of seal impressions on clay have survived. Most of these are tags which had been attached to bales of goods, for the reverse sides usually show traces of packing materials. Of special significance is the corpus of such clay bullas from the burnt-down warehouse or granary at Lothal. (Rao 1979: I, 111-14) Because all these tags were in use in the warehouse at one and the same time, an analysis of these seal impressions, presented here for the first time, reveals interesting facts about Harappan economic administration. Of the 77 tags, whose inscriptions are (at least partially) visible, 21 bear two, three or four seal impressions. Of the latter 21 tags, as many as 14 are interconnected, being linked with each other by shared seal impressions. Of the 9 different seal texts that can be read on these 14 tags, six recur also on tags with single seal impressions, the highest total frequency among them being occurrence on 15 different tags. Altogether, these 14 tags with multiple sealings and the related tags with single impressions (Fig. 1) constitute 37 tags, i.e. nearly 50 per cent of the total of 77. The remaining inscriptions (including those on the five other tags with multiple seal impressions: 6034, 6001, 6151, 6195, 6077) occur once only, with a few exceptions: two tags each bearing a double impression of one seal (different in each case, 6058, 6184); one text recurring twice on tags with single impressions (6155, 6170), and, strikingly, one seal text, U ('"•') "0 (pictorial motif: elephant), which is repeated 11 times (more than any other text) on tags with single impressions, but not once on tags with multiple seal impressions. Other types of Indus inscriptions Another major category of Indus texts consists of small tablets of various forms and materials: the texts were engraved on flat plates of steatite (at Harappa) or of copper (at Mohenjo-daro), or they were moulded in terracotta or faience. Such tablets may have text, and occasionally pictorial motifs, on one, two, three or four sides. All are likely to have had a religious or ritual function, and are conventionally called 'amulets'. There are inscriptions on about two dozen of the short bone or ivory sticks decorated with geometrical motifs. Their function is unknown, but the occurrence of concentric circles in differing numbers on various sides suggests that they may have been used as dice or as mantic sticks in divination; calendric interpretations have also been proposed. Current surface surveys at Mohenjo-daro have increased the number of inscribed bangles to forty (Franke in press). Just a few years ago this inscriptional type was almost non-existent. Pottery vessels form one important category of inscribed object. Several subgroups may be distinguished. Stamped seal impressions were undoubtedly made by the potter
But did the seal belong to him or to the person or institution for whom the pot was made? Some large storage jars have at their base inscriptions raised in relief, formed by negative incisions in the open mould on which they were made (Dales 1967: 36). Signs incised into the wet clay before the pot was fired are also to be ascribed to the potter, likewise the rare painted inscriptions. Most of the pottery inscriptions are graffiti scratched on the surface after firing. The majority of those drawn on the rim are numbers only (repeated strokes and, probably for the tens, semicircles); such marks occur also on the body (cf. Koskenniemi and Parpola 1980; 1982: 86-95). Numbers - alone or with other text - are also usually found in the inscriptions on axes and other copper or bronze weapons or tools {ibid. 95). Are they inventory numbers or do they record the weight of the precious object? In one interesting case, a distinctive sequence of seven signs occurs (besides numbers) on two blade axes (2796 and 2798) found at Mohenjo-daro, as well as on a fragmentary seal (2119) discovered nearby (Parpola 1975: 185); in 1984, a clay impression of another seal with the same sequence was found (Franke, in press). In addition to these types, there are a few miscellaneous objects bearing inscriptions, all rather small. With the possible exception of one broken stone slab (3599), there are no monumental royal edicts. Economic accounts almost certainly existed once, but they must have been written on perishable materials such as palm leaves, the traditional writing material of India. No bi- or multi-lingual texts have been found, either. Origins of the Indus script Recent archaeological researches in eastern Iran, southern Turkmenia and Afghanistan have been filling in that great gap which has long separated the Greater Indus valley from the Near East. It is now known that the Early Harappan cultures maintained overland contacts with Afghanistan, southern Turkmenia and northeastern Iran (Mughal! 1970; Kohl 1981). The coastal traffic between the Indus area and Western Asia represents a later development, undoubtedly connected with Harappan urbanization. The excavations at Tepe Yahya and Shahr-i Sokhta have shed much light on cultural interactions on the Iranian plateau and the dominant role played in them by the enormously extensive Proto-Elamite culture (conventionally dated to c. 3200-2900 B.C., though calibrated radiocarbon determinations suggest some 500 years earlier) (e.g. Tosi 1979). The neolithic societies on the threshold of urbanization had created communication methods that served their still relatively modest needs. It is legitimate, I think, to consider the 'potter's marks' on neolithic ceramics as important forerunners of real writing. In Susiana, in western Iran, pot-marks are found already in the 5th millennium, thus clearly antedating the Proto-Elamite writing; it is curious, though, that these potmarks do not continue into the 4th millennium, being thus separated from the ProtoElamite script by a considerable span of time (Dollfus and Encreve 1982). Comparable though much later 'potter's marks' are known from Tepe Yahya in southeastern Iran (Potts 1981), the Lut desert ofahe Iranian plateau (M. Ali Hakemi cited in Mughal 1970: 336), and Shahr-i Sokhta in Seistan. Also relevant for the study of the Early Harappan
marks (Potts 1982) are the incised marks on the terracotta cult statuettes in the bronze age culture of southern Turkmenia (Masson and Sarianidi 1969; Antonova 1972). Thousands of Early Harappan potsherds with 'potter's marks' (Fig. 2) have been discovered, usually on utilitarian unpainted ware. The marks were incised on the wet clay before firing by means of a sharp instrument (at Mehrgarh 57.40 per cent) or with a finger-nail (28.60 per cent) or both (14 per cent) (Quivron 1980: 274; Fairservis 1956: 328); marks painted before firing are much rarer. Graffiti scratched on pottery after firing also occur. Principal Early Harappan sites that have yielded pot-marks are detailed in the following. In their dating I have followed the chart of de Cardi (1984: 63), which places Mehrgarh III and KGM III between the Ubaid 4 and Uruk periods, c. 3700-3500 B.C. These are conventional dates: Mehrgarh III and KBM III are dated by Jarrige (1984: 22) to c. 4500-4000 B.C.; he cites an uncalibrated C14 date of 4745±90 b.c. (half-life 5730 years). In Baluchistan, Mehrgarh IV-VII (c. 3500-2500 B.C.) has a total of about 850 marks, of which 150 to 50 are different (Quivron 1980); Damb Sadaat I-III and other sites in the Quetta valley (including KGM IV, c. 3600-2600 B.C.) have yielded 362, with about 50 different marks, of which 18 were made with the finger-nail (Fairservis 1956; 328-35 and PI. 14); and from Periano Ghundai in the Zhob valley come comparable surface finds (Fairservis 1959: 359 and Fig. 59). Such marks occur also in southeastern Afghanistan, in Mundigak I (4th mill.) (Casal 1961: II, Fig. 49: 13) and period IV, dating from the first half of the 3rd millennium [IV.l (Fig. 87: 366-71), IV.2 (Fig. 93: 421-25), IV.3 (Fig. 105: 511-21)], and at Balakot (Early Harappan period A) in the Las Bela District of southern Baluchistan, probably a seaport in Harappan times (Dales 1974: 16 ff.). Near Balakot in southwestern Sind, incised 'potters' marks' have been found at Amri from periods I A to IIB, dated to c. 3500-2700 B.C. [Casal 1964, II: I A (Fig. 46: 63-67), I B (Fig. 48: 83, 49: 88, 51: 111, 53: 132-36), I C (Fig. 56: 152, 61: 193-98), I D (Fig. 66: 248-49), II A (Fig. 68: 268-70), II B (Fig. 72: 305-06)]; a few are known from nearby Kot Diji (Khan 1965: PI. XXIV: 1, 2, 6). The excavations at Rahman Dheri in the western plains of the northern Indus valley have not yet been properly published; calibrated radiocarbon samples of this large site date it to c. 3300-2500 B.C. 'Very large numbers of graffiti occur on the pottery of all periods at Rahman Dheri' (Allchin and Allchin 1982: 150-1). Sarai Khola II near Taxila has also yielded 68 sherds of Early Harappan ceramics with either pre-firing incised 'potter's marks' or post-firing graffiti (Halim 1972; 95-99 with table 11, Fig. 18: 101, 106; 19: 111, 115-120; 22: 134, 140; 23: 148; pi. XX B.I, XXII A-B). A few pot-marks are known also from Early Harappan Jalipur II in the Punjab (PA 4/1967: 8 f., Fig. 2.5; Mughal 1974: 110, A). A great number of further sites with Early Harappan pot-marks are likely to be added when the detailed results of recent explorations and excavations, notably in northern Baluchistan (Mughal 1972), in Bahawalpur (Mughal 1981) and in India (e.g., Kalibangan) are fully published. Many of the 'potter's marks' seem to be numbers, at least the repeated strokes which evidently stand for ones. The same system of marking numerals is used in the Indus script, and a large part of the graffiti on potsherds of the Indus civilization itself consist of nothing but such numbers. Did the (usually repeated) semicircles, which in the Indus script represent larger numbers (tens?), develop from the (often repeated) finger-nail impressions of the 'potter's marks'? (The Egyptian script has a similar semicircle for 'ten', but it developed from the picture of 'hobble for cattle', the corresponding Egyptian word being a homophone of the word for 'ten'; Gardiner 1957: 524 no. V19-20.) Similarities in relatively simple linear signs cannot be given too much weight. For example, the signs of the Vinc,a culture in Southeast Europe (Winn 1973: 196-204) offer parallels to the Indus script that are at least as close as those found among the Early Harappan pot-marks (Potts 1982: 517 Fig. 4), and, on the whole, signs really characteristic of the Indus script are not present among the latter. However, there are a few interesting exceptions that deserve special attention. One (unfortunately slightly broken) Early Harappan graffito from Balakot (Fig. 2: 5) [and possibly another from Rahman Dheri (Durrani 1981: PI. xix)] appears to be identical with the most common and most characteristic Indus sign 1 J" , usually found at the end of an inscription. Another Indus sign characteristically taking this position, the 'spear-head' T , is also attested in a slightly broken graffito from Balakot (Fig. 2: 2). However, in view of its absence elsewhere in the relatively abundant material, it is strange that these Balakot graffiti come from the lower levels (IV and I respectively) of the Early Harappan occupation (subdivided into 11 phases, I-IX corresponding to Nal polychrome andTogau C, and X-XI to Amri I C-D, Dales 1974: 11, i.e., roughly, to the middle and end of the 4th millennium in de Cardi's 'conventional' dating). The common Indus sign of 'two intersecting circles', (JP , often occurring in the earliest tablets (Koskenniemi and Parpola 1982: 182-4), is known as a painted mark from Gumla (Dani 1970-71: PI. 83), Sarai Khola II (Halim 1972: PI. XX B:l) and Rahman Dheri (Durrani 1981: PI. xix). This survey shows that 'potter's marks' existed throughout a vast area in the Greater Indus valley many centuries if not a millennium before the creation of the Indus script. Conditions for a relatively independent creation exist, and at least some of these earlier local symbols seem to have been adopted as graphemes into the Indus script. Moreover, the Indus script cannot be directly derived from any other known script, not even from the vaguely similar Proto-Elamite, which is closest to it. On present evidence, the similarities between the Indus script and Proto-Elamite are not sufficiently complex and extensive to warrant a conclusion of genetic relationship. Perhaps this question can be definitely solved when clear examples of the early formative stage of the Indus script become known, but it seems most likely that the Indus script was created autonomously and within a short period of time. The process of urbanization created the need for a writing system for effective communication and record keeping in administration and trade. This internal requirement was in all likelihood combined with the stimulus of idea diffusion from the west. The.strong influence of Proto-Elamite art on Harappan seal glyptics (Parpola 1984) supports the natural and often voiced view (inter alia, Potts 1982) that the indirect model was the geographically and temporally adjacent and vaguely similar Proto-Elamite script. At the moment, the Indus script appears fully developed and almost entirely standardized already in the earliest available examples, the 'miniature tablets' of the lowest levels of Harappa dating perhaps from the 26th century B.C. [on the basis of MASCA calibrated C14 dates, B. and R. Allchin (1982: 218 f.) place the beginning of the Mature Harappan period at c. 2550 B.C.]. Only a few signs differ slightly from their later forms. The direction of writing, too, is not yet standardized, while later there are very few exceptions to the rule, which is right to left. Characteristically, these earliest examples of the Indus script bear short inscriptions, with large numbers of duplicates. There are cogent reasons for thinking that they are associated with religious worship. One of the usually two inscribed sides normally bears only two signs: a numeral (1, which may be omitted, 2, 3 or 4) plus a U-shaped sign. The U-sign clearly represents a sacrificial vessel: in some cases the accompanying scene shows a kneeling worshipper extending a vessel of this shape towards a sacred tree or an anthropomorphic deity sitting on a throne. (Parpola 1981; Knorozov et al. 1981: 44, 106; Hrozny 1942: 36) The recent suggestion that these 'miniature tablets' are 'accounting tablets' comparable with those of Proto-Elamite Susa (Shendge 1985) is inadmissible. Difficulties and pitfalls of decipherment Several respected experts (e.g. Friedrich 1966: 145) have denied every possibility of deciphering the Indus script. Such pessimism is understandable, for none of the crucial keys that have opened other unknown scripts is available. We have no Indus inscriptions with parallel versions in some other readable script; nor do we, in the absence of any political history, know the names of the Harappan gods, kings, cities or villages. Even the affinity of the Harappan language and the type of writing system represented by the Indus script are much debated. The texts are short and limited in nature, and there are no clearly recognizable word dividers such as have been very helpful in the analysis of many other scripts. Considering these formidable difficulties it is hardly surprising that the decipherment of the Indus script remains a vexed problem. Many of the proposed solutions can be dismissed immediately as uncritical and/or uninformed in some field of knowledge vital for a successful decipherment, such as the study of writing systems and the methods of their decipherment; or archaeology, linguistics and philology, especially of the Indian subcontinent and Western Asia. For example, when Sir Alexander Cunningham studied the first Indus seal more than a century ago, he knew nothing about its antiquity; he was reasonable, but uninformed, in assuming that it might represent an early form of the Brahmi script, dating from the times of the Buddha. It is now common knowledge that Brahmi (first attested in Asoka's edicts c. 250 B.C.) is derived from the Semitic consonantal alphabet, which in turn is derived from Egyptian hieroglyphs. And yet some researchers still insist on deriving the Brahmi and with it the modern Indian scripts from the Indus script, basing further hypotheses on this premise. It seems that the Indus script was forgotten after the collapse of the urban civilization which created it - a situation comparable with most of postMycenaean Greece. The latest certain attestation of the Indus script is from Daimabad, dating from c. 1800-1600 B.C. (Sali 1982). Some of the later pottery graffiti in the Deccan may be interpreted as reminiscences (Lai 1960) - but these are no longer real writing. One mistaken method is so common that it deserves special mention. The Indus signs have been equated with similar-looking symbols in other scripts and read with their phonetic values. (An even cruder 'method' consists of interpreting the Indus signs pictorially and assigning them syllabic values from the first syllables of the corresponding words of an arbitrarily chosen language.) A comparison of signs that look alike in different pictographic (non-alphabetic) scripts will demonstrate, however, that the phonetic values are different, since the different scripts for the most part reflect different languages (Parpola 1979: 171). It is possible to advance knowledge only by making hypotheses, which, to be fruitful,must be informed guesses grounded on acceptable knowledge. Wrong guesses are unavoidable, but the scientific method tries to detect and rectify them by demanding that all hypotheses be subjected to adequate tests before they can be accepted as valid. The problem of the Indus script can be compared to a crossword puzzle, where individual solutions have to conform to each other as well as to certain specified requirements. Every hypothesis must be rigorously examined. With such an insistance on quality, every step forward should increase the means of testing. It is symptomatic of the uncritical and erroneous attempts at deciphering the Indus script that any possibility of internal control of the proposed readings is invalidated: the liberty is taken of positing one and the same value for many signs, and of allowing one and the same sign to have many different values (an extreme example is Hrozny 1942). As a result, all the texts can be 'read' yet the solutions usually fail to make good sense. The capacity to explain well external evidence is the other touchstone. The type of writing system represented by the Indus script What is the nature of the Indus script? This is one of the crucial questions which must be solved if there is ever to be a real decipherment. The theory of writing and decipherment provides us with several criteria. The number of distinct signs (graphemes) is one touchstone. Widely different estimates have been given, ranging from about 400, which is the usual figure, to just 52 (Rao 1979: I, 170). The reason for such a variation is twofold. One and the same sign may appear in several variant forms (allographs). Moreover, many signs are clearly combinations of one or more elements, and it can be difficult to decide whether a sign is a composite one or not. Even so, it is possible to state that there are at least 200 basic signs. This already precludes the possibility of an alphabetic script; even for a syllabic script this number is too high. 'Word' length is another criterion. Many texts consist of just one sign, which therefore necessarily must represent a complete 'word'. Moreover, most Indus texts can be divided into units of one to three signs by means of internal comparison, which can be carried out by computer methods (Fig. 3). It seems, indeed, that the Indus script represents the same kind of'nuclear' writing as the early Sumerian texts of Fara (26th century B.C.), in which only the lexical meaning of each word is expressed, while all or most grammatical affixes are omitted and left to be understood from the context (Civil and Biggs 1966). These considerations, based on the number of graphemes and the 'word' length, are in agreement with the third criterion, time. Harappans of the mature Indus civilization seem to have been in direct sea contact with Mesopotamia in the 24th century B.C. The urbanization process - and the creation of the Indus script - is likely to have taken place at least by the 26th century, if not earlier. At this time the only existing type of writing was a logo-syllabic or morphemographic type of script, in which each sign stands for a word or a morpheme of one or more syllables. The oldest known essentially syllabic script dates from around 2350 B.C. (this is Linear Elamite, assumed to have comprized about 80 signs in all, 58 only being attested in the 19 texts, cf. Hinz 1964: 27, 1975: 106 f.; also Vallat, this volume, 343), while alphabetic writing was created by the Semitic peoples only around 1600 B.C.
It seems indeed more than likely that the Indus script belongs to the earliest type of real writing and is a morphemographic script. The primary means of representing a concept in a script of this type is to draw a picture of it. But abstract concepts and grammatical elements are difficult if not impossible to represent pictorially, if another person is to recognize the intended meaning. For this reason, and to reduce the number of graphemes, recourse was taken to the rebus principle, the method of expressing the phonetic shape of the intended word or morpheme through another, similarly sounding word which was easier to depict. All people have word plays based on homonymy as an integral part of their folklore, and punning must have been applied in communication long before the creation of real writing. In any case, the early Sumerian and Egyptian scripts, which can be read, make much use of the rebus principle, and this must have been so in the Indus script also. This means that, depending on the context, a pictographic sign could denote the object represented by it, or it could denote any of the words or morphemes of the language underlying the script that were (within certain strict limits) phonetically similar to the word naming the depicted object. Basic methodology of decipherment On the premise that the Indus script represents morphemographic (logo-syllabic) writing, it is possible to devise a basic methodology for its decipherment. Homonymies are language-specific: puns cannot easily be translated into unrelated languages. Therefore, the use of the rebus principle provides the possibility of identifying the underlying language and deciphering individual signs. This requires, however, that four conditions are simultaneously fulfilled:
1 the object depicted by a given pictogram can be recognized; 2 the said pictogram has in a given context been used phonetically, i.e. in a meaning different from the pictorial meaning; 3 the intended (phonetically used) meaning can be deduced from the context; and 4 a linguistically satisfactory homonym fitting the posited meaning can be found in a likely language. It is inherent in the nature of a morphemographic script that it cannot be deciphered all at once, as can alphabetic and syllabic scripts. In alphabetic and syllabic scripts each sign (as a rule) has only one value, and these values are all phonetic and interrelated, forming a closed system. Once the phonetic value of a few signs has been correctly determined, the contexts will suggest the values of the remaining signs. By contrast, a morphemographic script is an open system, the signs may have several values, and they represent larger linguistic units whose interrelations are not so clearly circumscribed. In other words, a deciphered morphemogram will provide clues for other signs, but the signs will have to be deciphered one by one. For the Indus script, the pictorial meaning of a sign can be determined by various means. One is a comparative study of all its variant forms (allographs). Harappan art and culture generally can provide clues; thus some decorative motifs on early Harappan pottery seem to have been models for Indus pictograms (Parpola 1975: 199 f.). Similarlooking signs from other pictographic scripts provide one of the most useful external parallels. Unfortunately, most of the Indus signs are so simplified and stylized that their pictorial meaning - and therewith their decipherment - is likely forever to remain problematic and controversial. Those few signs whose pictorial form is identifiable are, therefore, crucial for the decipherment of the Indus script. The contextual meaning of the signs can be established in two ways. One is an internal analysis of the inscriptions by linguistic and statistical methods, in which automatic data processing is of great assistance (Koskenniemi 1981; and Fig. 3). This process must be distinguished from the use of external contexts, which include the pictorial form of the sign, and such things as the iconographic motifs that may accompany the text, the nature of the object on which it has been inscribed, and its archaeological find place. An especially precious clue is provided by the nature of most Indus inscriptions: they were engraved on seals used in administration and trade. Because Harappan seals have been found in the Near East, their inscriptions are likely largely to parallel the Mesopotamian seal inscriptions, which can be read: the latter contain mainly proper names and official titles, with or without attributes and dedicatory formulas (Gelb 1977). The affinity of the language underlying the Indus script The question of the actual language is, besides that of the nature of the script, the other great unknown to be solved, and without doubt the most interesting and debated issue. It cannot be decided without taking into account three major factors: * 1 the archaeological and culture-historical context,2 the findings of historical linguistics, especially with regard to the Indian subcontinent, and 3 the typological structure of the Indus language, established by the application of linguistic and statistical methods. Archaeological research has proved beyond doubt that the Indus civilization is not a Sumerian colony, and that the Sumerians and the Harappans do not share a common descent. Besides, the Sumerians had developed a distinctive script of their own long before the Harappans. Moreover, the sign sequences of the Indus script are uniform throughout the area of the Indus civilization and during its entire duration, thus indicating the use of just one language in the script, but Indus seals found in the Near East have totally different sign sequences. This is true especially of those Indus seals from Western Asia whose round or cylindrical form accords with the local seal forms. These seals bear common Indus signs, but the order is unique: the only conclusion seems to be that they record Sumerian or Semitic names of acculturated Indus merchants. According to this evidence, the languages they had adopted in the Near East differed radically from the native Harappan language. (S. Parpola et al. 1977; Brunswig et al. 1983) A Sumerian solution (Kinnier Wilson 1974), therefore, appears unlikely. The area of the Indus civilization has been inhabited by speakers of Aryan languages since at least 1000 B.C. The hypothesis that an early form of Indo-Aryan was spoken already by the Harappans has its supporters. This brings us to the vexed issue of the arrival of the Aryan speakers in India. Archaeology in India, the Indo-Iranian borderlands, Central Asia and elsewhere, has accumulated a wealth of new data on this question, and some tentative statements can already be made. There is no escape from the fact that the horse played a central role in the Vedic and Iranian cultures, as it undoubtedly did already in the culture of the people speaking Proto-Indo-European. The horse appears to have been domesticated in the South Russian steppes c. 5000 B.C. (inter alia, Piggott 1974: 17), and this region is focal in attempts to trace the dispersal of the Proto-Indo-European-speaking community archaeologically (Thomas 1982). In the Near East horse-drawn light war-chariots were an important asset of the Aryan-speaking nobles, who ruled Mitanni in Northern Mesopotamia c. 1500-1300 B.C. (for divergent views, see Mayrhofer 1974; Diakonoff 1972). It is generally believed that the earliest Indo-Aryan texts, the hymns of the Rgveda, were composed c. 1300-1000 B.C. in the area bounded by Afghanistan and the Punjab, and the later Vedic texts c. 1000-500 B.C. in the 'middle region' from the Punjab to the upper Ganges valley (Gonda 1975: 22 f, 360 f.). In regard to geographical and chronological span as well as material content, the later Vedic culture closely corresponds with the archaeological culture characterized by the use of Painted Grey Ware (Tripathi 1976). Many scholars have assumed that the Indus cities were meant in Rgvedic references to the forts of the Dasas and Dasyus, the dark-skinned enemies vanquished by the Aryans. Apart from the chronological gap, this is unlikely for several reasons. The form of the Dasa citadels, three concentric circles, differs from that of the usually square Harappan cities but exactly parallels the cultic centre with three concentric fortification walls at Dashly-3 in North Afghanistan (Parpola 1985: 76-78). The Dashly-3 citadel has been compared to the Old Iranian var and the cult places of the 'Kafirs' (Jettmar 1981). The Dashly-3 culture is related to the Northeast Iranian Gray Ware complex of the third and second millennia B.C. (represented by Tepe Hissar II-HI), which had spread to the borders of the Indus valley already in the Mature Harappan period (Francfort 1984; Santoni 1984). The people who ruled the North Iranian Gray Ware culture used horsedrawn war-chariots (Ghirshman 1977: 14 ff.). It has long been suspected that they represent an early wave of Aryan speakers (Thomas 1982: 64-7). This hypothesis is in agreement with the Aryan etymology of the ethnic name Dasa, and the merger of different linguistic and religious traditions of Indo-European origin in the early Vedic period (Parpola 1983). A major reason against assuming that the Harappans spoke an Indo-European language is that the horse is not represented among the many realistically depicted animals of the Harappan seals and figurines. Nor is there clear osteological evidence of the horse {Equus caballus) in the Indian subcontinent prior to c. 2000 B.C. (Jarrige and Santoni 1979:1,404 f.). In the 2nd millennium, however, a number of clay models of the horse and the Bactrian camel (but not of the bull which is prominent in Early Harappan and Indus art) are found at Pirak, a site on the border of Sind and Baluchistan (ibid.). The population of the Indus civilization is estimated at c. 5 million (McEvedy and Jones 1978: 182). The number of the invading Aryans can have been only a fraction of this. It is clear that the Harappans were not all killed, and that their language could not have disappeared overnight: they were subdued and linguistically Aryanized in the course of many centuries. Such a process involves bilingualism and leaves evidence of the disappearing substratum language in the structure and vocabulary of the surviving language. It is now almost unanimously accepted that there are clear traces of Dravidian substratum influence even in the earliest Vedic texts (Emeneau 1980: 85 ff., 167 ff.). Moreover, a Dravidian language, Brahui, is still spoken in the mountains of Baluchistan and Afghanistan, a core area of the 'Early Harappan' culture. Shared innovations prove that Brahui forms a North Dravidian subgroup together with Kurux and Malto, languages spoken in the hills of east central India and Nepal (ibid.: 314 ff.). After Aryan, Dravidian, now mainly spoken in South India, is the second largest linguistic family in the Indian subcontinent. Earlier this century it was spoken by one quarter of the total population; at present its speakers number some 130 million. In comparison, the other Pre-Aryan languages are little spoken and their geographical distribution and external relatives fall outside the Harappan cultural sphere: the TibetoBurman languages of the Himalayas and the Northeast, distantly related to Chinese; the Austro-Asiatic languages (Santali, Munda, etc.) spoken in central and eastern India and related to the languages of Southeast Asia; and the few isolated tongues, Burushaski high in the Hindu Kush, Nahali in central India, and Vedda and Rodiya in Sri Lanka (Parpola 1974; Zograph 1982). An impressionistic analysis of the Indus inscriptions suggests that structurally the Harappan language is of a fairly pure 'left-branching' and agglutinative type. Dravidian is the only match in the recorded linguistic history of the Indian subcontinent (Knorozov et al. 1981: 13 ff.; Koskenniemi and Parpola 1982: 11 f.). However, it must be emphasized that a proper linguistic analysis of the Indus texts has yet to be carried out
The present state of deciphermentOn the basis of external and internal criteria, then, it seems most likely that the Harappans spoke a Dravidian language. Can this be substantiated? By applying the methodology outlined above, some Indus signs can be read in Dravidian. The fish-like pictograms have a great frequency, and they occur in contexts suggesting that they stand for names of deities. The Indus seals are likely to contain mainly proper names and occupational titles, like those in Mesopotamia, and divine names have formed an important element of proper names and priestly titles in Mesopotamia as well as later in India. The pictorial interpretation of the basic sign ty> as 'fish' is verified in Harappan art, and the homonymy between the Dravidian words mln 'fish' and min 'star' makes an astral interpretation possible. The two words can safely be reconstructed for Proto-Dravidian, and are both probably derived from the verbal root min 'to glitter'. That such a homonymy existed in the Harappan language as well is suggested by the Harappan painted pottery combining star and fish motifs (Fig. 4)
In Mesopotamia, some divinities were represented by particular heavenly bodies, and in the cuneiform texts the determinative of divine names is a sign depicting a star (see page 330). That the Harappan gods also had an astral aspect is shown by an Indus seal, in which a star is placed on either side in the loops of the horned headdress worn by an anthropomorphic deity (Plate 2 and Parpola 1984). Moreover, the trefoils of the famous 'priest-king' statue from Mohenjo-Daro are assumed to have an astral meaning (Marshall 1931, 356 f.); for this there is now new evidence (Parpola 1985). It is possible to crosscheck this interpretation of the 'fish' pictogram by studying other signs which have been combined with it in the Indus script. Numbers preceding the fish sign give such readings as, for example, '(constellation consisting of) six stars', which are actually attested as star names in the most ancient Dravidian texts, the Old Tamil poems of the first centuries of the Christian era: aru-mln 'six-star', for example, denotes the Pleiades. Unfortunately many of the 'diacritical' signs that often (but never after numerals) modify the basic fish signs are open to various interpretations
One such sign, however, seems fairly unambiguous: placed above the fish sign, it looks like a 'roof: -^ (Plate 1). Through the homophony between Proto-Dravidian *vey/mey 'to cover (as a roof), to roof, thatched roof and *may 'black' (with the alternation of vand m- at the beginning of the word, and of a and e before -y, both of which are assumed to have existed already in Proto-Dravidian), we get the compound mai-m-min 'black star', which in Old Tamil denotes the dimly visible planet Saturn. In Sanskrit sources, too, the planet Saturn is invariably associated with the colour black. Its presiding deity is Yama, the god of death, who is usually depicted riding on a dark buffalo. Several iconographic texts mention the buffalo also as the vehicle of the deified planet Saturn. Interestingly, the mount of the planet Saturn (usually called sani or sanais-cara, 'slowly moving', in Sanskrit) in the Buddhist and Jaina sources is the tortoise, which is not only a slow creature but also a kind of 'fish' (aquatic animal) that carries a 'roof over itself! Some further interlocking interpretations of this kind are available. External support to such astral interpretations is given, inter alia, by the importance of astrological considerations in Indian name giving and the fact that, according to astronomical evidence, the Vedic star calendar was created c. 2300 B.C. This dating coincides with the peak of the Indus civilisation, which, like other early urbanized cultures, needed some form of calendar. The practice of astronomy and its importance for religion is shown by the orientation of Harappan cities according to the cardinal directions (Parpola 1975). There is, however, a limit to this approach. The highly simplified form of most of the signs makes their pictorial interpretation difficult and thus increases the element of uncertainty. It is to be frankly admitted that the prospects of ever fully understanding the Indus script are meagre. Moreover, it must be added that the validity of the above interpretations, developed from those presented by a Finnish research team more than 16 years ago, is an open question. It is symptomatic of the present state of decipherment that, although several scholars in recent years have been working on the Indus script with largely similar methodology and basic assumptions, including the Dravidian hypothesis, there is, as yet, little agreement on individual interpretations. Notable exceptions are the readings of the 'fish' sign and its combinations with the numeral signs: it was first suggested by Henry Heras (1953: 127 f.) that '6 + fish' stands for the Dravidian name of the Pleiades, and, along with other similar occurrences ('7 + fish', '3 + fish'), this interpretation is included in the extensive solution proposed by Soviet scholars (Knorozov et al. 1981: 93). These readings are also accepted by Mahadevan (1970: 179 ff., 255 f.), but not by Fairservis, who disagrees with the pictorial identification of the 'fish' sign (1984: 156, 158). I think this unusual agreement has to be taken seriously. The reason why few people have yet done so seems to be that really significant interpretations have been obscured among many obviously implausible or at least highly subjective and insufficiently established readings, which fail to carry conviction. Is it likely, for example, that a seal text reads, 'The red (or great) glittering monsoon's power'? Although the published 'solutions' offer many valuable observations and seminal ideas, it is clear that a lot of washing has to be done in order to obtain the gold dust, and that a critical reader is inclined to throw that gold dust away with the wash water. As John Chadwick saw, the only remedy is to lay stress on quality instead of quantity: What we shall need is not more possible or even plausible interpretations of signs based upon this theory, but the clearest possible demonstration that these meanings, and only these meanings, are correct. Once this foundation is unshakeable, it will be safe to build upon it, but we must not be led into admiring a house of cards which, elegant as it seems, will collapse if one prop is withdrawn (Clauson and Chadwick 1969: 207).
A considerable number of seal impressions on clay have survived. Most of these are tags which had been attached to bales of goods, for the reverse sides usually show traces of packing materials. Of special significance is the corpus of such clay bullas from the burnt-down warehouse or granary at Lothal. (Rao 1979: I, 111-14) Because all these tags were in use in the warehouse at one and the same time, an analysis of these seal impressions, presented here for the first time, reveals interesting facts about Harappan economic administration. Of the 77 tags, whose inscriptions are (at least partially) visible, 21 bear two, three or four seal impressions. Of the latter 21 tags, as many as 14 are interconnected, being linked with each other by shared seal impressions. Of the 9 different seal texts that can be read on these 14 tags, six recur also on tags with single seal impressions, the highest total frequency among them being occurrence on 15 different tags. Altogether, these 14 tags with multiple sealings and the related tags with single impressions (Fig. 1) constitute 37 tags, i.e. nearly 50 per cent of the total of 77. The remaining inscriptions (including those on the five other tags with multiple seal impressions: 6034, 6001, 6151, 6195, 6077) occur once only, with a few exceptions: two tags each bearing a double impression of one seal (different in each case, 6058, 6184); one text recurring twice on tags with single impressions (6155, 6170), and, strikingly, one seal text, U ('"•') "0 (pictorial motif: elephant), which is repeated 11 times (more than any other text) on tags with single impressions, but not once on tags with multiple seal impressions. Other types of Indus inscriptions Another major category of Indus texts consists of small tablets of various forms and materials: the texts were engraved on flat plates of steatite (at Harappa) or of copper (at Mohenjo-daro), or they were moulded in terracotta or faience. Such tablets may have text, and occasionally pictorial motifs, on one, two, three or four sides. All are likely to have had a religious or ritual function, and are conventionally called 'amulets'. There are inscriptions on about two dozen of the short bone or ivory sticks decorated with geometrical motifs. Their function is unknown, but the occurrence of concentric circles in differing numbers on various sides suggests that they may have been used as dice or as mantic sticks in divination; calendric interpretations have also been proposed. Current surface surveys at Mohenjo-daro have increased the number of inscribed bangles to forty (Franke in press). Just a few years ago this inscriptional type was almost non-existent. Pottery vessels form one important category of inscribed object. Several subgroups may be distinguished. Stamped seal impressions were undoubtedly made by the potter
But did the seal belong to him or to the person or institution for whom the pot was made? Some large storage jars have at their base inscriptions raised in relief, formed by negative incisions in the open mould on which they were made (Dales 1967: 36). Signs incised into the wet clay before the pot was fired are also to be ascribed to the potter, likewise the rare painted inscriptions. Most of the pottery inscriptions are graffiti scratched on the surface after firing. The majority of those drawn on the rim are numbers only (repeated strokes and, probably for the tens, semicircles); such marks occur also on the body (cf. Koskenniemi and Parpola 1980; 1982: 86-95). Numbers - alone or with other text - are also usually found in the inscriptions on axes and other copper or bronze weapons or tools {ibid. 95). Are they inventory numbers or do they record the weight of the precious object? In one interesting case, a distinctive sequence of seven signs occurs (besides numbers) on two blade axes (2796 and 2798) found at Mohenjo-daro, as well as on a fragmentary seal (2119) discovered nearby (Parpola 1975: 185); in 1984, a clay impression of another seal with the same sequence was found (Franke, in press). In addition to these types, there are a few miscellaneous objects bearing inscriptions, all rather small. With the possible exception of one broken stone slab (3599), there are no monumental royal edicts. Economic accounts almost certainly existed once, but they must have been written on perishable materials such as palm leaves, the traditional writing material of India. No bi- or multi-lingual texts have been found, either. Origins of the Indus script Recent archaeological researches in eastern Iran, southern Turkmenia and Afghanistan have been filling in that great gap which has long separated the Greater Indus valley from the Near East. It is now known that the Early Harappan cultures maintained overland contacts with Afghanistan, southern Turkmenia and northeastern Iran (Mughal! 1970; Kohl 1981). The coastal traffic between the Indus area and Western Asia represents a later development, undoubtedly connected with Harappan urbanization. The excavations at Tepe Yahya and Shahr-i Sokhta have shed much light on cultural interactions on the Iranian plateau and the dominant role played in them by the enormously extensive Proto-Elamite culture (conventionally dated to c. 3200-2900 B.C., though calibrated radiocarbon determinations suggest some 500 years earlier) (e.g. Tosi 1979). The neolithic societies on the threshold of urbanization had created communication methods that served their still relatively modest needs. It is legitimate, I think, to consider the 'potter's marks' on neolithic ceramics as important forerunners of real writing. In Susiana, in western Iran, pot-marks are found already in the 5th millennium, thus clearly antedating the Proto-Elamite writing; it is curious, though, that these potmarks do not continue into the 4th millennium, being thus separated from the ProtoElamite script by a considerable span of time (Dollfus and Encreve 1982). Comparable though much later 'potter's marks' are known from Tepe Yahya in southeastern Iran (Potts 1981), the Lut desert ofahe Iranian plateau (M. Ali Hakemi cited in Mughal 1970: 336), and Shahr-i Sokhta in Seistan. Also relevant for the study of the Early Harappan
marks (Potts 1982) are the incised marks on the terracotta cult statuettes in the bronze age culture of southern Turkmenia (Masson and Sarianidi 1969; Antonova 1972). Thousands of Early Harappan potsherds with 'potter's marks' (Fig. 2) have been discovered, usually on utilitarian unpainted ware. The marks were incised on the wet clay before firing by means of a sharp instrument (at Mehrgarh 57.40 per cent) or with a finger-nail (28.60 per cent) or both (14 per cent) (Quivron 1980: 274; Fairservis 1956: 328); marks painted before firing are much rarer. Graffiti scratched on pottery after firing also occur. Principal Early Harappan sites that have yielded pot-marks are detailed in the following. In their dating I have followed the chart of de Cardi (1984: 63), which places Mehrgarh III and KGM III between the Ubaid 4 and Uruk periods, c. 3700-3500 B.C. These are conventional dates: Mehrgarh III and KBM III are dated by Jarrige (1984: 22) to c. 4500-4000 B.C.; he cites an uncalibrated C14 date of 4745±90 b.c. (half-life 5730 years). In Baluchistan, Mehrgarh IV-VII (c. 3500-2500 B.C.) has a total of about 850 marks, of which 150 to 50 are different (Quivron 1980); Damb Sadaat I-III and other sites in the Quetta valley (including KGM IV, c. 3600-2600 B.C.) have yielded 362, with about 50 different marks, of which 18 were made with the finger-nail (Fairservis 1956; 328-35 and PI. 14); and from Periano Ghundai in the Zhob valley come comparable surface finds (Fairservis 1959: 359 and Fig. 59). Such marks occur also in southeastern Afghanistan, in Mundigak I (4th mill.) (Casal 1961: II, Fig. 49: 13) and period IV, dating from the first half of the 3rd millennium [IV.l (Fig. 87: 366-71), IV.2 (Fig. 93: 421-25), IV.3 (Fig. 105: 511-21)], and at Balakot (Early Harappan period A) in the Las Bela District of southern Baluchistan, probably a seaport in Harappan times (Dales 1974: 16 ff.). Near Balakot in southwestern Sind, incised 'potters' marks' have been found at Amri from periods I A to IIB, dated to c. 3500-2700 B.C. [Casal 1964, II: I A (Fig. 46: 63-67), I B (Fig. 48: 83, 49: 88, 51: 111, 53: 132-36), I C (Fig. 56: 152, 61: 193-98), I D (Fig. 66: 248-49), II A (Fig. 68: 268-70), II B (Fig. 72: 305-06)]; a few are known from nearby Kot Diji (Khan 1965: PI. XXIV: 1, 2, 6). The excavations at Rahman Dheri in the western plains of the northern Indus valley have not yet been properly published; calibrated radiocarbon samples of this large site date it to c. 3300-2500 B.C. 'Very large numbers of graffiti occur on the pottery of all periods at Rahman Dheri' (Allchin and Allchin 1982: 150-1). Sarai Khola II near Taxila has also yielded 68 sherds of Early Harappan ceramics with either pre-firing incised 'potter's marks' or post-firing graffiti (Halim 1972; 95-99 with table 11, Fig. 18: 101, 106; 19: 111, 115-120; 22: 134, 140; 23: 148; pi. XX B.I, XXII A-B). A few pot-marks are known also from Early Harappan Jalipur II in the Punjab (PA 4/1967: 8 f., Fig. 2.5; Mughal 1974: 110, A). A great number of further sites with Early Harappan pot-marks are likely to be added when the detailed results of recent explorations and excavations, notably in northern Baluchistan (Mughal 1972), in Bahawalpur (Mughal 1981) and in India (e.g., Kalibangan) are fully published. Many of the 'potter's marks' seem to be numbers, at least the repeated strokes which evidently stand for ones. The same system of marking numerals is used in the Indus script, and a large part of the graffiti on potsherds of the Indus civilization itself consist of nothing but such numbers. Did the (usually repeated) semicircles, which in the Indus script represent larger numbers (tens?), develop from the (often repeated) finger-nail impressions of the 'potter's marks'? (The Egyptian script has a similar semicircle for 'ten', but it developed from the picture of 'hobble for cattle', the corresponding Egyptian word being a homophone of the word for 'ten'; Gardiner 1957: 524 no. V19-20.) Similarities in relatively simple linear signs cannot be given too much weight. For example, the signs of the Vinc,a culture in Southeast Europe (Winn 1973: 196-204) offer parallels to the Indus script that are at least as close as those found among the Early Harappan pot-marks (Potts 1982: 517 Fig. 4), and, on the whole, signs really characteristic of the Indus script are not present among the latter. However, there are a few interesting exceptions that deserve special attention. One (unfortunately slightly broken) Early Harappan graffito from Balakot (Fig. 2: 5) [and possibly another from Rahman Dheri (Durrani 1981: PI. xix)] appears to be identical with the most common and most characteristic Indus sign 1 J" , usually found at the end of an inscription. Another Indus sign characteristically taking this position, the 'spear-head' T , is also attested in a slightly broken graffito from Balakot (Fig. 2: 2). However, in view of its absence elsewhere in the relatively abundant material, it is strange that these Balakot graffiti come from the lower levels (IV and I respectively) of the Early Harappan occupation (subdivided into 11 phases, I-IX corresponding to Nal polychrome andTogau C, and X-XI to Amri I C-D, Dales 1974: 11, i.e., roughly, to the middle and end of the 4th millennium in de Cardi's 'conventional' dating). The common Indus sign of 'two intersecting circles', (JP , often occurring in the earliest tablets (Koskenniemi and Parpola 1982: 182-4), is known as a painted mark from Gumla (Dani 1970-71: PI. 83), Sarai Khola II (Halim 1972: PI. XX B:l) and Rahman Dheri (Durrani 1981: PI. xix). This survey shows that 'potter's marks' existed throughout a vast area in the Greater Indus valley many centuries if not a millennium before the creation of the Indus script. Conditions for a relatively independent creation exist, and at least some of these earlier local symbols seem to have been adopted as graphemes into the Indus script. Moreover, the Indus script cannot be directly derived from any other known script, not even from the vaguely similar Proto-Elamite, which is closest to it. On present evidence, the similarities between the Indus script and Proto-Elamite are not sufficiently complex and extensive to warrant a conclusion of genetic relationship. Perhaps this question can be definitely solved when clear examples of the early formative stage of the Indus script become known, but it seems most likely that the Indus script was created autonomously and within a short period of time. The process of urbanization created the need for a writing system for effective communication and record keeping in administration and trade. This internal requirement was in all likelihood combined with the stimulus of idea diffusion from the west. The.strong influence of Proto-Elamite art on Harappan seal glyptics (Parpola 1984) supports the natural and often voiced view (inter alia, Potts 1982) that the indirect model was the geographically and temporally adjacent and vaguely similar Proto-Elamite script. At the moment, the Indus script appears fully developed and almost entirely standardized already in the earliest available examples, the 'miniature tablets' of the lowest levels of Harappa dating perhaps from the 26th century B.C. [on the basis of MASCA calibrated C14 dates, B. and R. Allchin (1982: 218 f.) place the beginning of the Mature Harappan period at c. 2550 B.C.]. Only a few signs differ slightly from their later forms. The direction of writing, too, is not yet standardized, while later there are very few exceptions to the rule, which is right to left. Characteristically, these earliest examples of the Indus script bear short inscriptions, with large numbers of duplicates. There are cogent reasons for thinking that they are associated with religious worship. One of the usually two inscribed sides normally bears only two signs: a numeral (1, which may be omitted, 2, 3 or 4) plus a U-shaped sign. The U-sign clearly represents a sacrificial vessel: in some cases the accompanying scene shows a kneeling worshipper extending a vessel of this shape towards a sacred tree or an anthropomorphic deity sitting on a throne. (Parpola 1981; Knorozov et al. 1981: 44, 106; Hrozny 1942: 36) The recent suggestion that these 'miniature tablets' are 'accounting tablets' comparable with those of Proto-Elamite Susa (Shendge 1985) is inadmissible. Difficulties and pitfalls of decipherment Several respected experts (e.g. Friedrich 1966: 145) have denied every possibility of deciphering the Indus script. Such pessimism is understandable, for none of the crucial keys that have opened other unknown scripts is available. We have no Indus inscriptions with parallel versions in some other readable script; nor do we, in the absence of any political history, know the names of the Harappan gods, kings, cities or villages. Even the affinity of the Harappan language and the type of writing system represented by the Indus script are much debated. The texts are short and limited in nature, and there are no clearly recognizable word dividers such as have been very helpful in the analysis of many other scripts. Considering these formidable difficulties it is hardly surprising that the decipherment of the Indus script remains a vexed problem. Many of the proposed solutions can be dismissed immediately as uncritical and/or uninformed in some field of knowledge vital for a successful decipherment, such as the study of writing systems and the methods of their decipherment; or archaeology, linguistics and philology, especially of the Indian subcontinent and Western Asia. For example, when Sir Alexander Cunningham studied the first Indus seal more than a century ago, he knew nothing about its antiquity; he was reasonable, but uninformed, in assuming that it might represent an early form of the Brahmi script, dating from the times of the Buddha. It is now common knowledge that Brahmi (first attested in Asoka's edicts c. 250 B.C.) is derived from the Semitic consonantal alphabet, which in turn is derived from Egyptian hieroglyphs. And yet some researchers still insist on deriving the Brahmi and with it the modern Indian scripts from the Indus script, basing further hypotheses on this premise. It seems that the Indus script was forgotten after the collapse of the urban civilization which created it - a situation comparable with most of postMycenaean Greece. The latest certain attestation of the Indus script is from Daimabad, dating from c. 1800-1600 B.C. (Sali 1982). Some of the later pottery graffiti in the Deccan may be interpreted as reminiscences (Lai 1960) - but these are no longer real writing. One mistaken method is so common that it deserves special mention. The Indus signs have been equated with similar-looking symbols in other scripts and read with their phonetic values. (An even cruder 'method' consists of interpreting the Indus signs pictorially and assigning them syllabic values from the first syllables of the corresponding words of an arbitrarily chosen language.) A comparison of signs that look alike in different pictographic (non-alphabetic) scripts will demonstrate, however, that the phonetic values are different, since the different scripts for the most part reflect different languages (Parpola 1979: 171). It is possible to advance knowledge only by making hypotheses, which, to be fruitful,must be informed guesses grounded on acceptable knowledge. Wrong guesses are unavoidable, but the scientific method tries to detect and rectify them by demanding that all hypotheses be subjected to adequate tests before they can be accepted as valid. The problem of the Indus script can be compared to a crossword puzzle, where individual solutions have to conform to each other as well as to certain specified requirements. Every hypothesis must be rigorously examined. With such an insistance on quality, every step forward should increase the means of testing. It is symptomatic of the uncritical and erroneous attempts at deciphering the Indus script that any possibility of internal control of the proposed readings is invalidated: the liberty is taken of positing one and the same value for many signs, and of allowing one and the same sign to have many different values (an extreme example is Hrozny 1942). As a result, all the texts can be 'read' yet the solutions usually fail to make good sense. The capacity to explain well external evidence is the other touchstone. The type of writing system represented by the Indus script What is the nature of the Indus script? This is one of the crucial questions which must be solved if there is ever to be a real decipherment. The theory of writing and decipherment provides us with several criteria. The number of distinct signs (graphemes) is one touchstone. Widely different estimates have been given, ranging from about 400, which is the usual figure, to just 52 (Rao 1979: I, 170). The reason for such a variation is twofold. One and the same sign may appear in several variant forms (allographs). Moreover, many signs are clearly combinations of one or more elements, and it can be difficult to decide whether a sign is a composite one or not. Even so, it is possible to state that there are at least 200 basic signs. This already precludes the possibility of an alphabetic script; even for a syllabic script this number is too high. 'Word' length is another criterion. Many texts consist of just one sign, which therefore necessarily must represent a complete 'word'. Moreover, most Indus texts can be divided into units of one to three signs by means of internal comparison, which can be carried out by computer methods (Fig. 3). It seems, indeed, that the Indus script represents the same kind of'nuclear' writing as the early Sumerian texts of Fara (26th century B.C.), in which only the lexical meaning of each word is expressed, while all or most grammatical affixes are omitted and left to be understood from the context (Civil and Biggs 1966). These considerations, based on the number of graphemes and the 'word' length, are in agreement with the third criterion, time. Harappans of the mature Indus civilization seem to have been in direct sea contact with Mesopotamia in the 24th century B.C. The urbanization process - and the creation of the Indus script - is likely to have taken place at least by the 26th century, if not earlier. At this time the only existing type of writing was a logo-syllabic or morphemographic type of script, in which each sign stands for a word or a morpheme of one or more syllables. The oldest known essentially syllabic script dates from around 2350 B.C. (this is Linear Elamite, assumed to have comprized about 80 signs in all, 58 only being attested in the 19 texts, cf. Hinz 1964: 27, 1975: 106 f.; also Vallat, this volume, 343), while alphabetic writing was created by the Semitic peoples only around 1600 B.C.
It seems indeed more than likely that the Indus script belongs to the earliest type of real writing and is a morphemographic script. The primary means of representing a concept in a script of this type is to draw a picture of it. But abstract concepts and grammatical elements are difficult if not impossible to represent pictorially, if another person is to recognize the intended meaning. For this reason, and to reduce the number of graphemes, recourse was taken to the rebus principle, the method of expressing the phonetic shape of the intended word or morpheme through another, similarly sounding word which was easier to depict. All people have word plays based on homonymy as an integral part of their folklore, and punning must have been applied in communication long before the creation of real writing. In any case, the early Sumerian and Egyptian scripts, which can be read, make much use of the rebus principle, and this must have been so in the Indus script also. This means that, depending on the context, a pictographic sign could denote the object represented by it, or it could denote any of the words or morphemes of the language underlying the script that were (within certain strict limits) phonetically similar to the word naming the depicted object. Basic methodology of decipherment On the premise that the Indus script represents morphemographic (logo-syllabic) writing, it is possible to devise a basic methodology for its decipherment. Homonymies are language-specific: puns cannot easily be translated into unrelated languages. Therefore, the use of the rebus principle provides the possibility of identifying the underlying language and deciphering individual signs. This requires, however, that four conditions are simultaneously fulfilled:
1 the object depicted by a given pictogram can be recognized; 2 the said pictogram has in a given context been used phonetically, i.e. in a meaning different from the pictorial meaning; 3 the intended (phonetically used) meaning can be deduced from the context; and 4 a linguistically satisfactory homonym fitting the posited meaning can be found in a likely language. It is inherent in the nature of a morphemographic script that it cannot be deciphered all at once, as can alphabetic and syllabic scripts. In alphabetic and syllabic scripts each sign (as a rule) has only one value, and these values are all phonetic and interrelated, forming a closed system. Once the phonetic value of a few signs has been correctly determined, the contexts will suggest the values of the remaining signs. By contrast, a morphemographic script is an open system, the signs may have several values, and they represent larger linguistic units whose interrelations are not so clearly circumscribed. In other words, a deciphered morphemogram will provide clues for other signs, but the signs will have to be deciphered one by one. For the Indus script, the pictorial meaning of a sign can be determined by various means. One is a comparative study of all its variant forms (allographs). Harappan art and culture generally can provide clues; thus some decorative motifs on early Harappan pottery seem to have been models for Indus pictograms (Parpola 1975: 199 f.). Similarlooking signs from other pictographic scripts provide one of the most useful external parallels. Unfortunately, most of the Indus signs are so simplified and stylized that their pictorial meaning - and therewith their decipherment - is likely forever to remain problematic and controversial. Those few signs whose pictorial form is identifiable are, therefore, crucial for the decipherment of the Indus script. The contextual meaning of the signs can be established in two ways. One is an internal analysis of the inscriptions by linguistic and statistical methods, in which automatic data processing is of great assistance (Koskenniemi 1981; and Fig. 3). This process must be distinguished from the use of external contexts, which include the pictorial form of the sign, and such things as the iconographic motifs that may accompany the text, the nature of the object on which it has been inscribed, and its archaeological find place. An especially precious clue is provided by the nature of most Indus inscriptions: they were engraved on seals used in administration and trade. Because Harappan seals have been found in the Near East, their inscriptions are likely largely to parallel the Mesopotamian seal inscriptions, which can be read: the latter contain mainly proper names and official titles, with or without attributes and dedicatory formulas (Gelb 1977). The affinity of the language underlying the Indus script The question of the actual language is, besides that of the nature of the script, the other great unknown to be solved, and without doubt the most interesting and debated issue. It cannot be decided without taking into account three major factors: * 1 the archaeological and culture-historical context,2 the findings of historical linguistics, especially with regard to the Indian subcontinent, and 3 the typological structure of the Indus language, established by the application of linguistic and statistical methods. Archaeological research has proved beyond doubt that the Indus civilization is not a Sumerian colony, and that the Sumerians and the Harappans do not share a common descent. Besides, the Sumerians had developed a distinctive script of their own long before the Harappans. Moreover, the sign sequences of the Indus script are uniform throughout the area of the Indus civilization and during its entire duration, thus indicating the use of just one language in the script, but Indus seals found in the Near East have totally different sign sequences. This is true especially of those Indus seals from Western Asia whose round or cylindrical form accords with the local seal forms. These seals bear common Indus signs, but the order is unique: the only conclusion seems to be that they record Sumerian or Semitic names of acculturated Indus merchants. According to this evidence, the languages they had adopted in the Near East differed radically from the native Harappan language. (S. Parpola et al. 1977; Brunswig et al. 1983) A Sumerian solution (Kinnier Wilson 1974), therefore, appears unlikely. The area of the Indus civilization has been inhabited by speakers of Aryan languages since at least 1000 B.C. The hypothesis that an early form of Indo-Aryan was spoken already by the Harappans has its supporters. This brings us to the vexed issue of the arrival of the Aryan speakers in India. Archaeology in India, the Indo-Iranian borderlands, Central Asia and elsewhere, has accumulated a wealth of new data on this question, and some tentative statements can already be made. There is no escape from the fact that the horse played a central role in the Vedic and Iranian cultures, as it undoubtedly did already in the culture of the people speaking Proto-Indo-European. The horse appears to have been domesticated in the South Russian steppes c. 5000 B.C. (inter alia, Piggott 1974: 17), and this region is focal in attempts to trace the dispersal of the Proto-Indo-European-speaking community archaeologically (Thomas 1982). In the Near East horse-drawn light war-chariots were an important asset of the Aryan-speaking nobles, who ruled Mitanni in Northern Mesopotamia c. 1500-1300 B.C. (for divergent views, see Mayrhofer 1974; Diakonoff 1972). It is generally believed that the earliest Indo-Aryan texts, the hymns of the Rgveda, were composed c. 1300-1000 B.C. in the area bounded by Afghanistan and the Punjab, and the later Vedic texts c. 1000-500 B.C. in the 'middle region' from the Punjab to the upper Ganges valley (Gonda 1975: 22 f, 360 f.). In regard to geographical and chronological span as well as material content, the later Vedic culture closely corresponds with the archaeological culture characterized by the use of Painted Grey Ware (Tripathi 1976). Many scholars have assumed that the Indus cities were meant in Rgvedic references to the forts of the Dasas and Dasyus, the dark-skinned enemies vanquished by the Aryans. Apart from the chronological gap, this is unlikely for several reasons. The form of the Dasa citadels, three concentric circles, differs from that of the usually square Harappan cities but exactly parallels the cultic centre with three concentric fortification walls at Dashly-3 in North Afghanistan (Parpola 1985: 76-78). The Dashly-3 citadel has been compared to the Old Iranian var and the cult places of the 'Kafirs' (Jettmar 1981). The Dashly-3 culture is related to the Northeast Iranian Gray Ware complex of the third and second millennia B.C. (represented by Tepe Hissar II-HI), which had spread to the borders of the Indus valley already in the Mature Harappan period (Francfort 1984; Santoni 1984). The people who ruled the North Iranian Gray Ware culture used horsedrawn war-chariots (Ghirshman 1977: 14 ff.). It has long been suspected that they represent an early wave of Aryan speakers (Thomas 1982: 64-7). This hypothesis is in agreement with the Aryan etymology of the ethnic name Dasa, and the merger of different linguistic and religious traditions of Indo-European origin in the early Vedic period (Parpola 1983). A major reason against assuming that the Harappans spoke an Indo-European language is that the horse is not represented among the many realistically depicted animals of the Harappan seals and figurines. Nor is there clear osteological evidence of the horse {Equus caballus) in the Indian subcontinent prior to c. 2000 B.C. (Jarrige and Santoni 1979:1,404 f.). In the 2nd millennium, however, a number of clay models of the horse and the Bactrian camel (but not of the bull which is prominent in Early Harappan and Indus art) are found at Pirak, a site on the border of Sind and Baluchistan (ibid.). The population of the Indus civilization is estimated at c. 5 million (McEvedy and Jones 1978: 182). The number of the invading Aryans can have been only a fraction of this. It is clear that the Harappans were not all killed, and that their language could not have disappeared overnight: they were subdued and linguistically Aryanized in the course of many centuries. Such a process involves bilingualism and leaves evidence of the disappearing substratum language in the structure and vocabulary of the surviving language. It is now almost unanimously accepted that there are clear traces of Dravidian substratum influence even in the earliest Vedic texts (Emeneau 1980: 85 ff., 167 ff.). Moreover, a Dravidian language, Brahui, is still spoken in the mountains of Baluchistan and Afghanistan, a core area of the 'Early Harappan' culture. Shared innovations prove that Brahui forms a North Dravidian subgroup together with Kurux and Malto, languages spoken in the hills of east central India and Nepal (ibid.: 314 ff.). After Aryan, Dravidian, now mainly spoken in South India, is the second largest linguistic family in the Indian subcontinent. Earlier this century it was spoken by one quarter of the total population; at present its speakers number some 130 million. In comparison, the other Pre-Aryan languages are little spoken and their geographical distribution and external relatives fall outside the Harappan cultural sphere: the TibetoBurman languages of the Himalayas and the Northeast, distantly related to Chinese; the Austro-Asiatic languages (Santali, Munda, etc.) spoken in central and eastern India and related to the languages of Southeast Asia; and the few isolated tongues, Burushaski high in the Hindu Kush, Nahali in central India, and Vedda and Rodiya in Sri Lanka (Parpola 1974; Zograph 1982). An impressionistic analysis of the Indus inscriptions suggests that structurally the Harappan language is of a fairly pure 'left-branching' and agglutinative type. Dravidian is the only match in the recorded linguistic history of the Indian subcontinent (Knorozov et al. 1981: 13 ff.; Koskenniemi and Parpola 1982: 11 f.). However, it must be emphasized that a proper linguistic analysis of the Indus texts has yet to be carried out
The present state of deciphermentOn the basis of external and internal criteria, then, it seems most likely that the Harappans spoke a Dravidian language. Can this be substantiated? By applying the methodology outlined above, some Indus signs can be read in Dravidian. The fish-like pictograms have a great frequency, and they occur in contexts suggesting that they stand for names of deities. The Indus seals are likely to contain mainly proper names and occupational titles, like those in Mesopotamia, and divine names have formed an important element of proper names and priestly titles in Mesopotamia as well as later in India. The pictorial interpretation of the basic sign ty> as 'fish' is verified in Harappan art, and the homonymy between the Dravidian words mln 'fish' and min 'star' makes an astral interpretation possible. The two words can safely be reconstructed for Proto-Dravidian, and are both probably derived from the verbal root min 'to glitter'. That such a homonymy existed in the Harappan language as well is suggested by the Harappan painted pottery combining star and fish motifs (Fig. 4)
In Mesopotamia, some divinities were represented by particular heavenly bodies, and in the cuneiform texts the determinative of divine names is a sign depicting a star (see page 330). That the Harappan gods also had an astral aspect is shown by an Indus seal, in which a star is placed on either side in the loops of the horned headdress worn by an anthropomorphic deity (Plate 2 and Parpola 1984). Moreover, the trefoils of the famous 'priest-king' statue from Mohenjo-Daro are assumed to have an astral meaning (Marshall 1931, 356 f.); for this there is now new evidence (Parpola 1985). It is possible to crosscheck this interpretation of the 'fish' pictogram by studying other signs which have been combined with it in the Indus script. Numbers preceding the fish sign give such readings as, for example, '(constellation consisting of) six stars', which are actually attested as star names in the most ancient Dravidian texts, the Old Tamil poems of the first centuries of the Christian era: aru-mln 'six-star', for example, denotes the Pleiades. Unfortunately many of the 'diacritical' signs that often (but never after numerals) modify the basic fish signs are open to various interpretations
One such sign, however, seems fairly unambiguous: placed above the fish sign, it looks like a 'roof: -^ (Plate 1). Through the homophony between Proto-Dravidian *vey/mey 'to cover (as a roof), to roof, thatched roof and *may 'black' (with the alternation of vand m- at the beginning of the word, and of a and e before -y, both of which are assumed to have existed already in Proto-Dravidian), we get the compound mai-m-min 'black star', which in Old Tamil denotes the dimly visible planet Saturn. In Sanskrit sources, too, the planet Saturn is invariably associated with the colour black. Its presiding deity is Yama, the god of death, who is usually depicted riding on a dark buffalo. Several iconographic texts mention the buffalo also as the vehicle of the deified planet Saturn. Interestingly, the mount of the planet Saturn (usually called sani or sanais-cara, 'slowly moving', in Sanskrit) in the Buddhist and Jaina sources is the tortoise, which is not only a slow creature but also a kind of 'fish' (aquatic animal) that carries a 'roof over itself! Some further interlocking interpretations of this kind are available. External support to such astral interpretations is given, inter alia, by the importance of astrological considerations in Indian name giving and the fact that, according to astronomical evidence, the Vedic star calendar was created c. 2300 B.C. This dating coincides with the peak of the Indus civilisation, which, like other early urbanized cultures, needed some form of calendar. The practice of astronomy and its importance for religion is shown by the orientation of Harappan cities according to the cardinal directions (Parpola 1975). There is, however, a limit to this approach. The highly simplified form of most of the signs makes their pictorial interpretation difficult and thus increases the element of uncertainty. It is to be frankly admitted that the prospects of ever fully understanding the Indus script are meagre. Moreover, it must be added that the validity of the above interpretations, developed from those presented by a Finnish research team more than 16 years ago, is an open question. It is symptomatic of the present state of decipherment that, although several scholars in recent years have been working on the Indus script with largely similar methodology and basic assumptions, including the Dravidian hypothesis, there is, as yet, little agreement on individual interpretations. Notable exceptions are the readings of the 'fish' sign and its combinations with the numeral signs: it was first suggested by Henry Heras (1953: 127 f.) that '6 + fish' stands for the Dravidian name of the Pleiades, and, along with other similar occurrences ('7 + fish', '3 + fish'), this interpretation is included in the extensive solution proposed by Soviet scholars (Knorozov et al. 1981: 93). These readings are also accepted by Mahadevan (1970: 179 ff., 255 f.), but not by Fairservis, who disagrees with the pictorial identification of the 'fish' sign (1984: 156, 158). I think this unusual agreement has to be taken seriously. The reason why few people have yet done so seems to be that really significant interpretations have been obscured among many obviously implausible or at least highly subjective and insufficiently established readings, which fail to carry conviction. Is it likely, for example, that a seal text reads, 'The red (or great) glittering monsoon's power'? Although the published 'solutions' offer many valuable observations and seminal ideas, it is clear that a lot of washing has to be done in order to obtain the gold dust, and that a critical reader is inclined to throw that gold dust away with the wash water. As John Chadwick saw, the only remedy is to lay stress on quality instead of quantity: What we shall need is not more possible or even plausible interpretations of signs based upon this theory, but the clearest possible demonstration that these meanings, and only these meanings, are correct. Once this foundation is unshakeable, it will be safe to build upon it, but we must not be led into admiring a house of cards which, elegant as it seems, will collapse if one prop is withdrawn (Clauson and Chadwick 1969: 207).
Indus script corpora are metalwork catalogs in Meluhha (Mleccha) language. Rebus-metonymy layered cipher. Kalyan http://bharatkalyan97.blogspot.in/2015/05/decipherment-of-cholistan-and.html
ReplyDelete