Ancient Texts from Indus Civilization: Machine Learning to Decifer

1. Meeting the Challenge of the Indus Script

In 1920, excavations at Harappa brought to light the ruins of a large brick-laid city, and soon a whole unknown civilization was uncovered in and around the Indus valley. 1 The Indus civilization, now dated to c. 2600-1800 B.C., collapsed some 500 years before the composition of most of the hymns collected in the Rgveda-Samhita, the oldest historical document of India.
No unambiguous information has been preserved to tell us the names of the Indus kings or their subjects, the names of the gods worshiped by them, or even what language they spoke.

The Indus (or Harappan) people used a pictographic script. Some 3500 specimens of this script survive in stamp seals carved in stone, in moulded terracotta and faience amulets, in fragments of pottery, and in a few other categories of inscribed objects. In addition to the pictographic signs, the seals and amulets often contain iconographic motifs, mostly realistic pictures of animals apparently worshipped as sacred, and a few cultic scenes, including anthropomorphic deities and worshippers. This material is of key importance to the investigation of the Harappan language and religion, which continue to be among the most vexing problems of South Asian protohistory.

The Indus script is an unknown writing system, and the inscriptions discovered are very short, comprising no more than five signs on the average. With good reasons, the prospects of a successful decipherment have been considered meagre at best. But great problems wield a power of attraction, and many attempts have been made to crack that posed by the Indus script. Most of the tries, however, have been quite uncritical and poorly informed.

Generally speaking, the standard of earlier research was rather low in 1964, when three young Finnish students entered the field. This encouraged us to believe that something worthwhile might be achieved in the study of the Indus script, especially as we could use the computer as an aid. After all, there were a fair number of texts. We were inspired by the phenomenal success of Michael Ventris’ systematic decipherment of the Mycenaean syllabic script, which at that time started being universally accepted. Our Finnish team has collaborated in producing several research tools, in developing a methodology for the decipherment of the script and in putting this methodology into practice. While the participants in the project have changed in part from time to time, this work is still going on. The present article reflects my own work chiefly, but its basic ideas stem from collective work some two decades ago.

2. What type of writing does the Indus script represent?

How, then, is it possible to decipher an unknown system of writing? Confronted with this primary question we are doubly fortunate in comparison to the decipherers of the Egyptian hieroglyphs more than 150 years ago. In the first place, we have a number of successful decipherments to look back to, both as potential models and as sources of inspiration, which reassure us in the indispensable belief that seemingly impossible feats can be achieved. Yet none of the earlier decipherments is directly comparable to the problem of the Indus script: most of them were based on a translation of a text in the unknown script into a known script and language, or at least the historical context provided crucial clues in the form of proper names. In the absence of such aids, we must look for a different approach.

Here we come to the second advantage we have over Champollion, whose decipherment was greatly hampered by his mistaken preconceptions about the nature of the hieroglyphs. Although the early history of writing is still lacunary, the great lines in the evolution of writing are now fairly well understood. This is important, because it enables us to get an at least approximate answer to one of the two fundamental questions about the Indus script, namely the type of writing represented. For in this case, as elsewhere in scientific research, a big and difficult problem becomes more manageable if it is broken into two or more smaller problems. We shall soon return to the other great unknown, the problem of the underlying language.

The human ability to analyze language and to represent it with written symbols has evolved gradually. In the first stage of “pre-writing”, pictures stood for whole sentences or narrations. The next step was to break sentences into separately written words, or rather, morphemes (that is, the smallest meaningful units) which include not only lexical roots but also grammatical markers. In the beginning it was almost exclusively the root morphemes carrying the lexical meaning that were marked in writing, while the grammatical affixes which the root happened to have in any given context were omitted in the script, being left to be guessed and supplied by the reader. This form of writing has been called “nuclear.”

Occasionally, however, some basic grammatical affixes were marked even in the earliest script, the archaic Sumerian. Gradually all the grammatical affixes became represented in writing. This complex “logo-syllabic” script demanded many hundreds of graphemes, each of which could have several different meanings. The next major step in the history of writing was the simplification of the system into a “syllabic” script, where only about one hundred signs, each with just one fixed phonetic value, could be used to write about 90% of the texts. The syllabic script was still somewhat clumsy, but eventually, with the successive emergence of the consonantal alphabet and the full alphabet, human speech could accurately and economically be mirrored in writing.

Several criteria taken together enable us to place the Indus script in its proper place within this evolutionary scheme (Fig. 2). There are about 450 different signs in the Indus script. Comparing this figure with the number of graphemes in the other early scripts, and also taking into regard the age of the Indus script, the conclusion of its belonging to the logo- syllabic type seems inevitable. It is likely that in most cases the Indus signs stand for nuclear words, i.e. the lexical roots. This is undoubtedly so in the case of texts comprising one single sign only. Important grammatical affixes may also have been marked. Their fundamentally different sign sequences suggest that some Indus texts discovered in West Asia contain local proper names, written in the Indus script with signs having a syllabic value.

3. The Principle of Picture Puzzles

The early logo-syllabic scripts functioned like picture puzzles, based on the rebus principle. Each sign was originally a picture, denoting the object represented by it. For example, the picture of an arrow meant ‘arrow.’ But because many things, especially abstract concepts, are not easy to draw as pictures in a simple and unambiguous manner, another expedient was resorted to. The meaning of the pictograms was extended from the name of the depicted object to all of its homonyms, that is, words with a similar phonetic shape but a different meaning.

For example, in the Sumerian script the picture of arrow meant ‘arrow’ as well as ‘life’ and ‘rib,’ because all three words were pronounced alike in the Sumerian language: ti . Homophonies like this have often given rise to etiological myths. Thus the identity of the Sumerian words for ‘life’ and ‘rib’ is behind that part of the Sumerian paradise myth in which the goddess Nin-ti ‘Mistress of Life’ heals the rib of the sick god Enki.

Puns are language specific, but etiological myths based on them can travel from one folk to another. If the Sumerian myth just mentioned was not known, it would be difficult to understand why Eve was created from Adam’s rib. The Hebrew name of Eve, Hawwa , means ‘life’ (cf. Genesis 3:20, “And Adam called his wife’s name Eve; because she was the mother of all living”), but it has no resemblance to the Hebrew word for ‘rib’, sela. On the other hand, we know from the Bible that the Virgin Mary did not become pregnant from a berry eaten in the forest. Undoubtedly, this peculiar version of the Biblical story in the ancient Finnish folklore owes its origin to the homonym marja ‘berry’ that the Finnish language has for the proper name Maria .

4. Pictorial Interpretation of the Indus signs

In most of the unsuccessful attempts at interpreting the Indus script, the “method” has consisted of comparing other pictographic scripts and supposing that the Indus signs have been pronounced like the similar- looking foreign signs. However, as a rule, different pictographic scripts have been used to write different languages, and their similar-looking signs have denoted words that have little in common in pronunciation (Fig. 3).

But if one abstains from unlikely conclusions regarding the phonetic value of Indus signs, comparisons of this kind are by no means useless. Irrespective of whether there is any genetic connection or not, similar- looking signs of other pictographic scripts provide valuable clues to the pictorial meaning of the Indus signs (Fig. 3).

In this respect still more important hints are given by the archaeological remains of the Indus civilization and its antecedent neolithic cultures, including such things as architecture, tools and weapons, and art. Artistic expression on painted pottery (Fig. 4) and in the iconographic motifs of the seals and amulets are of particular significance, since they let us see directly how the Indus people represented various objects.
Why all this concern about finding the pictorial meaning of the Indus pictograms? Because their pictorial shapes constitute one of the chief keys to their interpretation. If a pictogram has been used in its primary, pictorial meaning, its meaning can be understood directly from the iconic2 sign itself, irrespective of how the corresponding word was pronounced in the language underlying the Indus script.

Unfortunately, the pictorial meaning of most Indus signs is difficult to recognize. As in many other scripts, the demand for fluency in writing has led to a radical simplification of their shape. The characteristic elements have been reduced to the barest essential. Fortunately this tendency has not gone as far as in the Sumerian script, where the pictorial shapes of the earliest phase soon gave way to abstract cuneiform symbols: if we did not have an unbroken chain of successive forms, it would be impossible to reconstruct their pictorial prototypes.

But however important insights into the contents of the documents they might give, purely pictorial interpretations are hard to check and cannot alone constitute a decipherment of the script. By definition, this term implies a plausible identification not only of how the script functions but also of the underlying language, together with actual sign interpretations that are sufficiently cross-checked to carry conviction.

5. Do the ‘fish’ signs denote dieties?

Consideration of the picture puzzle principle employed by the early scripts suggests a way to recognize the identity of the underlying language. We should try to find a specific context where a given, pictorially identifiable sign apparently has been used in a meaning different from the primary pictorial meaning, and more particularly a context that with fair accuracy enables us to determine what that intended phonetic meaning was. In the language underlying the script, the pictorial meaning and the intended meaning were expressed with words pronounced alike.

The function of the object on which a given Indus inscription occurs can be an important clue to the intended meaning of the signs. The Indus texts are mostly inscriptions on carved seal stones intended for stamping clay seals. More than a hundred clay tags with ancient seal impressions come from a burnt-down grain warehouse at the Harappan port town of Lothal. Many of these tags also bear impressions of woven cloth, reed matting or other packing material. This shows that the tags were once attached to bales of goods, and that the seals were used, as in ancient West Asia, for controlling economic transactions.

Indus seals coming from West Asian sites testify to trade relations entertained by the Indus civilisation with Mesopotamia. It seems most likely, therefore, that the contents of the Indus seal inscriptions are analogical to the contemporaneous Mesopotamian ones, which can be read and understood. These latter seals chiefly contain proper names, with or without attributes of various kinds, and titles of the priestly elite. Proper names were mostly theophoric, i.e. they contained names of divinities as their components. In priestly titles, too, the deity served by the priest is usually specified. In classical Indian civilization, too, proper names of human individuals usually mention divinities.

Names of divinities, therefore, can reasonably be expected to occur in the inscriptions of the Indus seals in great numbers and in definite positions suiting both proper names and priestly titles. A large group of signs meeting these conditions has the basic appearance of ‘fish’. This pictorial interpretation can hardly be questioned, for in the iconography of the Indus amulets a similarly drawn fish is shown in the mouth of the fish-eating gavial (Fig. 5, 6).

If we assume the pictorial meaning to be ‘fish’ and the intended meaning to be ‘god’, are these meanings linked by a homonymy in any historically plausible language? Before replying to this question we must first consider the language problem.

6. Sanskrit or Dravidian?

Many hypotheses have been put forward about the affinity of the Indus language, but only two alternatives have had wider support.

Indo-Aryan languages have been spoken in the area once occupied by the Indus civilisation and gradually all over North India since at least 1000 B.C. It is natural to assume that they were spoken there even earlier. Speakers of Hindi, Bengali and other Neo-Indo-Aryan languages especially have been prone to interpret the Indus texts as Sanskrit (understood in the broad sense of Old Indo-Aryan), from which their own mother tongues have evolved.

The Sanskrit hypothesis, however, is difficult to reconcile chronologically with the date of the Indus civilisation (about the second half of the third millennium B.C.) and antecedent Early Harappan neolithic cultures which were responsible for its creation. Comparison of the Vedic texts with the Avesta and with the West Asian documents relating to the Aryan kings of Mitanni suggests that the Vedic Aryans entered the Indian subcontinent from Northeast Iran and Central Asia in the second millennium B.C.

Moreover, it is abundantly clear that the early Aryans were nomads and that the horse played a dominent role in their culture, as it did in the culture of their Proto-Indo-European-speaking ancestors. The horse is conspicuously absent from the many realistic representations of animals in the art of the Indus civilisation. Comprehensive recent bone analyses have yielded the conclusion that the horse was introduced to the subcontinent around the beginning of the second millennium B.C.

Horse-drawn chariots made the Aryan-speaking nomads a superior military force which gradually subdued all of North India. Numerically the early Aryans can have been only a fraction of the Indus population, which is estimated to have been about five million. Obviously these millions of people were not all killed; they were made to acknowledge the Aryan overlordship and to pay taxes. In the course of time and through gradually increasing bilingualism, the earlier population eventually became linguistically assimilated. It is most unlikely that this process of linguistic Aryanization happened without leaving clear marks of the earlier substratum language upon Indo-Aryan.

There are several structural and lexical Dravidisms even in the Rgveda, the earliest preserved text collection, pointing to the presence of Dravidian speakers in Northwest India in the second millennium B.C. The 25 Dravidian languages spoken at present form the second largest linguistic family of South Asia. Until recently, about one quarter of the entire population has spoken Dravidian, while the speakers of Austro-Asiatic, the third largest linguistic family of long standing in South Asia, numbered just a few per cent. The Indus language is likely to have belonged to the North Dravidian sub-branch represented today by the Brahui, spoken in the mountain valleys and plateaus of Afghanistan and Baluchistan, the core area of the Early Harappan neolithic cultures, and by the Kurukh spoken in North India from Nepal and Madhya Pradesh to Orissa, Bengal and Assam.

7. Fishes and Stars: evidence for astral divinities

An early form of Dravidian, then, emerges as the historically most likely language to have been spoken by the Indus people. The uniformity of the sign sequences in Indus inscriptions coming from all parts of the large area occupied by the Indus civilisation precludes the possibility that widely different languages were used, at least as far as the literate elite is concerned. With this conclusion, we may now return to interpreting the ‘fish’ pictograms discussed earlier.
In most Dravidian languages the usual word for ‘fish’ is meen. This phonetic shape can also be reconstructed for the mother language, Proto- Dravidian. A homonym meen denoting ‘star’ has also existed in Proto- Dravidian. Both words refer to a glittering object, and appear to be derivatives from the Proto-Dravidian root meen ‘to glitter, to sparkle’.

The ‘fish’ pictograms of the Indus script, then, can be interpreted as denoting gods, if stars were used as symbols of deities. This is indeed most likely, for in the cuneiform script the pictogram of ‘star’ is prefixed to every divine name as a symbol of divinity. In the times of the Indus civilisation, the Mesopotamians associated their divinities with specific heavenly bodies. For example, Inanna-Ishtar, the goddess of love and war, was symbolized by the planet Venus. A similar system was later adopted by the Greeks and by the Romans, after whose deities the planets have their English names. In India, too, the planets have been worshipped as minor divinities for the past two millennia at least, and each of them has one of the principal gods of Hinduism as its overlord.
In the Near East, the ‘star’ symbol distinguished divinities not only in the script but also in pictorial representations (Fig. 6). Significantly, a seal from Mohenjo-daro depicts an Indus deity with a star on either side of his head in the Near Eastern fashion.

The interpretation of the ‘fish’ signs as symbols for astral divinities is further supported by different kinds of proofs for the practice of astronomy by the Indus priests. The straight streets of the Indus cities are oriented towards the cardinal directions, which presupposes astronomical observations and the use of the sun-stick, the gnomon.

The star-calendar used by the Vedic ritualists was adopted by the Aryans in India, for there are no references to it in the Avesta or in the oldest books of the Rgveda.
On the other hand, astronomical evidence dates the compilation of this calendar at around the 23rd century B.C., when the Indus civilisation flourished at its peak. Like other urban civilisations, it undoubtedly needed a calendar that adjusted the lunar and the solar time-reckoning.

It is in accordance with the traditions of Hindu name-giving as well to expect names of astral and planetary deities to be mentioned in seal texts likely to contain proper names. Astral names were given to Indian children as early as in the Vedic period, at least from 1000 B.C., and they continue to be given today. Brahmans are given two names, one derived from the child’s birth asterism. Planetary names, too, can be cited from rather early sources.

8. Fish and the God of Waters

If the ‘fish’ pictograms of the Indus script generally meant ‘star’, why was the meaning ‘star’ not expressed directly with a ‘star’ pictogram as in the ancient Near East? Why did the Harappans in a more complicated way use a ‘fish’ pictogram and the rebus principle? One answer could be that though this method may appear more complicated to us, it probably did not appear so to the Indus scribes, in whose language the same word denoted both ‘fish’ and ‘star’. But there is also another satisfactory answer to this question.
In the Sumerian script, the ‘star’ pictogram means not only dingir ‘god’ but also anu ‘sky’, and it seems that the ‘star’ pictogram originally was the exclusive symbol and attribute of the sky-god Anu. As Anu was the leading divinity of the Sumerian pantheon, his symbol started also meaning ‘godhead’, and then ‘god’ in general.

The ‘fish’ pictogram of the Indus script appears to have had a somewhat similar background. The reason why ‘fish’ and not ‘star’ was selected to represent the concept of ‘god’ seems to be that in the Early Harappan religion the fish occupied a central position: fish is one of the most popular motifs of the Early Harappan painted pottery (Fig. 9). As the aquatic animal par excellence, it appears to symbolize the God of Waters. The importance of this deity in the Harappan pantheon is proved by his popularity in the Harappan iconography.

The famous “Proto-Siva” wears the horns of a water-buffalo, another animal closely associated with water (Fig. 10).
In one Indus-type cylinder seal from the Near East, this buffalo-horned deity is surrounded by a pair of buffaloes, a pair of snakes, and a pair of fish (drawn exactly like the ‘fish’ pictograms of the Indus script) (Fig. 11).
In a triangular amulet from Mohenjo-daro, the same deity, squatting in “Proto-Siva’s” “yoga posture”, is flanked on either side by a fish, an alligator, and a snake. (Fig. 12).
In a comparable fashion, the fish (sometimes alone, but usually placed in a river flowing out of the god’s shoulders) characterizes the Sumerian water god Enki in the Mesopotamian seal glyptics. The water-god Varuna of the later Indian mythology is likewise said to be the lord of all sorts of aquatic monsters. We can be sure that, like Enki and Varuna, the Harappan “Proto-Siva” was not merely the god of waters, but had other important aspects as well. It seems probable that he was also the god of death and fertility: the buffalo, fish and alligator associated with the Indus deity had such associations later.

9. Saturn and the tortoise

One further way to check the correctness of the above proposed interpretation of the ‘fish’ pictograms is to study Indus signs that occur next to the ‘fish’ pictograms. The task is to test whether any of these signs can be interpreted by applying the same methods and hypotheses, and to see if something reasonable emerges from the sign combinations. The Indus script is like a crossword puzzle: the individual sign interpretations are like the word guesses which have to fulfill certain strictly defined conditions and which are kept in control by examining how well they fit together.

A structural analysis of the Indus inscriptions shows that the ‘fish’ pictograms often form stable combinations with the immediately preceding signs. Moreover, many of the ‘fish’ signs themselves are combinations of the basic ‘fish’ pictogram and various diacritic marks, that is, differentiating or modifying additions. It is most likely that such compounded and diacritic signs in the Indus script have had the same function as in the readable pictographic writing systems of ancient Sumer and Egypt, where they have been used either to express words compounded with the basic sign or to clarify its meaning by quoting one of its homonyms or synonyms. Compounded words, of course, were meant to be read in addition to the basic sign, while phonetic or semantic complements just functioned as reading aids, helping to recognize the intended word, especially in ambiguous cases where the reader had a choice between several options. Now the question is, can these signs occurring next to the basic ‘fish’ sign be interpreted systematically and in such a way that they smoothly fit their context and the assumed reading of the ‘fish’ sign?

Unfortunately most of the Indus signs are so stylized and simplified that their original pictorial meaning cannot be recognized unambiguously. This means that normally the pictorial interpretations are bound to remain sub- jective and unreliable guesses. There are, however, some Indus signs whose pictorial meaning is beyond any reasonable doubt.

Among the diacritical marks added to the basic ‘fish’ sign is one that has been placed over the ‘fish’ sign and looks like a ‘roof’ (Fig. 13, b ). The most wide- spread root from which words denoting ‘roof’ are derived in the various Dravidian languages is *vay- : *vey- : *mey- ‘to cover a house with a thatched roof’, in which etymon the alternations *v- / *m- and *-ay- / *-ey- can be reconstructed for Proto-Dravidian3. Thus in Proto-Dravidian the root *vey- / *mey- ‘to roof’ was partially homophonous with the root *may- ‘black’. The sequence of the pictograms for ‘roof’ and ‘fish’ in the Indus script can be read in Proto-Dravidian as *mey-meen in the sense of *may-meen ‘black star’. What makes this reading really significant is that the last-mentioned compound is factually attested as the name of the planet Saturn in the oldest available Dravidian texts, the poems of the Sangam literature written in Old Tamil in the early centuries of our era (Purananuru 117). Such a name is natural for Saturn, which is a dim planet, the root *may- (Tamil mai- ) meaning not only ‘to be or become black’ but also ‘to be dim’.

The oldest Sanskrit texts dealing with the worship of the planets also associate Saturn with the colour black. They further mention Yama, the Hindu god of death, as the deity presiding over this dark planet. Yama is associated with the colour black in the Brahmana texts of the Veda (cf. e.g. Maitrayane Samhita 3,14,11 yamaya krishna). In classical Hinduism, Yama’s colour is black and his vehicle usually the dark water buffalo. The planet Saturn, too, is said to ride the water buffalo in some texts.

But Saturn is not only a dark but also a slow planet; in fact, for this reason it is usually called sani or sanaiscara ‘slowly-going’ in Sanskrit. In the iconography of the Buddhists and the Jains, the planet Saturn rides the tortoise, whose slowness is proverbial. It seems quite likely to me that this association of the planet Saturn with the tortoise may go back to the Harappan times: the compounded Indus sign depicting a fish with a roof over it could, then, symbolize the deified planet Saturn even pictorially through his vehicle, for the tortoise is an aquatic animal (i.e. a kind of ‘fish’) covered with a shell (i.e. a kind of roof)!

10. Other planets: examples of cross-checked readings

Another diacritic sign is drawn either directly or obliquely across the body of the ‘fish’ pictogram (Fig. 14, 13c). It could denote ‘halving’ or ‘dividing into two parts’: the corresponding Proto-Dravidian root, *pacu , is homophonous with Proto-Dravidian *pacu ‘green’. The resulting compound, * pacu-meen ‘halved fish’ & ‘green star’, is not known as such from Dravidian languages, but in Old Tamil the word paccai literally meaning ‘greenness’ is attested as the name of the planet Mercury, and simultaneously as the name of the green-hued pastoral god Krsna.
The Indus script is normally to be read from right to left. (This is usually the direction of writing in the seal impressions, while in the original seals – such as those illustrated in Fig. 14 – the direction of writing is reversed.) The sign of ‘two long vertical strokes’ (on the right) followed (on the left) by the plain ‘fish’ pictogram can be shown to form one syntactic unit by means of a structural analysis of the Indus inscriptions. They are likely to have formed a compound (Fig. 13 d). But how to read the pictogram consisting of ‘two long vertical strokes’? Such a simplified symbol can be pictorially interpreted in many different ways. For example, it could be assumed to denote ‘two’, or ‘pair’, or ‘path’, or ‘parallel’. In the absence of any idea about the intended meaning of the ‘fish’ pictogram, it would be difficult to decide which, if any, of these possible interpretations is correct. The above suggested ways of reading the ‘fish’ pictogram and its attributes, however, make it possible to approach the present case from a different angle, and to use it as a test. Supposing that we are right in assuming that many of the stable combinations of Indus pictograms ending in the ‘fish’ sign denote Dravidian planetary names of the type ‘COLOUR’ + (‘fish’ =) ‘star’, do any of the actually attested Dravidian compounds of this type provide a colour term having a homonym that would adequately fit the pictorial form of the sign ‘two long vertical strokes’?

The word for ‘white’ that has the widest distribution among in the Dravidian languages is vel. The compound vel + meen (assimilated into vel-meen ) ‘white (or bright) star’ is known from Old Tamil as the name of the planet Venus, that brightest star of the morning and evening sky; and the noun vell-i , derived from the root vel , denotes ‘planet Venus’ in a number of Dravidian languages. The meaning of its homophone veli ‘enclosed or intervening space, open space’ could hardly be expressed better pictorially by any other symbol than ‘two long vertical strokes’.

In Tamil, at least, the word velli is used not only in the meaning of ‘planet Venus’ but in the general meaning of ‘star’ as well.
It is highly significant to observe that in the Indus script, the pictogram of ‘two long vertical strokes’ is used not only as an attribute of the ‘fish’ pictogram (in the above discussed compound read as vel + meen ‘white star = Venus’) but also as its synonym: both pictograms, the plain ‘fish’ as well as the ‘two long vertical strokes’, occur as if they were the latter member of a compound in the Indus texts in identical contexts, namely, immediately after a ‘fig tree’ pictogram (Fig. 13e, 15). We shall now consider the significance of this first member and of the resulting compound.

11. The sacred fig tree and the North star

The pictogram just mentioned has several variant forms in the Indus texts (Fig. 16, 17). Their comparison with the motifs of Early Harappan painted pottery (Fig. 7) suggests that this pictogram represents the large Indian fig tree, Ficus Indica. Except when combined with another sign (depicting the ‘crab’), which is placed inside it omitting the central “branch” (Fig. 15), the tree is shown as three-branched (Fig. 16), just as in the painted pottery. In one variant of the combined sign, the branches end in fig leaves as in the painted pottery (Fig. 15), but in other variants the branches seem to bear either fig fruits or hanging aerial roots or both (Fig. 17).
The rope-like hanging aerial roots are characteristic of the Indian fig tree; indeed, it seems that one of the Sanskrit names for this tree, vata , is of Dravidian origin (it is called vatam in a number of Dravidian languages), and that this name is ultimately derived from the Dravidian word vatam ‘rope, cord’.

This etymology also appears preferable to the current explanation, which derives vata ‘Indian fig’ from Middle Indo-Aryan (Prakrit vata Sanskrit vata ‘covered, surrounded’), because the Aryan nomads did not bring this tree to the subcontinent, but encountered it first there. They could, of course, give this tree an Aryan name: nyag-rodha ‘downwards growing’ is one, based on the tree’s habit of growing new trunks from the hanging aerial roots. But adoption of a tree’s earlier native appellation would be an even more natural process.
The Dravidian explanation of the Indian fig’s name vata as ‘rope-tree’ makes it possible to find a Dravidian homophone eminently fitting the astral context assumed in the Indus texts where the ‘fig tree’ pictogram is followed by the ‘fish’ or ‘two long vertical strokes’ (Fig. 18).
In Dravidian, the word vata also means ‘north’, and in Old Tamil vata-meen is the name of the North star. This Dravidian homophony, linking vata ‘Indian fig’ and vata ‘north’, suggests an earlier Indian background also for the classification of trees found in Sanskrit texts which make the Indian fig the “tree of the northern direction”.

Moreover, even as early as in the Rgveda (1,24,7), mention is made of an Indian fig whose roots are kept up in the middle of the sky by the god Varuna. Later cosmological descriptions seem to associate this often mentioned heavenly fig tree with the north star; for in reply to the question why do the stars remain in the sky and not fall down, the Puranas offer an explanation that reminds us of the Indian fig’s aerial roots: it is maintained that the stars and planets are fixed to the north star with invisible ropes.

Above, Harappan painted pottery and graphic variants were invoked to justify the pictorial interpretation of the ‘Indian fig tree’ pictogram. A confirmation of this interpretation is provided by two Indus amulets, one side of which shows nothing but a clearly recognizable fig leaf. The other side shows a short inscription of three pictograms, the middlemost being that read as vata (Fig. 19). Besides the ‘single fig leaf’ pictogram on the reverse, these amulets offer another contextual proof for the ‘Indian fig tree’ pictogram.
The last of the three signs (Fig. 19) is not so significant in the inscription on the obverse, since most Indus texts end in it.
But the first sign (on the right), which precedes the ‘Indian fig tree’ pictogram, is very important for two reasons: first, because it can be assumed to be either an attribute of the word expressed by the ‘Indian fig tree’ pictogram or the first member of a compound formed with it., and second, because its pictorial shape can be interpreted with minimal ambiguity. There can be no reasonable doubt that this attributive sign represents the numeral ‘four’ through its four short vertical strokes (Fig. 19). Read in Dravidian, does it make any sense in this context? ‘Four’ is nal in Dravidian, and there is a closely homophonous Dravidian root nal meaning ‘to hang, be suspended’, used with special reference to hanging ropes and vines. The compound formed by the two pictograms would mean, then, ‘(tree characterized by) hanging ropes’, as the Indian fig is.

The latter root, nal ‘to hang’, originally starts with a palatal nasal, but it is assumed that the common alternation between na and na and the loss of the nasal in the initial position in Dravidian languages already existed in Proto-Dravidian. This phonetic alternation not only makes nal ‘to hang’ fully homophonous with nal ‘four’, but also makes it possible to suggest here an etymology (i.e. derivation from nal ‘to hang; hanging rope’) for one of the principal Dravidian names of the Indian fig tree, namely al.4

12. The Pleiades and the Seven Sages

In the Indus script, numerals are marked by repeating a short vertical stroke the required number of times. The pictograms of ‘six’ (six short strokes divided to two lines) and (on its left side) ‘fish’ together form a syntactic unit (Fig. 13f). It corresponds to the compound aru-meen ‘six-star’ occurring in Old Tamil texts and denoting the asterism of the Pleiades. This constellation was the first one in the ancient Indian star calendar whose conjunction with the sun at the vernal equinox marked the new year around the 23rd century B.C.
The Pleiades hold a prominent place as the mothers or wet nurses of the newborn infant in one of the most ancient and central Hindu myths, that of the birth of the war-god Rudra/Skanda, who evidently represents, among other things, the victorious rising sun (and as vernal sun the new year). The Pleiades are said to have been the wives of the seven sages, who are identified with the seven stars of the Great Bear.
The Great Bear’s Old Tamil name elu-meen ‘seven-star’ corresponds to the combination of the pictograms ‘7’ + ‘fish’, which alone constitutes the entire text of one finely carved Indus seal (Fig. 21).
The Satapatha-Brahmana (2,1,2,4) states that the six Pleiades were separated from their husbands on account of their infidelity; other texts specify that only one of the seven wives, Arundhati, remained faithful and was allowed to stay with her husband: she is the small star Alcor in the Great Bear, pointed out as a paradigm of marital virtue to the bride in the Vedic marriage ceremonies.
Evidence for the Harappan origin of this myth is provided, among other things, by Indus seals which show a row of six or seven human figures (Fig. 20, 22); their female character is suggested by the one long plait of hair, which to the present day has remained characteristic of the Indian ladies.

13. Conclusion

The Indus script represents logo-syllabic writing. This means that it does not constitute such a closed system of single-valued graphemes as the syllabic and alphabetic scripts, which could be cracked as wholes. Rather, individual signs may be interpreted one by one, and many of the graphemes are likely to remain eternal mysteries.

The interpretations presented above, few in number but cross-checked, suggest that the Indus script was essentially similar to the other pictographic scripts created before the middle of the third millennium B.C., that the language of the Indus people was Dravidian, and that they professed a religion that was genetically related to the religions of both the ancient West Asia and the later India.

The Harappan religion emerging from these interpretations is in an interesting way reflected in the Indus pictograms. As iconic signs making use of the picture puzzle (or rebus) principle, they can simultaneously communicate two separate messages, one pictorial, one phonetic. It seems to me that the creators of the script were at pains to invent such iconic symbols that the two messages would be in harmony with each other. Witness the ‘roofed fish’ (pictorial message) as the rebus for the ‘black star’ (phonetic message), both symbols for the deified dark planet Saturn, conceived as riding the slow-creeping tortoise.

THE BOOK

[Originally published as Parpola, Asko (1988) Religion reflected in the iconic signs of the Indus script: penetrating into long-forgotten picto+graphic messages. Visible Religion 6: pp. 114-135.]

Using Machine Learning to Help Digitize Ancient Texts from Indus Civilization