Is the Indus script indeed not a writing system? By Asko Parpola

Is the Indus script indeed not a writing system? By Asko Parpola

Is the Indus script a writing system or not? I represent the traditional view
that it is, and more accurately, a logo-syllabic writing system of the
Sumerian type. This paper is an enlarged version of the criticism that I
presented two years earlier in Tokyo, where it was published soon
afterwards (Parpola 2005). What I am criticizing is "The collapse of the
Indus script thesis: The myth of a literate Harappan Civilization" by Steve
Farmer, Richard Sproat and Michael Witzel (2004), where the authors
categorically deny that the Indus script is a speech-encoding writing
system.
Farmer and his colleagues present ten main points or theses, which
according to them prove that the Indus script is not writing:
1. Statistics of Indus sign frequencies & repetitions
2. ―Texts‖too short to encode messages
3. Too many rare signs, especially ―singletons‖4. No sign repetition within any one text
5. ―Lost‖longer texts (manuscripts) never existed
6. No cursive variant of the script developed, hence no scribes
7. No writing equipment has been found
8. ―Script‖signs are non-linguistic symbols
9. Writing was known, but it was consciously not adopted
10. This new thesis helps to understand the Indus Civilization better than
the writing hypothesis.
I shall take these points up for discussion one by one.
Statistics of Indus sign frequencies & repetitions
Firstly, Farmer and his colleagues claim that comparison of the Indus sign
frequencies ―can show that the Indus system could not have been a
Chinese-style script, since symbol frequencies in the two systems differ
too widely, and the total numbers of Indus symbols are too few‖ (Farmer
& al. 2004: 29). They also point out that signs are repeated within a single
inscription much more often in Egyptian cartouches than in Indus seals of
a similar length.
112 Airāvati
There is no difficulty to agree with these observations. There is a vast
difference between the Chinese script with its theoretically nearly 50,000
signs (and even in practice about 5000 signs) and the Indus script with
only about 400 known graphemes.
―But [as Farmer and his colleagues themselves conclude,] studies of
general sign frequencies by themselves cannot determine whether the
Indus system was a ‗mixed‘ linguistic script [that is, a logo-syllabic script
of the Sumerian type]... or exclusively a system of nonlinguistic signs‖
(Farmer & al. 2004: 29).
As this is an important point, my colleague Dr Kimmo Koskenniemi, who
is Professor of Computer Linguistics at the University of Helsinki,
verified from Dr Richard Sproat by e-mail in April 2005 that they both
agree on the following: ―Plain statistical tests such as the distribution of
sign frequencies and plain reoccurrencies can (a) neither prove that the
signs represent writing, (b) nor prove that the signs do not represent
writing. Falsifying being equally impossible as proving.‖
Rebuses were used very much from the earliest examples of the Egyptian
writing. Around 3050 BC, the name of King Narmer was written with the
hieroglyphs depicting ‗catfish‘ (the Egyptian word for 'catfish is n'r) and
‗awl‘ (the Egyptian word for 'awl' is mr). (cf. Gardiner 1957: 7). Egyptian
rebus-punning ignored wovels altogether, but the consonants had to be
identical (cf. Gardiner 1957: 9). Other early logo-syllabic scripts too,
allowed moderate liberties, such as difference in vowel and consonant
length. The Egyptian words represented by the hieroglyphs could contain
three or two consonants or just one (cf. Gardiner 1957: 25). Eventually
only the one-consonant signs were selected by the Egyptian-trained
Semitic scribes for writing their own language, but they were used
copiously also in Egyptian-language texts, and not only for writing
foreign proper names. This easily explains the difference in the statistics
between Egyptian cartouches and Indus seal inscriptions.
―Texts‖ too short to encode messages
The second argument of Farmer and his colleagues is that ―Indus
inscriptions were neither able nor intended to encode detailed ‗essages‘,
not even in the approximate ways performed by formal mnemonic
systems in other nonliterate societies‖ (Farmer et al. 2004: 42). One of the
two reasons adduced in support of this thesis is that the Indus inscriptions
are too short.
Parpola, Asko 113
But although the Indus texts have as their average length five signs, this is
quite sufficient to express short noun phrases in a logo-syllabic script of
the Sumerian type. We cannot expect complete sentences in seals and
other types of objects preserved (cf. Parpola 1994: 87). But even written
noun phrases qualify as language-based script — I shall return to this
point later.
The Mesopotamian seal inscriptions typically contain: a proper name ±
descent ± occupation (cf. e.g. Edzard 1968). In the most elaborate seals of
the high officials, this information is couched in an invocation addressed
to the King or other dignitary. Here are two examples of Mesopotamian
seal inscriptions: ―Adda the Scribe‖ ―O Sharkali- sharri, King of Akkad:
Ibni-sharri the Scribe (is) your servant‖. These Akkadian seals are
contemporary with the heyday of the Indus Civilization, and the latter one
in fact attests to contacts with it. The water-buffalo depicted in it was
imported to Mesopotamia from the Indus Valley during the rule of Sargon
the Great, King of Akkad (2334-2278 BC) and entered Mesopotamian
iconography towards the end of his 60 year long rule, to disappear from
the iconography and the faunal remains in the beginning of the second
millennium BC when the Indus Civilization collapsed (Cf. Boehmer
1975).
Not all Indus texts 2 are so short — for instance the one-line seal
inscription M-355 from Mohenjo-daro has 14 signs. But even a single
sign of a logo-syllabic script can convey a message. The single-sign seal
inscription H-94 from Harappa probably renders the occupational title of
the seal owner. Single-sign texts may consist of non-composite signs, but
here this single sign is a composite sign consisting of two component
signs. Many composite signs (like the one in the text H-94) have ‗man‘ as
the final component and may denote occupational titles such as ‗oliceman‘
or ‗ilk-man‘. Partially identical sequences show a functional
correspondence between compound signs and their component signs (cf.
Parpola 1994: 80-81 with fig. 5.3). The Egyptian script around 3000 BC
was used in a number of inscriptions, most of which were very short,
often consisting of just two or three signs. They recorded proper names
with a high percentage of rebus signs and thus qualify as writing.
Too many rare signs, especially ―singletons‖
The third argument of Farmer and his colleagues has been put into words
as follows: ―Further evidence that clashes with the Indus-script thesis
shows up in the large number of unique symbols (or ‗singletons‘) and
other rare signs that turn up in the inscriptions ... A number of inscriptions
114 Airāvati
also contain more than one singleton in addition to other rare signs,
making it difficult to imagine how those signs could have possibly
functioned in a widely disseminated ‗script‘‖(Farmer & al. 2004: 36).
It is true that around 25% of the about 400 graphemes of the Indus script
are attested only once (cf. Mahadevan 1977: 17; Parpola 1994: 78, table
5.1).
But if more texts are excavated, many of these ‗singletons‘ will occur
more than once; there will also be new rare signs. Many of the Indus
‗singletons‘ occur in the midst of more frequently occurring signs, which
helps their understanding. All logo-syllabic scripts had rarely occurring
signs, some of these scripts quite many. Chinese has very many rare signs,
and some of them do occasionally occur even in newspapers.
No "random-looking" sign repetitions within any one text
Although Farmer and his colleagues in passing refer to logosyllabic
writing systems of the Sumerian type and their functioning, their
argumentation implies that in order to represent a language-based script
the Indus signs should largely be phoneticized in the manner of the
Egyptian cartouches. However, in early logosyllabic scripts one sign
often stands for a complete word. Even a seal with a single sign can
express its owner, and there is mostly little reason for sign repetition in
short seal texts written in an early logosyllabic script of the Sumerian type.
The alleged lack of what they call random-looking type of sign repetition
is mentioned as the fourth and most important and critical evidence
against the thesis that the Indus script is a writing system: ―Most
importantly, nowhere in Indus inscriptions do we find convincing
evidence of the random-looking types of sign repetition expected in
contemporary phonetic or semi-phonetic scripts‖ (Farmer & al. 2004: 29-
30).
Farmer and his colleagues themselves admit that ―some Indus signs do
repeat in single inscriptions, sometimes including many repetions in a
row‖ (p. 31). However, they do not accept the evidence of such
duplications: ―Whatever the origins of these different types of
duplications, all that is critical for our purposes is to note again the lack of
any suggestions in them of the random-looking repetitions typical even of
monumental scripts like Luwian or Egyptian hieroglyphs‖ (p. 36).
The hieroglyphic signs drawn in black in fig. 1 mark the repetitions in the
cartouches of Ptolemy and Cleopatra; they were crucial in the
Parpola, Asko 115
decipherment of the Egyptian script. But these are the repetitions when
both of the two cartouches are taken into consideration. Farmer and his
colleagues speak of sign repetitions limited to single cartouches, in which
case Ptolemy‘s cartouche has only one sign repetition, namely the
duplication of the sign E, one after the other in a row, which according to
Farmer & al. does not count as a "random-looking" repetition. Within
Cleopatra‘s cartouche, there is likewise only one sign repetition, namely
that of the eagle-shaped sign for A. This latter case would qualify as an
example of a ―random-looking‖ sign repetition.
Fig. 1 : Cartouches of Ptolemy and Cleopatra: the Egyptian hieroglyphs and their
transliterations (with repetitions shown in bold). (After Parpola 1994: 41, fig. 3.1.)
But sign repetition within single Indus inscription DOES occur, also of
the ―random-looking type" completely missed by Farmer and his
colleagues. Such repetition occurs even in the ―bar-seals‖, which Farmer
and his colleagues (2004: 33) consider particularly crucial for the Indus
script thesis. The following counter examples by no means exhaust the
material.
In the 10-sign text M-682 from Mohenjo-daro, one sign is repeated three
times, two other signs are repeated twice, and all in different places, that
is, not in a row.
In M-634 from Mohenjo-daro one sign is repeated in three different
places. Farmer and his colleagues have noticed this case, but disqualify it
because in their opinion the ―sun symbol‖shows that non-linguistic
symbols are involved. Of course this sign can very well depict the solar
wheel with rays, as I have myself proposed on the basis of Near Eastern
and later Indian parallels (cf. Parpola 1994: 104, 106 fig. 7.5; 110; 116-
116 Airāvati
117). But, how do Farmer and his colleagues know that this sign has not
been used phonetically as a rebus sign: after all, it is precisely this type of
―random repetition‖that they consider as proof for phonetic usage!
In M-1792 (Marshall 1931: III, pl. 106 no. 93) from Mohenjo-daro one
sign (different from that of M-634) is also repeated in three different
places.
The seal K-10 from Kalibangan has ten signs. One and the same two-sign
sequence is repeated in two different places.
In the 11-sign text M-1169 from Mohenjo-daro, one sign is repeated in
two different places.
In the 8-sign "bar-seal" M-357 from Mohenjo-daro, one sign is repeated
in two different places.
I agree with Farmer and his colleagues that some of the sign duplications
in row imply quantification (cf. Farmer & al. 2004: 31). I shall come back
to the probable function of the small bifacial tablets later on. The
inscription on one side of them usually has just the U-shaped sign,
preceded by one to four vertical strokes for the numbers 1 to 4: UI, UII,
UIII, UIIII. In some tablets, such as H-764 from Harappa, the U-shaped
sign is repeated three times: UUU, obviously an alternative to UIII, where
III = number 3 is a numeral attribute (cf. Parpola 1994: 81). Farmer and
his colleagues want to deny the use of number signs as numeral attributes
of following signs; according to them they are independent symbols for
fixed conceptions: thus seven strokes should denote ―THE seven‖
However, different numbers clearly alternate before certain signs, among
them the U-shaped sign, clearly suggesting attributive use (cf. Parpola
1994: 81-82; 88; 120, fig. 7.21, I).
Farmer and his colleagues (2004: 31) surmise that the duplication of other
signs may emphasize their magical or political power. They do not
mention that such sign reduplications can reflect emphasizing linguistic
reduplications common in Dravidian (and other Indian languages)
especially in onomatopoeic words, or as grammatical markers, such as
Sumerian nominal plurals (cf. Parpola 1994: 82). There are also cases like
the reduplication of the sign ―dot-in-a-circle‘ that could depict the ‗ye‘.
Comparing the Dravidian words kaN ‗eye‘ and ka:N ‗to see‘, I have
proposed reading their reduplication as a compound word, namely kaNka:
Ni attested in Tamil in the meaning ‗overseer‘, a meaning that would
Parpola, Asko 117
suit very well for instance its occurrence on an ancient seal-impression on
a potsherd from Mohenjo-daro (M-1382) (cf. Parpola 1994: 215; 275).
"Lost" longer texts (manuscripts) never existed
All literary civilizations produced longer texts but there are none from the
Indus Valley — hence the Indus ―script‖is no writing system: Farmer and
his colleagues reject the much repeated early assumption that longer texts
may have been written on ―birch bark, palm leaves, parchment, wood, or
cotton cloth, any of which would have perished in the course of ages‖as
suggested by Sir John Marshall in 1931 (I, 39). Farmer and his colleagues
are ready to believe the ―Indus script thesis‖only if an Indus text at least
50 signs long is found.
But even though Farmer and his colleagues speak as if our present corpus
of texts was everything there ever existed, this is not the case. More than
2100 Indus texts come from Mohenjo-daro alone, and yet less than one
tenth of that single city has been excavated. Farmer and his colleagues do
not know what has existed and what may be found in the remaining parts
of the city, even if it is likely that only imperishable material of the kinds
already available continue to be found. The Rongo-Rongo tablets of
Easter Island are much longer than 50 signs. But does this make it certain
that they represent writing in the strict sense?
Seed evidence shows that cotton has been cultivated in Greater Indus
Valley since Chalcolithic times, and cotton cloth is supposed to have been
one of the main export item of the Harappans. Yet all the millions of
Harappan pieces of cotton cloth have disappeared for climatic reasons,
save four cases where a few microscopic fibers have been preserved in
association with metal (cf. Possehl 2002: table 3.2, with further
references). Alexander‘s admiral Nearchus mentions ―thickly woven
cloth‖used for writing letters in the Indus Valley c 325 BC. Sanskrit
sources such as the Ya:jñavalkya-Smrti (1,319) also mention cotton cloth,
(ka:rpa:sa-)paTa, as writing material around the beginning of the
Christian era. But the earliest preserved examples date from the 13th
century AD (cf. Shivaganesha Murthy 1996: 45-46; Salomon 1998: 132).
Emperor Asoka had long inscriptions carved on stone (pillars and rocks)
all around his wide realm in 260 to 250 BC. They have survived. But also
manuscripts on perishable materials must have existed in Asoka‘s times
and already since the Achaemenid rule started in the Indus Valley c 520
BC. This is suggested among other things by the mention of lipi ‗script‘
in Pa:Nini‘s Sanskrit Grammar (3,2,21) which is dated to around 400-350
118 Airāvati
BC. Sanskrit lipi comes from Old Persian dipi ‗script‘. The earliest
surviving manuscripts on birch bark, palm leaves and wooden blocks date
from the 2nd century AD and come from the dry climate of Central Asia
(cf. Shivaganesha Murthy 1996: 24-36; Salomon 1998: 131). We can
conclude that manuscripts on perishable materials have almost certainly
existed in South Asia during 600 years from the start of the Persian rule
onwards, but they have not been preserved; this period of 600 years with
no surviving manuscripts corresponds to the duration of the Indus
Civilization.
No cursive variant of the Indus script developed — hence no scribes
The sixth argument of Farmer and his colleagues is based on the
observation that everywhere scribes writing manuscripts tended to
develop a cursive style. From the fact that the Indus script changed very
little during its 600 years of existence they conclude that there were no
longer texts nor any scribes.
But the Egyptian hieroglyphs preserved their monumental pictographic
shapes for 3000 years.The Egyptian cursive hieratic style of papyrus
manuscripts does not differ so very much from the monumental
hieroglyphs. The difference between Maya manuscripts and monumental
inscriptions is not all that great, either.
Actually there is quite a lot of graphic variation in the Indus signs (see the
sign list in Parpola 1994: 70-78, fig. 5.1), and in my opinion this variation
provides also an important key to their pictorial or iconic understanding.
On the other hand, the Indus script emerges in the Mature Harappan
period already more or less fully standardized, and by this time a lot of
shape simplification or creation of a more cursive script had already taken
place.
No writing equipment has been found
No writing equipment has been found, hence Farmer and his colleagues
conclude that there were no scribes nor any manuscripts. Four
archaeologists specializing on the Indus Civilization have interpreted
some finds as writing equipment, but their suggestions ―are no longer
accepted by any active researchers‖(Farmer et al. 2004: 25).
But thin metal rods, such as used in South India to incise palm leaf
manuscripts, could have early on corroded away or beyond recognition.
From painted Indus texts on Harappan pots (e.g. Sktd-3 from Surkotada
Parpola, Asko 119
in CISI 1: p. 392) and bangles (cf. Blk-6 from Balakot in CISI 2: p. 432)
we know that Indus people used brushes to write, although such brushes
have not survived or have not been recognized — and in North India
palm leaf manuscripts have been painted with brushes. For the record,
some of the provisional identifications for Harappan writing equipment
(Mackay 1938; Dales 1967; Konishi 1987; Lal 2002) were published
fairly recently, and two of these scholars are still themselves "active
researchers".
The Indus "script" signs are actually non-linguistic symbols
Instead of a language-based writing system, Farmer and his colleagues
(2004: 45) see in the Indus signs ―a relatively simple system of religiouspolitical
signs that could be interpreted in any language‖ The nonlinguistic
symbols of Mesopotamian iconography are said to be a
particularly close and relevant parallel, as they may be arranged in regular
rows with a definite order like the Indus signs.
But in Mesopotamian seal iconography, the non-linguistic symbols
usually occur as isolated signs, for instance near the gods they belong to.
Arranged in longer rows and with a definite order they occur only in very
limited contexts: mainly on stelae and boundary stones (kudurru) between
1600 and 600 BC. Mesopotamia was a literate civilization, and the
symbols on the boundary stones followed the order of divinities in curse
formulae written down in cuneiform texts — the symbols represented
deities invoked to protect the boundary stone (cf. Black & Green 1992:
15-16; 113-114).
Writing was known to the Indus people from Mesopotamia, but it was
consciously not adopted
Finally, Farmer and his colleagues ask themselves: ―The critical question
remains of why the Harappans never adopted writing, since their trade
classes and presumably their ruling elite were undoubtedly aware of it
through their centuries of contact with the high-literate Mesopotamians‖
(Farmer et al. 2004: 44). Their answer is that the Harappans intentionally
rejected writing for some such reason as the Celtic priests of Roman
times: for the druids were averse to encode their ritual traditions in
writing like the Vedic Brahmins of India (ibid.).
But it is not likely that the Harappans would have rejected writing for
such a reason because: adopting writing did not oblige them to divulge
their secret texts, which could be guarded in an esoteric oral tradition. In
120 Airāvati
any case literacy must have been fairly restricted. Even in Mesopotamia
literary texts were written down only long after the invention of writing.
It is true that some complex societies did prosper without writing — the
Incan empire for example used instead a complex communication system
of knotted strings. But writing does offer advantages not easily discarded.
We can indeed ask a counter question: Why was the Indus script created?
In my opinion for economic and administrative reasons, like the Archaic
Sumerian script. This is strongly suggested by the fact that the majority of
the surviving texts are seal stamps and seal impressions quite clearly used
in trade and administration (cf. Parpola 1994: 113-116). But proper
judgement requires acquaintance with the evolution of the Indus
Civilization. (The following short overview is mainly based on Possehl
2002).
The Indus Civilization came into being as the culmination of a long
cultural evolution in the Indo-Iranian borderlands. From the very
beginning, this was the eastern frontier of a large cultural area which had
Mesopotamia as its core pulsating influence in all directions. In Western
Asia, the domestication of animals and plants started by 8000 BC. This
revolution in food production reached the mountain valleys of western
Pakistan by 7000 BC. From the Neolithic stage, about 7000-4300 BC,
some twenty relatively small villages are known, practically all in
highland valleys. People raised cattle, sheep and goats. They cultivated
wheat and barley, and stored it in granaries. Pottery was handmade, and
human and bovine figurines reflect fertility cults. Ornaments reflect
small-scale local trade.
During the Chalcolithic phase, about 4300-3200 BC, the village size grew
to dozens of hectares. Settlements spread eastwards beyond the Indus up
the ancient Sarasvati river in India, apparently with seasonal migrations.
Copper tools were made, and pottery became wheel-thrown and
beautifully painted. Ceramic similarities with southern Turkmenistan and
northern Iran also suggest considerable mobility and trade.
In the Early Harappan period, about 3200-2500 BC, many new sites came
into existance, also in the Indus Valley, which was a challenging
environment on account of the yearly floods, while the silt made the
fields very fertile. Communal granaries disappeared, and large storage
jars appeared in house units. Potter‘s marks suggest private ownership,
and stamp seals bearing geometrical motifs point to development in
administration. Irrigation canals were constructed, and advances were
made in all crafts. Mastery of air reduction in burning enabled making
Parpola, Asko 121
high quality luxury ceramics. Similarities in pottery, seals, figurines,
ornaments etc. document intensive caravan trade with Central Asia and
the Iranian plateau, including Shahr-i Sokhta in Seistan, where some
Proto-Elamite accounting tablets have been discovered. There were
already towns with walls and a grid pattern of streets, such as Rahman
Dheri. Terracotta models of bullock carts attest to improved transport in
the Indus Valley, which led to considerable cultural uniformity over a
wide area, especially where the Kot Diji style pottery was distributed.
The relatively short Kot Diji phase between 2800 and 2500 BC turned the
Early Harappan culture into the Mature Indus Civilization. During this
phase the Indus script came into being, as the recent American
excavations at Harappa have shown. Unfortunately we still have only few
specimens of the early Indus script from this formative phase (see CISI 3:
pp. 211-230). At the same time, many other developments took place. For
instance, the size of the burned brick, already standardized during the
Early Harappan period, was fixed in the ratio 1:2:4 most effective for
bonding.
During the Indus Civilization or Mature Harappan phase, from about
2500 to 1900 BC, the more or less fully standardized Indus script was in
use at all major sites. Even such a small site as Kanmer in Kutch, Gujarat,
measuring only 115 x 105 m, produced during the first season of
excavation in 2005-2006 one clay tag with a seal impression and three
carefully polished weights of agate (Kharakwal et al. 2006: figs. 11-12).
During the transition from Early to Mature Harappan, weights and
measures were standardized, another very important administrative
measure suggesting that economic transactions were effectively
controlled. Weights of carefully cut and polished stone cubes form a
combined binary and decimal system. The ratios are 1/16, 1/8, 1/6, 1/4,
1/2, 1 (= 13 g), 2, 4, 8, 16, ... 800.
By about 2500 BC, the Harappan society had become so effectively
organized that it was able to complete enormous projects, like building
the city of Mohenjo-daro. The lower city of at least 80 hectares had
streets oriented according to the cardinal directions and provided with a
network of covered drains. Many of the usually two-storied houses were
spacious and had bathrooms and wells. The water-engineering of
Mohenjo-daro is unparallelled in the ancient world: the city had some 700
wells constructed with tapering bricks so strong that they have not
collapsed in 5000 years. Development of water traffic made it possible to
transport heavy loads along the rivers, and to start direct trade with the
122 Airāvati
Gulf and Mesopotamia. Over thirty Indus seals and other materials of
Harappan origin, such as stained carnelian beads, have been found in
Western Asia.
That the numerous Indus seals were used to control trade and economy is
certified by the preservation of ancient seal impressions on clay tags that
were once attached to bales of goods and otherwise to safeguard property.
There are impressions of clothing and knotted strings on the reverse of
these clay tags, such as the one found at Umma in Mesopotamia (cf.
Parpola 1994: fig. 7.16). Almost one hundred such clay tags come from
the port town of Lothal on the coast of Gujarat (see CISI 1: pp. 268-289).
A warehouse had burned down and therewith baked and preserved these
tags. Many of them bear multiple seal impressions, some involving four
different seals, as does the clay tag K-89 from another site, Kalibangan.
The practice suggests the use of witnesses. Such bureaucratic procedures
imply keeping records comparable to the economic tablets of
Mesopotamia. Registers and other official documents — the kind of
longer texts that I miss — are likely to have been written on palm leaves,
cotton cloth or other perishable material that has not survived for climatic
reasons.
I spoke earlier of sign duplications that imply quantification. The small
bifacial tablets mainly known from Harappa had some economic and
ritual function. At the right end of the tablet M-478 from Mohenjo-daro
(cf. CISI 1: p. 115 & Parpola 1994: 109 fig. 7.12), we see a worshipper
kneeling in front of a tree, undoubtedly sacred, and extending towards the
tree what looks like a pot of offerings shown in profile. The
accompanying inscription, read from right to left, begins with a U-shaped
sign similar to the assumed pot of offerings, preceded by four strokes that
represent number four. One side of most tablets from Harappa usually has
nothing but this pot-sign, preceded by one to four vertical strokes for the
numbers 1 to 4. In some cases, as in the tablet H-247, the pot-sign is held
by a kneeling worshipper, as in the scene of the tablet M-478. In Harappa,
many identical tablets have been found in one and the same location.
They may have been distributed by priests to people who brought a given
amount of offerings, either as receipts that dues had been paid to the
temple, or as protective amulets in exchange of offerings. In either case,
the priests probably kept some kind of log of the transactions. In a South
Indian village where I have done field work (Panjal in Kerala), I have
witnessed how each house brings one or more vessels full of paddy to the
local shrine at festivals, to be managed for common good by temple
priests.
Parpola, Asko 123
Conclusion: Is the Indus script writing or not?
So is the Indus script writing or not? We have seen that all evidence
adduced by Farmer and his colleagues is inconclusive: none of it can
prove their thesis that the Indus script is not writing but only nonlinguistic
symbols, "a relatively simple system of religious-political signs
that could be interpreted in any language‖ (Farmer & al. 2004: 45).
The question requires the consideration of some further issues. One of
these is the fact that non-linguistic symbol systems (―potter‘s marks‖and
iconographic symbols) existed as early as since 3300 BC not only in
northern Indus Valley but also in Baluchistan, Seistan & Kerman on the
Iranian Plateau and in southern Turkmenistan, a circumstance not
mentioned by Farmer and his colleagues (cf. Vidale 2007).
In contrast to these relatively simple systems of non-linguistic pot-marks,
the Indus script has a great number of different signs, around 400, and
they have been highly standardized. Moreover, the signs are usually
neatly written in lines, as is usual in language-bound scripts. The normal
direction of writing is from right to left; this is the direction of the
impressions made with seal stamps, which were carved in mirror image.
Occasionally the seal-carver ran out of space, and in such cases he
cramped the signs at the end of the line to preserve the linear order. For
instance in the seal M-66 from Mohenjo-daro, the single sign of the
second line is placed immediately below the space which had proved too
small. The three last signs thus have the same sequence as the last three
signs in the seal M-12 from Mohenjo-daro.
But the most important characteristic of the Indus texts from the point of
view of speech-encoding becomes evident if we do not limit the
observation of sign repetition to single inscriptions as Farmer and his
colleagues do. The fact is that the Indus signs form a very large number
of regularly repeated sequences. The above discussed sequence of the
three last signs in the seals M-66 and M-12 occurs in Indus inscriptions
about 100 times, mostly at the end of the text. The order of these three
signs is always the same, and this sequence is recorded from nine
different sites, including two outside South Asia, one in Turkmenistan
and one in Iraq (see fig. 2). If the Indus signs are just non-linguistic
symbols as Farmer and his colleagues maintain, for what reason are they
always written in a definite order, and how did the Indus people in so
many different places know in which order the symbols had to be written?
Did they keep separate lists to check the order? And one should note that
there are hundreds of regular sequences that occur several times in the
124 Airāvati
texts. The text of eleven signs written on top of fig. 2 (attested in several
identical tablets from Harappa: H-279 through H-284, see CISI 1: p. 222-
223; and H-871 through H-873, see CISI 2: p. 335) can be broken into
smaller sequences all of which recur at several sites (see fig. 2). As this
small example shows, the texts even otherwise have a regular structure
similar to linguistic phrases. The Indus signs do not occur haphazardly
but follow strict rules. Some signs are usually limited to the end of the
text, and even when such a sign occurs in the middle of an inscription, it
usually ends a recurring sequence. Some other signs are limited to the
beginning of the text, but may under certain conditions appear also in
other positions. And so forth. (See Parpola 1994: 86-101).
Fig. 2 : Indus signs occur in strictly ordered sequences that recur at many different sites.
Table compiled by AP for this paper
The unrelated graffiti scratched on pots at the Megalithic site of Sanur in
South India (see fig. 3) offer a contrasting example. Three signs occur
many times together, but their order varies. It does not matter in which
order they are placed. This is what one normally expects from nonlinguistic
symbols. I do not believe that these Megalithic graffiti represent
real writing in the sense of speech-encoding, but are non-linguistic
symbols.
Parpola, Asko 125
The Indus sign sequences are uniform all over the Harappan realm in
South Asia, suggesting that a single language was used in writing. By
contrast, both native Harappan and non-Harappan sign sequences occur
on Indus seals from the Near East, the sequences usually being in
harmony with the shape of the seal: square seals are typical of South Asia,
round seals are typical of the Gulf and cylinder seals are typical of
Mesopotamia. One would expect that the most frequently attested Indus
sign would very often occur next to itself, but this is never the case in the
Indus Valley. The combination is however attested on a round Gulf-type
seal coming from the Near East, now in the British Museum (BM
120228). This seal contains five frequently occurring Indus signs but in
unique sequences (cf. Parpola 1994: Fig. 8.6). This suggests that
Harappan trade agents who resided in the Gulf and in Mesopotamia
became bilingual and adopted local names, but wrote their foreign names
in the Indus script for the Harappans to read. The cuneiform texts in fact
speak not only of a distant country called Meluhha which most scholars
equate with Greater Indus Valley, but also of a village in southern
Mesopotamia called Meluhha whose inhabitants had purely Sumerian
names.
Farmer and his colleagues claim that the Indus script is a system of nonlinguistic
symbols that can be understood in any language. They suggest
that it belongs to the category which Andrew Robinson (2002: 30)
proposes to call ―proto-writing‖, and to which he assigns ―Ice Agecave art,
Amerindian pictograms, many modern road signs, mathematical and
scientific symbols and musical notation‖. The speech-bound scripts or in
Robinson‘s terms ―full writing ― came into being with the phonetization
of written symbols by means of the rebus or picture puzzle principle.
Let us consider the rebus principle utilized in logo-syllabic scripts. Most
signs were originally pictures denoting the objects or ideas they
represented. But abstract concepts such as ‗life‘ would be difficult to
express pictorially. Therefore the meaning of a pictogram or ideogram
was extended from the word for the depicted object to comprise all its
homophones. For example, in the Sumerian script the drawing of an
arrow meant 'arrow', but in addition 'life' and 'rib', because all three words
were pronounced alike in the Sumerian language, namely ti. Homophony
must have played a role in folklore long before it was utilized in writing.
The pun between the Sumerian words ti 'rib' and ti 'life' figures in the
Sumerian paradise myth, in which the rib of the sick and dying water god
Enki is healed by the Mistress of Life, Nin-ti. But the Biblical myth of
Eve's creation out of Adam's rib no more makes sense because the
original pun has been lost in translation: ‗rib‘ in Hebrew is Sela:c and has
126 Airāvati
no connection with Eve's Hebrew name H‘awwa:, which is explained in
the Bible to mean ―mother of all living‖ (Cf. Parpola 1994: 102.) The
point is that homophony usually is very language-specific, and rebuses
therefore enable language identification and phonetic decipherment.
Fig. 3 : Pottery graffiti from the Megalithic site of Sanur in TamilNadu, South India. After
Banerjee & Soundara Rajan 1959: 32, fig. 8.
Since the appearance of my criticism in 2005, Farmer and his colleagues
have underlined that the rebus principle is occasionally used also in
symbol systems not so tightly bound to language3. As an example they
mention the use of rebus puns to express proper names in the otherwise
Parpola, Asko 127
clearly non-linguistic communication system of heraldry. But by
definition any ancient or modern symbol system which consciously uses
rebuses and which therefore at least partially can be read phonetically
counts as full writing.
Even short noun phrases and incomplete sentences qualify as full writing
if the script uses the rebus principle to phonetize some of its signs. (Cf.
Robinson 1995: 12.) Archaic Sumerian is considered a full writing system,
because it occasionally uses rebus puns, for instance on a tablet, where
the single word gi ‗reimburse‘ (expressed by the sign depicting 'reed' = gi
in Sumerian), constitutes the very incomplete phrase in its own
compartment that constitutes a text unit (cf. Robinson 2002: 26). Even in
later times, the Sumerian script had more logograms than syllabic signs,
although with time the number of phonetic signs increased. When the
cuneiform script was adapted for writing the Akkadian language, the
system could be improved upon, and the script became almost fully
phonetic.
The Egyptian script around 3100-3000 BC was used in a number of very
short inscriptions, often consisting of just two signs, which recorded
proper names but with a very high percentage of the signs used as rebuses
(see e.g. Schott 1951). The famous palette of King Narmer with an
inscription already quoted above is a good example. This is definitely
already a writing system, even if the texts are on average shorter than the
Indus texts! Here two rebus signs express the proper name of King
Narmer, whose feats are related in a non-linguistic way in the pictures
taking up the rest of the palette, yet with many formalized conventions.
This is fully parallel to the use of rebus symbols to express proper names
in the non-linguistic communication system of heraldry or coats of arms.
The new thesis helps to understand the Indus Civilization better than
the writing hypothesis
As to the very last point raised, and claim made, by Farmer and his
colleagues in their 2004 paper, I honestly cannot understand how the
hypothesis that the Indus signs are non-linguistic symbols helps us to
understand the Indus Civilization much better than the hypothesis that the
Indus script is a logo-syllabic writing system. In a logo-syllabic script the
signs may denote what they depict, or they may be used as rebuses.
Before we can even start pondering their use as rebuses, we must clear up
their iconic meaning. This necessary first step is identical with the efforts
of Farmer and others to understand the Indus symbols as pictograms.
128 Airāvati
As an example of my own efforts to understand the pictorial shapes of the
Indus signs, I would like to mention my interpretation of one particular
sign as depicting the palm squirrel (Parpola 1994: 103 with fig. 7.1): the
sign clearly represents an animal head downwards, tail raised up and four
legs attached to a vertical stroke representing tree trunk. The palm
squirrel spends long times in this pose, wherefore it is called in Sanskrit
‗tree-sleeper‘. In seal texts, the sign is more likely to have been used as a
rebus rather than in its iconic meaning (for my interpretation see Parpola
1994: 229-230). Could the non-linguistic approach of Farmer and his
colleagues offer a better explanation for the meaning of this sign?
References
Banerjee, N. R., and Soundara Rajan, K. V.
1959. Sanur 1950 & 1952: A Megalithic site in District
Chingleput. Ancient India 15: 4-42.
Black, Jeremy, and Green, Anthony
1992. Gods, demons and symbols of ancient Mesopotamia: An
illustrated dictionary. Illustrations by Tessa Rickards. London:
The British Museum Press.
Boehmer, R. M.
1975. Das Auftreten des Wasserbüffels in Mesopotamien in
historischer Zeit und seine sumerische Benennung. Zeitschrift
für Assyriologie un Vorderasiatische Archäologie 64: 1-19.
CISI = Corpus of Indus Seals and Inscriptions.
1987. Vol. 1: Collections in India, edited by Jagat Pati Joshi and
Asko Parpola
1991. Vol. 2: Collections in Pakistan, edited by Sayid Ghulam
Mustafa Shah and Asko Parpola.
2008. Vol. 3: New material, untraced objects, and collections
outside India and Pakistan. Edited by Asko Parpola, B. M. Pande
and Petteri Koskikallio. Part 1: Mohenjo-daro and Harappa, in
collaboration with Richard H. Meadow and Jonathan Mark
Kenoyer. (Annales Academiae Scientiarum Fennicae, B 239-
241.) Helsinki: Suomalainen Tiedeakatemia.
Dales, George F., Jr.,
1967. South Asia's earliest writing - still undeciphered.
Expedition 9 (2): 30-37.
Parpola, Asko 129
Edzard, Dietz Otto,
1968. Die Inschriften der altakkadischen Rollsiegel. Archiv für
Orientforschung 22: 12-20.
Farmer, Steve, Richard Sproat & Michael Witzel
2004. The collapse of the Indus script thesis: The myth of a
literate Harappan Civilization. Electronic Journal of Vedic
Studies 11 (2): 19-57.
Gardiner, Alan
1957. Egyptian grammar, being an introduction to the study of
hieroglyphs. Third edition. London: Oxford University Press,
Kharakwal, J. S., Y. S. Rawat and Toshiki Osada
2007. Kanmer: A Harappan site in Kachchh, Gujarat, India. Pp.
21-137 in: Toshiki Osada (Ed.), Linguistics, archaeology and the
human past. (Occasional paper 2.) Kyoto: Indus Project,
Research Institute for Humanity and Nature.
Konishi, Masatoshi
1987. Writing materials during the Harappan period. Pp. 213-
217 in: B.M. Pande & B.D. Chattopadhyaya (eds.), Archaeology
and History: Essays in memory of Shri A. Ghosh. Delhi: Agam
Kala Prakashan.
Lal, B.B.
2002. The Sarasvati flows on: The continuity of Indian culture.
New Delhi: Aryan Books International.
Mackay, E.J.H.
1938. Further excavations at Mohenjo-daro, I-II. Delhi:
Manager of Publications, Government of India.
Mahadevan, Iravatham
1977. The Indus script: Texts, concordance and tables.
(Memoirs of the Archaeological Survey of India, 77.) New
Delhi: Archaeological Survey of India.
Marshall, John (Ed.)
1931. Mohenjo-daro and the Indus Civilization, I-III. London:
Arthur Probsthain.
130 Airāvati
Possehl, Gregory L.
2002. The Indus Civilization: A contemporary perspective.
Walnut Creek, CA: Alta Mira Press.
Parpola, Asko
1994. Deciphering the Indus script. Cambridge: Cambridge
University Press.
2005. Study of the Indus script. Transactions of the International
Conference of Eastern Studies 50: 28-66.
Robinson, Andrew
1995. The story of writing. London: Thames and Hudson.
2002. Lost languages: The enigma of the world‘s undeciphered
scripts. New York: McGraw Hill.
Salomon, Richard
1998. Indian epigraphy. New York: Oxford University Press.
Schott, Siegfried
1951. Hieroglyphen: Untersuchungen zum Ursprung der Schrift.
(Akademie der Wissenschaften und der Literatur [in Mainz],
Abhandlungen der Geistes- und sozialwissenschaftlichen Klasse,
Jahrgang 1950, Nr. 24.) Mainz: Verlag der Akademie der
Wissenschaften und der Literatur.
Shivaganesha Murthy, R. S.
1996. Introduction to Manuscriptology. Delhi: Sharada
Publishing House.
Vidale, Massimo
2007. The collapse melts down: A reply to Farmer, Sproat &
Witzel. East and West 57 (1-4): 333-366.
Parpola, Asko 131
Notes
1 This paper was written for, and presented at, the workshop on ―Scripts,
non-scripts and (pseudo-)decipherment‖ organized by Richard Sproat and
Steve Farmer at the Linguistic Society of America's Linguistics Institute
on the 11th of July 2007 at Stanford University
(http://serrano.ai.uiuc.edu/2007Workshop/). It was also read as a public
lecture at the Roja Muthiah Research Library, Chennai, on the 16th of
February 2008. I thank the organizers of both events for this opportunity
to participate in the debate on the nature of the Indus script, and am glad
to publish the paper in honour of my old friend Iravatham Mahadevan, a
great epigraphist.
2 The Indus texts are cited in this paper with their labels in the CISI (see
references).
3 From the abstracts of the Stanford workshop papers
(http://serrano.ai.uiuc.edu/2007Workshop/abstracts.html), I got the
impression that at least one of the three authors wants to back out from
their original thesis and change it into something else. While Farmer
repeats the claim that ―the so-called Indus script was not a speechencoding
or writing system in the strict linguistic sense, as has been
assumed‖, Witzel writes as if he and his colleagues had only claimed that
the Indus script does not SYSTEMATICALLY encode language in the
sense that ―Indus signs do not encode FULL phrases or sentences‖(my
emphasis, AP). Witzel also admits that ―Indus symbols... may... contain
occasional puns‖ Or maybe, when speaking of recent studies which
suggest this, he is referring to me, since these have been my very
assumptions, namely that the Indus seals hardly contain complete
sentences and that they contain puns. In any case, I am happy if Witzel
has changed his previously more radical view and now agrees with me.
When I mentioned these impressions of mine at the Stanford workshop,
Michael Witzel assured me that he was not backing out from the original
claim but continues to maintain that the Indus script does not encode
language.