~& )

P&?l!h.e
0'Vl\'} W ~)

JOURNAL
OF

THE INSTITUTE OF

ASIAN STUDIES

Vol. VII No. 1 September 1989

WHAT DO WE KNOW ABOUT THE INDUS SCRIPT?

NETI NETI ('NOT THIS NOR THAT')
IRAVATHAM M A HADEVAN

INSTITUTE OF ASIAN STUDIES

THIRUVANMIYUR, MADRAS-600 04.1
INDIA
 What do we know about the Indus Script?
Neti neti ('Not this nor that').
Iravatham Mahadevan

It is doubtful whether there is any other undeciphered script with the

possible exception of the Phaistos Disk, which has drawn as much attention
. from would-be decipherers.as the Indus S<!ripL i There are presently more than
forty c1aims.of decipherment and the number is steadily increasing. 2 However
as no two.c1aims have anything in common , it is hardly !:iurprising that scholars
in general have remained sceptical of all the attempts. It is not pos ible within
the brief span of a lecture to discuss any of these claims in depth or even refer
to all of them. In any case, I shall be concerned here with methods rather than
results some of which I may cite only by way of illustration.

This paper is in three parts. I begin with a brief summary of the

universal features of writing as observed in the ancient Oriental scripts to
p~ovide the background to the search· for possibly similar features io the
contemporary Indus Script. I then proceed to a short survey of the known facts
about the Indus Script and · some reasonable inferences we can draw from
them. This. part also contains a report on the recent structural and analytical
studies of the script and some significant results from such studies. In the final
part, I try to look beyond the structural studies and discuss the possibility of
acquiring a broad comprehension of the contents of the Indus inscriptions even
before decipherment, and conclude with an assessment of future prospects in
the field.

Paper presented at the 49th Session of the Indian History Congress at Dharwar
in November 1988.
Journal of the Institute of Asian Studies 2

2. Oriental Pictographic Scripts

The Indus Script is one of the seven pictographic writing systems
developed in the ancient Orient during the Bronze Age (ca. 3000-1500 B.C.)
(TabJe 1).

No. Script Area Earliest occurrence

(ca.)
------------------------.--------~--- .. ---------....-----. ----------------- ..... ------------------------------
l. Sumerian Mesopotamia 3100 B .C.
2. Egyptian Egypt 3000 B.C.
3. Proto-Elamite Elarn 3000 B.C.
4. Indus Indus Valley 2500 B.C. (?)3
S. Cretan Crete 2000 B.C.
6. Hittite Anatolia 1500 B.C.
7. Chinese China 1500 B:C.

Table 1. The Seven Ancient Oriental Scripts

The Harappans had cultural and trading contacts with contemporary

0

West'Asian cultures. 4 Seals with Harap.pan motifs and writing have been found
in Babylonian , Elamite , Persian Gu.U oa nd' Central Asian sites 5 West Asian
influence on the Harappan Culture is evidenced by the occurrence or Imported
seals as well as by rnanY ocorrespondenc~s between their glyptic art (e.g. 'trefoil'
motifs, Gilgamesh-like figure on othe oseals etc.). 6 We also know from a
comparative study of the decipnered and k'fIOWn Onental scripts (Sumerian,
Egyptian, Hittite and Chinese) that they shared many common structural
features and followed very similar hnes of evolutlOn. 7 ·lt is therefore reasonable
to begin with the working hypothesis that tO n e Indus Script, occupying almost
the mid:dle position both spatially and temporally in this °group, woulo also
share the universal features of its contemporaries.

3. Universal Features of Writing

( 1) Typology of Signs
All ancient systems of writing employed basically only tnree types
of signs or characters, viz. word-signs for whole words : syllabic signs for
phonetic syllabl~,,'and alphabetic characters for single sounds (consonants
or vowels).
 3 What to We know about the Indus Script? Neti Neti ('not this nor that')

(i) Word-Signs:
Word-signs (logograms) developed from pictures. 8 Pictograp~ic
writing consisted essentially of three types. At the earliest stage each sign
literally represented the object pictorially depicted. Very soon word-signs
were also used "to represent not only the objects shown in the pictures but
also any idea associated with such Objects. Thus the SUN sign could also
stand for 'light', 'day' , etc. These two types of "signs are also called
ideograms as they convey the meanings and frot the sounds of words. 9 It
was soon discovered that a word-sign could also be used to represent any
other word with the same sound (homonym) through the technique of
rebus writing (phonetic transfer) . Thus the SUN sign could also stand for
'son' . Word-signs used in this manner are called phonograms. It is this
development which led to phonetic writing at the next (syllabic) stage .

(ii) Syllabic Signs: "

A fundamental discovery in the history of writing was the use of
signs to represent sounds without meaning .. This was achieved by forming
phonetic syllables from word-signs. Thus the SUN sign could be used in the
word "'sundry' as a mere phonetic syllable. Syllabic signs could represent
compound syllables (CVC) , closed syllables (VC), open syllables (CV) or
vowels (C = consonant: V = vowel).

(iii) Alphabetic Cha"racters :

The Egyptian Script developed at a very early stage 'alphabetic' or
uni-literal consonantal signs by ignoring the vQwels in the corresponding
syllabic words. However Egyptian continued to be mainly an ideographic
script till the end. " True alphabetic writing began with ~he Semitic
consonantal scripts (ca. 1500 B.C.).lO The vowels were added by the
Greeks (ca. 800 B.C.) to complete the development of alphabetic writing.
Vowels were also represented by 'diacritical' marks (as in the Semitic
scripts) or by 'medial' signs attached to the consonants (as in the Indian
scripts).

(iv) Ancillary Signs

(a) Determinatives are ideograpmc signs added to phonograms to.
determine the intended meaning. For example, the STAR ideogram may
be added to the SUN sign to indicate that the intended meaning is 'sun'
Journal of the Institute of Asian Studies 4

and not 'son'. Determinatives also served to indicate the broad class or
category of the words to which they , were added.

(b) Phonetic Complements are phonetic signs added to ideograms

as a guide to pronunciation where more than one reading is possible. For
example, the phonetic syllable lanl may be added to the SUN sign to
indicate that the intended reading is 'sun' and not. say, 'light'.

(c) Conventional Signs were also employed in a few cases, for

example, plural markers, word-dividers etc.

(2) Typology of Scripts

The ancient oriental· scripts may be classified notionally into three
types based on the types of signs employed by them, .viz. logograptiic
scripts, syllabaries or alphabets, comprising respectively logograms,
syllabic signs or alphabetic characters. In fact however there were no pure
systems, and even at the earliest stage, the Egyptian and the Sumerian
scripts are found to be logo-syllabic, that is. a mixture of word-signs and
phonetic syllables. The Chinese Script is unique in that it has preserved its
logographic character. In general, each sign in this script is a compound ot
ideographic and phonetic elements.

(3) Evolution of Writing

Historically the Oriental scripts evolved ffom logographic to
syJlabic and finally to the alphab~tic stages. However the scripts were
conservative and retained some ideographic elements till the end even
while increasingly employing phonetic syllabic signs. An important result
of the transition was a sharp reduction in the total number of signs in a
script. To cite one telling example: Early Pictographic SumeriC:Ul had about
2000 signs. This was reduced to about 900 in .Sumerian Cuneiform and
further to about 600 in Akkadian and 450 in Hittite Cuneiform (all
logo-syllabic systems). The Elamite Cuneiform syllabary (with closed and
open syllables) used 163 signs while in tbe Old Persian Cuneiform
syllabary (with open syllables) the number was further reduced to 4l.
Finally the Ugaritic" Cuneiform alphabet used only 30 signs. In general the
total number of sign~ is specific to each type of writing within a range
(Table 2).
 What to We know about the Indus Script '1 Neti Neti ('not this nor that')
5
------------------------------------------------------------------------.. -----------------------------:-
No. Ty~ of Scripts Type of Signs Total No. Examples
of signs
(ca.)

I. LOGOGRAPHIC Word-signs Thousands Chinese

II. LOGO-SYLLABIC Word-signs and 900-400 Sumerian,

Phonetic syllables Egyptian,
Hittite

Ill. SYLLABIC (A) Closed and 200-100 Elamite

open syllaQles Cuneiform

(B) Open syllables 100-40 Linear-B,

Old Persian
IV. ALPHABETIC Single-sound signs Below 40
Semitic,
Ugaritic,
Greek,
Latin

Table 2. Types and Number of Signs in Ancient Scripts

(4) External Developments in writing

In course of time the pictorial signs became simplified and
increasingly unrecognizable. The shape of signs was influenced by the
nature of the writing surfaces (stone , clay, cloth , paper etc.) and the tools
for writing (chisel, stylus, brush , pen etc.) . In Mesopotamia the use of the
reed stylus on soft clay transformed the Sumerian pictographic into the
totally different cuneiform script. It is however important to note that
there is no correlation between the internal and external developments in
writing systems. The Egyptian Hieroglyphic and Demotic scripts look very
different, but have the same internal structure. The Ugaritic Script is
cuneiform in appearance but its internal structure is patterned after the
semitic alphabets.

4. The Indus Script: Facts.

We may now turn our attention to the Indus Script and proc~ed to
examine the nature of the script .in th(' light of the foregoing s\lmma;y of the
universal features of the ancient writing systems. I shall begin with a brief
Journal of the Institute of Asian Studies 6

recital of the few facts that we know ab tit thl 'cript from the archaeological
context and preliminary in pection of the in cription . II

The Indu inscriptions are found only on rna]] objects like ston seals.
terracotta sealings, stone and faience tablet , pottery, copp r tablet , bronz
implement , ivory and bone rod and a few other m; cellaneous art facts.
About 3500 inscriptions are known, mo tly occurring n seal. No long
in cription on tone, clay, papyru or other mat rial ha a far be n
discovered. Nor are there any acco~nting tablet, 0 abundant ly found at
Babylonian and Elamite ites. The in cripti n aTe extremely bri f, th av rage
length being \e than four ign in a line and five ign in a text. The lange t
in cription ha only 26 igns in 3 line occurring on the ide of two terracotta
prism (1623' 2847).

The Indus Script is mainly pictographic in character (PI. 1-3). Many of

the signs ciearly depict men, animal, insects, 'fi h, bird, implements.
structures, ve sels etc. But many other si.gns are too styli eeJ to be identified
pictoTially. Two main characteristic of the cript are modification of the igns
by strokes or other mark and combination of two or m re igns. The number
of signs in the Indus Script is about 400-450. 12 There are minor graphic variant
for many of .the sign . l3 However the script did not develop any cur i e or
linear variety, but remained 'frozen' in a tandard form throughout it
&'xistence. 14

No hi-lingual inscription has a far turned up to ai.d decipherment. The

only external dues we have are tho e provided by th~ archaeological
excavation (site stratigraphy, a ociated artefacts and Location of the find ).
the type of objects carrying the in criptions and, in the ca e of eats, eating
an,d copper tablets, the accompany-ing pictorial motif .15 'The e are rna tly
animal (the o-called 'unicorn', short-horned bull, humped bull. buffalo,
elephant, tiger, rhinoceros , antelope, goat, gharial and mythical beasts), anq
religious or mythological . cenes'. It is noteworthy that among . th ev ra1
animal portrayed on the eal , horse., lion and camel are absent. Preliminary
e
in pection does not reveal any cia. link between the in criptions and th
pictorial motifs except in the case of the c pper tablets fr m Mohenjodaro. lo

5. Direction of writing:
One of the few well-e tabli hed fact about the Indus ~cript j that j[ is
generally written from the right, though there are exceptional ca es of lines
 7 What to We know about the Indus Script? Neti Neti ('not this nor that')

running from the Idt. I I The ~eneral direction of writing has been established
on the basis of many simple observations like the overflow of the last sign at the
left end to a lower I~rie. cramping of signs .for want of space towards the left
end. writing of a text along the top ~ left and bottom edges of a square seal
(4254) leaving the right edge blank (as se'en in impression) showing that the
writing is in an anti-clockwise (that is right to left) direction etc. B.B. Lal has
demon trated from a study of overlapping incisions on pottery graffiti that the
inscriptions must have been incised from the right (Pi. 4.1). I have drawn
attention to pairs of identical texts occurring in single lines and also in two 'lines
one below the other thus indicating the real se~uence
; .
of signs and the direction
of writing (PI. 4.2). ..

It . is necessary to emphasise that the value of the ' external evidence

mentioned above is limited by the fact that the inscriptions run in either
direction as seen by simple inspection. Hence the external characteristics are
reversible and 'will lead to the wrong conclusion unless the results are
controlled by the intenWlI evidence provided by the sequence. of signs. I. may
refer here to the curious case of the inscription on pottery from Kalibangan
published by B.B. Lal. IX The inscription (8221) is written from the right as
proved by ·the overlapping incisions pointed out by Lal, but is to be read from
the left as proved by the sign sequence (PI. 4.1).19

The .statistical study published in the Indian Concordance shows that

about 83 percent of the lines included in the Corpus run from the right..,.and
about 7 percent from the left. 20 (The rest are single-sign lines or doubtful lines
due to ·damage). Where there are two or three lines on a side, the normal
practice is for each line to s.tart at the right. Writing in the boustrophedon mode ·
(alternate lines in opposite directions) is rare . (Only 9 examples are listed in
the Concordance). 21 The evidence seems to suggest that the second or ' third
line runs in the rev~rsed direction only when the previous line has an
incomplete sequence (e.g. 1247, 6402).
It is therefore surprising that some scholars still attempt to decipher the
Indus script on the assumption that the writing is from the left. 22 Applying the
test of direction of writing we can safely ignore all such attempts as I ~ot
deserving serious consideration. We also come across attempts to read all the ·
.lines mechanically from the right. 2.- It is a matter of simple observation that'
there are cases of bi-directional writing of identical texts. It is possible to
recognize reversed writing by observing the following 'direction-markers
(PI. 4.3):
Journal of tho Institute of Asian Studies 8

{a) the asymmetric igns (with re peel to the vertical axis) will
appear rever ed (a in the case of the Egyptian Script);

(b) the most frequent r\ght-end and left-end signs will exchange
posit)<H1and orienlat\On'

(c) the rno t frequent jgn-groups (pair and triplets) will appear in
the reversed oreJer.

A question may however arise whether there are not genuine ca es of

reversed peHing forming different word or phra es (as for example, GOD
and DOG in Engh h) . it is true that there are such e~ceptlooal cases of
reversed pelling in the Indu tex.t . There 1 a imple test by which we can
d teet uch cases. Thu XY and YX are genuinely reversed pelling forming
different words only jf it can be hown that both spenings occur w;thin a tong.er
lext (e.g . ABCXY and ABCYX). But a complete text, say XYZ, cannot be
read as ZYX (even when the Jatter is jn the rjght-to-Jeft order) it the reversal i
not found within a longer text. Uncontrolled reading of uch texts by orne
cholars from the right or in eilher direction have to be disregarded as
arb; trary. 2.$

6. Structure of the Indus Script

(1) Number of Signs
Th simple t and one of the rno ·t d ' cisive tests for th typoJogy of a
script i ' ju t to count the number of ign in it (e Table 2). Thi j how
hampollion concluded that the Egyptian Script with about 700 ign could not
be pur ~y lcgcgraphtc hke foe Chflte e; Michael Ve ntris proce d d on (he
as umptlon that the Mycenaec\fl Linear-S Script with about 90 sign coulu only
be a yHabary of ~pen syHabie ~ and ViroHeaud recogn; 'ed instantly that th
Ugaritic Script with ju t 3 sign could be, nothing but an a1phabet inspite of its
deceptive cuneiform appearance. Th 'eventual decipherment proved all them
right. 2:'

It i difficult to be pred e about the total number of ign .n an

undeciphered script because it i not easy to distingulsh between independent
sign and mere graphic var\(i(\ts, or el.Jen hetween theign and Qth r pi<::t riat
m tif ace mpanying them . There 1 a(~O the po lb)\ity lila\ there may 5.til~ b
me undiscovered 19n . The latest 'ign-iist in the Finnish Concordance and
the IndIan Concordance put the total numberof sign in the Indus Script at 396
y What to We know about the Indus Script? Neti Neti ('not this nor that')

and 419 respectively. Combining the two Jist and allowing a margin for
variants and undi covered igns, the present best e timate· for the total number
is 425 ± 25 signs. This number, falling. in the range of the mid-hunqreds, is too
small for a fully logographic script (like the Chinese) and far too large for a
p~rely alp~abetic cript (like the Semitic), or even for a simple open syllabary
(hke the Lmear-B). The evidence of the sign-count is compelling that the Indu
Script is , like its West Asian contemporaries, a logo-syllabic script posses 'ing
word-signs and phonetic syllables.

A legitimate question to ask is whether the number of signs in the Indus

Script cannot be reduced by excluding combinations and modifications, both of
which are known to occur in this script. 26 Firstly, it is exceedingly difficult to
segment the signs in an unknown script, as we have nothing to go by except the
external forms, and any such procedure is bound to be arbitrary and
ubjective. 27 Secondly, even if we can find a way. to split the combinations and
modifications, it would be an unprofitable exercise ciS we would thereby lose
much of the information contained in the texts. We know from contemporary
cripts that sign A + sign B may not mean AB, but C, that is, something wholly
different .

e,g. MAN + BREAD = 'to eat' (Sumerian)

SKY + DROPS = 'rain ' (Chine e)

Combinations of ideograms may also yield a totally different phonetic

word- ign.

e .g. HAND + LEG = 'horn' (phonetic, in Egyptian)

Similarly modifications of a sign by strokes may radically alter its meaning or \

ound.

e .g. MOUTH + 2 strokes 'Two-thirds' (Phonetic in Egyptian)

MOUTH + 3 stro kes - 'Three-Fourths' (ibid)

Even assuming the script to be purel y phonetic (for which there is no evidence)
and treating the sign combination A + B as AB, t~e frequency-distribution
characteristics of AB are likely to be quite different from tho e. of A or B
Hence the only sound approach to an unknown script is to regard each sign
(separated by blank space on either side) as integral until we learn to
distinguish its component parts after decipherment.
Journal of the Institute of Asian Studies

Another question rai ed in the Indian context about the significance of

the ign-count may a1 0 be considered here. In Indo-Aryan we have 45 sounds
(10 vowels, 2 emi-vowels and 33 con onants). if we add 3<)6 (33 x 12)
character for consonants with attached medial vowel marks , we can easily
match the sign-count in the Indus Script. even without taking into account th
almost unlimited number of compound signs (samYllktakshara) . Howe ve r
structural analysis of the script disproves this possibility for th e reasons
ummarised below.

It i a matter of common observation that a ,word requires mar

characters to write as scripts evolve from word-sign through yllabic to
alphabetic stage

e.g. WOMAN 1 word- ign

wo-man 2 syllable
w-o-m-a-n 5 alphabetic characters

As mentioned earlier, the averag length of the Indus inscription is le s

than 4 sign in a line, and less than 5 igns in a text . There iue many inscriptions
with just one or two signs only. It is difficult to think of a complete sentence
made up of only one or two alphabetic characters. Ewen syllabic writing,
especially with open syllables as in the Indian scripts, seems unlikely with such
extreme brevity. If we 100k at comparable inscriptions on seal ,coins or votive,
objects from the Hi.s torical Peri.od in India, we find that the average number of
syllable required to compose these very short legends is more than the average
number of signs in an Indus inscription. It appears much more probable that
the average lndus text has a few words rather than syllables, especially in the
shorter in. criptions. .

(2) Segmentation of Words and Phrases

An even more telling evidence against a simple syllabic model for the
Indus Script come from word segmentation analysis. Se~eral analytical studies
have established that it is possible to segment the Indu texts into constituent
words and phrase ·through simple frequency-distribution analysi as we ll a by
sophi ticated cryptanalytical and computer studies. 211 They have proved that
the sign of the Indus Script are mostly word-signs and cannot be r garded as
phonetic units (syllabic or alphabetic). In view of the importance ·of this
conclu ion, l hall briefly summarise some of the simple technique for
word-divi ion, which can easily be verified from the Concordances:
 11 What to We know about the Indus Script? Neti Neti ('not this not tHat')

(i) Well-preserved and complete inscriptions with single signs constituting

the whole text are compiled as these must represent single words. They
are then compared with two-sign texts in which one sign is already
known to be an independent word from the single-sign texts, thus
proving the other sign to be independent. The search is then
progressively extended to longer; texts following the same procedure
(PI. 5.1).

(ii) Pairs of texts are compiled, which are identical but for the presence of
one additional sign at either end, proving these to be independent words
(including grammatical morphs) . The procedure is then extended to
cover pairs of near-identical texts varying only by the substitution of one
sign by another in the other text , proving the independent character of
both the signs (PI. 5.2).

(iii) Texts which are built up by the progressive addition of one sign at a time
at either end or in the middle prove the independent word-values of the
signs occurring in therr (f!1. 5.3).

(iv) A longer text can be shown to consist of two or more Sllorter texts
occurring as complete texts elsewhere indicating clearly the boundaries
(PI. 5.4).

(v) A few signs occur with very high frequencies and form stable pairs with a
large number of other signs (e.g. 'JAR and the two-stroke superscript
signs). It follows that the signs in such pairs must be separate words or
gI;ammatical morphs. This is a particularly productive method in view of
the very high frequency of the terminal and the superscript signs and the
large number of stable pairs formed by them (PI. 6. 1) .

(vi) Comparison of the frequencies of succe sive adjacent pairs of signs (e.g.
AB , BC, CD and DE in the text ABCDE) reveals the word boundaries
at the 'weakest junctions'. By this method almost all the long texts can
be segmented into constituent phrases and words (PI. 6.2).

(vii) Numerals form a natural indicator of word-boundaries especia lly in

two-sign texts . Thus the texts IV CUP, III CUP and II CUP occurring
on the miniature tablets at Harappa show the CUP sign to be an
independent word (PI. 0.3).
 Journal of the Institute of Asian Studies 12

The e methods are ov r app in g a nd the r suIt s are cumubti c, pro lin g thl'
essentially logographic cha racter of the In dus Script. It has not so far been
po ible to identify any phon tic syllabl es by such ana l., tiet! procedures th o ugh
they do probably exi t in th script. An interesting result of the segmentation
analysis is that ' p~rase ' (by which I mea n int gra l lin guistic units co nsisting of
more than one word) consist m stly of only two or three ·igns . Th e longest
ingle phrase (e.g. 1013) do s not 'ee m to be more th all 4 signs ~n length . ::? )

(3). Frequency-Distribution Analysis of Signs

(a) Frequency: Frequency analysis of the Indus signs is quite instruc-
tive. The frequency statistic recorded in the Indian Concorda nce (w ith a
C;orpu of 2906 texts and 13,372 ign-occurrences) is umm arised in Table 3: 30

Frequency Range No. of signs Total sign Percent (of total

occurrences occurrences)

1000 or more 139 10.43

999-500 649 4 .85
499-100 31 6344 47.44
99-50 34 2381 17 .81
49-10 86 1833 13.71
9-2 152 658 4 .92
Only once 112 11 2 0.84

Total 417 13372 100.00

Table 3. Frequency Analysi of sign in the Indus Texts

It will be seen that only 67 igns account for over 80 percen t of th ign
occurrences. These signs constitute the core of the Indu Script. as presently
kn wn and it is obviou that attempts to decipher the script must tart with
the e signs . More than half the number of signs in the sign-Ii t ccur less than
10 time each and 112 of them occur only once . These I w-frequency igns
offer no scope for comparative study and attempt to ' decipher' them will lead
us nowh reo

(b) Distribution: PositIOnal or di tributional analy i indicates the

tendency of certain frequent signs to occur in the initial or final positions by
themselves or in clusters of two or sometimes three signs. An important result
of the positional analysis is th at the signs occur generally in a fixed order and in
13 What to We know about the Indus Script'" Net; Neti ('not this nor that')

fixed positions. This is unlikely to be the case in a syllabic script in which the
d, ·tribution of ~yi t a hks may be expected to be much freer, subject only to
certain phollo)ogical constraints in the underlying language . The observed
pattern of occurrence of the Indus signs is best explained by the model of a
yntactic arrangement of words.
By combining the results of frequency-distribution and segmentation
GlJ)a\yses we can further narrow down the area of search to locate the most
common words and phrases and the position they occupy in the texts. Thus
the 67 frequent signs form only about 50 frequent 2-sign 'phrases' occurring at
least 25 times each, (of which only 6 occur more than ]00 times and 12 other5
more than 50 times each) . The number of stable 3-sign combinations is much
ie S, and only one stable 4-sign combination seems to occur as an integra)
phrase. }t is therefore profitable to concentrate on these relatively few but very
frequent words and phrases and attempt to determine their probable function
and meaning by relating them to their archaeological context and through
ideographic parallels .

7. Some Results of Structural Studies

After the publication of the Concordance, J have been pursuing
structural and anatyticaf studies of the Indus Texts ytith the aim of building a
model for an objective comprehension of the contents of the inscriptions even
before linguistic decipherment. in structural studies 'involving computer.
appllcatioo, ,I have had the benefit of collabQration with Mythih Ranga Rao of
TIFR .. Bombay. { should also acknowledge that we ~re indebted to many other f
scho\ars in the fietd pursuing similar lines of investigation. especially the Soviet
and the Finnish Groups, Gift Siromoney and Abdul Huq .31 Without going into
too many details I· shan briefly sum up our main conclusions.

(1) The Indus Script consists mainly of word-signs which appear to be of the
following types:

{a} Ideograms: These are the dear, ' transparent' signs whose ideog-
raphic signif1canc~ is apparent. These signs can be understood but
not 'read'

e.g.

MAN, HORNED PERSON, ARCHER

 Journal of the Instit ute of Asian Studies \4

(b) Phonograms: The 'i mprobable' pictograms like fi h, birds , insects,

animals etc., in what are most likely to be names and titles in the
seal-texts, can be explained only on the ba is of their being
phonograms formed by rebus writing. These s'gns cannot be
under tood or read without making an assumption about the
underlying language , as puns are language-specific.:'2
(c) Conventional Signs: These include the superscript signs, ' bracket'
signs and other 'stroke' signs. While it is po sible to determine their
function by structural analysis, there is no method presently to
discover their phonetic values.
(d) Numeral Signs: Numeral have been identified by their logical.
sequence and their use on pottery and bronze implements, obviously
for enumeration. Numbers precede the objects en umerated. The
sy tern appears to be deci~al .. The units are represented by short
strokes and the tens by inverted semi-circles~ both as in the Egyptian.
Numerals from 4 to 10 are also found written with two-tiered trokes.
The long strokes do not seem to repre ent ordinary numbers (except
probably on the miniature tablet froin Harappa) . The hort
superscript suffixes are certainly not number. The sign with 12
strokes arranged in three tiers does not function as a numeral as the
number of strokes is found to be variabl~ and the occasional zig-zag
arrangement of the tiers a nd doubling of the sign are feature ' not
shared by the numeral igrts. Numerals al 0 appear to be used in
ideographic (non-numeral) function especially when they appear as
fixed numbers in set combinations (e.g) Vll-CITY , Ill-FENCE. The
largest number identified so far are 35 and 76 occurring on two
bronze axes (6306, 2925). Signs for higher numbers, especially for
100 and 1000, may exist as still un-identified word-signs . The
numerals ar~ iIJustrated (in PI. 7.1).
(e) Phonetic syllables: They probably do exist, as a developed writing
system cannot manage without them .. But structural studies have not
so far helped in their identification.

(2) Ligatures and modifications: Compounding and modification of sign

appear to be ideographic and not phonetic in character. This inference is based
on the observation that in most ca es the ligatured or modified signs have the
same distributional pattern as the basic or unmodified signs in question. For
example, any modified FISH sign can be substituted for anv other sign in the
 15 What to We know about the Indus Script? Neti Neti ('not this nor that')

, group in almost all contexts. 'S uch a pattern is who ll y inconsistent with pholletlC
combinations or modifications .

(3) Words: Word-signs appear to r~prescnt hasically only two types of words
(or morphs) viz., roots/stems and suffixes.

(A) The root sign functions either as a substantive or an attribute. The

same sign can serve in either function. The vast majority of the signs
belong to this category of vocabu lary items in the language.
Attributes precede the substantive they qualify.
:8) Suffixes fall into two main groups.
(i) Terminal Suffixes: Five very frequent terminal signs (and their
ligatures and combinations) appear to function as nominal
suffixes (PI. 7.2).
(ii) Grammatical Suffixes: Another set of six suffixes comprising
three superscript and two middle-register stroke signs and a
4-stroke bracket sign appear to be grammatica l morphs
(PI. 7.3):
(a) The superscript suffixes function like case-marke rs, most
probably for the locative, possessive or oblique cases.
(b) The middle-register suffixes appear to function like
conjunctions as they generally serve to join two parts of a
text appearing as separate texts elsewhere. These stroke
signs also appear to denote numerals 1 and 2.
(c) Plural-marker: The four-stroke bracket sign functions like
a grammatical suffix replacing the terminal suffixes. This is
probably the plural-marker as originally suggested by
Heras.

(4) Syntactic Order in the Texts:

(a) Substantive Phrase: The core of a text is the root/stem morpheme. It
may be preceded optionally by one or more root/stem morphemes
functioning as att ributes qualifying the substantive. The substantive
may be followed by one to three nominal suffixes . The whole
sequence constitutes the main or substa ntive phrase of the text.
(b) Introductory Phrase: The substantive phrase may be preceded
(optionally) by one (or more) 'introductory' phrases qualifying the
Journal of the Institute of Asian Studies 16

substantive phrase. The introductory phrase con ists of a root/stem

as the substantive, preceded (optionally) by one or more attributes
qualifying the substantive, and followed by a superscript case-
marker suffix.

The following diagram illustrates the general syntax of an Indus Text:

I
Introductory Phrase( s) Substantire Phrase
I

Attri- . . Substantive Suffix Attri- Substantive Suffix( es)

bute(s) (root/stem) (case bute(s) (root/stem) (Nominal)
marker)
A provisional analysis of three specimen texts is givqt below to illustrate
the · syntactical pattern of the Indus Texts:

2476 V ~ '1 II .' 0

~ ~ .~
~
*
3091

5261
V D q ~ ~
E n C B A

AB Introductory (A ttributrve) Phrase

A Substantive
B Suffix (case-marker)
CDE Main (Substantive) Phrast"
C Attribute
D Substantive
E Suffix (nominal)
17 What to We know about tnt Indus Script? Neti Neti ('not this nor that')

8. Methods to test the validity of decipherment:

Though none of the claims of decipherment has received general
acceptance, it would be unwise to ignore them or dismiss them out of ha nd . At
least some of the attempts are based on years of serious study and may provide
valuable clues or insights even if they are not successful. It is th ere fore
necessary to evolve some objective criteria to assess the various claims of
decipherment. The foregoing discussions leap us to three simple but decisive
tests for a preliminary s<..Teening of the c1aims. ·33

(1) Test of Direction 34

The general direction of r:eading the Indus inscnptions from the right is
now so we ll established that we can safely leave out of serious consideration
any attempt to read the script generally from the left. A claim for
decipherment will also be suspect if the decipherer mechanically reads all the
lines from the right and is unable to identify th e occasional reversal of direction
in the inscriptions, which can be done quite easily in most cases with the help of
the sign sequences.

(2) Test of word segmentation 35

As a result of the analytical and structural studies, we can now
confidently demarcate word boundaries in the Indus inscriptions. A proposed
readi ng is suspect if it does not match word boundaries indicated by
segmentation analysis. For example, if a text ABCDEF is segmented as
AB/CD/EF by structural analysis, a linguistic re ading ABC/DEF wiH be
unacceptable . If several such cases of mismatch occur in a decipherment
model, the whole olaim is suspect.

(3) Test based on Frequency-Distributioh analysis 36

Since we know the freque ncy-distribution pattern of the signs in the
Indus inscriptions, we can match the data with those for the sounds in the
language proposed by a would-be decipherer. The readings are suspect if there
is no reason able match . For example, the phonetic values ala proposed for the
JAR sign do not seem to be possible since the vowel signs are expected to occur
initially in a syllabary of open syllables while the JAR sign avoids the initial
position altogether. :n Another value proposed viz. sa has a better fit. especially.
because, as a gramatical morph, it is both final and a separable suffix like the
JAR sign. 38 But since the JAR sign never occurs initially, a different sigri for sa
has to be postulated for this position, which is unlikely iQ the phonetic script
assumed by the model . 39
Journal of the Institute of Asian Studies J8

The tests mentioned above are of general applicability. In other words ,

any proposed decipherment will have to satisfy these criteria irrespective of the
methods adopted. However the tests .are negative in character. They can
invalidate a claim as not being consistent with the criteria, but they cannot
prove that a proposed decipherment which passes the tests must necessarily be
correct. It can only be·said that such a decipherment appears to be prima tacit
sound and deserves serious study. The results of the tests can therefore be
summed up as neti neti ('not this nor thaf). However they do serve to warn us
of the pitfalls ahead and to point towards the . likely direction of fruitful
research.

9. Beyond the structural studies:

Alphabetic and syllabic phonetic scripts can be deciphered without
bi-lingual records provided sufficient material for analysis is available , as
proved by the ' spectacular success of Michael Ventris in deciphering the
Linear-B Script. A mainly logogra'phic system like the Indus Script is unlike.ly
to yield itS' secrets in this manner. In the absence of hi-lingual records the
possibility of a complete decipherment of the Indus Script is quite remote. In
fact even a bi-Iingual text may not help in determining the phonetic values of
word-signs not present in that record. As far as I can see at present, further
progress in understanding the Indus Script beyond the structural studies can
come only through a study of the ideograms in the Indus Script utilising -
(1) The archaeological context,
(2) Parallels from contemporary pictographic scripts , and
(3) Parallels from survivals of the Indus traditions in the later Historical
Period in India .
The purpose of such a study is not to 'read' th~ script but to achieve a broad
comprehension of the contents of the inscriptions through 'the context of ·
situation'

(1) Archaeological Context

Attempts to decipher the Indus Script have been based mostly on
linguistic ana analytical techniques, and very little attention has been paid to
the archaeological context of the inscriptions. This is unfortunate as analysis of·
the inscriptions with reference to the archaeological context is likely to yield
valuable clues to their contents, even before the script is deciphered. For
example, the starting point of Virolleaud's decipherment of th~ Ugaritic Script
was the discovery of the words for 'axe' and 'owner' found on a series of small
bronze axes. 40 Before one tries to read an . unknown inscription, one must
19 What to We know about the Indus Script? Neti Neti ('not this nor that')

surely know where it comes from, on what type of object it is inscribed and
what other clues are available from the circumstances of its ~iscovery . A good
example of the use of archaeological context is the recognition of the
HORNED PERSON sign (No . 8) as an ideogram representing a divine,
prie tly or noble personage by comparing it with the horned personages
depicted in the Harappan pictorial motifs (e.g. 2420,2430). This interpretation
is corroborated by a nother sign (No. 171) which often precedes this ideogram
and i identical in shape to the Sumerian sign for 'great' .

Unfortunat e ly the stratigraphic data for Mohenjodaro . and Harappa

cannot be easily interpreted or re lat e d to the inscriptions. The other sites
(except for Lothal and Oaimabad rece ntly) lack full publication. I hope that
with the publication of the original Fieldbooks of Mohenjodaro now being
undertaken by th e German 'Proj ect Moh e njodaro' and fuller publication of
data from other sites, researchers will turn their attention more to the
archaeological context of the I ndus inscriptions. ~ I By way of illustration, I have
reproduced here our study of the distribution of the most frequent 'phrases' at
Mohenjodaro with refere nce to locations (PI. 8). ~2

(2) Parallels from Contemporary Pictographic Scripts

The Indus Script appears to be an independent invention, judging from
the distinctive character of its signs. The invention must have come from the
diffusion of the idea of writing rather than direct borrowings. ~-' However given
the nature of pictographic writing, it is not unlikely that such scripts may have
similar signs with similar meanings (but not similar sounds). Langdon, Hunter
and Heras have published lists of correspondences between the Indus and the
Sumerian, Egyptian and Proto-Elamite scripts.~~ However it should be
emphasised that similarity in form does not guarantee similarity in meaning as
we know that the same mea ning was expressed by differe':lt ideograms in
various scripts and that the s{lme ideogram may have different meanings.
Notwithstanding ·these limitations, ideographic parallels from corrtemporary
cripts may provide valuable clues provided they are supported by contextual
and other evidence. I shall illustrate the possibilities with an example.

'SEVEN CITIES' 1111

III
The interpretation of a pair of Indus signs as SEVEN CITIES has
gained wide acceptance from scholars. Waddel (I (25) was the first scholar to
identify the pair with the Sumerian equivalents imina bara and to suggest the
Journal of the Institute of Asian Studies 20

meaning 'Heavenly House' (by aBoting the SEVEN sign, not the numeral but
an ideographic value).45 He was followed by F W. Thomas (1932) who pointed
out that the Indus ideogram in question is "too similar to the Sumerian sign for
CITY WALL and at the same time tOQcomplex to be otherwise than identical
with it. ,,46
After a long interval waddel's almost-forgotten theory of 'Indo-
Sumerian' has been resurrected by Kinnier Wilson (1974) in a more
sophisticated attempt. 47 Kinnier Wilso'n equates , the Indus sign-pair with
Sumerian imin bad meaning 'Seven (walled) Cities'. He has drawn attention to
the occurrence of the phrase bad'imin in a Sumerian epic as the name of a place
to the east of Sume.r. Bailey (1975) pointed out the apparent equivalence of
Sumerian bad imin with sapta sindhavas in the Rigveda and hapta hindu in the
Avesta. 48 He also suggested that sapta sindhavas could be interpreted as
'Seven High Places' , prol5ably the Harappan name for the Indus region , which,
was later adopted by the incoming Aryans into their language. Mitchiner
(1978) and Atre (1983) have also supported this identification. 49
This example ,is quite instructive. What has made the identification
attractive and acceptable is th~ independent corroboration it has received from
the near-identical signs in the Sumerian Pictographic Script as well as
attestation of the name from ,a ncient Sumerian, Vedic and ,Avestan sources.
The example also illustrates how Harappan names can be recovered from
survivals in the later Indian tradition if only we know where to look for them.

(3) Bi-lingual Parallels from Indian . Historical Tradition

The example cited above leaves one question unanswered. As pointed
out by Bailey sapta sindhavas could not have bee,n the original name of the
Indus region in the Harappan times. The linguistic diversity of the Sumerian,
Vedic and Avestan names shows them up to be no more than loan translations.
What was the original name then?
The Indian historical tradition has come down to us in two main
linguistic streams, viz., Indo-Aryan and'Dravidian. It is likely that due to
prolonged bi-lingualism and racial fusion in the Indian sub-continent,
Harappan names passed into the Indo-Aryan as loan-words and translations. It
is therefore useful to search for bi-lingual parallels from both Indo-Aryan and
Dravidian sources while attempting to interpret the ideographic signs. The
adyantage of the method of bi-lingual parallels is that it is not necessary to
mflke any a priori assumption about the linguistic affinity of the Harappan
lahguage~ even while hoping that accumulation of evidence would ultimately
help to resolve this question. 50

•
21 What to We know about the Indus Script? Neti Neti ('not this nor that')

These theoretical considerations led me to look for a Dravidian parallel

for the SEVEN CITY ideograms of the Indus Script. To my pleasant surprise I
found an exact equivalent, viz., e!.-eyil meaning 'Seven (walled) City ' occuring-
in Pura-nanuru (33:8), a compilation of bardic poetry dating from about the
beginning of the Christian Era, but probably incorporating much earlier
traditions. (d. eyil: fortification, wall, city, town; DEDR 808).51 The poem
prai es the valour of a Chola prince who sacked the Pandyan city of El-eyil
after breaking through the city gates (kadavam} .52 Further search turned up
another variant el-iL. meaning 'Seven House', a name which appears to be
equally ancient a~d occurs seven times in five anthologies of old Tamil Sangam
poetry belonging to the same age . 53 EI-il is identified (Narr .. 391 :6-7) as a city
situated on a hill in the Konkan region to the north-west of the Tamil Country.
(It i significant that Tamil tradition refers to the migration of Tamil tribes
from the Konkan and Tulu regions and further north-west. suggesting that the
name is a dimly-remembered tradition ante-dating the migration).5-l Another
interesting point is that since the place-names occur in the singular in the Old
Tamil sources, it appears likely that e!.-Ie!.u was not used in the literal numerical
sense of 'seven' (DEDR. 910), but as a homonym meaning 'high, elevated' (d.
c/.u-: height, elevation, eminence etc. DEDR. 851), reminding us of Bailey's
'High Places'. And finally, the availability of two similar, but not identical,
names even within the Dravidian tradition should caution us against regarding
either of them as th e actual 'reading' of the Indus ideograms.

10. Future Prospects

Even though no attempt to decipher the Indus Script has succeeded so
far , I feel that the re is no cause for undue pessimism. More textual material is
being continually added from fresh excavations. Publication of the critical
ditions of the T e xts, computerised Concordances and statistical data has laid a
firm foundation for further progress in the study of the script. In particular the
two Concordances, Finnish and Indian, have triggered a spate of analytical and
structural studies which are rapidly advancing our understanding of the script.
The recent publication of the first volume of the Corpus of Indus Seals and
Inscriptions (Joshi and Parpola 1987) with excellent reproductions of both the
original seals and impressions from the Indian collections is most welcome, and
I look forward to the quick publication of the next two volumes comprising the
material available in Pakistan and elsewhere. ·( have no doubt that with the
availability of this magnificent edition of the originals and the two Concord-
ances, studies on the Indus Script would gain greater momentum in days to
come .
Journal of the Institute of Asian Studies 22

NOTES AND REFERENCES

1. A clay disc from Phaistos in Crete dated to ca. 17 Cent. B.C. It i stamped on
both ·side with pictographic signs in a piral arrangement. The cript is still
undeciphered. A.J. Evans, Scripta Minoa, vol. I , Oxford , 1909, pI. xii.

2. The more recent studie on the Indus Script (including claims of decipherment
and their reviews) ' ar~ listed in -
I. Mahadevan and Mythili Ranga Rao , 'The Indus Script and Related
Subject: A Bibliography of Recent Studie (1960-86)" Tamil Civilization,
vol. 4, no. 3-4, 19 6 , pp . 214-37.
Four other claims of decipherment have appeared ub equently:
Subha h C. Kak, 'The Study of the Indus Script: General Considerations',
Cryptologia, vol. XI, no. 3, 19 7 , pp. 1 2-191' and 'A Frequency Analysi of
the Indu Script' , Cryptologia, vol. XII, no. 3, 1988, pp . 129-142'
B.V. Subbarayappa, 'Indus Script: the Womb of Numbers' , Qty. ll. of the
Mythic Society, vol. LXXVIII , no . 1 & 2, 19 7, pp. 126-63;
S. Kalyanaraman , 'The Indus Script: An Economic Ghronicle', (Paper 1-6;
unpubl.) , 1988, A ian Development Bank , Manila ;
K.K. Raman , ' Key to Harappan Script' , The Week , ochin" Apr. 3 19 ,pp.
34-38.
This Ii t does not include the most recent analytical or descriptive tudie not
claiming to be decipherments.

3. This is the conventional dating (after Wheeler) . However the chr.onology of

the Indus Civilization 'i still a matter of debate . The mo t recent estimate
based on Radio-carbon dating are: ca. 2900-2100 B.C. for the Early Indus ,
2200-1 00 B .C. for the Mature Indus and 1 00-1300 B.C. for the Late lndu
Periods. Applying the MASCA correction, the dat rang for Early and
Mature' lndus periods are ca. 3200-22 0 B . . and 2700-21 0 B.C. re pective-
Iy. K . . Ramachandran, 'Dating the Indu ivilizati o n', Frontier of the Indus
Civilization, ed. B.B . Lal and S.P. Gupta, New Del hi, 1984, pp. 538-39.
Recent excavation, especially at Mehrgarh , Pakistan, have added a new
dimension to the prehistoric equenc of the Greater Indus a rea, progressi e-
ly extending the beginning of ' ettled life to the commencement of the 7th
Mill. B. . J .F. Jarrige , 'Chr no logy of the Earlier Periods of the Gr ater
Indu a een from Mehrga rh, Pakistan', South Asian Archaeology 19 I.
ambridge, 19 4, pp. 21-2 .

4. S. P. Asthana , History and Archaeology of India's contacts with other countries

from earliest Times to 300 B. c.. D Ihi , 1976;
Shereen Ratnagar. Encounters: The Westerl Trade of the Harappa Civiliza-
tion , New Delhi 19 1;
. . Lamberg-Karlovsky, 'Trade Mechani m in Indu -Me op tamian Inter-
relations', JAOS, 92, 1972, pp. 222-229.
23 What to We know about the Indus Script? Neti Neti ('not this nor that')

5. Scattered notices on the occurrence of Indus sea ls in West and .Central Asia n
sites are brought together a nd di scussed in O.K . Chakrabarti, ':Sea ls as a n
evidence of Indus-West Asia Interrel ations', History and Society, Essays in
honour of Prof. Niharranjan Ray, ed. D .P. Chattopadhyaya, Calcutta, 197X, ,
pp.93-116.
Unpublished as well as mo re recent finds are listed in R .H. Brunswig Jr. ,
Asko parpola and Daniel Potts, 'New Indus Type and Related Sea ls from th e
Near East', Dilmun : New Studies in the Archaeology and Early History of
Bahrain, ed. D.T. Potts, Be rlin , 19X3, pp . 101-115 ;
V.M . Masson , 'Seals of a Pro-Indian T ype from Altyn-depe ', Th e Bron ze-Age
Civilization of Central Asia: Recent Soviet Discoveries, ed . P.L. Ko hl, New
York, 1981, pp. 149-162.

6. Seals imported from West Asian region into th e Indus are discussed in O .K.
Chakrabarti (see n . 5) and, J.P. Joshi and Asko Parpol a, Corpus of Indus
Seals and Inscriptions, vol. I, Helsinki a nd New De lhi , 1987, pp . xii-xv. The
authors suggest th at the stepped-type a nd T-shape sea ls found at Harappa
(H-165-66 in this vol. ) and a cylinder seal from Mohenjod aro (M-4l9) sho w
NE Iranian influence . They suggest that " there a re weighty reasons to ass um e
that the NE Iranians represent the first wave of Aryan-speaking immi gra nts in
South Asia"
For cultural contacts and interrelationship betwee n the Indus a nd West Asian
regions see -
E.C.L. During Ca~pers , 'Some motifs as evide nce for maritime contact
between Sumer a nd the Indus Vall ey', Persica, 5, 1971 , pp . 107-18 , pI. viii-xi;
and 'Sumer., Coastal Arabia a nd the Indus Valley in Proto-literate a nd Ea rly
Dynastic Eras: Supporting Evidence for a cult ural linkage ' , JESHO , 22,1979 ,
pp . 121-35;
S. Parpol a, A. parpol a and R .H. Brunsw ig Jr. , 'The Meluhha Vill age ,
Evidence of acculturatio n of Hara ppan T raders in Third Millennium
Mesopota mia ?', J ESHO , 20, 1977, pp . 129-165;
Asko Parpola , 'New Correspondences between Harappan a nd Nea r-East e rn
Glyptic Art', ·South Asian Archaeology 1981, ed. B . Allchin, Ca mbridge,
1984, pp. 176-195;
Asko Parpola , The Sky-garment (A study of the Harappan reli gion and its
relation to the Mesopotamian and late r religions) , Helsinki, 1985 . The
'trefoil' pattern decorating the robe of the priest-king stat ue from Mo he n-
jodaro is identified with Mesopotami an 'sky-garment' motifs (with astr;.o '
significance), and the 'tarpya' garme nt of the Vedic ritual.

7. I.J. Gelb , A Study of Writing, (revised ed .), C hicago , 1963 . The best
introduction to the theory of writing.
Two other works (more descriptive) which may be usefully consulted are -
David Diringer, The Alphabet (2 vols.), London, 1968;
Hans Jense n , Sign , Symbol and Script, tr. from G e rmfln . G eorl!e Unwin,
London , 1970.
Journal of the Institute of Asian Studies 24

8 For an alternative theory that writing originated from reckoning , see Denise
Schamandt-Besserat, 'Reckoning before writing', Archaeology, 32:3, 1979,
pp. 23·31. Her theory that writing originated from Neohthic day tokens used
for reckoning \s persuas\ve~ but these may be regarded as complementing
pictures as sources for the devetopment of writing.

9. Gelb (n. Tabove) prefers the term 'logogram '. tn my view 'ideogram' should
be retained as a useful word to connote sem;e-signs as distinguished from
sound-signs (phonograms) formed by rebus , both of which are induded in the
term 'logogram'. The term:pic(ographic' refers to the external pictorial aspect
while 'ideographic' refers to the internal structure of a script.

10. G.R . Drive'f. Semitic Writing. rev. 3rd ed . by S.A. Hopkins, London, 1<)76~
Joseph Naveh, Early History of the Alphabet: An IntroductionJo WeJt Semitic
Epigraphy and Paleography, Leiden, 1982;
J. D. Hawkins, 'The origin and dissemination of writing in Western Asia ' , The
Origin of Civilization, ed. P.R.S . Moorey, London, 1978.

11. The best sources for facts about the Indus Script are stilJ the original
excavation reports, espeCially -
J. Marshall~ Mohenjodaro and the Indus Civilization, 3 vots .• London, 1931;
E.J.H. Mackay, Further Excavations at Mohenjodaro, 2 vots. , New Delhj,
1937-3R;
M".S. Vats, Excavations at Harappa. 2 vols .•. Calcutta. t940.
For scattered and later hnds , the introduction and documentation In the
foHowing works may be consulted:-
I. Mahadevan, The Indus Script: TeXIS, Concordance and Tables, New Delhi,
1977 (cited as the [ndian Concordance). References to sign and text nu'mbers
are from "this volume;
Kimmo Koskenniemi, and Asko Parpola, A Concorda.nce 10 the Texts in the
Indus Script, Helsinki, 1982 (cited as the Finnish ConcordanceL
J.P. Joshi and Asko Parpola, Corpu5 of Indus Seals and Inscriptions, vol. I. ,
Helsinki and New Delhi, 1987.
[2. The eartier sign lists have been superceded by the comprehensive lists
published in the [ndial) Concordance (pp. 32-35) and the Finnish Concord-
ance (pp. 20-2t). For further discussion on the signs see para 6(\) be\ow.

13 . Out of 419 signs listed in the Indian Concordance, 179 signs have variants
totalling 641 forms recorded separately in the List of Sign Variants (Appendix
1, pp. 785-792). The cr'ticism that the Concordances which use ' normalized'
s;gnary may result in loss of data for further research is ba~d -on a
misunderstanding of the ~urpose of a concordance. which is a reference tool
to locate readily the occurrences of each sign in every context. The serious
researcher should have no diffculty in looking up the actual forms in the
originals especiaUy now with the availability of the photogntphic edition of the
onginals (Joshi and Parpo}a 1987).
25 What to We know about the Indus Script? Neti Neti ('not this nor that')

14. I do not" accept S.R. Rao's claim about the evolution of a Late Harappan
linear and alphabetic script of 20 signs occurring in graffiti on pottery. For a
discussion on this question, see S.R. Rao, Tl;ze Decipherment of the Indus
Script, Bombay, 1982; and I. Mahadevan, 'S.R . Rao's Decipherment of the
Indus Script', The Indian Historical Review, vol. 8. no. 1-2, 1981-82, pp.
58-73 .

15. The object-types and pictorial motifs are catalogued and illustrated in the
Indian Concordance (App. Ii-III, pp. 793-813).

16: There are two good studies on the copper tablets from Mohenjodaro:
B.M . Pande, 'Inscribed Copper Tablets from Mohenjodaro: A Preliminary
analysis' , Radiocarbon and Indian Archaeology, TIFR. Bombay. 1973, pp.
305-322;
Asko Parpola, 'Tasks, Methods and Results in the Study of the Indus Script',
JRAS, 1975, pp. ;78-209. .

17. Important studies on the direction of the Indus Script include:

1. Marshall, 1931, pp. 409. 427-28;
G.R. Hunter, The Script of Harappa and Mohenjodaro, London, 193.4, pp.
37-43;
A.S.C. Ross, 'The Direction of the Mohenjodaro Script', New Indian
Antiquary, 2, 1939-40, pp. 554-558;
B.B. Lal, 'The direction of writing in the Harappan Script', Antiquity, 40.
1966, pp. 52-55;
G. V. Alekseev, 'The characteristics of the Proto-Indian Script', Preliminary
Report on the Investigation of the Proto-Indian Texts (in Russian). ed. Y .V.
Knorozov, tr. by A.R.K. Zide and K.V. Zvelebil in The Soviet Decipherment
of the Indus Valley Script, Mouton. 1976. pp. 17-20;
Indian Concordance, pp. 10-14.

18. B.B. Lal, 'A Further Note on the Direction of Writing in the Harappan
Script, Puratattva, vol. 1, 1968, pp. 15-16, pI. 1. .

19. I. Mahadevan , 'Recent Advances in the study of the Indus Script', Puratattva,
no. 9, 1980, pp. 34-42. As mentioned here, Text 8221 has been copied in the
wrong direction in the Indian Concordance erroneously. It has to be corrected'
and read from the left for reasons discussed in the cited paper.

20. Indian Concordance, p . .14 .

21. {bid, p. 26, n. 16;

H. Heras, Studies in Proto:lndo-Mediterranean Culture, vol. I, Bombay,
1953. He attempted to read all the second lines from the left in the
boustrophedon mode, while most of them have to be read from the right (pp.
98. 104, 106, 107).
Journal of the Institute of Asian Studies 26

22. E.g. Clyde Ahmad Winters, 'The Harappan Script', JI. of Tamil Studies, 30,
1986, pp. 89-111. He states that the inscriptions should be read from right to
left, but proceeds to do 0 from the original se'als (not from the impressions),
thus reading the inscriptions in effect from the left. It is not perhaps surprising
that he should find the signs to have the sound values of the African Manding
Script!

23. E.g. S.R. Rao, The Decipherment of the Indus Script, · 1982, Figs. 17.44,
17B. 66, 17C.107, etc.

24. E.g~ S.R. Rao, ibid, Figs . 14.1 and 2; 23C.65 and 73; 26.14 and 15 etc., where
he reads the same texts in either direction.

25. Maunce Pope , The Story of Decipherment from Egyptian Hieroglyphic to

Linear 8, London , 1975;
E.1.W . Barber, Archaeolo.f!ical Decipherment: A Handbook, Princeton, 1974.

26. S.R. Rao advocates segmentation of signs on this basi into basic igns and
auxiliary mark (S. R. Rao, 1982, Fig . 47-61) .

27. See my criticism of this technique in the review article cited in n.14 above:

28. Sydney Smith (in Marshalt, 1931), pp . 415-422;

Papers by M.A. Probst, A.M. Kondratov and Y .V. Knorozov in Soviet
Decipherment, pp . 23-29;
Seppo Koskenniemi, A ko Parpola and Simo Parpola, 'A Method to clas ify
characters of Unknown Ancient Script', Linguistics, 61, 1970, pp. 65-91;
W. A . Fairservi Jr., Excavations at Allahdino II I: The Graffiti, A model in the
Decipherment of the Harappan Script, papers of the Allahdino Expedition,
New York, 1977 Fig. 1 (the grid);
I. Mahadevan, ' Recent Advances in the Study of the Indus Script', Puratallva,
9, 1980, pp_ 34-42 ;
Kimmo Koskenniemi , 'Syntactic Methods in the Study of the Indus Script',
Studia Orientalia, SO, 1981, pp. 125'-136'
I. Mahadevan , Toward a Grammar of the Indus Texts: Intelligible to the.
eye, if not to the ears', Tamil Civilization, .vol. ( no. 3-4, 1986, pp. 15-30;
Gift Siromoney and Abdul Huq , 'Segmentation of Indus Texts: A Dynamic
Programming Approach' , Computers and Humanities, 22, 1988, pp. 11-21;
Abdul Huq, Computer Analysis of the Indus Script, Madras, 1988, (unpub!.
Ph.D . thesis).

29. Even this 'phra e ' qn be further segmented . See 8001 for the last two igns of
thi 'phrase a a complete text , 4334 in which the first three signs of the phrase
can be isolated as a segment, and 2549 in which the first two sign of the
phrase can be segmented, ultimately leading'to the conclusion that each of the
4 igns is a word forming integral phra es with 2, 3 or 4 'signs.
27 What to We know about the Indus Script? Neti Neti ('not this nor that')

30. Indian Concordance, p. 17 .

31. See references in n. 28 above; However I am responsible for the interpreta-

tion of the results of the analyses.

32. E.g. the famous suggestion of Heras that the FISH sign, min in Dravidian ,
stands for 'star, planet' from a homonym .

33. L Mahadevan , 'Claims of decipherment of the Indus Script: Some objective

method s to test their validity', SAARC Workshop on Epigraphy, Mysore,
1985 (unpub!.) .

34. See para 5 and notes 17-24 abo-ve.

35. See para 6(2) and notes 28-29 above.

36. See para 6(3) above .

37. S.R. Rao , 1982, p. 32.

38. John E. Mitchine r, 5wdies in the Indus Valley Inscrip rio ns , New Delhi. 1<}7X ,
p. 76 , Table 7:
Subhash C. Kak, 1987 , p. 191, Table 3.
39. Subhash Kak (1988) makes an interestin g comparison between the ten most
frequent signs of the Indus Script and the most frequent sounds of Indo-A rya n
as reco rded in the Brahmi script. [n my view the attempt is not successful
because (1) It is an established rule in the field of decipherment that
comparisons of the external shapes of signs betwe'en two scripts. one of thcm
being unknown, will be misleading as similarity in form may not mean
similarity of sound. For, after all, linear scripts can all be made up by .\
combination of a few elemc:nts like circle, square. triangle and curve . This
rule applies even to scripts close to each other in .time like the Cypriote,
Linear A and B Scripts (Gelb , 1963, p. 144; Barber, 1974. pr . 97-9~):
(2) Kak's list excl udes some out of the ten most frequent signs and includes
others much less frequent. As a result of this, there is no convincing match
betwee n the Indus signs and the Brahmi letters and their sounds in ·
Indo-Aryan: (3) Many of the comparisons of shape are far-fetched , c . f.
MAN with to. JAR with so etc.

40. For an acco unt of the decipherment of the Ugaritic Script, see Pope, 197,), pp.
117-22.

41. M. Jansen and G. Urban (ed.), Interim Reports, vol. I, Reports on field wlnk
carried out at Mohenjo-Daro, Pakistan, by the ISMEO-Aachcn University
Mission, 19H2-~n, Aachen, 19H3 .
Journal of the Institute of Asian Studies 28

42. Reproduced from 1. Mahadevan and Mythili Ranga Rao, ' ~rchaeologic(ll
Context of Indus Texts at Mohenjodaro', il. of the Institute of Asian Studies.
voL 4, no . 2, 1987, pp. 25-56.

43. It is becoming increasingly clear from recent studies that the Indus Script is an
indigenous invention with a long period of gestation and growth as revealed
by pre-Harappan pottery graffiti from many sites, especially Rehman Dheri in
Pakistan. This of course does not exclude similar pictographic signs with
similar meanings in contemporary scripts, derived probably from a common
stock from Neolithic times.
Recent tudies include:
B.B. La!, 'From the Megalithic to the Harappa: Tracing back the graffiti on
pottery', Ancient India , 16, L960 , pp . 4-24;
W.e. Brice, The structure of Linear A. with some Proto-Elamite and
Proto-Indic Comparisons', Europa: Festschrift Ernst Gwmach. Berlin, 1967.
pp. 32-44;
Farzand Ali Durrani. 'Indus Civilization: Evidence ' West of Indus' , Indus
Civilization: New Perspectives. ed . A.H. Dani, Islamabad. 1981. pp. 133-37.
(esp. pI. xvii-xix of a seal and pottery graffiti from Rehman Dheri) ;
D . Potts, The potter's marks of Tepe Yahya', Paleorient. vol. 7. no.!, 1981.
pp . 107-122;
D. Potts , The Role of the Indo-Iranian Borderlands in the formation of the
Harappan Writing System', Annali dell' Istiluto Orientale di Napoli. vol. 42,
1982, pp. 513-519.

44. Langdon (in Marshall 1931) pp. 434-455 ;

G.R. Hunter, 1934, pp. 203-210;
H. H'eras, 1953, pp . 248-278 .

45 . L.A. Waddel, Indo-Sumerian Seals Deciphered, 1925, Reprint (Delhi, 1972),

pp. 92, 95, 98.

46. F.W. Thomas , Review of Mohenjodaro and lhe Indus Civilization by

Marshall , JRAS, 1932, pp. 453-66.

47. J. V. Kinnier Wilson, Indo-Sumerian: A New Approach to lhe problems of the

Indus Scrip;, Oxford, 1974, pp. 3-5.

48. H.W. Bailey , <[ndian Sindhu. Iranian Hindu '. BSOAS. vol. 38, no . 3, 1975.
pp . 610-611.

49. John E. Mitchiner, 1978, pp. 19-33;

Shubhangana Atre, 'Bad-imin: the Union of Indu Cities', Bull. Deccan
College Research Institute, vol. 42, 1983, pp. 18-24.
 2g What to We know about the Indus Script" Neti Neti (,not this nor that')

50. I. Mahadevan, 'Study of the Indus Script through Bi-lingual Parallels',

Procds. 11 Ann. Con/ceo 0/ Dr. Linguists (1972). Trivandrum, 1975;
(Reprinted in) Ancient Cities of the Indus. ed . G. Possehl, New Delhi, 1979,
pp. 261-267;
1. Mahadevan, 'Study of the Indus Script: A Bi-lingual Approach', SOllth
Asian Languages: Structure. Conllergence and Diglossia. Hyderabad. 1986.
pp. 114-119.
Implicit in the search for Dravidian pa~aJle]s for the Indus ideograms is the
hypOthesis that the Harappan language, not being Indo-Aryan. could only be
Dravidian. The Indo-Aryan hypothesis seems to be ruled out on the basis of
(a) chronological considerations, as the incoming Aryans reached the Indus
Region only when the Mature Phase of the Indus Civilization was ending
around the beginning of the 2nd Mill B.C. (b) difference in culture. as the
Indus Civilization was largely urban wh~e the Aryans were nomadic and
pastoral; and (c) the absence of the horse and the chariot with spoked wheels
in the Harappan glyptic~rt. Thus Harappan parallels in Indo-Aryan culture is
best explained by the substratum Dravidian influence after the fusion of the
two cultures.
Also see I. Mahadevan. ' Dravidian Models of Decipherment of the Indus
Script: ~ Case Study', 'Ta":'iI .Civilization. vol. 4. no. 3-4. 1986. pp. 133-143.

51. T. Burrow and M.B. Emeneau. A Dravidian Etymological Dictionary. 2nd

rev. edn ., Oxford, 1984 (DEDR).

52. The city has not been identified. It is hkely that the bard's praise is
conventional, recalling a past exploit of the prince's an~esto;;. d. praise of
Cheral Atan, a Chera prince, for feeding both the armies during the Bharata
War (Puram. 2).

53. Akam: 152, 345. 349~ Puram: 166~ Kurlllll. : 138: arr. : 39J : POrllll: 63.

54. M. Raghava Ayyangar: Velir Vara{Qrlt (in Tamil). 1913. 3n:1 edn .. Madras.
1964, pp. 1-37. His 'main thesis is that the Velie. a land-owning agricultural
Tamil community, led by their Chief Agastya. migrated from Dvaraka to tbe
Tamil country, passing through the Konkan and Tufu regions, many centuries
before the Sangam Age. The Velir shared several myths and traditions with
the Yadavas of the North and the Cbalukyas, Hoysalas and other Dynasties of
the Deooan.
See also I. Mahadevan. 'Agastya and the Indus Civilization'. JI. of Tamil
Studies, 30, 1986, pp. 24-37.
 t f t r
t
It

t 11
1*1
2t
i
~m(to ~
12t 14t
3

1St 13
"4 5 6

m 16
7

~
17t
8t

A
18
)~)

~
9t

19t
1:
10

t'
20

r r
21 22
~
23
~ 1j24 25
';\1' ~ ~ ~~
I
26
I
29t Xl 28t 30

~ ~ ~~ td ~ ~ k t
~
31

41
32t

42
33

l- i 43
fiU
34

44
JSt

ti)
45
36

1AV:J ;:; 1
46
J7

47
*
J8t

48t
39

49t
«>t

~
sot

1 If OC
Sit 52 S3t
~ S4t
~
SSt
[
56t
:
57t
XX
58
~
59t
I~'I
•
60t

:~ft: (~1 (~~) (~i) ~ '~.

I
66
1 ~ ,I~I, V ~
61 62 63 64 65 67t 68t 69t 70t

'~.
, I ~ t~'
I I
{f t .~ i i f ~
7l nt 73t 74t 75 76t :n 78t 79 80

(i) ~
8lt 82
+ &3
~
84t
I~\ 85 86t
I II
87t
:II:
88
III
89t
1I~
90t

'" I II
IIr 91t
III
92
III III
94t
93
1111
9S
11111
96t 97 98t 99 100
II

Plate 1. Signs of the Indus Script

(Source: Sign~ 1-100 Sign List, Indian Concordance, p.32)
 U A 'm "i ffi1 0
~002
4003
4004
* «< § §. I::: ~ CJ C,
,UJ~~ 119 ~c
4005 V )~ ! I ~ .Q ~ ~ "0
4006. ~D'II(vO
%, ~
4007
4008 ,1 ~ 1l><J~
~ 1111
40-09 III

4010 ®6(1~
401 1 . / I" 0
~~
~~:

4012 :~: V ~ U 1~fm

1..01J V Iffi1 ~ IIi ill ~
1..014 V ~ ml· .Q 'i oc ~ U ~ ffil
4015 tllllU.Q'i~
4016 V~~~OCI®
4017 ~(f)~'V~1I
4018 V»
4019 V ;} ~ ! I .Q II (f) ~[>(
4020 V A 1m III UJ (i) ".Q tt 9
4021 V A 1m IlllU ,.. ~.
4022 r l ~ ~ t ~ (i) "~
4023 V , ffi( t (1) r »n( II

Plate 2. Texts in the Indus Script

(Source: Text from Harappa , Indian Concordance. p .98)
 A
4619 00 V· A 'm ~ 11«< II ) •
1803 10 r II t A'A 1m ~
'291 1 10 ill f f A 'm ~
3303 ,0 jilt 1: A 'm ~
1557 00 ,U A 1m III UJ 11 OC "~
2446 00 U A 1m III UJ t oc ®II~ 8
4021 00 U £ 1m III UJ ,t- ~ •
4020 00 U A 'm III UJ ®"~ ~,
8019 00 U A 1II111~
III
~ ~

4002 00 U A'm'~mlO
2478 00 U A 1I1111~
III ~

ullm A 'm"~
1 146
1 3/~ [.
2012 00
00
00
VA
A

A
*
U A 1111III 11
1111)
III

1553 00 UA 'm v 1] If~ 1: +

9022 1 a U4)lIi
20 U AUII
A 111

30 «<~
~.

8051 00 U AllUUCl
2032 00 ~ III WA 'm UtI-
7016 00 ~AA
2301 00 rIIllIU~AA

Plate 3. Concordance of Indus Texts

(Source: Indian Con rdance. p.454)
 8221
(as in the original)

1. Unique pottery graffiti from Kalibangan

incised from R .. but to be read , from L.

1/0
~J. D~ VOX)1I0
6112 2618

2. 'Split sequence' indicating Direction

(a) Normal orientation

of signs ~ r ~ :> ex
Reversed orientation ~ '1 ~ ~ 'JJ
I
(b) ~reque~t Right-end II~ \)~
sign-pairs "®
Frequent Left-end
sign-pairs ~ ·V \} U t lJ
(c) Frequent sign-pairs mw ~) III V')
Their reversed order IWIII III lJ ~V
3. 'Direction-Markers' in the Indus Texts.

Plate 4. Direction of the Indus Texts.

 I~I
1. 3157 I •

5477 ~ :~
1177 :~:'i
2. 4289
U ~ II
4143 ;:U~II

3103 t.Qi
2183 't.Q'iOC
3. 4632 U~
6122 tV~
2380 U~I
2444 ~ V~j'
4325 ~~U~'i
4. 2461 tHllDl A1 ~ II oXO @ A-
1437 tmffil1\1
2039 ~(j) ~

4254 ffJJ ItI t 1IIl!) U IJ ~ 'i @

2371 ml~1
2015 tlllW
2605 VIxJ~i'

Plate 5. Procedures for Word Segmentation - I

1. 2476 V r.llI~
2168 U 1.,"*
5069 VU1
2201 VU~
5031 UU~

2. 1
93
, Ii
83

Pairwise
Frequencies V 'f I ;:; " ®
(1010)
40 17

Segmented
Text
V -r i/~/ "®

3 1411 Till 4548 J) iI

2008 filii 4387
~m
1422 r 11111 4508 U 1111
1243 rillIU

1025 filiiIII

Plate 6. Procedures for Word Segmentation - II

 II III llil 11111
1 2 3 4 5

III 1111 1111 11111

III III 1111 1111
6 7 8 9

0 R ~
10 20 "30

1. Numerals in the Indus Script (Variants not shown)

u
2. Terminal suffixes in the Indus Texts

II
...
If

3. Grammatical suffixes in the Indus Texts

Plate 7. Numerals and Suffixes in the Indus Script

FREQ. CITADEL LOWER CITY OTH TOTAL
PAIRS SO SO L HR VS OK OK MO CORP Of.
10
(ML) (MY) (ML) (MY)

4 0 21 15 28 101 7 177 ~91

" ~
1 60.82

1 0 0 4 0 2 17 2 26 184 14.13
~ V
4 0 2 9 11 15 46 7 94 126 74.60
III UJ
0 0 0 0 0 3 0 4 124 3.23
U III
V ,) 0 0 4 4 6 25 2 42 114 36.84

V U 2 0 0 6 9 43 4 65 110 59.09

V ~ 1 0 0 12 4 9 34 2 62 93 66 .67

A V 0 3 7 34 4 51 87 58.62
It

* 0 1 8 2 8 29 3 52 83 62.65

U If 0 0 0 0 0 0 0 0 0 78 0.00

~ I' 0 5 10 20 3 41 76 53.95

1m Im 0 1 4 4 6 23 8 47 70 67 .14

~ II 0 1 0 8 5 26 5 46 67 68.66

I i O· 3 2 4 12 3 26 58 44.83

U 1111 0 0 0 0 0 3 0 4 58 6.90

t ~ 0 0 ' 5 5 6 14 32 55 58 .18

T 6 0 0 0 4 6 14 0 25 54 46.30

1 ) 0 0 4 4 6 19 2 36 54 66.67

TOTAL 15 6 8 100 57 128 463 53 830 1782 46.58

Plate 8. Distribution of Frequent 'Phrases' at Mohenjodaro

(I. Mahadevan and Mythili Ranga Rao, 1987, p.S1)