1

The Scope and Significance of Epigraphy

in Indological Studies

Though it has been frequently stated, the importance of epigraphy in Indology can
hardly be overemphasized. The primary reason for the particular importance of epig-
raphy in the study of traditional India (as compared, e.g., with that of classical Europe
or China) is the extreme paucity, especially in the ancient period, of the type of his-
torical data from literary sources which is available for other major civilizations of
the ancient and medieval world. This situation is a reflection of what might be called
the "ahistorical" orientation of traditional Indian culture. Traditional India, with its
strongly idealistic and theoretical orientation, had little interest in what we in the
modern world think of as "history," and except for a few outstanding exceptions
Sanskrit and classical literature include little in the way of "historical" texts in the
stricter sense of the term. As a result, the history of ancient and early medieval (i.e.,
pre-Islamic) India must for the most part be reconstructed from incidental sources;
that is, sources whose original intent was something other than the recording of his-
torical events as such. Such sources include many branches of literature, both secu-
lar and religious, but these typically give us little more than tantalizing tidbits of
historical information, often distorted, and out of context at that. Far more revealing
are the archaeological sources, including not only the results of formal stratigraphic
excavations but also numismatic and above all epigraphic materials.
It has been authoritatively estimated1 that something like 80 percent of our
knowledge of the history of India before about A.D. 1000 is derived from inscrip-
tional sources.2 Without inscriptions, for example, we would have only the vagu-
est notion of the history of the Gupta dynasty, the greatest northern Indian empire
of the classical period. But the importance of epigraphy goes beyond historical
studies in the narrower sense of the term, that is, political history. The modern study
of most aspects of the cultural history of traditional India, such as the arts, litera-
ture, religion, and language, are also heavily indebted to inscriptions for their basic
chronological and geographical framework. Again reflecting the non-historical

1. D. C. Sircar, EINES 91.

2. For further details, see section 7.1.

3
4 Indian Epigraphy
orientation of traditional Indian culture as a whole, traditional literary and cultural rel-
ics typically seem to exist in a chronological vacuum; for example, the date of original
composition of literary texts, especially in the earlier centuries, is rarely recorded in
the works themselves (although later manuscript copies are often dated). Thus the dates
and even the relative chronologies of major cultural developments are often uncertain
or totally unknown. But here again inscriptional material, with its vast volume and di-
versity of contents, frequently comes to the rescue. In the history of literature, for ex-
ample, inscriptional allusions to and imitations of great classical poets, as well as original
compositions preserved in epigraphic form, provide a bedrock of evidence for the chro-
nological development of Sanskrit poetic literature.3 The same holds true, to a greater
or lesser extent, for nearly all branches of Indology ;4 in the words of D. C. Sircar, "there
is no aspect of the life, culture and activities of the Indians that is not reflected in in-
scriptions."5 Thus epigraphic materials, directly or indirectly, provide almost the only
solid chronological foundation for modern historically oriented studies. This is true
primarily because inscriptions, unlike literary sources, which almost always come to
us only after being copied and recopied through the centuries, are inherently datable,
either by an explicit date or by paleographic estimate.6 A reference to a particular legal
principle, religious sect, philosopher, poet, and so on, in an inscription thus gives at
least an approximate terminus ante quern for a person or event whose date might other-
wise (i.e., from literary sources alone) be impossible to determine even in the broadest
estimate.
It is mainly for these reasons that epigraphy is a primary rather than a secondary
subfield within Indology. Whereas in classical studies or Sinology, for example,
epigraphy serves mainly as a corroborative and supplementary source to historical
studies based mainly on textual sources, in India the situation is precisely the oppo-
site. There, history is built upon a skeleton reconstructed principally from inscrip-
tions, while literary and other sources usually serve only to add some scraps of flesh
here and there to the bare bones. There are, of course, some exceptions to the rule,
most notably in Kashmir, where unlike nearly everywhere else in India a sophisti-
cated tradition of historical writing flourished, best exemplified by Kalhana's
Rajatarahgim. Nonetheless, the general pattern remains essentially valid for politi-
cal, and to a lesser extent for cultural, history.7
The paucity of historical materials in the usual sense for India is balanced, as it
were, by the relative abundance of inscriptions.8 According to Sircar, "About 90,000

3. For further discussion, see section 7.2.

4. For examples and details, see sections 7.3-7.
5. EINES 102.
6. See section 5.4.
7. In this respect, the study of the history of traditional India is methodologically more compa-
rable to the study of the early urbanized cultures such as those of Egypt and Mesopotamia, for which
the principal sources are epigraphic rather than literary, rather than to that of the contemporary classi-
cal cultures of Europe and the Near and Far East.
8. The abundance of inscriptions in India may be attributed, in part at least, to the desire of the
issuing authorities to preserve their records in a form which would survive the rigors of the Indian cli-
mate, where documents on perishable materials such as palm leaves or paper tend not to last more than
a few generations. In particular, the widespread practice of recording land grants and other transac-
tions on copper plates seems to reflect such concerns (see section 4.1.2).
 The Scope and Significance of Epigraphy in Indological Studies 5

inscriptions have so far been discovered in different parts of India."9 These estimated
90,000 inscriptions come from virtually every corner of India and from every cen-
tury from the third (or possibly fourth) century B.C. up to modern times, though their
distribution is by no means equal in terms of antiquity and geographical and linguis-
tic distribution. For example, early inscriptions, that is, those from before about the
beginning of the Christian era are relatively rare, and inscriptions only become very
common in medieval times, from about the eighth century onward. In general, the
bulk of inscriptions in the later period are from southern India and the majority of
these are in Dravidian languages.10
Sircar's estimate of 90,000 inscriptions from India presumably does not include
inscriptions in the neighboring South Asian countries of Pakistan, Nepal, Bangladesh,
and Sri Lanka, which would add many thousands more to the total. Inscriptions in In-
dian languages, moreover, are also found in large numbers over a vast area of Asia and
even beyond Asia, for instance, in Africa.'' The comments made earlier about the prime
importance of Indian inscriptions apply equally to these extra-Indian inscriptions, many
of which constitute our principal, sometimes even our sole, source for the historical
study of the Indianized civilizations of ancient Southeast and central Asia.
The number of known inscriptions, moreover, is constantly growing as new
records, including many of considerable importance, continue to be discovered within
and outside of India every year.12 A complete history of India written today, taking
into account all recent epigraphic discoveries, would be significantly different from
and more complete than one written, for example, only twenty years ago; there is no
reason to think that this pattern should change in the foreseeable future. Moreover,
as also noted by Sircar,13 a great many of the estimated 90,000 inscriptions already
found in India have not yet been published. While it is true that many of the still un-
published records are either of minimal importance or too badly damaged to be deci-
phered, no doubt some of them will, when they are finally published, add significantly
to our historical and cultural knowledge. Sircar thus decries the incorrect notion "that
all important inscriptions have already been discovered and utilized in the reconstruc-
tion of the lost history of ancient India and that there is little else to do."14 Epigraphy is
indeed still a living and developing field, even if it has in recent years suffered some-
thing of a decline in standards in India15 and a general neglect elsewhere.16
The decline of Indian epigraphic studies may be attributed, in part at least, to the
special difficulties and problems involved in this field.17 Not only is the material vast,

9. EINES 91. It has recently been estimated that some 58,000 Indian inscriptions have been pub-
lished (R. Garbini, JESI 19, 1993, 64).
10. According to Sircar (EINES 91), "the largest numbers come from the Tamil- Kannada- and
Telugu-speaking areas—about 35,000, 17,000, and 10,000 respectively."
11. On Indian-language inscriptions from outside of South Asia, see section 4.3.7.
12. Cf. EINES 91 and SIE 13.
13. EINES 91.
14. EINES 109; cf. also 91.
15. See EINES 91-2.
16. Cf. SIE 13 n. 1: "The study of Indian inscriptions . . . is no longer popular among European
scholars." But see also section 6.5.
17. EINES 92.
6 Indian Epigraphy
voluminous, and inherently difficult; it also requires a command of a range of lan-
guages, dialects, and script forms far greater than that needed for epigraphic studies
in most other parts of the world. The inscriptions include materials in virtually all
the major languages of the Indo-Aryan and Dravidian families and also involve sev-
eral non-Indian languages of Southeast Asia and other regions. The variety of scripts
is enormous, these being subject to the same pattern of extensive local and chrono-
logical variation as are Indian languages and dialects. Indian epigraphy is thus a
subject of vast complexity which requires many years of study, but its inherent im-
portance and the rewards to be gained from it more than justify the effort.
 2
Writing and Scripts in India

2.1 General Introduction

The history of writing in India, insofar as it applies to the material treated in this book,1
is essentially coextensive with the history of the Brahml script and its many deriva-
tives, and of the Kharosthi script. While the latter, though important in its day, was
a regional script which died out without any descendants, the former is the parent of
one of the major script families of the world, comprising not only all of the indig-
enous scripts of South Asia (as opposed to those of Perso-Arabic or European ori-
gin) but also several other major scripts of central and especially Southeast Asia. While
the processes of development of these scripts will not be presented in detail in this
book (this being, as explained in the preface, primarily a work on epigraphy rather
than paleography), the general outlines of these developments, particularly as they
relate to epigraphy proper, are presented in the context of a general discussion of the
history of epigraphic writing in India.

2.1.1 Writing in traditional India

On the whole, traditional (i.e., pre-Islamic) India was much less oriented toward the
written word than many other ancient and traditional cultures such as those of clas-
sical China and Japan or of the Islamic world. From Vedic times on, and even to the
present day in some cultural contexts, it is oral rather than written learning that has
always been esteemed in India as true knowledge, an attitude reflected in such prov-
erbs as pustakastha tu yd vidya parahastagatam dhanam, "Knowledge in a book [is
like] money in someone else's hand."2 There is a pervasive, if unspoken, attitude

1. As explained in the preface, the protohistoric Indus Valley writing itself will not be discussed
here except insofar as it relates to the study of the writings of the historical period.
2. Quoted by Ojha, BPLM 14 n. 6 (attributed to the Canakya-mti). The esteem accorded to the
spoken as opposed to the written word should not, however, be overstated. In this connection, Kalidasa's
Raghuvamsa 3.28cd, liper yathavadgrahanena vanmayam nadlmukheneva samudram avisat, "As one
enters the ocean through the mouth of a river, so did [Raghu] enter into literature by learning to write
correctly," has been cited (e.g., in PIP 4 and Falk, SAI 252 n. 2) as an indication of the respect ac-
corded to the written word, at least in the classical (as opposed to the Vedic) tradition. This status is
also reflected in the attribution of the invention of writing to Brahma himself (see later discussion).

7
8 Indian Epigraphy
that the written word is essentially a reflection rather than a true manifestation of
language, and this may explain the relative lack of attention to the aesthetic aspects
of writing in traditional India (2.5.3), as well as the carelessness and imprecision which
characterize many, though by no means all, of the written documents, both epigraphic
and non-epigraphic, of ancient India (2.5.1).
This same orientation no doubt also explains the paucity of descriptions of writing
as such in Sanskrit and related literatures. Discussions of the origin and history of writ-
ing, its varieties, styles and methods, and practical instruction therein are surprisingly
rare in Indie texts, though we do find a few more or less incidental references in some
relatively late texts to the invention of writing by the creator god Brahma.3 This tradi-
tion is also reflected iconographically, Brahma and also his wife (or daughter) Sarasvatl,
the goddess of learning, being regularly depicted in sculpture with a book in hand. But
beyond legendary accounts such as these, the literature of the Brahmanical-Hindu tra-
dition has, on the whole, little to say about writing as such; this, in striking contrast
with its profound fascination with (spoken) language and grammar.
The picture is somewhat different in the heterodox traditions of Buddhism and
Jainism, which (especially the former) exhibit a higher esteem for the written word.4
Moreover, it is only in the texts of these traditions that we find lists of Indie scripts.
Most important among these is the list in the tenth chapter (Lipisalasamdarsana-
parivarta) oftheLalitavistara5 of the sixty-four scripts (lipi), beginning with Brahml
and KharostI [sic], which the future Buddha knew as a child.6 The historical value of
3. E.g., nakarisyad yadi srasta likhitam caksur uttamam / tatreyam asya lokasya nabhavisyac
chubha gatih, "Had the Creator [Brahma] not invented writing, the supreme eye, the course of this
world would not have gone well" (Richard W. Lariviere, ed., The Namdasmrti, University of Pennsyl-
vania Studies on South Asia, vol. 4 [Philadelphia: Department of South Asia Regional Studies, Uni-
versity of Pennsylvania, 1989], 1.78); and a verse attributed in several sources (e.g., Mitramisra's
Vyavaharaprakasa; see BIP 1 n. 3 and PIP 3 n. 2 for further references) to Brhaspati: sanmasike' pi
samaye bhrantih samjayate nrnam/dhatraksaranisrstanipatrarudhanyatahpura, "Because men forget
things within six months, of old the Creator invented letters [to be] set down on leaves."
4. It is thus not a coincidence that the art of calligraphy is more highly developed in Buddhist and
Jaina manuscript traditions than in Brahmanical circles.
5. The Mahavastu (1.135 in E. Senart's edition [Paris: L'Imprimerie Nationale, 1882]) also has a
similar list of thirty-two scripts.
6. The list, according to the edition of S. Lefmann (Halle: Verlag der Buchhandlung des Waisen-
hauses, 1902), 1.125-6 reads:
(1) Brahml (2) KharostI (3) Puskarasarl (4) Angalipi (5) Varigalipi (6) Magadhalipi (7) Maiigalyalipi
(8) Arigullyalipi (9) Sakarilipi (10) Brahmavalilipi (11) Parusyalipi (12) Dravidalipi (13) Kiratalipi
(14) Daksinyalipi (15) Ugralipi (16) Samkhyalipi (17) Anulomalipi (18) Avamurdhalipi (19) Daradalipi
(20) Khasyalipi (21) Clnalipi (22) Lunalipi (23) Hunalipi (24) Madhyaksaravistaralipi (25) Puspalipi
(26) Devalipi (27) Nagalipi (28) Yaksalipi (29) Gandharvalipi (30) Kinnaralipi (31) Mahoragalipi
(32) Asuralipi (33) Garudalipi (34) Mrgacakralipi (35) Vayasarutalipi (36) Bhaumadevalipi (37) Antarik-
sadevalipi (38) Uttarakurudvlpalipi (39) AparagaudanTlipi (40) Purvavidehalipi (41) Utksepalipi
(42) Niksepalipi (43) Viksepalipi (44) Praksepalipi (45) Sagaralipi (46) Vajralipi (47) Lekhaprati-
lekhalipi (48) Anudrutalipi (49) Sastravarta [*lipi] (50) Gananavartalipi (51) Utksepavartalipi
(52) Niksepavartalipi (53) Padalikhitalipi (54) Dviruttarapadasandhilipi (55) Yavaddasottarapadasand-
hilipi (56) Madhyaharinllipi (57) Sarvaruta-samgrahanTlipi (58) Vidyanulomavimisritalipi (59) Rsita-
pastapta rocamana [*lipi] (60) Dharanipreksinllipi (61) Gaganapreksimlipi (62) Sarvausadhinisyanda
[*lipi] (63) Sarvasarasamgrahani [*lipi] (64) SarvabhutarutagrahanT [*lipi]. The names of the scripts
vary considerably in different text editions; compare, e.g., the list as given in Rajendralala Mitra's edition
(Bibliotheca Indica no. 15 [Calcutta: The Asiatic Society, 1877], 143-4).
 Writing and Scripts in India 9
this list, however, is limited by several factors. Its date, first of all, is uncertain. While
the LaUtavistara as such is known to be fairly old (a version of it had already been
translated into Chinese in A.D. 308),7 different versions of the text must have existed
at various times, and certain indications in the extant script list (notably the inclu-
sion of no. 23, Hunalipi, "the script of the Huns") suggest a date for the list in its
present form not earlier than the fourth century A.D. The several geographical names
such as Arigalipi, Vaiigalipi, and Purvavidehalipi (nos. 4, 5,40) also suggest a rela-
tively late period by which the BrahmT script had become extensively differentiated
into local varieties.
Second, the names of many of the scripts are difficult to evaluate. Only the first
two, Brahml and KharostI (i.e., KharosthI; see 2.2.2 and 2.3.5), can be positively
identified with known scripts. As for the rest, they consist mainly of geographical
terms which can presumably be connected with the appropriate local derivatives of
BrahmT (though in view of the chronological questions they cannot be specifically
identified),8 and of terms which are apparently descriptive of graphic or calligraphic
characteristics (e.g., no. 25 Puspalipi, "the flower script," no. 41 Utksepalipi, "the
upward-flowing script," etc.). Some of the latter group might be identified with vari-
ous calligraphic scripts preserved in inscriptions (see 2.5.3), but this is hardly more
than a guess. Other names, such as nos. 26-8, Devalipi "script of the gods," Nagalipi
"script of the Nagas," and Yaksalipi "script of the Yaksas," are presumably fanciful,
and the total number of scripts (64) is a conventional one, which also must give rise
to suspicions as to the historicity of the list as a whole.
Similar but much briefer lists of eighteen (again, a stereotypical number) scripts
are also preserved in several Jaina canonical Prakrit texts. The oldest form of the list,
which appears in the Pannavana-sutta and the Samavayanga-sutta, includes (like
the Buddhist lists) Bambhl (= BrahmT, no. 1) and KharotthT (= KharosthI, no. 4).9
According to Btthler (OIBA 25-7), the Jaina list is probably independent from and
"in all probability is considerably older than that of the Buddhists" (26). Particularly
notable is the inclusion of the "Javanaliya" (v. 1. Javananiya) script (no. 2), presum-
ably referring to the Greek alphabet (Sanskrit yavanarii). But otherwise, as in the
Buddhist lists, most of the scripts mentioned cannot be clearly identified with forms
of writing known from epigraphic remains. It is also interesting to note that the list

7. John Brough, BSOAS 40, 1977, 85.

8. The twelfth script, the Dravidalipi or "Dravidian" script, was identified by Blihler (BIP 2; OIBA
24) with the distinct southern variety of Brahml represented by the Bhattiprolu inscriptions ([Link]),
but in view of the probable late date of the list this specific identification is questionable (cf. the re-
marks in 1C 11.673).
9. The Jaina lists are cited and discussed by Albrecht Weber in Indische Studien 16, 1883, 280
and 399-401. The earliest version in Pannavana-sutta as cited by Weber reads (with variant readings
from the Jaina-Agama-Series edition [no. 9, part 1; ed. Muni Punyavijaya, Dalsukh Malvania, and
Amritlal Mohanlal Bhojak (Bombay: Shri Mahavlra Jaina Vidyalaya, 1969), 38, 1.107]):
(1) Bambhl (= BrahmT) (2) Javanaliya ("Greek") (3) Dasapurlya [Dosapuriya] (4) Kharotthi
(= KharosthT) (5) Pukkharasariya (6) Bhogava'iya (7) Paharaiyau [Paharalyao] (8) Amtarikariya
[Amtakkhariya] (9) Akkharaputthiya (10) Venaiya (11) NinhaTya (12) Amkalivi (13) GanitalivT
(14) GamdhavvalivT (15) Ayasalivl [AyamsalivT] (16) Mahesarl (17) Damill (18) Polimda [PolimdT]
(399). The list in Samavayanga-sutta 18 is essentially similar but has several variants. Other, evidently
later, Jaina script lists are cited by Weber, ibid., 400-1.
10 Indian Epigraphy
is presented with the introductory remark bambhie nam livle attharasaviha-
likkhavihdne pannatte, "18 different forms of writing of the Brahmi script are known."
This evidently means that the term "Brahmi" applies to writing as such (presumably
referring to its legendary creation by Brahma, as discussed earlier; cf. OIBA 25), as
well as to a particular script, no. 1 in the list which follows, which may or may not
actually refer to what modern scholars call "Brahmi" script (2.2.2).
Other than these, very few references to script types and names have been
located in Sanskrit and related literatures, so that the further study of the origin and
development of the Indie scripts must be pursued from the evidence of the docu-
ments (almost exclusively epigraphic for the earlier phases) themselves.

2.1.2 The antiquity of writing in India of the historical period

O. von Hiniiber,DerBeginn derSchrift... [BS]; H. Falk,Schrift im alten Indien [SAI].
After more than a century of study, the early history of writing in India remains prob-
lematic. It begins with the still undeciphered script found on the seals and other relics
of the Indus Valley civilization, which flourished, according to recent estimates, around
the second half of the third and first half of the second millennium B.C. But after the
decline of the Indus Valley culture, the graphic record of India is virtually a total blank
for well over a thousand years until the time of the Asokan inscriptions, the earliest
definitely datable written records of the historical period, around the middle of the third
century B.C. From this time on, written records (epigraphic and, in later centuries, manu-
script and other nonepigraphic sources) become increasingly common, so that the devel-
opment of the Indian scripts and their many derivatives both in and outside of India can,
for the most part, be traced in considerable detail from Asoka's time to the present day.
But practically nothing is known of what might have happened in the long
period between (very roughly) 1750 and 260 B.C. Certain bits of evidence have been
proposed as missing links between the protohistoric and historical writings. For ex-
ample, graffiti found on megalithic and chalcolithic pottery from southern and west-
ern India were discussed by B. B. Lai,10 who noted (23-4) resemblances of some of
the shapes found in these graffiti to both Indus Valley script characters and letters of
the Brahmi script. However, Lai concluded with the suitably cautious note that "to
stress the point that the symbols do have a phonetic, syllabic or alphabetic value would
indeed be presumptuous in the present state of our knowledge" (24). A few inscrip-
tions, or pseudo-inscriptions, have been proposed as specimens of "prehistoric"
writing or "missing links" between the Indus Valley script and Brahmi, for instance,
the Vikramkhol "inscription," but none of these examples of supposed prehistorical
writing is convincing.'' In short, the evidence currently available does not permit us
to determine conclusively whether the art of writing simply died out in the second
millennium B.C., to be replaced much later by unrelated systems, or whether (though
this seems much less likely) some form of it, as yet undiscovered, somehow survived
and reemerged in the form of the scripts of historical times.

10. "From the Megalithic to the Harappa: Tracing Back the Graffiti on the Pottery," Ancient
India 16, 1960, 4-24.
11. Possible connections between the Indus Valley writing and Brahmi are further discussed in
section [Link].
 Writing and Scripts in India 11
Because of the virtual absence of actual documentary evidence for writing in the
blank period, many historians and epigraphists have addressed the question of the
possibility of literacy in pre-Mauryan India through the examination of literary and
other evidence. The literature on the subject is far too extensive to be presented in
detail here. The discussions by Burnell (SIP 1-11), Buhler (OIBA 5-35 and BIP 3-6),
Ojha (BPLM 1-16), Pandey (PIP 1-22), and Sircar (HEP 104-6) may be taken as
more or less representative of earlier opinions, of which a complete survey and up-
to-date analysis is provided in Falk's SAL To begin with Vedic literature, certain
authors (notably Ojha in BPLM 9-13 and, mostly following him, Pandey and oth-
ers) have claimed evidence for a literate culture in the later Vedic texts, and even in
the Rg Veda, on the grounds of references therein to poetic meters, grammatical and
phonetic terms, very large numbers, and relatively complex arithmetic calculations.
But it is by no means certain that such cultural phenomena presuppose literacy, and
it may be argued to the contrary that the absence of a single explicit and indisputable
reference to writing anywhere in early Vedic literature suggests that the Vedic cul-
ture was a preliterate one.12
The testimony of Greek and Latin authors on writing in early India has been
studied by many scholars13 but remains somewhat inconclusive. For example,
Nearchos, who visited northwestern India around 325 B.C., explicitly mentions that
Indians wrote letters (ETUOTOA.^) on cotton cloth. This observation is usually as-
sumed to refer to the Kharosthi script, but it has recently been suggested14 that
Nearchos may actually have been referring to writing in Aramaic. Megasthenes,
who lived in northeastern India some two decades after Nearchos, stated that the
Indians "did not know written characters" (ox>8e yap ypduuata eiSevai amotx;),
but it is not entirely certain whether this is a blanket statement15 or refers only to
the immediate context of legal procedures in which the comment was made. The
confusion on this issue among the classical historians seems to be reflected in
Strabo's comment, with reference to Nearchos' report, that "others opine that they
did not make use of written characters."
The Pali Buddhist canon, especially the Jdtakas and the Vinaya-pitaka, contains
numerous explicit references to writing and written documents, particularly to "private
and official correspondence by means of letters" (OIBA 7). But it is uncertain whether
any of these references can really be taken to represent the state of things in pre-Mauryan
India, as Buhler and others have claimed, since all or most of them seem to belong to
the later strata of the canon (BS 22-54; SAI270-83). In Panini's Astadhyayi (3.2.21)
we seem to have a clear reference to early writing in the term lipillibi 'script'.16 Panini's
date is a matter of complex controversy, but an authoritative recent opinion is that "the
evidence available hardly allows one to date Panini later than the early to mid fourth

12. See the classic presentation of this position in F. Max Muller's History of Sanskrit Literature
So Far as It Illustrates the Primitive Religion of the Brahmans (London: Williams and Norgate, 1859),
455-80. See also BIP 5 and, more recently, BS 18 and SAI 240-56.
13. Most recently and authoritatively by von Hinuber (BS 19-21) and Falk (SAI 290-7).
14. By von Hinuber, BS 21; cf. SAI 290.
15. So von Hinuber, BS 20; cf. SAI 291-3.
16. Although even this has been questioned; see BS 57. The term grantha (Astadhyayi 1.3.75,
4.3.87, 4.3.116), though normally rendered "book," need not necessarily refer to a written composi-
tion (SAI 261-2, 298-9).
12 Indian Epigraphy
century B.C.,"17 which would seem to provide us a strong indication of writing before
the Mauryan period. But here too (as in the case of Nearchos), it has recently been sug-
gested18 that the script known to Panini might be Aramaic rather than any Indian script.
As to archaeological and epigraphic evidence for the antiquity of writing in the
historical period, we have only a small handful of brief archaic inscriptions (see
[Link]) which could conceivably be somewhat older than the Asokan inscriptions.
But the weight of scholarly opinion nowadays19 is in favor of dating such early records
as the Piprawa, Sohgaura, and Mahasthan inscriptions as contemporary with or later
than Asoka.20 Certain paleographic characteristics of the early inscriptions, such as
the absence of distinction in some records (e.g., Mahasthan) of vowel length—mostly
for vowels other than a and a—have sometimes been taken as an indication of their
pre-Asokan antiquity; yet the very same phenomena have been invoked by others in
support of arguments for a recent origin and short period of development for the
Brahml script. Equally inconclusive is the matter of alleged regional variations in
early Brahml. Biihler (OIBA 40; cf. BIP 6-8) took this as evidence that "the letters
of the [Asokan] edicts had been used at least during four or five hundred years," but
more recent studies by C. S. Upasak and A. H. Dani (see 2.2.4) have largely discred-
ited the supposed regional variants of Asokan Brahml, and some (e.g., S. R. Goyal
in OBS 7-10) have taken the geographical unity of the Asokan script as evidence
that it must have been a recent invention or development.
However, a new body of material has recently come to light that seems to sup-
port the older theory that Brahml existed before Mauryan times, that is, in the fourth
century B.C. or possibly even earlier. This is a small group of potsherds bearing short
inscriptions, evidently proper names, which were found in the course of excavations
at Anuradhapura, Sri Lanka in strata which are said to be securely assigned by radio-
carbon dating to the pre-Mauryan period.21 Various dates have been proposed for
these graffiti, ranging from the sixth to the early fourth century B.C. The more recent
publications on the subject have tended to favor the later date within this range, but
in any case, these inscriptions still seem to show that BrahmT did indeed predate the
Mauryan period. Some doubts remain, however, as to the chronological significance
of these inscriptions; it is possible, for example, that the inscribed potsherds were
intrusive in the strata concerned, and actually date from a later period, although the
excavators have argued against this scenario.

17. George Cardona, Panini: A Survey of Research (Mouton: The Hague/Delhi: Motilal Banarsidass,
1976), 268.
18. BS 58; SAI 258.
19. Reviewed in SAI 177-88.
20. All attempts to attribute specific dates to certain of the early inscriptions have been decisively
discredited. For example, Ojha's (BPLM 2-3) dating of the Barll inscription (SI 89-90) to the year 84
of a supposed Jaina era of 527 B.C., i.e., to 443 B.C., is out of the question; Sircar dates the inscription
to the late second century B.C. on paleographic grounds. The alleged specimens of Kharosthi and Brahml
letters on coins of the Achaemenian emperors of Iran are also doubtful (see 2.3.2 n. 132).
21. See S. U. Deraniyagala, The Prehistory of Sri Lanka: An Ecological Perspective (Archaeo-
logical Survey of Sri Lanka, 1992), 2.739-50; F. R. Allchin, The Archaeology of Early Historic South
Asia: The Emergence of Cities and States (Cambridge: Cambridge University Press, 1995), 163-81 and
209-16; R. A. F. Coningham, F. R. Allchin, C. M. Batt, and D. Lucy, "Passage to India? Anuradhapura
and the Early Use of the Brahmi Script," Cambridge Archaeological Journal 6, 1996, 73-97 (esp. 76-
7); and further references provided in these sources.
 Writing and Scripts in India 13
In conclusion, both the literary and the epigraphic evidence for the antiquity of
historical writing in India are disappointingly inconclusive, since virtually all of the
testimony is in one way or another vague or ambiguous. Probably the most cogent single
piece of literary evidence for writing before the Mauryan period is Panini's reference
to script (lipi), although the uncertainties as to his date partially vitiate the value of this
testimony. Moreover, although it seems clear that this proves the existence of some
form of writing in Panini's home region of northwestern South Asia in or before the
mid-fourth century B.C., there is no explicit indication as to what type of script22 he is
referring to. Although, as noted earlier, Falk and von Hiniiber have recently suggested
that may have been Aramaic, there is actually no cogent reason to rule out KharosthT,
which therefore may well date back to the mid-fourth century B.C. or quite possibly
even earlier (see 2.3.2 and 2.3.6). But even if this is so, there is no direct evidence for
the use of Brahml before the time of Asoka. Thus, the trend of recent writings such as
BS23 and SAI has been to emphasize an empirical interpretation of the actual surviving
data, which leads to the conclusion that at least Brahml, and probably KharosthT as
well, did not exist before the Mauryan period, and hence that (leaving aside the proto-
historic Indus script) writing originated in India no earlier than the late fourth century
B.C. However, the recent discovery, mentioned above, of apparently pre-Mauryan graffiti
in Sri Lanka has cast some doubt upon this point of view.
Moreover, many other authorities have found it difficult to imagine that the evi-
dently high level of political organization and cultural complexity that had been
reached in the pre-Mauryan period could have existed without writing. Thus various
estimates for a hypothetical prehistory of Indian writing have been proposed. Biihler's
suggestion (OIBA 84; also BIP 16) of an early date of ca. 800 B.C., or possibly even
earlier, for the "introduction of the prototypes of the Brahma letters" into India is
hardly plausible in light of modern knowledge, but more cautious estimates such as
that of A. B. Keith that "the real development of writing belongs in all likelihood to
the fifth century B.C." are not unreasonable.24 This more traditional point of view has
recently been maintained by, among others, K. R. Norman.25 Like other proponents
of pre-Mauryan writing, Norman (279) attributes the absence of any surviving writ-
ten records before the time of Asoka to the fact that early writing was primarily used
for ephemeral documents. The practice of writing monumental inscriptions on stone
was presumably an innovation of Asoka himself, possibly under the inspiration of
the Achaemenian empire of neighboring Iran. Before Asoka, writing was probably

22. As has been noted in various sources, most recently SAI 259—61, the interpretation of Panini's
prescription (4.1.49) of the feminine adjective yavanam as referring to lipi, i.e., the Greek script, is
based on a vartika of Katyayana and hence does not prove anything about Panini's own knowledge of
Greek script.
23. E.g., BS 22: "Fremde Beobachtungen sprechen also in Ubereinstimmung mit den Zeugnissen
aus Epigraphik und Numismatik eindeutig dafiir, daB es in Indien vor Asoka keine Schrift gegeben
hat, wenn man von den indischen Provinzen des Achamenidenreiches absieht." ("Thus, foreign obser-
vations clearly agree with the testimony of epigraphy and numismatics that there was no writing in
India before Asoka, with the exception of the Indian provinces of the Achaemenian empire.")
24. In E. J. Rapson, ed., Cambridge History of India. Vol. 1: Ancient India (Cambridge: Cam-
bridge University Press, 1922), 126.
25. In his review of von Hiniiber's BS in JRAS, ser. 3, vol. 3, 1993, 277-81. See also R. Salomon,
"On the Origin of Early Indian Scripts" [review article on von Hiniiber's BS, Falk's SAI, etc.], JAOS
115, 1995, 271-9, esp. 278-9.
14 Indian Epigraphy
used principally, if not exclusively, for economic and administrative, as opposed to
literary and monumental, purposes; perishable materials such as palm leaves, tree
bark, and (according to Nearchos) cloth, which have little chance of surviving the
rigors of the Indian climate, were used. Thus, according to this view, we need not be
surprised that no early specimens of Indian writing have survived, and their absence
does not prove that they never existed.
Such hypothetical pre-Mauryan writing would presumably (though not neces-
sarily; see Goyal, OBS 2,30) have been some protoform of BrahmT and/or Kharosthl
script. The nondifferentiation of vowel length in KharosthT generally, and also, as
noted earlier, in some early specimens of Brahml, suggests that a pre-Asokan Brahml
would not yet have distinguished vowel quantity. Similarly, indications from the
extant early specimens suggest that the notation of consonant groups would have been
rudimentary or even totally absent in their presumptive prototypes. The protoscripts
may also have had an incomplete array of phonemes vis-a-vis Sanskrit, since they
were presumably developed for recording early MIA dialects rather than Sanskrit.
The confusion between s and s in some early BrahmT inscriptions (see 2.2.4), for
example, suggests that the earliest forms of the script did not have distinct characters
for the three sibilants of Sanskrit. The script that we actually have in the Asokan
inscriptions thus could be a refined and standardized "national" script, developed
under Asoka for purposes of governing his vast pan-Indian empire, on the basis of
an earlier form of Brahml or proto-Brahml script, of which the Anuradhapura graf-
fiti may be a specimen.
In short, two schools of thought are dominant with regard to the problem of the
antiquity of writing in historical India. One side sees no cogent archaeological or liter-
ary evidence for the existence of writing, and particularly of BrahmT script, before
the Mauryan period. The other camp finds this hard to accept on pragmatic grounds,
and moreover now sees archaeological evidence of pre-Mauryan Brahml in the
Anuradhapura graffiti, which are allegedly datable to the early fourth century B.C. at
the latest. The issue remains unresolved, though it may be hoped that further discus-
sions and examinations of the new evidence may ultimately lead to a consensus.

2.1.3 Characteristics of Indie writing

With only a few partial exceptions (see n. 37 and [Link]), all of the Indie scripts
(that is, BrahmT and its derivatives, and Kharosthl) follow the same basic principles
of graphic representation. This system is, in historical terms, a modified consonant-
syllabic script whose basic unit is the graphic syllable (which by definition ends with
a vowel) typically consisting of a consonant with or without diacritic modification
to indicate the following vowel, and/or other modifications such as nasalization in-
dicated by the anusvara sign. Various diacritic strokes are added to the top or bot-
tom of the consonantal character to indicate the vowels a, /, T, u, u, r, and so on, but
the vowel a is treated as inherent in the consonant. That is to say, an unmarked con-
sonant is to be read with the vowel a; thus early BrahmT -f" = modern DevanagarT
aand so on.
This system, and particularly the inherent vowel rule, leads to certain complica-
tions. First, to represent a syllable consisting of a vowel only, without a preceding
 Writing and Scripts in India 15

consonant,26 these scripts require a set of full vowel signs in addition to the more
frequently used diacritic vowels; thus Brahml H = Devanagarl 3{ = a; H = 3fT = a, and
so on. Second, the representation of consonants which are not immediately followed
by vowels, that is, of consonant clusters and of word, line, or sentence final conso-
nants, also requires special techniques. Consonant clusters are represented by vari-
ous types of ligatures of the consonants concerned, joined together or abbreviated in
such a way as to make it evident that the prior consonant(s) are to be pronounced
together with the following one(s), without an intervening vowel. Thus in Brahml
the syllable iva must be written as ^ (= DevanagarT r? ) in order to distinguish it
from Ai = tava (fR). Final consonants are indicated either by a reduced form of the
normal consonant or by a special diacritic sign (called virama or halanta) indicating
cancellation of the inherent vowel (e.g., Devanagarl $ = k, not ka).21
The Indie system of writing is difficult to classify in terms of the traditional ty-
pology of writing systems which recognizes three main script types, namely, logo-
graphic, syllabic, and alphabetic.28 The Indian system is syllabic in the sense that its
basic graphic unit is the syllable (aksara), but it differs from a pure syllabary in that
the individual phonetic components of the syllable are separately indicated within
the syllabic unit. It thus resembles an alphabet insofar as the vowels have a separate
and independent notation but cannot be called a true alphabet in that the vowels do
not have a fully independent status equal to that of the consonants (this being the
defining characteristic of an alphabet in the strict sense of the term). Although the
Indie scripts do have alphabetic symbols for the vowels in the "full" or initial vowel
characters, these were never extended beyond their restricted use for vowels not pre-
ceded by a consonant, and thus did not attain full alphabetic status.
Various terms have been suggested for this type of script, intermediate between
syllabary and alphabet, such as "neosyllabary,"29 "pseudoalphabet," or "alphabetic
syllabary."30 There is as yet no commonly accepted term in the relatively rudimen-
tarily developed field of grammatology for scripts of this type. Thus, for the present
it may be best to refer to it by general descriptive terms such as semisyllabary or
semialphabet (since it partakes of significant features of both systems), or, more pre-

26. Such occurrences are relatively rare in Sanskrit, which in general disallows vowel hiatus within
and between words, so that in Sanskrit texts the full or initial vowels mostly occur at the beginning of
a sentence or line of verse. They are commoner in Prakrit and other languages which do permit vowel
hiatus, but even here they are often avoided by writing a consonant such as v or v to represent an
intervocalic glide.
27. The notation of final consonants is a relatively late development, first appearing in the earlier
centuries of the Christian era ([Link]). This is due to the fact that Prakrit, in which all of the early
inscriptions are written, does not permit word final consonants, so that their notation only became
necessary later when inscriptions began to be written in Sanskrit.
28. Ignatz Gelb (A Study of Writing [Chicago: University of Chicago Press, 1953], 188) admits to
"disturbing problems" in connection with the typological classification of the Indie and similar script
types, and does not reach a definite conclusion on the point.
29. James G. Fevrier, Histoire de I'ecriture (Paris: Payot, 1959), 333.
30. Fred W. Householder Jr., Classical Journal 54, 1958-59, 382. Recently Peter T. Daniels (JAOS
110, 1990, 730) has suggested the name "abugida," "from the Ethiopian word for the auxiliary order
of consonants in the signary." This term, however, is more likely to appeal to Semitists than to
Indologists, who might prefer a name like "aksara script," this being the Indie term which would come
closest to a designation of this type of writing.
16 Indian Epigraphy
cisely, as "diacritically (or alphabetically) modified consonant syllabary." Which-
ever terminology one might prefer to adopt, the Indie script family, comprising all
of the derivatives and modifications of Brahml script in India and many other parts
of Asia, is the principal representative of its type among the scripts of the world. The
only other major example of this type is the Ethiopian script, which employs a simi-
lar system of vowel notation by diacritic signs attached to the preceding consonant,
and which has, like the Indie scripts, an inherent or neutral vowel a but does not regu-
larly form consonantal conjuncts.31
The process of development of the Indie system is for the most part not directly
traceable in the extant materials, wherein the system appears almost fully developed32
from the beginning. But parallels from the history of script developments elsewhere in
the world make it a priori likely that such a modified consonantal syllabary developed
from a pure consonantal syllabary, whose prototype presumably (and in the case of
Kharosthl virtually certainly) would be a Semitic script of this type.33 If this assump-
tion is correct, the addition of diacritic vowel markers to the basic consonants would
be a natural way to adapt the Semitic system to Indie languages, which, unlike Semitic
languages, could not be conveniently and economically represented without vowels.34
The development of the inherent a and of consonantal conjuncts can be readily ex-
plained by reference to the phonetic characteristics of Sanskrit and related Indie lan-
guages. The vowel a is statistically strongly predominant in these languages, so that it is
both natural and economical to assign it as inherent in all consonants.35 Since the early
Indie scripts seem to have been originally used exclusively or principally for MIA lan-
guages, which unlike Sanskrit have few clusters of nonhomorganic consonants, the in-
herent vowel system did not at first cause any significant complications. For such con-
sonantal groups as do occur in MIA consist mainly of nasal plus homorganic stop, easily
represented by theanusvara, and geminates, which were simply represented by the single
consonant. It was presumably only later, when these scripts began to be applied to for-
mal literary uses and to the Sanskrit language, that the complications necessitated by
the inherent vowel—mainly the frequently cumbersome consonantal conjuncts—arose.
From a practical point of view, one could easily imagine a system wherein a simple vowel
cancellation marker would be put to use to eliminate the need for all conjuncts. But, in
fact, this was not done, whether because the conventions of the system had become so

31. These striking similarities have led some scholars to posit a historical connection between the two
script groups, suggesting that the Ethiopic may have developed under the influence of an Indie model (e.g.,
Suniti Kumar Chatterji, India and Ethiopia from the Seventh Century B.C., The Asiatic Society Monograph
Series, vol. 15 [Calcutta: Asiatic Society, 1968], 49-56; see also Gelb, op. cit., 188). This is plausible on
historical grounds, in view of well-attested trade contacts between India and East Africa in ancient times.
Nevertheless, other authorities are inclined to attribute the similarity to parallel but separate developments,
or to Greek influence on the Ethiopic script; see, e.g., David Diringer, The Alphabet, 231-2.
32. However, see section 2.2.4 for some important exceptions.
33. See further discussions in sections 2.2.3 and 2.3.6.
34. This process is in principle not different from various other secondary devices developed to
indicate vowels in other Semitic and Semitic-derived scripts, such as the pointing systems used in Hebrew
and Arabic, or even the fully alphabetized vowels of Greek. It is not possible to say for certain why
this particular type of modification arose in India, but it may be that an already established linguistic
concept of the aksara or syllable as the essential unit of language favored a system which preserved
the syllabic unit over other theoretically possible developments, such as full alphabetization.
35. O. von Hiniiber, BS 16; but see also n. 170 for a different interpretation.
 Writing and Scripts in India 17

firmly entrenched that they were not subject to fundamental reform, or through the in-
fluence of the previously mentioned linguistic principle of the aksara, or for other rea-
sons; the vowel-cancelling marker alluded to earlier is, in general, used only when there
is no other way to represent a vowelless consonant, such as at the end of a word, sen-
tence, or line.36 In any case, the result was that the Indie system, remarkable for preci-
sion if not for simplicity, has been preserved virtually unchanged in nearly all of the
many scripts used for Sanskrit and other Indo-Aryan languages.37

2.2 The Brahmi Script and Its Derivatives

2.2.7 Geographical and chronological range
Unlike KharosthI, which was always geographically limited and died out at a rela-
tively early period (see 2.3.1 and 2.3.2), the BrahmT script (see table 2.1) appeared
in the third century B.C. as a fully developed pan-Indian national script (sometimes
used as a second script even within the proper territory of KharosthI in the north-
west) and continued to play this role throughout history, becoming the parent of all
of the modern Indie scripts both within India and beyond. Thus, with the exceptions
of the Indus script in the protohistoric period, of KharosthI in the northwest in the
ancient period, and of the Perso-Arabic and European scripts in the medieval and
modern periods, respectively, the history of writing in India is virtually synonymous
with the history of the Brahmi script and its derivatives.

2.2.2 The name of the script

Until the late nineteenth century, the script of the Asokan (non-Kharosthl) inscrip-
tions and its immediate derivatives was referred to by various names such as "lath"
or "Lat,"38 "Southern A£okan," "Indian Pali," "Mauryan," and so on. The applica-
tion to it of the name Brahmi [sc. lipi], which stands at the head of the Buddhist and
Jaina script lists (see 2.1.1), was first suggested by T[errien] de Lacouperie,39 who
noted that in the Chinese Buddhist encyclopedia Fa yuan chu lin the scripts whose
names corresponded to the Brahmi and KharosthI of the Lalitavistara are described
as written from left to right and from right to left, respectively. He therefore sug-
gested that the name Brahmi should refer to the left-to-right "Indo-Pali" script of the
Asokan pillar inscriptions, and KharosthI to the right-to-left "Bactro-Pali" script of

36. This restriction has, however, been somewhat relaxed in modern usage due to the exigencies
of printing, in which some consonantal ligatures are impractically cumbersome and are therefore some-
times split up by use of the halanta sign.
37. Note that some Indie scripts which have been adapted for non-Indo-Aryan languages have
introduced simplifications that eliminate the need for consonantal conjuncts. Such is the case, e.g., in
Tamil (see also [Link]), which regularly writes vowelless consonants in their full syllabic form and
marks them with a dot (pulli) to indicate that no vowel follows (e.g., £>«n1i.u> kantam, 'piece').
38. This name arose from the occurrence of Asokan inscriptions on pillars known colloquially in
northern India as lath (< Skt. yasti).
39. "Did Cyrus Introduce Writing into India?" Babylonian and Oriental Record 1,1886-87, 58-64
(esp. 59-60). As noted by Falk (SAI 84, 106), Terrien de Lacouperie evidently got the references in
question from the note of T. Choutze (pseudonym of Gabriel Deveria) in the Revue de I 'Extreme-Orient
1, 1882, 158-9, though he does not mention him.
18 Indian Epigraphy

TABLE 2.1. Early BrahmT Script

VOWELS
Full or initial forms

Medial (postconsonantal) forms

»
ka ki ku kc ko • •> kaip
ka a ku u kau

CONSONANTS
Unvoiced Voiced Voiced Voiced Nasal Semi- Sibilant
unaspi- aspirated unaspi- aspirated vowel
rated rated
Guttural ka kha
*"
8111
* 118

Palatal ca cha ja jha

88 fa
*
Retroflex
* *• * dh,
" M
58
Dental dha na
* *• da ta
xf "
Labial pa pha te bha
"

Conjunct consonants
(representative examples)

khya tva ^ .
tnha sta sya
-

Note: These are normalized forms, based on letters found in ASokan inscriptions, where
available; otherwise from other earliest citations elsewhere.

the rock inscriptions from the northwest. Lacouperie's suggestion was adopted by
his contemporaries, most significantly by Buhler40 in his influential works, and thereby
became the accepted term.
While the name BrahmT for the ancient Indian national script is no doubt in a gen-
eral sense correct, it should be kept in mind that we do not really know precisely what
form or derivative of the script the authors of the early script lists were referring to as
40. The expression "Brahma alphabet" used by Buhler in OIBA has, however, been supplanted in
modern usage by "Brahmi" [lipi].
 Writing and Scripts in India 19

"Brahml," nor whether this term was actually applied to the script used in the time of
Asoka.41 The name Brahml is thus used loosely, as a matter of convenience, by mod-
ern scholars to refer to the Asokan script and to its varieties and earlier derivatives
(distinguished by regional or dynastic terms such as "early southern Brahmi" or "east-
ern Gupta Brahml") until about the end of the Gupta period in the sixth century A.D.
After this time, the scripts have for the most part differentiated into distinct regional
and local varieties, and are conceived as separate scripts denoted by descriptive or,
more commonly, geographical terms (e.g., Siddhamatrka [post-Gupta northern script]
or proto-Kannada). The terminology for the various premodern Brahml-derived scripts
is, however, largely unstandardized and typically made up ad hoc, due mainly to the
lack of attested indigenous terms for many of them (2.1.1). D. C. Sircar broadly cate-
gorizes the stages of development into "Early," "Middle," and "Late Brahmi" peri-
ods, corresponding (in northern India) to the third through first centuries B.C., the first
century B.C. through third century A.D., and the fourth through sixth centuries A.D.,
respectively (HEP 113), though others refer to his "Late Brahmi" as "Gupta script"
(cf. Gai, HE, 34). A. H. Dani, however (in DIP), considers such dynastic terminology
misleading and prefers to use only regional and geographical categories (see [Link]).

2.2.5 The origin of Brahmi

Biihler, OIBA; S. P. Gupta and K. S. Ramachandran, eds., OBS; Falk, SAI 109-67.
The origin of the Brahml script is one of the most problematic and controversial prob-
lems in Indian epigraphy. Most opinions on the question fall into one of two camps: the
proponents of an indigenous Indian origin for Brahml, and those who see it as a bor-
rowing or adaptation from some non-Indian (usually Semitic) prototype.42 Within each
camp there are also several versions of the two principal theories; in the following two
subsections the major representatives of each theory will be briefly summarized.43

[Link] Theories of indigenous Indian origin

Some early scholars, most prominent among them Alexander Cunningham,44 proposed
theories of the origin of BrahmT from a pictographic-acrophonic system based on
the model of Egyptian hieroglyphic writing, wherein, for example, the Brahmi letter
kha would be derived from a pictorial representation of a hoe or mattock by associa-
tion with the root khan 'to dig'. Such theories are purely speculative and imaginative,
and do not merit further consideration; in the words of Isaac Taylor, "Such an elastic

41. On this point, see the warnings in 1C 11.667 and OBS 99.
42. It has been noted (e.g., S. R. Goyal in OBS 6-7) that nowadays most of the proponents of the
theory of indigenous origin are Indians, while nearly all Western scholars subscribe to the theory of
Semitic borrowing; and there is no doubt some truth in Goyal's comment that some of their views have
been affected by "nationalist bias" and "imperialist bias," respectively.
43. The literature on the origin of BrahmT is too vast to be presented in full detail here. Moreover,
a great deal of what has been written on the subject is trivial, presenting no real new material or in-
sights. An extensive though not complete bibliography of the subject is given in N. P. Rastogi's Ori-
gin of Brahmi Script, 141-61. Almost all of the relevant materials are included in the more general
bibliography in SAI (15-66).
44. Inscriptions of Asoka (CO 1), 49-63.
20 Indian Epigraphy

method [as Cunningham's] may establish anything, or—nothing."45 Other early specu-
lative theories attributed the origin of Brahml in the protohistoric period to various
"races," such as the Dravidians46 or the "enlightened Aryans."47
Somewhat more sophisticated than these superficial theories are the arguments
of G. H. Ojha in BPLM, who was highly critical of Buhler's Semitic derivation (see
[Link].4) and was inclined to doubt any foreign derivation, though he avoided deny-
ing the possibility altogether. Ojha concluded (30) that an indigenous origin is most
likely, although the precise source and development cannot be specified. R. B. Pandey
(PIP 35-51) argued more categorically in favor of an indigenous origin, concluding
that "the BrahmT characters were invented by the genius of the Indian people who
were far ahead of other peoples of ancient times in linguistics and who evolved vast
Vedic literature involving a definite knowledge of alphabet" (51).
Since the discovery in the 1920s and subsequent decades of extensive written arti-
facts of the Indus Valley civilization dating back to the third and second millennia B.C.,
several scholars have proposed that the presumptive indigenous prototype of the BrahmT
script must have been the Indus Valley script or some unknown derivative thereof. This
possibility was first proposed48 by S. Langdon in 1931,49 supported by G. R. Hunter,50
and endorsed by several later authorities,51 most significantly by D. C. Sircar.52
It has already been mentioned (2.1.2) that such a connection between the pro-
tohistoric Indus writing and the later Brahml script should not be taken for granted,
that is, it should not be assumed a priori that two scripts of the same cultural area but
different periods must be historically or genetically connected. It is all too easy, given
the large number of characters in the Indus script (over four hundred), to find super-
ficial connections between similar shapes of some characters in the two scripts, but
these are of little value unless and until the Indus script itself is convincingly deci-
phered and the alleged graphic similarities can be correlated to phonetic values.
Various claims to decipherment of the Indus script based on such superficial com-
parisons with Brahml, such as that of Langdon and many since him, are not at all
convincing and have little if any scholarly value. Equally unconvincing are the vari-
ous claims put forward for the Vikramkhol and other pseudoinscriptions (see 2.1.2)
as constituting a link between the Indus Valley and Brahml scripts.53
45. The Alphabet, 1.307 n. 1.
46. Edward Thomas, JRAS, n.s. 5, 1871, 421-2 n. 2. See also, more recently, T. N. Subramanian
in SITI III.2, 1587-1608.
47. John Dowson, "The Invention of the Indian Alphabet," JRAS, n.s. 13, 1881, 102-20 (quoted 118).
48. It is, however, interesting to note that as early as 1877 Cunningham (Inscriptions of Asoka,
61) proposed as a possible ancestor of Asokan BrahmT a single inscribed seal from Harappa, which
was then an isolated find whose significance had not yet become clear; Cunningham estimated its date
as not later than 500 or 400 B.C.
49. "The Indus Script," ch. 23 of John Marshall, Mohenjo-daro and the Indus Valley Civilization
(London: Arthur Probsthain, 1931), 11.423-55.
50. The Script of Harappa and Mohenjodaro and Its Connection with Other Scripts, Studies in
the History of Culture, no. 1 (London: Kegan Paul, Trench, Trubner & Co., 1934).
51. E.g., Pandey (PIP 51), who suggests a connection with the Indus Valley script ("The BrahmT
was derived from pictographs, ideographs and phonetic signs, the earliest specimens of which are to
be found in the Indus Valley inscriptions"), but does not elaborate any further.
52. SI 1.242 n. 1; HEP 107-8, 111; EINES 85. See also OBS 70-1 and 84-5.
53. E.g., K. P. Jayaswal, "The Vikramkhol Inscriptions," IA 62, 1933, 58-60.
 Writing and Scripts in India 21
There is, however, at least one feature of the Indus script which could in fact
indicate a systemic connection with the historical scripts of India. The former script
has a large number of what appear to be compounded and/or diacritically modified
forms of the basic characters, which are reminiscent of the characteristic patterns of
the Indie scripts of the historical period (2.1.3). Hunter54 hypothesized that this sys-
tem functioned to indicate, among other things, vowel variations, exactly as in Brahml,
and ventured (92-3, 102-3) to directly derive some of the Brahml vowel diacritics
from Indus Valley signs. He also pointed out (54) a possible relationship with the
Brahml system of conjunct consonant formation. These parallels, noted by Hunter
and others since him (e.g., Dani, DIP 16-7), are certainly intriguing, and could actu-
ally reflect some historical connection, direct or indirect, between the scripts. Never-
theless, in view of the still undeciphered status of the Indus script and the huge chro-
nological gap between it and the earliest attested scripts of the historical period, it
would be premature to try to explain and evaluate the significance of the apparent
typological similarities.
Recently several writers have put forth theories to the effect that Brahml was
purposefully invented ex nihilo in or around the time of Asoka. S. R. Goyal, for ex-
ample,55 argued that the phonetically logical structure, primary geometric forms, and
geographical uniformity of early Brahml show that it was "an invention of the gram-
marians" (10) of the time of Asoka. Similar arguments were presented by T. P. Verma,
who proposed an origin in Buddhist circles.56 N. P. Rastogi, in Origin of Brahml Script,
offered a purely formal presentation of "the origin of the Brahml script from the
geometric signs in the Vedic period" (139).
The strongest point in favor of the invention theory is the stiffly symmetrical,
geometric appearance of Asokan BrahmT, which does indeed give the impression of
an arbitrarily created script. But this superficial feature does not necessarily prove
anything about the history of the script; Asokan BrahmT as we have it could be a
formalized monumental version of a preexisting script of more cursive aspect. More-
over, despite its superficially regular and standardized appearance, closer analysis
reveals variations in individual letter forms and systemic features (see 2.2.4) which
are suggestive of a reformed version of a preexisting script. For other reasons as
well, the invention theories are not persuasive.57 The comparative study of the his-
tory of the evolution of writing systems worldwide shows that the invention ex nihilo
of a highly sophisticated script would be unusual, if not completely unique. Arbi-
trarily created scripts are generally based, to a greater or lesser degree, on preexist-
ing scripts which the creators of the new script know or at least have seen. More-
over, as will be seen later ([Link].4 and [Link].5), there are clear indications that a
Semitic script, presumably Aramaic, served, in part at least, as a model for BrahmT.
In any case, all of the invention theories are purely speculative, lacking any hard
historical or documentary support.

54. Op. cit. (n. 50), 1 and 51-8.

55. "Brahml—An Invention of the Early Mauryan Period," in OBS 1-53.
56. The Palaeography ofBrahmi Script in North India, 8, and "Fresh Light on the Origin of BrahmT
Alphabet," JOI 13, 1964, 360-71 (esp. 367).
57. See also the generally critical discussions of Goyal's invention theory in OBS 67—125.
22 Indian Epigraphy
[Link] Theories of non-Indian (Semitic) origin

[Link].1 Greek. James Prinsep, the decipherer of Brahml (see 6.2.1), was the first
to suggest a possible connection between Greek and the ancient Indian scripts, namely,
that "the oldest Greek... was nothing more than Sanskrit [sic] turned topsy turvy !"58
This relationship was reversed by K. Ottfried Miiller,59 who proposed that Brahmi
was derived from Greek after the invasion of Alexander the Great. A modified ver-
sion of the Greek theory was proposed by J. Halevy,60 who derived six of the Brahml
characters (a, ba, ga, dha, tha, andna) from the corresponding Greek letters, and the
rest of the characters from Kharosthl (see 2.3.7) and Aramaic. While composite al-
phabets are not in principle impossible, as in the example of the Coptic script which
supplements the Greek alphabet with six characters from demotic Egyptian, Halevy's
derivation from three separate prototypes is obviously forced.
Although the theory of Greek origin won some early followers,61 it had fallen
out of favor until recently, when it was revived by Falk in SAI (109-12 and 338-9).
Falk, influenced by the arguments of Halevy, sees Brahml as an intentional creation
of the time of Asoka, created on the model of Kharosthl and Greek. According to
him, the formation of Brahml was influenced by Greek, particularly in respect to
its direction of writing (cf. [Link].4), its monumental ductus, the differentiation (in
contrast to Kharosthl) of short and long vowels, and the specific formation of some
characters, especially 0 tha, which he connects with Greek dtheta (SAI 111). Al-
though the arguments for a Greek influence on the general ductus of Brahml (cf.
n. 61) are plausible, it would seem that Greek had little influence on the specific for-
mation of the Brahml characters. The example of tha and theta, stressed by Falk, is
an isolated and exceptional case for which other possible explanations are available
(see [Link]). The development in BrahmT of a system for the differentiation of vowel
quantity can more easily be seen as indigenous, in light of the long-standing Indian
tradition of sophisticated phonetic analysis, than as an influence from Greek. Whereas
the short/long vowel pairs in Brahml are indicated by a complete and regular system
of variations of a basic form or diacritic for each vowel, in Greek script they were
represented by an entirely different (and defective) system involving distinct and un-
related alphabetic signs (e.g., e epsilonl r\ eta). In light of these very different ap-
proaches to the notation of vowel quantity, it is doubtful whether Brahml derived
even the basic concept from a Greek prototype. Thus although it is possible that
Brahml was influenced by Greek in the formation or development of its superficial
aspect, the evidence for an underlying role in the formation of the script itself is not
strong.

58. JASB 6, 1837, 390.

59. Gijttingische Gelehrte Anzeigen 1838, 252.
60. JA, ser. 8, vol. 6, 1885, 290ff.; see also Revue Semitique 3, 1895, 256ff. For full references,
see n. 77 and section 2.3.6.
61. See, for example, SAI 109-11. Diringer's account in The Alphabet, 335, of the adherents of
the derivation of Brahml from Greek is inaccurate and misconstrues the views of some of the persons
cited, notably of Emile Senart, who really suggested (in his review of Cunningham's Inscriptions of
Asoka, JA, ser. 7, vol. 13, 1879, 535) nothing more than an "influence grecque . . . exterieure et
secondaire" on the monumental character of the script.
 Writing and Scripts in India 23

[Link].2 Kharosthi. As noted earlier, Falk (SAI338-9), influenced by Halevy (see

also 2.3.8), saw Brahml as essentially based on Kharosthi with a strong secondary
influence from Greek; according to him, "haben wir es bei der Brahml mit einer
Neuentwicklung auf der Basis zweier Schriften zu tun"62 (338). If it is true, as now
seems nearly certain (see 2.1.2), that KharosthT was older than Brahml, it can cer-
tainly be accepted that the overall graphic system which characterizes both (2.1.3)
was adapted from the former to the latter, and in this limited sense Brahml can be
said to be "derived" from KharosthT. But in terms of the actual forms of the charac-
ters, the differences between the two Indian scripts are much greater than the simi-
larities (see 2.3.7), and many more of the forms of the Brahml characters can be
explained by reference to Aramaic and/or other Semitic scripts ([Link].3-[Link].5)
than to Kharosthi. Although some of the Brahml forms, such as A ga and Q tha, are
explained by Falk (SAI 111) by reference to the corresponding Greek letters (y gamma
and Q theta; see earlier), the great majority of them can be accounted for by neither
Greek nor Kharosthi.

[Link].3 South Semitic. An apparent relationship between Brahml and the ancient
South Semitic scripts was noted by several early scholars63 who noted similarities
between certain Brahml letters and the corresponding characters of the Himyaritic
inscriptions of South Arabia. A South Semitic prototype for Brahml was first pro-
posed by Frangois Lenormant in 1875,64 and at greater length by W. Deecke in 1877.65
A more plausible argument for a South Semitic (Sabaean) derivation was presented
by Isaac Taylor66 with a comparative chart (320) of the alphabets concerned.
Among the points argued in favor of the South Semitic hypothesis is, first of all,
the direction of writing. South Semitic inscriptions were written from right to left,
like other Semitic scripts, but also sometimes in boustrophedon, or, like Brahml, from
left to right. Second, and more important, a South Semitic prototype for Brahml, or
a hypothetical common ancestor, would provide plausible prototypes for several of
the Brahml letters which are most problematic in the North Semitic derivation
([Link].4). Thus Brahml D ba, for which a North Semitic prototype is problematic,
is very much like some Sabaean forms of bet such as D , as given by Taylor in his
comparative chart. Likewise Brahml £> da, which also presents serious problems in
the North Semitic derivation, has a much closer resemblance to such Sabaean forms
of dalet as ^; Brahml y ma looks like Sabaean ^ mem; and so on. According to Tay-

62. "In Brahml we are dealing with a new development on the basis of two scripts."
63. References given by Albrecht Weber, "Ueber den semitischen Ursprung des indischen Alpha-
bets," ZDMG 10, 1856, 389-406 (esp. 402-5) = Indische Skizzen (Berlin: F. Dummler, 1857), 127-50
(esp. 145-8).
64. Essai sur la propagation de Valphabet phenicien dans Vancien monde (Paris: Maisonneuve,
1875), 1.152.
65. "Ueber das indische Alphabet in seinem Zusammenhange mit den iibrigen siidsemitischen
Alphabeten," ZDMG 31, 1877, 598-612. Deecke proposed to derive both BrahmT ("Indisch") and
Himyaritic from a hypothetical Ur-South Semitic script, which, in turn, was derived from Assyrian
cuneiform. A similar idea was espoused by T. W. Rhys Davids in Buddhist India (London: T. Fisher
Unwin, 1903), 114.
66. The Alphabet, 11.314-23.
24 Indian Epigraphy

lor, the South Semitic derivation can also be justified on historical and geographical
grounds: "[F]rom the 10th to the 3rd century B.C. Yemen was the great central mart
in which Indian products were exchanged for the merchandize of the West" (314),
so that "there was . . . ample opportunity for the transmission to India of the Sabean
alphabet" (315).
Despite these advantages, the South Semitic hypothesis also has several impor-
tant weaknesses. First, the argument based on the common direction of writing in
BrahmT and some South Semitic inscriptions does not carry great weight; for, as will
be discussed shortly ([Link].4), the overall importance of the factor of direction of
writing has been exaggerated in much of the discussion of this subject. Second, as
for the derivation of individual characters, for every case in which South Semitic
offers a better prototype, there is at least one other where a North Semitic model is
preferable. Brahml (,pa, for instance, is certainly more readily derivable from North
Semitic ? pe than from the South Semitic forms of the letter. Finally, the historical
arguments are similarly inconclusive, in view of the lingering uncertainty as to the
antiquity both of the South Semitic inscriptions67 and of BrahmT itself.
For these reasons, the South Semitic hypothesis, though not completely without
merit, no longer enjoys much support,68 and Biihler's criticisms (OIBA 53-5) of
Taylor's South Semitic models and arguments in favor of North Semitic prototypes
for the BrahmT letters have been generally accepted. The undeniable similarities
between some South Semitic letters and their BrahmT correspondents are nowadays
usually considered by the proponents of a Semitic origin to be due to parallel but
independent developments from an ultimate common prototype, conditioned, in part
at least, by their common monumental form which was conducive to the develop-
ment of plain geometric forms such as the square bet/ba.69

[Link].4 Phoenician. The possibility of a connection of the Indie scripts with North
Semitic writing (see table 2.2) was noted as early as 1821 by Ulrich Friedrich Kopp.70
But Kopp's comparisons (which, to his credit, were presented with the greatest hesi-
tancy and restraint) were based on modern forms of the Indie scripts and hence did
not permit of anything like reliable conclusions. The first comprehensive and authori-
tative treatment of the Semitic hypothesis was Albrecht Weber's influential "Ueber
den semitischen Ursprung des indischen Alphabets" (see n. 63), in which he presented
the first detailed comparison of the Phoenician script with early Brahml. Although
the results were inevitably imperfect,71 on the whole the argument is reasonably

67. See, e.g., J. Naveh, Early History of the Alphabet, 2d ed. (Jerusalem: Hebrew University, 1987),
43-4.
68. R. N. Cust, e.g., at first favored this theory but later retracted his support for it ("On the Ori-
gin of the Indian Alphabet," JRAS, n.s. 16, 1884, 325-59; see especially 351-4 and 359). Although
Cust's article does not contribute much in the way of original ideas on this subject, it does provide
useful summaries and evaluations of previous discussions, as well as bibliographic information. (See
also Cust's article "The Origin of the Phenician and Indian Alphabets," JRAS 1897, 49-80.)
69. Cf. [Link].1, n. 61. According to Senart, cited there, the common factor would be the second-
ary influence of Greek on the outward monumental form of both Brahml and South Semitic.
70. Bilder und Schrifien der Vorzeit (Mannheim: privately printed, 1819-21), 11.367-75.
71. Some of the identifications, such as that of Phoenician sade with BrahmT jha, were obviously
wrong.
 TABLE 2.2. Comparison of the North Semitic and Early Indian Scripts

Semitic Indie
Value Phoenician Aramaic Value Kharosthi Brahmi
• a

b ba

g ga
d da

h ha

w va

z ja
fc
h gha

t tha

y ya

k ka

i la

m ma
n na

s sa
t e

P pa

s ca

q kha

r ra

S Sa

t ta
26 Indian Epigraphy
convincing, and Weber's conclusion is not unjustified: "Sollte im Einzelnen die
palaographische Identification Widerspruch und Berichtigung erfahren, so wiinsche
ich dergl. selbst auf das dringendste herbei, das allgemeine Resultat aber wird
schwerlich angefochten werden konnen"72 (401).
Weber's pioneering efforts were refined in the definitive statement of the North
Semitic hypothesis by Georg Biihler in 1895, namely, in his classic On the Origin of
the Indian Brahma Alphabet (OIB A).73 Biihler compared the early forms of BrahmT
from the Asokan and other early inscriptions, including the aberrant forms of the
Bhattiprolu inscriptions, with the corresponding letters of the North Semitic alpha-
bet of the tenth to eighth centuries B.C. (to which period he attributed the Indian bor-
rowing; see 2.1.2) as found in the early Phoenician inscriptions and other contempo-
rary documents which were then known. He was thereby able to formulate a derivation
which improved in several respects upon that of Weber. For example, Biihler gave
BrahmT ca as the derivative of sade, which is obviously preferable to Weber' sjha.
Likewise Biihler's gha from het is clearly better than Weber's ca, as is his kha in-
stead of Weber's ka as the derivative of qoph.
By no means, however, were all of Biihler's derivations beyond question. In
order to account for graphic differences between many of the BrahmT characters
and their supposed Semitic prototypes, Biihler had to invoke certain general deri-
vational principles (OIBA 56-8). For example, the "appendages" which constitute
the distinctive portion of the letters are generally attached at the bottom or middle,
rather than at the top as in Semitic; therefore, "a number of the Semitic signs had
to be turned topsy-turvy or to be laid on their sides, while the triangles or double
angles, occurring at the tops of others had to be got rid of by some contrivance or
other" (57-8). But even if these principles are accepted, fewer than half of the
equations of the twenty-two original Semitic characters with their alleged Brahml
derivatives can be considered beyond doubt. Among these, only seven—a, ga, tha,
pa, ca, kha, and sa—can be derived from their proposed prototypes with a mini-
mum of manipulation, for example, by simple inversion as in the case of (, pa < 7
pe and 9 kha < <? qoph. The remaining fourteen derivations range from likely but
paleographically uncertain, for example J, ya < 3. yod, to highly problematic, as in
« ma < "1 mem, ^ ra < ^ resh and nba<3 bet. In his explanations of some of the
more difficult derivations Biihler could well be charged with abusing his rather
vague principle of "some contrivance or other" (58) in the adaptation of the Semitic
forms to the Indian pattern.
Moreover, the direct derivations from the twenty-two consonantal characters of
the supposed North Semitic prototype account for less than half of the full Indie al-
phabet, which comprises at least forty-six74 distinct characters. Biihler thus derived
the remaining Brahml characters by a process of secondary development from the
primary, Semitic-based characters. Most of his secondary derivations, such as those

72. "Though the paleographic identification may be subject to refutations and correction in de-
tails, I would most insistently maintain that the overall result can hardly be challenged."
73. Here Biihler acknowledged his debt to his predecessor with the comment that "the identifica-
tions agree for the greater part with Professor Weber's, whose important essay . . . very nearly solved
the problem of the origin of the Brahma alphabet" (55).
74. On the probable number of characters in the primitive form of Brahml, sec OIBA 27—35.
 Writing and Scripts in India 27

of the aspirates \apha and <fc cha from the corresponding nonaspirates (, pa and <j ca
are obvious, but others, such as bha and dha, are less convincing.75 The retroflex
consonants, which have no direct prototypes in Semitic, are derived, according to
Biihler, by a similar secondary process, whereby, for instance, O tha is a reduction
of O tha and C ta a further reduction of tha.
Thus although Biihler's North Semitic hypothesis has much merit, it is far from
perfect, and has not surprisingly been subjected to extensive criticism on various
methodological, procedural, and historical grounds. Ojha, for example (BPLM 24-6),
was strongly critical of Biihler's manipulations of the forms of the supposed
Phoenician prototypes, alleging that by such methods one could "derive" virtually
any script from any other.76 J. Hale"vy77 objected, with good reason, to Biihler's cita-
tion of the aberrant forms of several letters of the Bhattiprolu script as intermediate
forms between those of the Semitic prototype and their Asokan Brahml reflexes.78
He argued that since the Bhattiprolu inscriptions are at least fifty years later than the
Asokan,79 Buhler's use of them constitutes a violation of the first part of his own
guiding principle that "the comparison must be based on the oldest forms of the In-
dian alphabet and actually occurring Semitic signs of one and the same period" (OIBA
55).80 Although some of HaMvy' s other criticisms of Biihler are completely unjusti-
fied, and although his own conclusions (see [Link].1) are highly questionable,
Halevy' s methodological criticisms, especially those from the point of view of Semitic
epigraphy, do expose significant weaknesses in Buhler's arguments.
Finally, there is the much discussed matter of the direction of writing. In sup-
port of his argument for a Semitic derivation, Biihler cited an early coin from Eran
bearing a Brahml legend running from right to left instead of the normal direction
as a remnant of "a period during which the Brahma characters were written in both
directions" (OIBA 45), that is, of a transitional phase in the reversal of the original
direction of writing of the Semitic prototype. This particular piece of evidence has
been rejected by several scholars81 on the grounds that reversed coin legends are
not at all uncommon from various periods and regions. However, some more re-
cent discoveries, such as the Erragudi Asokan rock edicts written in semiboustro-

75. On the derivation of the aspirates kha, gha, and tha, see section [Link].
76. By way of illustration, Ojha shows how by such arbitrary alterations of forms one could for-
mulate a pseudoderivation of Brahml from modern Roman script (26). But while it is true that several
of Biihler's derivations did involve seemingly arbitrary alterations of the originals, Ojha's criticisms
are exaggerated and hence only partially justified, since, as we have seen, at least half of Buhler's deri-
vations are reasonably cogent. Thus when Ojha emphasizes (21) that only one Brahml letter (go) agrees
exactly with Phoenician, he hardly does justice to Buhler's carefully argued, if not invariably satisfac-
tory, formulation; insisting on exact identity between prototype and derivative is just as unreasonable
as arbitrarily concocting resemblances between dissimilar characters.
77. "Nouvelles observations sur les ecritures indiennes," Revue Semitique 3, 1895, 223-86 (dis-
cussed 234).
78. E.g., Biihler gives the Bhattiprolu form of/a, C ,as a link between Phoenician t zain and ASokan
Brahmi £70.
79. Actually, as it now appears, they may even be considerably later than that; see section [Link].
80. The second part of this rule is also violated, according to Halevy (242-3), by Buhler's choice
of North Semitic prototype letters from various inscriptions of widely differing regions and periods.
81. E.g., by E. Hultzsch in his otherwise favorable review of BIP in IA 26, 1897, 336. See also J. F.
Fleet in the introductory note (3-4) to his English translation of Buhler's Indian Paleography (IA 33).
28 Indian Epigraphy

phedon,82 have again raised the possibility that Brahmi might sometimes have been
written from right to left in the early phase of its development. The several early
Brahmi cave inscriptions from Sri Lanka written from right to left83 could also be
cited as a survival of the old Semitic direction of writing, but since they are prob-
ably considerably later than the Asokan inscriptions84 they may be nothing more
than an anomalous local development. The evidence on this point is thus incon-
clusive, but in any case its overall importance has perhaps been exaggerated in many
of the discussions of the origin of BrahmT. Instability in direction of writing is a
common phenomenon in ancient scripts generally, and in any case a reversal is
attested in various derivatives of Semitic scripts such as Greek and Ethiopic. Thus
the left-to-right direction of (most) early Brahmi is not in and of itself any strong
evidence against a Semitic origin.

[Link].5 Aramaic. A derivation of BrahmT from the later North Semitic Aramaic
script (see table 2.2) rather than the earlier Phoenician was proposed as early as 1874
by A. C. Burnell in SIP 8-9, and has more recently been endorsed by Diringer,85
who is inclined to consider Brahmi to be an adaptation rather than a "simple deriva-
tive" of early Aramaic. Paleographically the Aramaic derivation is plausible, since
many of the early Brahmi letters can be formally derived as well, or even better, from
Aramaic as from Phoenician prototypes. The paleographic ramifications of this theory,
however, have not yet been fully worked out. Historically and chronologically too,
the Aramaic theory is much preferable to the Phoenician derivation. The widespread
use of the Aramaic language and script as a lingua franca throughout the Near East
and the Iranian world and as a bureaucratic language of the Achaemenian empire
provides a ready explanation for its influence in India, in contrast to Biihler's weak
historical, geographical, and chronological justifications for a Phoenician prototype.
Moreover, the discovery since Biihler' s time of six Mauryan inscriptions in Aramaic
([Link]) strongly supports the hypothesis of an Aramaic connection.
However, a possible objection to the derivation of Brahmi from Aramaic86 is that,
since it has been established with virtual certainty that the KharosthI script is derived
from Aramaic (2.3.6), it is hard to see why another, very different, Indie script would
have developed from the same prototype in a contiguous region. If the hypothesis of
the invention of Brahmi under Asoka's sponsorship ([Link]) is correct, this re-creation
may be attributed to the emperor's desire to invent a distinct imperial script, perhaps
under the inspiration of Old Persian cuneiform, which would be suited to the pro-
mulgation of edicts in written form.87 But it must be admitted that there is no direct
statement to this effect in the edicts themselves.

82. Though the reversed writing of some of the lines at Erragudi may be due to the incompetence
or ignorance of the engraver; see the comments on this point in OBS 31.
83. Besides the well-known Duvegala cave inscription (SI 1.242), S. Paranavitana (Inscriptions
of Ceylon, [Link]) notes no less than thirty-nine such inscriptions.
84. Sircar (SI 1.242) dates the Duvegala (Duwe Gala) inscription to the first century B.C.
85. The Alphabet, 336.
86. As noted by Senart, JA, ser. 7, vol. 13, 1879, 534.
87. As proposed by, among others, Falk (SAI 338-9), although he sees KharosthI itself, rather
than Aramaic, as the principal prototype of Brahmi ([Link].1 and [Link].2).
 Writing and Scripts in India 29

[Link] Observations and conclusions on the probable origin of BrahmT

The various theories of an indigenous Indian origin for Brahml are all more or less
speculative, and hence impossible to either prove or disprove. Nonetheless, it would
be imprudent to dismiss them entirely. In particular, the possibility that there is ulti-
mately some historical connection between the Indus Valley script and Brahml can-
not be decisively ruled out, but unless and until some significant progress toward the
decipherment of the former is achieved, we can only treat the matter as an unproven
hypothesis.
The theory of a Semitic origin for Brahml, on the other hand, does have a strong,
if not entirely conclusive, body of concrete evidence in its favor. The derivation, along
the lines worked out by Biihler, has (as noted earlier, 2.2.3, n. 42) been generally
adopted by Western scholars; typical is the opinion of Hultzsch that the North Semitic
theory "is so well supported with facts that it cannot fail to meet with general accep-
tance."88 But in South Asia, as we have seen (see 2.2.3 and [Link].4), the Semitic
hypothesis is not widely accepted, though there too some scholars, notably A. H. Dani
in his influential work on Indian paleography (DIP 23-30),89 have cautiously sup-
ported some form of Semitic derivation.90 Part of the problem is that, despite the
defects in Biihler's methodology and data, no one since him has undertaken a com-
prehensive and careful paleographic reexamination of the Semitic hypothesis. Hence
evaluations of the theory have essentially consisted of discussions, supportive or
critical, of Biihler's work; but it should be kept in mind that the shortcomings of
Biihler's presentation do not necessarily discredit the Semitic hypothesis itself.
Moreover, searching for formal prototypes for each of the characters of early
BrahmT in one or the other alleged Semitic prototypes is one way, and not necessar-
ily the most effective way, to address the problem. For even many of the supporters
of the Semitic hypothesis concede that, in Dani's words, "[T]he BrahmT letters are
not literally 'derived' from the Semitic letters as is commonly understood, but are
only based on them" (DIP 29).91 In other words, the relationship of BrahmT to an
alleged Semitic parent script may be more analogous to, say, that of the Arabic script
and its presumptive Nabataean parent, which is widely accepted even though it can-
not be traced in every detail,92 than it is to better attested derivations such as that of
the Greek alphabet from Phoenician (which seems to have served as the model, con-
sciously or otherwise, for the studies of Biihler et al.).
Thus it may be more revealing to pay more attention to systemic features and
patterns instead of concentrating on the formal derivation of individual characters.
From this point of view, the Semitic hypothesis is more convincing. For instance,
the system of postconsonantal diacritic vowel indicators looks like a natural adapta-

88. IA 26, 1897, 336.

89. In the preface to the recent second edition of this book (1986), however, Dani expresses some
second thoughts on the matter, suggesting that "for the origin of Brahml we should look elsewhere
[than Semitic]" (ix).
90. Cf. also K. G. Krishnan in OBS 68: "The Semitic affiliation of some of the forms, as explained
by Biihler, cannot be ignored, though it is ... partial in its coverage."
91. See also Diringer, The Alphabet, 336.
92. See ibid., 270-1, and Naveh, Early History of the Alphabet, 160-1 and 221.
30 Indian Epigraphy
tion of the Semitic consonant-syllabic script for use in Indian languages (2.1.2 and
2.3.6). Similarly, the evident development of the retroflex consonants as modified
forms of the corresponding dentals suggests an adaptation of a non-Indie prototype,
since in an originally Indian system one would have expected independent signs for
the two classes from the very beginning. Particularly revealing is the pattern of for-
mation of the aspirate consonants. For while these are typically formed as modifica-
tions of the corresponding nonaspirates ([Link].4), there are some striking excep-
tions to this pattern, notably, 7 kha, b gha, and© tha,93 none of which bear any graphic
relationship to the corresponding nonaspirates (-r ka, A ga, and A to) but which can
be derived directly from Semitic prototypes, namely, f1 qoph, n het, and 6 tet, re-
spectively. This pattern can hardly be coincidental, and is perhaps the single stron-
gest systemic indication of a Semitic background.
In conclusion, there are strong systemic and paleographic indications that the
BrahmT script derived from a Semitic prototype, which, mainly on historical grounds,
is most likely to have been Aramaic. However, the details of this problem remain to
be worked out, and in any case it is unlikely that a complete letter-by-letter deriva-
tion will ever be possible; for Brahml may have been more of an adaptation and re-
modeling, rather than a direct derivation, of the presumptive Semitic prototype, per-
haps under the influence of a preexisting Indian tradition of phonetic analysis.94
However, the Semitic hypothesis is not so strong as to rule out the remote possibility
that further discoveries could drastically change the picture. In particular, a relation-
ship of some kind, probably partial or indirect, with the protohistoric Indus Valley
script should not be considered entirely out of the question.

2.2.4 Characteristics of Brahml in the Mauryan period (third century B.C.)

The Brahml script as it first appears in inscriptions of the Mauryan period (see table
2.1) is an almost fully developed writing of the characteristic Indian diacritically
modified syllabic type described earlier (2.1.3). The repertoire of the script as repre-
sented in the Asokan inscriptions consists of six full (initial) vowel signs (a, a, i, u,
e, o), thirty-two consonants, and eight medial vowel diacritic signs (a, /, T, u, u, e, o,
ai), plus the anusvara sign. The absence of the initial vowel signs T and u is no doubt
a statistical accident due to their rarity, but other vowels which are absent in both
full and diacritic form, such as r, f, and possibly au, apparently did not yet exist in
the script, as they were developed only later when Brahml began to be used for writ-
ing Sanskrit (see [Link]). Among the consonants only the sign for na is lacking, as
is the visarga, no doubt also because they are not needed for the Prakrits which BrahmT
was used to record at this time.
Asokan BrahmT displays on the whole a uniform aspect, and although it was for-
merly believed95 to have significant regional varieties, more recent studies have shown
that the script was in fact essentially uniform all over India.96 There are, however,
93. The cases of two other anomalous aspirates, dha and hha, seem to involve different circumstances,
94. Cf. DIP 28-30, and Diringer, The Alphabet, 336.
95. See, e.g., BIP 33-4.
96. C. S. Upasak, The. History and Palaeography of Mauryan Brahml Script, especially 29-32
and 193; DIP 35; and section 2.1.2.
 Writing and Scripts in India 31

variant forms for many of the characters, which, though evidently not geographi-
cally determined, are still significant for the development of the script; see, for ex-
ample, the several varieties of initial a noted in BIP 6. Also of interest are a few
letters such as dha (D or (J) and diacritic o (e.g., in 1 or i, no), whose directional
orientation has not been fixed (DIP 48).
The orthography of Brahml of the Mauryan period is also not fully developed or
standardized. The notation of the three sibilants in some of the Asokan inscriptions
is inconsistent and inaccurate (especially in Kalsl rock edicts XI-XIV where, for
example, sususa - Skt. susrusa, XI. 29). Vowel quantity, especially the distinctions
between ill and u/u, is not always consistently noted, again at Kalsl in particular.97
Consonantal conjuncts are in general constructed as in the later scripts, that is, by the
graphic combination of the component characters, sometimes with abbreviation of
one component; but the relative positioning of the components is not always pho-
netically accurate. For example, combinations with r are always constructed with the
r above, regardless of its actual phonetic position, so that the same character (£) rep-
resents both rva and vra.98
The other (non-Asokan) early Brahml inscriptions, such as Sohgaura, Piprawa,
and Mahasthan (2.1.2), have an even more primitive orthography. They have no
conjuncts, and do not distinguish i/i or u/u at all; Piprawa seems not to have long
vowels at all, except for one doubtful a. These inscriptions also have a few variant
letter forms, such as the ma (g) at Sohgaura; but due to the paucity of the materials it
is difficult to evaluate their chronological significance. Thus while some scholars
have attributed an early, pre-Mauryan date to some or all of this group, others (e.g.,
Dani, DIP 56-7) have dated them to the post-Mauryan era. All in all, it seems most
likely that they are roughly contemporaneous with the Asokan inscriptions. Their
primitive orthography does not prove an early date for the inscriptions themselves,
only the use of a less formal style than that of the Asokan inscriptions, which occa-
sionally show early signs of the influence of Sanskritic orthography.

2.2.5 The historical development and derivatives ofBrahmT

[Link] The "Sunga" period {second through first centuries B.C.)
After the Mauryan period, Brahml script began to undergo its gradual process of evo-
lution (see tables 2.3 and 2.4). But, with the exception of the Bhattiprolu and early Tamil
scripts (to be discussed shortly), the changes are at first mostly minor, so that BrahmT
of the first century B.C. is still not markedly different in appearance from its Mauryan
predecessor. Nonetheless, certain of the formal evolutions first noted in this period
(especially in the latter half thereof), such as the development of rudimentary head marks
and the equalization of the vertical portions of certain letters, are important in prefig-
uring more radical formal changes which are to appear in the following centuries.
The head marks which begin to appear in certain inscriptions of this period (e.g.,
in the Hathlgumpha inscription) in the form of a small triangle at the top of the ver-

97. Hultzsch, Inscriptions of Asoka (CII 1), Ixxi.

98. Other instances of apparently reversed conjuncts, such as ^ yva for vya, are also quotable (see
section [Link]).
 TABLE 2.3. The Development of Brahmi Script

Scythian-
Value Mauryan SuAga Ku?ana Gupta Siddhamatrka
(3rd c. B.C.) (2nd-lstc. B.C.) (lst-3rdc. A.D.) (4th-6th c. A.D.) (7th-9th c. A.D.)

ka

ga

ja

ta

da

pa

ya

la

sa

ha

Note: These are selected representative North Indian characters in normalized forms.
 TABLE 2.4. The Development of NA in Brahmi and Its Derivative Scripts

Northern Scripts Southern Scripts

3rd c. B.C.

IstC. A.D.

3rdc. A.D.

6th c. A.D.

8th c. A.D.

lOthc. AJX

12thc. A.D.

Modem Scripts Punjabi Devanagari Gujarat! Bengali Oriya Telugu Kannada Tamil Maiayalam Sinhalese

Note: This table presents in broad outline die development of a representative character, na, from early BrShmi to the major modem Indian scripts.
Detailed charts for the development of each aJksara are provided in Sivaramamurti, ffiSIS 57-153
34 Indian Epigraphy
tical stroke are no doubt due to the influence of pen and ink writing," in which such
a mark naturally tends to appear at the point where the scribe begins the letter. This
accidental formation eventually came to be perceived as a part of the letter form
itself, so that it began to appear in epigraphic writing by way of imitation of pen and
ink script. This head mark was to develop in the succeeding centuries into different
shapes in different regions, eventually resulting in the formation of such character-
istic features of medieval and modern Indian scripts as the square of the "box-headed"
script, the continuous top line of Devanagarl, the curved "umbrella" of Oriya, or the
"check-mark" of Telugu.
The tendency toward equalization of the two verticals in such characters as \jpa
and ;u sa, which begins to make its appearance in some inscriptions of this period,
also prefigures developments of later centuries wherein the letters tend to be remod-
eled into a square frame. This again is an influential development for the history of
the derivative scripts, many of which (notably Devanagarl) retain this pattern through-
out their history.
Notable developments of individual characters in this period include the elonga-
tion of the vertical of the old cross-shaped ka of Mauryan Brahml leading to the
"dagger-shaped" form of this letter (-f > -f); the rounding of the top of ga (A > n); the
reversal in direction of dha (D > C); a tendency toward the development of angular
forms in such letters as ma and va (V > ^; i > A); and the replacement of the old curved
ra with a straight vertical shape (f > J). Among the few new characters which make
their appearance in this period is the initial ai (a), first seen in the HathTgumpha in-
scription (1. 1).
While the changes which Brahml underwent in this period in northern India are
relatively minor and superficial, the Bhattiprolu inscriptions present a unique and
radically different southern variety of the script. This script, found in nine inscrip-
tions100 on relic caskets from the stupa at Bhattiprolu in Andhra Pradesh, differs from
standard early Brahml in two important respects. First, the formation of five of the
consonantal characters, namely gha,ja, ma, la, and sa (or sa)m is radically differ-
ent. The character ma, for example, (X) is upside-down compared with the standard
form (V) of the letter. Particularly interesting is the formation of gha (1), which can
be seen (El 2, 323-4) as a secondary derivative of the sign for ga, unlike standard
Brahml gha, which appears to be a distinct character probably derived directly from
Semitic het.m

99. See DIP 52-3.

100. A tenth inscription on a small piece of crystal is in a script which is much more like standard
early Brahml, though it has a few features, such as da with the opening to the right, which agree with
the aberrant local script rather than the standard.
101. The character in question (J-) was read by Biihler (El 2, 324) assa because of its paleographic
similarity to standard sa, but Liiders (Philologica Indica, 213-29) provided cogent linguistic arguments
(213—7) for reading it as sa. Although most modern scholars have followed Biihler, Lu'ders' reading
seems to be preferable. (Note that the relation and distinction between sa and sa in early BrahmT are
often problematic; see 2.2.4 and [Link] n. 5.)
It is worth noting that although Biihler's edition of the Bhattiprolu inscriptions is the one which is
usually referred to and followed in subsequent publications, Liiders' readings and interpretations are
very different from and much superior to Biihler's, and thus should be adopted as the standard.
102. OIBA 46; see section 2.2.4. Note, however, that Dani (DIP 70) sees the Bhattiprolu gha as a
 Writing and Scripts in India 35

The second peculiarity of the Bhattiprolu script is its system for notation of the
postconsonantal vowels a and a. Uniquely among all the early Indie scripts, at
Bhattiprolu the inherent vowel system is discarded, and a consonant followed by a
does have a vowel marker consisting of a short horizontal line at the upper right,
similar to the sign for a in standard BrahmT; e.g., -f = ka. (a in the Bhattiprolu script
is marked with a further downward extension of the a diacritic, as in -f = ka.) Biihler
(OIBA) explained that this device "seems to have been invented in order to avoid
the necessity of forming the ligatures, which make the ordinary Brahma alphabet
cumbersome and difficult to read . . . , and in order to express final consonants more
conveniently" (82). He evidently assumed that consonantal groups would have been
written in this script by putting the prior member(s) in the bare form, without any
diacritic, to indicate their vowelless state, thereby obviating the complication of con-
junct consonants. But, as noted by Mahalingam (ESIP 120), no such vowelless char-
acters or final consonants actually occur in the Bhattiprolu inscriptions, presumably
because they were written in a vernacular Prakrit dialect. Biihler's explanation seems
to presuppose that the script was also used for writing Sanskrit or more formal vari-
eties of Prakrit in this period, but there is no evidence that this was the case. Rather,
discoveries made subsequently to Biihler's writings, namely the Old Tamil Brahml
inscriptions discussed later, suggest that this aberrant vowel notation system reflects
innovations made in the environment of Dravidian rather than Indo-Aryan languages.
The significance of the anomalous consonant forms at Bhattiprolu is, however,
still not entirely clear, especially in light of the prevailing uncertainty as to the date
of the script; Biihler (El 2, 325) tentatively dated the Bhattiprolu inscriptions at
not later than 200 B.C., while Sircar (SI 224; cf. n. 5 there) places them at "about
the end of the 2nd century B.C." These nonstandard consonant characters can hardly
be dismissed as mere "mistakes" on the part of the engraver (so Dani, DIP 70),
though Biihler ([Link].4) seems to have overemphasized their significance for the
origin and history of Brahml. All in all, it seems more likely that the Bhattiprolu
script represents a provincial offshoot of early BrahmT in the south, rather than a
separate line of development from a hypothetical Semitic prototype itself, as Biihler
believed.
A second southern Indian variety of BrahmT attributable to this period103 is the
script used in a series of brief dedicatory cave inscriptions from Tamil Nadu in an
early form of the Tamil language. This alphabet has two notable peculiarities, which
in some respects resemble those of the Bhattiprolu script. First, it has four entirely

reformed variant of the standard Brahmi gha. See also section [Link].4, on Buhler's treatment of the
Bhattiprolu script in connection with his derivation of Brahml.
103. The dates of the Old Tamil inscriptions are uncertain and controversial. Iravatham Mahadevan,
in "Tamil-Brahmi Inscriptions of the Sangam Age" (in R. E. Asher, ed., Proceedings of the Second
International Conference Seminar of Tamil Studies [Madras: International Association of Tamil Re-
search, 1971], 73—106), 83, stresses their overall close paleographic resemblance to Asokan Brahmi
and hence dates the earliest specimens to the second and first centuries B.C. But Dani (DIP 72—4) places
them in the first century A.D., citing the evidence of potsherds from Arikamedu bearing inscriptions in
this script (AI2, 1946, 109—14) which are stratigraphically datable to this period. However, Mahadevan
points out that in some respects the Arikamedu inscriptions are paleographically more advanced than
the early cave inscriptions, and thus his dating seems preferable.
36 Indian Epigraphy
new characters, interpreted as na, ra,ra, and la, which were evidently created in order
to represent Dravidian phonemes not represented in standard (northern) BrahmT. (It
does not, however, share with Bhattiprolu the peculiar forms of the consonants gha,
ja, etc.) Second, its system of vowel notation differs from that of standard Brahml
and resembles that of Bhattiprolu. According to the authoritative interpretation of
this system by Mahadevan,104 in what appears to be the earlier form of this script a
consonant written without a vowel diacritic is to be understood as representing the
vowelless consonant itself, while the consonant written with the diacritic sign which
denotes long a in standard BrahmT is to be read with either the vowel a or a, depend-
ing on the context.105 In other words, the script in question has essentially abandoned
the "inherent vowel" principle.
Thus we find in the Tamil BrahmT inscriptions a system of representation of pure
(i.e., vowelless) consonants which seemed to be implied, but was not actually attested,
in the Bhattiprolu inscriptions. Therefore, although the two scripts do not appear to
be directly related paleographically, there evidently is a systemic relationship between
them. The presence in Bhattiprolu of explicitly noted postconsonantal a, but without
any examples of the expected "pure" or vowelless consonants, now looks, in light of
the Tamil BrahmT inscriptions, like the reapplication to Prakrit of a form of Brahml
which had been previously modified for the representation of a Dravidian language.
For, as explained by Mahadevan ("Tamil-Brahmi Inscriptions," 81-2), the systemic
features of these early southern varieties of Brahml would seem to reflect phonetic
features of Dravidian, such as the common occurrence of consonants in word final
position, rather than of Prakrit, which has no word final consonants.
The subsequent historical development of the modified southern system, how-
ever, is not at all what one would have anticipated by comparison with the usual
patterns of evolution of scripts. Surprisingly, the capability of the Bhattiprolu and
Tamil Brahml scripts to write vowelless consonants in a simple manner and thus to
obviate the need for consonantal conjuncts does not seem to have been continued in
the later forms of the southern scripts. Indeed, as shown by Mahadevan (op. cit., 82-3),
this innovative system actually is attenuated in some (perhaps later)106 Tamil Brahml
inscriptions, in which the unmarked consonant may represent either the pure conso-
nant or the consonant plus a, and in still later inscriptions the system of vowelless
consonants falls out of use entirely. Mahadevan attributes the failure of this tentative
experiment toward alphabetization to the influence of the scripts of adjoining regions
of India and Sri Lanka, which retained the standard Indie modified syllabic system;
104. In "Tamil-Brahmi Inscriptions of the Sangam Age"; see also his "Corpus of the Tamil-Brahmi
Inscriptions," in R. Nagaswamy, ed., Kalvettik Karuttaranku/Seminar on Inscriptions, 1966 (Madras:
Books [India] Private Ltd., 1968), 57-73; and T. V. Mahalingam, ESIP 120-2 and 140-1.
105. Mahadevan, "Tamil-BrahmT Inscriptions," 79. Mahalingam (ESIP 120) thinks that the long
vowel a is sometimes noted by a doubled form of the diacritic, but no clear examples of this are actu-
ally visible in the reproductions of the inscriptions.
106. In "Some Aspects of the Tamil-Brahmi Script" (JESI12, 1985, 121-8), Mahadevan expresses
some doubts about his own earlier formulation of the chronological relation of the two forms of the
script, considering it "more likely" that they were "more or less contemporaneous styles" (123). See
also R. Nagaswamy, "The Tamil, Vatteluttu and Grantha Script," in Proceedings of the Second Inter-
national Conference Seminar of Tamil Studies [see n. 103] 11.410—5 (esp. 413).
 Writing and Scripts in India 37

in his words, the southern innovation was "too radical a departure from all the other
systems of Indian writing" (83).107

[Link] The first through third centuries A.D.

In about the first three centuries of the Christian era, the gradual geographical differ-
entiation of the Brahml script continued to the point that we begin to discern several
distinct regional varieties. Although these local forms can still be considered essen-
tially as varieties of the same script, they clearly foreshadow the development of
separate local scripts which was to take place in the following centuries (cf. DIP 77-8).
While it was formerly the standard practice to classify the scripts of this and the suc-
ceeding periods in dynastic terms ("Kusana script," "Gupta script," etc.), Dani (DIP
78-9, 100-1; cf. 2.2.2) has argued that the traditional dynastic terminology is mis-
leading and inaccurate, and has preferred to treat the paleographic development of
Brahml in regional rather than dynastic terms. Thus, for the period in question, he
discerns several local varieties, among which the Kausambl, Mathura, western
Deccan, and eastern Deccan styles are of particular importance. Dani's geographi-
cal approach to historical paleography is undoubtedly a major improvement in this
field, and will in general be followed here.
Among the new developments in this period, several characters such as r (both
initial and postconsonantal), au, na, h (visarga), and halanta (vowelless consonants)108
first came into common use. These are, of course, the result of the growing popular-
ity in this period of Sanskrit or "Epigraphical Hybrid Sanskrit" (EHS) (3.2) as in-
scriptional languages (see 3.3.1). For the same reason, we now see a further devel-
opment of consonantal conjuncts, which become more frequent and more complex.
Particularly noteworthy is the development of special modified or abbreviated forms
of the semivowels r (in both preconsonantal and postconsonantal position, e.g., 6
rpa and y pro) and y (postconsonantal), reflecting their statistical frequency in con-
sonantal combinations in Sanskrit.109
The forms of the basic characters as well underwent significant modifications,
such that for the first time in this period (especially in the latter half thereof) the script
began to differ markedly in its overall appearance and ductus from the early forms
of Brahml. Among the most important changes was the further development of the

107. Later scripts used for Tamil did, however, develop a vowel cancellation marker consisting
of a dot above the consonant (pulli; cf. n. 37) which has the same effect as the vowelless consonant
system, i.e., eliminating the need for conjuncts, but which departs less radically from the standard Indie
graphic principles. The pulli is attested as early as the second century A.D. in Dravidian coin legends
(D. C. Sircar, "Silver Coin of Vasishthi-putra Satakarni," El 35, 1963-64, 247-52; see also R. Naga-
swamy, "A Bilingual Coin of a Satavahana," in Seminar on Inscriptions, 1966 [see n. 104], 200-2, and
P. Panneerselvam, "Further Light on the Bilingual Coin of the Satavahanas," IIJ 11, 1968-69, 281-8),
and is also referred to in the Tolkappiyam (Eruttadikaram, sutra 15). But it is difficult to know whether
pulli was a survival of the peculiar vowel notation system of the Tamil Brahml inscriptions, or evolved
separately as a result of the internal phonetic characteristics of Tamil.
108. This is used only in pausa (sentence final or verse final position), not within or between words
as in the Tamil systems discussed in the preceding section.
109. Note that a similar development takes place in Kharosthf; see section 2.3.4.
38 Indian Epigraphy
head mark, whose origins were first noticed in the preceding period. Now, the head
mark becomes firmly entrenched as an intrinsic rather than incidental part of the let-
ters, and is written as a distinct separate stroke. Moreover, we now begin to discern
the different regional treatments of the head mark (DIP 79-81), which are to have
such far-reaching consequences in later centuries (cf. [Link]).
Some of the consonantal characters were radically remodeled in this period. For
instance, the old form of da with the bulge to the right 0) was replaced by its mirror
image («(). The tendency toward the equalization of verticals in characters such as pa
and sa, first noted in the preceding period, is now strengthened so that these and oth-
ers tend to be molded into a square outline, which is reinforced by the angularized
forms of some letters such as Eja, X ma, and EJ sa. But in this period (especially in
its later phases) we also see in other characters the contrary effects of cursivization
(here, as elsewhere, reflecting the epigraphic imitation of pen-and-ink forms), which
result in rounded and looped forms of such characters as A ta, A na, c*> bha, and 34 sa.
Also in this period there first emerged a tendency toward calligraphic elabora-
tion. Thus characters such as a, ka, and ra, which end with verticals at the bottom,
developed ornamental loops to the lower left (e.g., $ ka). The diacritic vowel signs
were also often elaborated with extensions and flourishes. These more elaborate forms
are characteristic of the Deccan style, especially in the eastern Deccan. Here, in in-
scriptions from such sites as Nagarjunakonda, AmaravatT, and Jaggayyapeta, we find
extreme calligraphic developments with long vertical flourishes above and below
many of the letters (cf. [Link]). In general, we can begin to perceive in this period a
broad differentiation between the northern scripts, which are beginning to develop
their characteristic squarish and angular forms, and the southern scripts with their
typically rounded and flowing shapes.

[Link] The fourth through sixth centuries A.D.

During this period, conventionally referred to as the "Gupta era," the degree of
regional differentiation continued to increase. The northern and southern groups
were by now clearly differentiated, and other distinct regional scripts made their
appearance, for instance, in central India. In the north, the inscriptions of the Guptas
and their contemporaries were written in regional varieties of the northern script,
which may be called, following Sircar's terminology (HEP 113; cf. 2.2.2), "late
Brahml." The varieties of the northern Gupta script have traditionally been classi-
fied as western and eastern, the latter being characterized mainly by cursivized forms
of la (,0 and ha ft) and the looped sa (H). But Dani (DIP 102ff.) has shown that this
division is oversimplified, and he classifies the northern script into regional
subvarieties such as the Kausambi (roughly equivalent to the old "eastern" style),
Mathura, and Malwa styles. In general, the northern scripts of the Gupta era are
characterized by the continued extension of various forms of the head mark or, as
it now can be called, the top line. The shapes of the individual characters also con-
tinued to develop; among the changes which prefigure important characteristics of
later scripts are the prolongation of the right arm in f\ ga, *\ ta, § bha, and 1 sa', the
looped form of ^i na; and the development of the dot inside 6 tha into a horizontal
line.
 Writing and Scripts in India 39

In central India the peculiar "box-headed script" began to develop during this pe-
riod. The principal characteristic of this script, namely, the square head mark, is noted
in some northern Gupta inscriptions, but its full development, with the letters them-
selves molded into characteristically square, angular forms, first appeared in the in-
scriptions of the Vakatakas. This highly stylized script enjoyed a long period of popu-
larity in central India, where it continued to be used into the seventh century,110 and
also spread to the south, where it appears in some Kadamba and Pallava inscriptions.
In the south, we now begin to see clearly, for instance, in the inscriptions of the
Kadambas and early Calukyas, the strong preference for rounded forms and wavy
lines which is to characterize most of the southern scripts of subsequent centuries up
to modern times.111 In some inscriptions from the earlier part of this period we find
peculiar aberrant forms, apparently cursive modifications, of certain letters such as
* ta, 6 na, $ ma, and jj sa in the Mayidavolu (SI 1.457-61) and other early Pallava
copper plates. But in general the southern scripts of this and subsequent periods tended
to be more conservative as regards the basic shapes of the characters, which under-
went less radical reformation than in the north.

[Link] The development of the regional scripts in the early medieval

period (ca. seventh-tenth centuries A.D.)
Around the late sixth century, the so-called Gupta script of northern India evolved
into a distinct new script for which the preferred name is Siddhamatrka,112 which
was to have a profound effect on the subsequent development of the northern scripts.
Early specimens of this script include the Bodh-Gaya inscription of Mahanaman (A.D.
588/89; CII 3, 274-8) and the Lakkha Mandal prasasti (Appendix, no. 8). It contin-
ued to be used into the tenth century, undergoing a gradual transformation into
Devanagarl during the latter part of this period. Siddhamatrka was used as an epi-
graphic script not only in northern and eastern India but also in the west, where it
replaced the southern-style scripts which until then had predominated there, and
occasionally in the Deccan and even in the far south, for example, in the Pattadakal
biscript inscription (El 3, 1-7; see 2.5.4). In this respect it prefigured the role of its
daughter script, Devanagarl, as a quasi-national script which was sometimes preferred
to the local scripts for writing Sanskrit.
The Siddhamatrka script is principally characterized by a strongly angular aspect,
with a sharp angle (whence the term "acute-angled script") at the lower right corner of
each letter, reflecting the influence of pen-and-ink writing on the epigraphic script; by

110. So according to Sircar in HEP 113; according to C. Sivaramamurti (IESIS 202) it remained
in use into the eighth or ninth centuries,
111. This characteristic of the southern scripts is usually explained as a result of the exigencies of
writing with a stylus on palm leaves. This explanation has, however, been challenged in 1C II.680-1,
where it is attributed to the influence of calligraphic tendencies.
112. The script has also been referred to by several other names, such as "acute-angled," "early
NagarT," "Kutila," or "Vikata." (The latter two result from apparent misinterpretations of descriptive
terms found in various inscriptions; see Biihler, El 1, 1892, 75-6; Fleet, CII 3, 201; and Sircar, El 36,
1965-66, 50). The name Siddhamatrka is reported by Al-Birunl, and appears to be corroborated by the
term "Siddham" which is applied to it by Buddhist tradition in East Asia. This is thus one of the few
cases wherein we know the traditional name for one of the premodern scripts (cf. 2.2.2).
40 Indian Epigraphy

the extension of the head mark into wedgelike or triangular forms (whence it is some-
times referred to as "nail-headed"); and by a strong tendency toward calligraphic elabo-
ration, especially in the treatment of the vowel diacritics and subscript consonants (see
[Link]). Some of the vowel signs are highly developed; in particular, the curves of
diacritic i and T are extended downward to the point that they are equal to or even greater
in height than the consonant character to which they are attached, prefiguring the de-
velopment of vertical lines as vowel markers in Devanagarl. The shapes of certain let-
ters underwent major alterations, particularly in the later stages of the script wherein
the bipartiteya (u) and looped ka ($), which would become characteristic of Devanagarl
and other northern scripts, began to make their appearance.
In the far northwest, the Proto-Sarada script first emerged around the beginning
of the seventh century. This isolated variety is the ancestor of the later Sarada and
other scripts of the northwestern subgroup (see [Link]).
In the upper part of southern India, what may now be called the Telugu-Kannada
script continued its separate development. Changes in the characters were largely
determined by a strong tendency toward rounded and enclosed forms; thus 3 ka and
0 ra have fully enclosed shapes, and the left member of <* ya grew into an ovoid. The
base of several letters such as pa, da, and tha (©) developed a pronounced notch,
which was to become characteristic of the later forms of the script.
In the far south, three scripts, namely, Grantha, Tamil, and Vatteruttu, made their
appearance in the early medieval period; of these, the first was used for writing San-
skrit, the other two for Tamil. These scripts constitute a distinct subgroup, and while
there is no question that all three are Brahml-derived, the precise details of their evo-
lution are still somewhat unclear and controversial, due mainly to the paucity of in-
scriptional materials in the centuries immediately preceding their appearance. Although
Biihler (BIP 75) thought that the Tamil script was derived from a northern alphabet, it
is nowadays recognized that it, along with Grantha and Vatteruttu, developed from
earlier southern scripts (cf. HEP 124 and IESIS 222). Grantha and Tamil in their earlier
stages resembled the forerunner of the Telugu-Kannada script, and are presumably
derived from it. Vatteruttu, which Biihler (BIP 51-6) thought to be a cursive variety of
Tamil, is nowadays generally considered to be a separately developed script; some
scholars'13 have recently pointed out apparent affinities with the later stages of the Tamil
Brahml script ([Link]) and suggest that Vatteruttu is actually the descendant of it.

[Link] Developments in the later medieval period: Origins of the

modern Indie scripts (ca. A.D. 1000 onward)
In northern India, the late forms of Siddhamatrka gradually shaded into early ver-
sions of the modern northern scripts, the most important of which is the Nagarl or
Devanagarl script.114 Biihler (BIP 51) cited examples of inscriptions in Nagarl, or
showing Nagarl characteristics, as early as the eighth or even seventh centuries, but
V. S. Sukthankar115 argued that there are actually no authentic specimens of this script

113. Mahalingam, ESIP 299-307; K. G. Krishnan, El 40, 1973-74, 91-2.

114. On these terms, see 1C 11.678.
115. "Palaeographic Notes," in Commemorative Essays Presented to Sir Ramkrishna Gopal Bhan-
darkar (Poona: Bhandarkar Oriental Research Institute, 1917; reprint ed. as R. G. Bhandarkar Com-
memoration Volume [Delhi: Bharatiya Publishing House, 1977]), 309-22.
 Writing and Scripts in India 41
earlier than two Silahara inscriptions from Kanheri dated A.D. 851 and 877/78 (CII
6, 3-8). By the beginning of the eleventh century, Devanagarl has supplanted
Siddhamatrka for epigraphic and literary purposes and has essentially attained its
modern form, characterized by the horizontal top line across the entire character, rec-
tangular corners and generally square frame, and fully extended vowel diacritics. The
individual characters have by now undergone extensive, sometimes radical remodel-
ing, as in the case of ja, tha, and ha. Thus, although numerous subvarieties of (Deva-)
nagarl are known from both epigraphic and literary sources from different times and
regions, by about A.D. 1000 it can be said to have essentially achieved its standard form.
As such, it became the closest thing to a national script of India for writing Sanskrit;
although the regional scripts continued to be widely used for this purpose in medieval,
and to a considerable extent still in modern, times, Nagari was not rarely used for San-
skrit inscriptions outside of northern India, including the Deccan and the far south (the
so-called Nandi-nagarl variety; for examples see IESIS 184-93). "6
In the northeast, the local derivative of Siddhamatrka was the script known
as Proto-Bengali or GaudI, which was current from the tenth to the fourteenth
centuries.117 This, in turn, gave rise to the modern eastern scripts, namely, Bengali-
Assamese, Oriya,118 and Maithili, which became clearly differentiated around the
fourteenth and fifteenth centuries.
In the far northwest, Proto-Sarada developed into the Sarada script of Kashmir
and neighboring hill regions, which was important both as an epigraphic and literary
script in medieval times and as the forerunner of the Punjabi (GurmukhT) and other
scripts of the western Himalayas and adjoining areas.
In upper southern India, the Telugu-Kannada script began to approach the mod-
ern forms of Kannada and Telugu in the fourteenth and fifteenth centuries. These
two scripts are only minimally differentiated, the main contrast being the shape of
the head mark, which takes a form like a check mark in Telugu and a horizontal line
with a hook at the upper right in Kannada.
In the far south, the Tamil and Grantha scripts continued to develop and approxi-
mated their modern forms by about the fourteenth and fifteenth centuries, as seen,
for example, in the inscriptions of the Vijayanagara kings. At about the same period,
the Malayalam script developed out of Grantha.

[Link] Extra-Indian derivatives ofBrahmi

In ancient times, Brahml and its derivatives were in use over a vast area of the
Indian-influenced regions of Asia, and thus came to be the parent of several impor-
tant scripts of these regions. The Tibetan script is evidently derived from a form of
Siddhamatrka of about the seventh century, adapted to the phonetic structure of the
Tibetan language. According to some authorities,119 the Indian prototype came into
Tibet through a central Asian intermediary.
116. Various other local scripts of northern and western India, such as Modi and Kayethl, devel-
oped out of NagarT; of these, only the Gujarati script has attained any importance as an epigraphic script
(see 3.4.3), the others being used mainly for ephemeral documents in the local languages.
117. See, for example, the Nalanda inscription of Vipulasrlmitra, Appendix, no. 12.
118. The Oriya script actually appears to have had a composite origin, with secondary influences
from the southern scripts and Nagarl (1C 11.680).
119. E.g., A. H. Francke, El 11, 1911-12, 269.
42 Indian Epigraphy

The Sinhalese script is thought to be essentially a local development from an early

form of Brahml, which is well attested in the early cave inscriptions of Sri Lanka
([Link].2 and [Link]). But the evolution of the Sinhalese script over the centuries
has been strongly influenced by other Indian models, especially Grantha (see DIP
215-25, esp. 216).
In Southeast Asia, Indian scripts came into wide use from the early centuries of
the Christian era on ([Link]-[Link]). Although northern scripts were occasionally
found there, the predominant form was from the beginning a script which, according
to most authorities, is derived from the southern script used by the early Pallava kings,
often referred to as Pallava Grantha. Others, however (e.g., Sircar, HEP 132-3), have
noted paleographic similarities in the early Southeast Asian inscriptions to scripts of
the western coastal region of India. In view of the close relationship between the early
southern and western scripts, it is difficult to reach a definite conclusion as to the
exact geographical origin of the early Southeast Asian form of Brahml (cf. DIP
232-3), but the predominant influence does seem to have been from the south.
Whatever may have been the precise origin of the Southeast Asian form of Brahml,
it rapidly developed local characteristics of its own, which may be broadly charac-
terized as a tendency toward curved and wavy lines (somewhat like those of the
Telugu-Kannada script), calligraphic elaboration in the form of extended curves and
curls, and a precise and artistic ductus.120 Over the centuries this script gradually de-
veloped into local varieties, used both for Sanskrit and for the indigenous languages,
and these in turn eventually evolved, through various processes of differentiation and
phonetic adaptation, into the later and modern scripts used for Thai, Lao, Cambo-
dian, and Burmese.121

2.3 The KharosthI Script

S. Konow, Kharoshthi Inscriptions . . . (CII 2.1); A. M. Boyer, E. J. Rapson, and

E. Senart, KharosthI Inscriptions Discovered by Sir Aurel Stein ... [KI]; C. C. Das
Gupta, The Development of the KharosthI Script [DKS]; S. J. Mangalam, KharosthI
Script; J. Brough, The Gandharl Dharmapada [GD]; G. Biihler, "The Origin of the
Kharosthi Alphabet" [OKA]; SAI, 84-105.

2.3.1 Geographical range

The presumptive homeland and principal area of the use of KharosthI script (see table
2.5) was the territory along and around the Indus, Swat, and Kabul River Valleys of

120. See, for example, the V3t Ph'u inscription, Appendix, no. 9.
121. On the origin and development of Cambodian and other Southeast Asian scripts, see the re-
cent studies of J. G. de Casparis, "Palaeography as an Auxiliary Discipline in Research on Early South
East Asia," in R. B. Smith and W. Watson, eds., Early South East Asia: Essays in Archaeology, His-
tory and Historical Geography (New York: Oxford University Press, 1979), 380-94, and H. B. Sarkar,
"The Introduction of the Indian Art of Writing to Southeast Asia," chapter 7 of his Cultural Relations
Between India and Southeast Asian Countries (New Delhi: Indian Council for Cultural Relations/MoUlal
Banarsidass, 1985), 168-79.
 TABLE 2.5. Kharosthi Script

VOWELS
Full or initial forms

I 1 I am

Medial (postconsonantal) forms

ka 1 ku\ | kam

CONSONANTS

Unvoiced Voiced Voiced Voiced Semi-

unaspirated aspirated unaspirated aspirated Nasal vowel Sibilant
Guttural ka kha ga gha ha
Palatal ca cha ja jha fia ya sa
Retro flex ta tha da dha na ra sa
Dental ta tha da dha na la sa
Labial pa pha ba bha ma va -

Conjunct consonants (representative examples)

ksa tra tva pra

rva sta spa sya

Note: These are normalized forms based on inscriptions of ca. first century A.D.
44 Indian Epigraphy

the modern North-West Frontier Province of Pakistan; that is, the ancient Gandhara
and adjoining regions, with whose MIA language, nowadays usually called "Gand-
harl" (see [Link].1), the script is almost invariably associated (see map 1). In Konow's
words, "The KharoshthI area proper may be defined as extending from about 69° to
73° 30' E. and from the Hindu Kush to about 33° N., and there can be little doubt that
its place of origin was Gandhara, perhaps more especially Taxila" (CII 2.1, xiv).122
But although Konow's general comments about the KharosthI area "proper" still
hold true, numerous discoveries since his time have provided many examples of in-
scriptions and other documents in KharosthI script ranging over a much broader
area.123 To the far west and northwest, we can now add to the handful of inscriptions
known to Konow several more specimens (for details see [Link]) from sites along
the Kabul River in Afghanistan as far west as Wardak or Khawat,124 some thirty
miles west of Kabul. Recent archaeological excavations have also yielded numer-
ous KharosthI inscriptions from north of the Hindu Kush, in ancient Bactria, both in
sites in northern Afghanistan such as Qunduz and in several places in the former Soviet
republics of Uzbekistan and Tajikistan (see [Link] and [Link]).
In the far north, graffiti in KharosthI have recently been found in large numbers
at various sites around Chilas and other places along the upper course of the Indus
River (see [Link]). These new discoveries provide a connecting link between the cen-
tral KharosthI area and the previously isolated KharosthT inscription from Khalatse125
along the Indus in Ladakh.
To the south and southwest, sporadic examples of KharosthI inscriptions were
known to Konow (there are no major subsequent discoveries in this area) along the
lower Indus as far as Mohenjo-Daro, and in Baluchistan;126 and to the southeast, in
Kangra127 and in the region around Delhi and Mathura.128 Occasional stray finds of
KharosthI letters farther to the east and south are not indicative of significant use of
the script in those regions. Thus a plaque with a KharosthI inscription found at
Kumrahar (Patna; CII 2.1, 177-8) is most likely an import from the northwest (CII
2.1, xiv), and the apparent occurrence of a few KharosthI letters on the bases of pillars
from Bharhut (Satna Dist., MP) as reported by Cunningham129 is of uncertain signi-
ficance. Still farther afield, the well-known occurrence of the single word lipikarena
in KharosthI in three Asokan inscriptions from Karnataka (see [Link] n. 67) is evi-
dently only a scribe's flourish.130

122. See also Konow's map of the findspots of KharosthI inscriptions (CII 2.1, xiv), and the maps
in O. Fussman, "Gandharl ecrite, Gandharl parlee" (for reference see the bibliography), 434—5.
123. For a survey of these discoveries, see Fussman, "GandharT ecrite, Gandharl parlee," 444—51.
124. I.e., the Wardak vase inscription, CII 2.1, 165-70.
125. CII 2.1, 79-81.
126. Tor Dherai, CII 2.1, 173-6.
127. Both of the Kangra inscriptions, from Pathyar (CII 2.1, 178) and Kanhiara (ibid., 178-9),
are biscripts, with the text given (with some variation) in both KharosthI and Brahml.
128. Karnal, CII 2.1, 179; Rawal, ibid., 161-2; Mathura, ibid., 30-49.
129. The Stupa of Bharhut, 8 and pi. VIII.
130. Cf. Konow, CII 2.1, xiv. Reports by B. N. Mukhcrjcc (e.g., Quarterly Review of Historical
Studies 29.2, 1989-90, 6—14) of inscriptions in KharosthI or a mixed KharosthT-Brahml script from
West Bengal and even farther to the east are of questionable validity; see SAI 91—2 for evaluation and
further references.
 Writing and Scripts in India 45

MAPI
Extent of Kharosthl script

In the form of legends on the coins of the Indo-Greek and Scythian kings, Kharosthl
script was also in use over a wide area of northern India and adjoining regions, includ-
ing the western coastal region. Noteworthy in the latter region are the coins of the
early Western Ksatrapas Nahapana and Castana, whose coins had legends in three
scripts—Greek, BrahmT, and Kharosthl.
Finally, in and around the third century A.D. Kharosthl was a major script for
epigraphic and literary purposes in central Asia, as attested especially by the abun-
dant finds of official documents of the Shan-shan (Kroraina) kingdom on the south-
ern rim of the Tarim Basin written on leather and wood (see [Link]) as well as of
other inscriptions, and by the birch-bark manuscript of the Dharmapada in Kharosthl
script from the region of Khotan (Brough, GD). Some still unpublished Kharosthl
documents, apparently of later date, have also been found at sites on the north of the
Tarim Basin.131 A few Kharosthl inscriptions have even been found in China proper,
for instance, at Lo-yang (see [Link]).
Thus, while the KharosthT script developed and remained concentrated in the
Gandhara region, it also spread over a wide area within and far beyond the immedi-
ately adjoining portions of modern Pakistan and Afghanistan. The spread of KharosthT
was no doubt promoted by its use by the Indo-Scythian rulers, and especially by the
131. See n. 138.
46 Indian Epigraphy
Kusana kings, who seem to have been responsible for its introduction into the north
Indian heartland on the one hand and Bactria on the other. The expansion of Kharosthl
was also intimately connected with the spread of Buddhism under the patronage of
the aforementioned rulers, as is indicated by the content of the inscriptions from
outlying regions such as Bactria and China; these inscriptions, mostly dedicatory and
memorial in content, are often hardly distinguishable in form from similar inscrip-
tions from India, and presumably reflect the presence of GandharT-speaking Indian
monks in the Buddhist monasteries of these places.

2.5.2 Chronological range

The Kharosthl script first appears in inscriptions in the Shahbazgarhi and Mansehra
rock edicts of Asoka, datable to the middle of the third century B.C. In these early records
the script appears as a more or less fully developed system, which may be taken to
indicate a significant prehistory. But it is difficult to estimate how long such a pre-
sumed developmental period might have taken, though it is likely on historical and
paleographic grounds (see 2.3.6) that Kharosthl originated sometime during the
Achaemenian era.132 Biihler (OKA 50 = OIBA 98) was inclined on paleographic
grounds to date the origin of Kharosthl to the earlier phase of this period, that is, around
the beginning of the fifth century B.C., but in view of uncertainties about the details of
his proposed Aramaic derivation (see 2.3.6), Biihler's early dating can be accepted
only provisionally at best. On the other hand, late dates such as ca. 330 B.C. proposed
by Halevy133 or after 325 B.C. suggested by Falk134 are problematic, especially if, as
seems most likely, the lipi referred to by Panini is Kharosthl (2.1.2). In short, there is
no clear evidence to allow us to specify the date of the origin of KharosthT with any
more precision than sometime in the fourth, or possibly the fifth, century B.C.
KharosthT continued to be the principal script of the northwest in the following
centuries, flourishing during the so-called foreign period of the reigns of the Indo-
Greek, Indo-Scythian, Indo-Parthian, and Kusana kings from the first century B.C. to
the second century A.D. But the script appears to have fallen out of general use in
South Asia itself by sometime in the third century A.D. or shortly thereafter. The lat-
est datable inscription in Kharosthl is probably Skarah DherT (CII 2.1, 124—7), ap-
parently dated in the year 399 of an unspecified era, usually presumed to be the "Old
Saka" era, which would put the date of the inscription around the middle of the third
century A.o.135 Two other Kharosthl inscriptions from Jamalgarhl and Hashtnagar
(CII 2.1, 110-3 and 117-9) are dated in the years 359 and 384, presumably of the
same era, confirming that the script was still in use around the period in question.

132. E. J. Rapson ("Counter-marks on early Persian and Indian Coins," JRAS 1895, 865-77) thought
that some silver sigloi of the Achaemenian kings of Iran, dating from the fifth and fourth centuries B.C.,
bore countermarks consisting of aksaras in Kharosthl (and Brahml; cf. 2.1.2 n. 20) scripts. However,
Rapson's identification' of these countermarks as syllables of the Indian alphabets was called into ques-
tion by G. F. Hill ("Notes on the Imperial Persian Coinage," Journal of Hellenic Studies 39, 1919, i 16-29)
and is no longer generally accepted, so that they cannot be taken as specimens of pre-Mauryan KharosthT.
133. JA, ser. 8, vol. 6, 1885, 267;Revue Semitiquel, 1895, 378-80. (For full references see 2.3.6.)
134. SAI 104.
135. See section [Link]. Konow, following W. E. van Wijk, places the epoch of the era in ques-
tion in 84 B.C., so that the date of the inscription would correspond to A.D. 315.
 Writing and Scripts in India 47
A series of Kharosthl image inscriptions from Jaulian (CII 2.1, 92-7) were at-
tributed by John Marshall136 on archaeological grounds to a date as late as the sec-
ond half of the fifth century A.D., and taken as evidence that "Kharoshth! was still
the ordinary script of the townspeople of Taxila" at this period. But Konow sus-
pected that these inscriptions may have been copies of older ones, and thus do not
constitute evidence that Kharosthl was still in active use at this late date. It is thus
safest to conclude that KharosthT probably fell out of general use in South Asia in
or around the third century A.D., but it may have been used occasionally for another
century or more.
This conclusion is supported by the material from central Asia, where the
Kharosthl documents from Niya and other sites on the southern rim of the Tarim
Basin ([Link]) are also approximately datable to the third and early fourth centu-
ries, mainly on the grounds of their co-occurrence with a Chinese document dated
A.D. 269.137 It is possible, however, that Kharosthl continued to be used in the cities
of the northern rim of the Tarim Basin well after this time; documents in this script
were reported to have been found there together with others in the Kuchean language
datable to the seventh century.138 But since these documents are still unpublished
their significance remains uncertain, and there is still no firm evidence that Kharosthl
survived much longer than the third century.
The decline of Kharosthl around the third century A.D. can presumably be attrib-
uted to historical developments of this time. With the decline and fall of the Kusana
empire, the center of political power in South Asia began to shift from the northwest
toward northern and northeastern India. This development presumably reduced the
importance of Kharosthl as the regional script of the former area, with the eventual
result that it was supplanted entirely by the pan-Indian Brahml.

2.5.3 Uses of KharosthT

In South Asia, the Kharosthl script is preserved principally in inscriptions, mostly
on stone but also on metal and other materials, numbering in the hundreds. The vast
majority of these are Buddhist records, mostly concerning pious donations and foun-
dations. Kharosthl was also extensively used in the coin legends of the Indo-Greeks,
Indo-Scy thians, and other foreign rulers of India, as well as of some indigenous polities
such as the so-called tribal republics of the Audumbaras, Kunindas, and Rajanyas.
KharosthT was rarely used by itself on coins; most specimens are biscriptual (or even
triscriptual), in combination with legends in Greek and/or Brahml (2.3.1).
KharosthT is also well attested, in relics from central Asia, in nonmonumental
functions. In the Niya documents (2.3.1) it functions as an administrative script, while
the manuscript of the Dharmapada attests to its literary usage.139 Moreover, the fact
136. Excavations at Taxila: The Stupas and Monasteries at Jaulian (MASI 7, 1921), 10.
137. See John Brough, "Comments on Third-Century Shan-shan and the History of Buddhism,"
BSOAS 28, 1965, 582-612 (esp. 604).
138. 1C 11.672; H. W. Bailey in D. Winton Thomas, ed., Essays and Studies Presented to Stanley
Arthur Cook (London: Taylor's Foreign Press, 1950), 121.
139. This latter function is also reflected in occasional quotations from canonical texts in inscrip-
tions (see [Link]), and also by the use of Kharosthl for writing Sanskrit in a few of the central Asian
documents (e.g., KI no. 511).
48 Indian Epigraphy

that some Chinese Buddhist texts reflect originals in the Gandharl language (see
[Link].1 n. 23), and hence, presumably, in KharosthI script, confirms that KharosthI
was in its day a major vehicle of culture in the Buddhist circles of northwestern India.
Thus the judgments of earlier scholars such as Biihler, which were based mainly on
the somewhat misleading evidence of the bilingual coin legends and the then fairly
sparse repertoire of inscriptions and other KharosthI records, to the effect that
"KharosthI held always . . . only a secondary position by the side of the Brahma al-
phabet even in Northwestern India"140 are no longer valid in light of subsequent dis-
coveries which have vastly enhanced our understanding of the important historical
and cultural role of this script.

2.3.4 Paleographic features of KharosthI

Unlike Brahml and the many scripts derived from it, KharosthI, alone among Indie
scripts,141 is written from right to left. Unlike the highly symmetrical, "monumen-
tal" early Brahml script, KharosthI has a cursive ductus which would seem to reflect
an origin in a "clerk's" script, written with pen and ink.142 The KharosthI script also
contrasts with Brahml in that it is top-oriented; that is, the distinctive features of each
character tend to be at the top instead of at the bottom as in Brahml.
As noted earlier (2.1.3), KharosthT employs essentially the same diacritically
modified consonantal syllabic system of Brahml and its derivatives, despite some
important differences in detail. The script comprises a basic set of consonantal sym-
bols with inherent a, "full" vowel signs used for word initial and non-postconsonantal
vowels, and diacritic signs attached to the consonants to indicate postconsonantal
vowels. The main difference in graphic principle between Brahml and KharosthT is
that the latter for the most part143 does not differentiate long and short vowels. That
is to say, the vowel signs, both full and diacritic, designate vowel quality but not
quantity. For example, the same sign represents either i or r, and the inherent post-
consonantal vowel can be read as either a or a (to be understood by context); thus
7> = kaoT ka; b = ki or ki, and so on. The full or initial vowel signs further differ from
those of Brahml in that they are all constructed from the basic vowel sign for a to
which are affixed the postconsonantal vowel diacritics to form initial i, u, and so on;
thus 9 = initial a/a, while ?• = initial Hi.
As in BrahmT, the inherent vowel system necessitates the formation of conso-
nantal conjuncts or ligatures, though the methods and details of their formation in
KharosthI are somewhat different and sometimes problematic. The basic principle
is the same as that of Brahml, with the two or more consonants of a phonetic cluster
joined graphically from top to bottom, with or without abbreviation of one of them.

140. OKA 44 = OIBA 92; cf. also OIBA 51.

141. Except for occasional examples of early Brahml inscriptions written from right to left; see
2.2.3,2.4.
142. Cf. Biihler, OKA 49 = OIBA 97: "The KharosthI alphabet is not a Pandit's, but a clerk's,
alphabet."
143. For the development in the later period of a system for differentiating vowel quantity in
KharosthI, see section 2.3.8.
 Writing and Scripts in India 49

Some of the resulting conjuncts, for example, £ ska < T so. + 7i ka, are self-evident,
but many others involve alterations of the constituent parts to the point that they
may become unrecognizable. As in later forms of Brahml and its derivatives (see
[Link]), ra (which is especially common in KharosthT texts, since the Gandharl lan-
guage tends to retain it in consonantal groups; see [Link] and [Link].1) develops
special forms for both pre- and postconsonantal position: for example, *7 ra + 7 va >
^ rva; 1 = vra. Similarly (and again as in BrahmT),144 postconsonantal ya has a spe-
cial form, as in _/ sya.
Other consonantal combinations, particularly those involving stops and sibilants,
are also prone to take special and often obscure forms in KharosthI, for instance, ?
sa + ~) ta > ^ sta. Some special consonantal signs are problematic in that both their
graphic constituents and their phonetic value are uncertain, notably the common
character y, which regularly corresponds to Sanskrit ksaus but which can hardly be
derived from a graphic combination ofhka and T sa; Dani (DIP 259) considers it a
reduction of y cha, while Rapson (KI III.320 n. 3) suggests a connection with T sa.
Thus it is not at all certain that KharosthT characters of this type are really conjuncts in
the standard sense of the term; they may have arisen from modifications of single
consonants, or may have been created ad hoc to represent phonemes peculiar to the
Gandharl language.
Such developments are characteristic of KharosthI, which in general was more
flexible and adaptable to linguistic change than BrahmT, for instance, in creating
diacritically or otherwise modified characters to indicate new sounds such as fric-
ativized intervocalic consonants which developed in the Gandharl language (see
[Link].1). The virtual absence of such modified characters in Brahml and its pre-
modern Indian derivatives does not indicate that the other MIA languages did not
develop such new sounds, but rather that in the Sanskrit-dominated cultural milieu
in which these scripts functioned the phonetic repertoire of Sanskrit and its imme-
diate MIA derivatives was taken as definitive and unalterable. In the KharosthT/
Gandharl sphere, however, the influence of Sanskrit, while by no means totally
absent,146 was far less pervasive, so that the script was relatively free to grow and
change in accordance with phonetic changes in Gandharl.
On the whole, KharosthI orthography tends to be informal, approximative, and
often inconsistent. Geminate consonants, for instance, are normally indicated by the
single consonant, and unaspirate-aspirate clusters by the simple aspirate (e.g., budha
for buddha).141

144. These similarities between the two scripts need not result from a direct influence of one on
the other, but rather are presumably independent parallel developments conditioned by the statistical
frequency in Gandharl and Sanskrit of conjuncts with semivowels, which naturally led to the creation
of cursive ligatures.
145. On the phonetic value of this character, which is variously transliterated as cha (incorrectly),
ch'a, or ksa, see section [Link].
146. A clear example of Sanskritization in KharosthT/GandharT is the Sui Vihar copper plate in-
scription (CII 2.1, 138-41), whose orthography is strongly Sanskritized, e.g., using -sya (rather than
-sa) for the genitive singular ending.
147. These features are, however, not untypical of epigraphic Prakrits generally, especially in the
earlier period; see 3.1.1.
50 Indian Epigraphy
2.3.5 The name of the script
The script which is nowadays generally known as Kharosthl (or Kharosti, Kharostri,
etc.) was until the end of the nineteenth century referred to by various names such as
"Bactrian," "Indo-Bactrian," "Kabulian," "Bactro-Pali," "Ariano-Pali," and so on.148
It was T[errien] de Lacouperie who proposed in 1886-87l49 that the name Kharosthl,
the second in the list of sixty-four scripts in the Lalitavistara (2.1.1), should corre-
spond to the northwestern script, on the grounds that the corresponding script in the
Fa yuan chu lin is described as written from right to left. De Lacouperie's sugges-
tion was adopted by Biihler and others, and has won nearly universal acceptance.150
The correct form of this name and its meaning, however, are still uncertain. Vari-
ous spellings appear in different manuscripts of the various Buddhist and Jaina script
lists: kharosti, khalosti, and karotti in the Lalitavistara; kharosti and kharastrl in the
Mahavastu; and kharotthi and kharotthiya in the Ardha-magadhI dialect of the Jaina
texts. In the Fa yuan chu lin the name of the script is given as K'(i)a-lu-she-t'o (in
O. Franke's transliteration), which according to Franke151 would correspond to San-
skrit kharostha. These testimonia are not sufficient to definitively reconstruct the original
form of the name, and the now conventional spelling kharosthi, as adopted by Biihler
in his authoritative works on paleography (BIP) and on the origin of the script (OKA),
is by no means certain; some modern scholars prefer the forms kharosti orkharostrT.152
Since the word was very likely not originally Sanskrit or even Indie, variations in spelling
may have arisen partly from different Sanskritizations of the original name of the script.
As for the etymology of the name, in 1902 Sylvain Levi153 first proposed that the
name was actually a geographical term, reconstructed as *kharostra from the Chi-
nese K'ia-lu-shu-ta-le, supposedly a toponym for Kashgar. R. Pischel,154 however,
doubted that a remote central Asian region could have given its name to the Indian
script, and preferred the traditional attribution in the Fa yuan chu lin of the script (as
first pointed out by de Lacouperie) to a sage K'(i)a-lu-she-t'o, which name, according
to the accompanying gloss, means 'ass-lip', suggesting an Indie original kharostha.
Levi responded in 1904155 with an expanded and modified statement of his views in
which he reidentified the "Kharostra" country not specifically with Kashgar but rather

148. Cf. 6.2.2 n. 59. Alexander Cunningham (Coins of Ancient India from the Earliest Times . . .
[London: B. Quaritch, 1891], 31) proposed the more appropriate term "Gandharian," but this sugges-
tion did not win much acceptance (unlike H. W. Bailey's more successful suggestion many decades
later of "Gandharl" for the language with which the script was linked; see [Link].1).
149. Babylonian and Oriental Record 1, 59-60 (cf. 2.2.2 n. 39).
150. 1C (II.670-2) still holds to the term "arameo-indien" for the earlier forms of the script in India
on the grounds that the term "Kharosthl" properly applies to the central Asian form of the script; but
this is probably not correct.
151. IA 34, 1905, 21 (see n. 154 for full reference).
152. See, e.g., B. N. Mukherjee, "A Note on the Name KharoshthI," JAS 23, 1981, 13-15.
153. "L'ecriture kharostrl et son berceau," BEFEO 2, 1902, 246-53 (esp. 248) = "The Kharoshtri
Writing and Its Cradle," IA 33, 1904, 79-84 (esp. 81).
154. In O. Franke and R. Pischel, "Kaschgar und die Kharosthl," Sitzungsberichte der koniglich
Preussichen Akademie der Wissenschaften zu Berlin (Phil.-hist. Klasse) 1903, 184-96 and 735-45 =
"Kashgar and the Kharoshthi," IA 34, 1905, 21-7 and 41-6.
155. "Notes chinoises sur 1'Inde IV: Le pays de Kharostra et 1'ecriture kharostrl," BEFEO 4, 1904,
543-79 = "The Kharostra Country and the Kharostri Writing," IA 35, 1906, 1-30.
 Writing and Scripts in India 51

with the entire intermediate region between China and India, and endorsed the spell-
ing kharostn on the grounds that it could be connected with the "donkeys and camels"
(Sanskrit kharostra) which are characteristic of the region in question.
None of these explanations being particularly convincing, several other sugges-
tions of a non-Indie origin for the name Kharost(h)! have been offered. Albert
Ludwig156 proposed a derivation from an unattested Aramaic *harutta predicted on
the basis of the root HR§ (Hebrew)/ HRT (Aramaic) 'engrave', and Buhler (OIBA
114 n. 1) seemed willing to accept this as an alternative to the traditional etymology.
Others have looked for Iranian or semi-Iranian etymologies, most persuasive among
which is Jean Przyluski's suggestion157 connecting the name with that of Kharaposta,
given in the MahdmayurT as one of the yaksas of northwestern India and rendered
into Chinese as 'hide of donkey' (<khara 'donkey' + Iranian post 'hide'). Kharosthal
Kharosthl, and so on, would evidently result from an incorrect Sanskritization of the
Prakrit form of the word in question. According to Przyluski, the application of the
name to the script can be attributed to the practice (actually attested in some of
the central Asian Kharosthl documents) of writing on the hides of donkeys and other
animals.158 He also suggested a connection of the name Kharaposta with that of king
Kharaosta, known from coins and from the Mathura lion capital inscriptions. This
suggestion was taken up by H. Humbach,159 who proposed that the script was named
directly after King Kharaosta, with whom its importation into India proper may have
been associated. Humbach's explanation is endorsed by Falk (SAI90). More recently,
H. W. Bailey160 has proposed several possible Iranian etymologies for Kharosthi, of
which the most compelling is *xsadra-pistra 'royal writing'.
All in all, the question of the origin and meaning of the name Kharosthl remains
problematic, but it appears that the connections noted by Przyluski et al. with the
well-attested proper names Kharaposta and Kharaosta lie at the heart of the matter.
In all probability Kharost(h)T is a Sanskritization of an original Iranian name whose
etymology is uncertain. The connection wUhkhara 'donkey' is probably nothing more
than folk etymology, and the original term was most likely connected with Old Ira-
nian *xsa9ra 'sovereignty', as noted by Bailey and others,161 though the second part
of the word remains unclear, as does the process by which the name in question be-
came associated with the script.

2.3.6 The origin of Kharosthl

J. Halevy, "Essai sur I'origine des ecritures indiennes," JA, ser. 8, vol. 6,1885,243-301;
"Un dernier mot sur le kharosthi," Revue Semitique 3, 1895, 372-89; G. Buhler, OKA;
C. C. Das Gupta, DKS, ch. 17, "The Origin of the Script," 280-90; SAI 92-9.

156. "Uber den Namen der alien linkslaufigen Schrift der Inder," in Gurupujakaumudi: Festgabe
zum fiinfzigjahrigen Doctorjubilaum Albrecht Weber (Leipzig: Otto Harrassowitz, 1896), 68-71.
157. "Le nom de 1'ecriture kharosthi," JRAS 1930, 43-5 = "The Name of the Kharosthl Script,"
IA60, 1931, 150-1.
158. Compare Skt. pustaka 'book' from the same Iranian post.
159. Review of DIP, Orientalistische Litemturzeitung 63, 1968, 489-91.
160. "Kharostn," in Indo-Scythian Studies Being Khotanese Texts. Vol.7 (Cambridge: Cambridge
University Press, 1985), 46-9; see also JRAS 1978, 4.
161. For example, F. W. Thomas, El 9, 1907-8, 139.
52 Indian Epigraphy

A connection between KharosthT and the Semitic scripts (see table 2.2), particularly
Aramaic, has been evident to scholars from an early period. The Semitic connection
was first discussed in general terms by Edward Thomas162 as early as 1858, and the
particular connection with Aramaic was noted by Isaac Taylor.I63 The theory was
further developed by J. Halevy ("Essai sur 1'origine," 247-67), and was finally worked
out in what has become the definitive statement by Biihler in OKA in 1895.
It was presumably the right-to-left direction of KharosthT that first suggested a
Semitic derivation.164 This impression was reinforced by the generally Semitic ap-
pearance of the script, and especially by its obvious resemblance to Aramaic. This
presumptive relationship has been successfully worked out in detail by the previ-
ously mentioned scholars to the point of virtual certainty. First of all, the historical
circumstances of the development of Kharosthl from Aramaic are easily explained
by reference to the Achaemenian conquests in the western borderlands of India; as
noted by Biihler, "[T]he territory of the Kharosthl corresponds very closely with the
extent of the portion of India, presumably held by the Persians" (OKA 47 = OIBA
95). Since Aramaic served as the lingua franca of the Persian empire, it is easy to
imagine how the Aramaic alphabet could have been adapted to the local Indian lan-
guage, namely, Gandhari, as "the result of the intercourse between the offices of the
Satraps and of the native authorities" (OKA 49 = OIBA 97).165 Moreover, the recent
discoveries of Asokan edicts in Aramaic, including passages with Prakrit words
written in Aramaic script and provided with glosses in the Aramaic language (see
[Link]), have confirmed the close connection of Aramaic with Indie languages and
scripts in the Achaemenian period and following centuries.
Second, specific relationships between the individual characters of the two scripts
have been clearly established. In refining the only partially satisfactory formulations
of Taylor166 and Halevy,167 Biihler presented in OKA what has become the authori-
tative formulation of the derivation of Kharosthl from Aramaic. The derivations of
most of the characters are self-evident and beyond reasonable doubt; for instance,
Kharosthl 7 ba < Aramaic $ bet, or'jna<^ nun. In several cases certain alter-
ations of the Aramaic prototypes have to be posited, such as inversion,168 as in
Kharosthl ~l la< I lamed; cursivization and stroke reduction, as in f ga < \ gimel;
and other alterations such as the addition of an extra stroke to disambiguate charac-
ters of the new script, as in 7> ka < 'I kaph, with an extra stroke at the left to disam-
biguate the derived character from Kharosthl "7 ta. Since most of these alterations
follow regular patterns and are subject to logical explanation, they do not constitute
a weakness in the proposed derivation. As in the case of the proposed Semitic deri-
vation of Brahml (see [Link].4), the Kharosthl characters for aspirate consonants
are constructed either as secondary derivatives of the corresponding nonaspirates

162. In his editorial notes to James Prinsep's Essays on Indian Antiquities, 11.143—70.
163. In The Alphabet, 11.256-62.
164. But see the comments on direction of writing in section [Link].4.
165. In this connection Biihler also referred to the use of the Iranian word dipt 'writing' and vari-
ous derivatives thereof in the KharosthT versions of the Asokan edicts (OKA 46-7 = OIBA 95).
166. The Alphabet; in particular, see his chart of "The Iranian Alphabets," 11.236.
167. "Essai sur 1'origine"; see especially plate I, facing page 252.
168. This is in keeping with the top orientation of the derivative script; see section 2.3.4.
 Writing and Scripts in India 53

(e.g., f ga > % gha),m or by adaptation of a phonetically similar Aramaic character

(e.g., 9 kha < f1 qoph). The derivation of the KharosthI retroflexes is more problem-
atic; Biihler takes them all as secondary derivatives within KharosthT (e.g., ^ tha <
7 ta and f da < $ da), but the paleographic connections here are far from certain.
Third, as noted earlier (2.1.3), the diacritically modified syllabic system of
KharosthI can be best explained in historical terms as a refinement of the Semitic
consonant-syllabic system.170 The nondifferentiation of vowel length in KharosthI
can be seen as an intermediate stage of development between the consonant-syllabic
Semitic prototype and the fully vocalized Indie system as found in Brahmi.
Thus, the three main criteria—historical, paleographic, and systemic—for estab-
lishing genetic connections between scripts are satisfied in the derivation of KharosthI
from Aramaic. The theory has accordingly been accepted by nearly all authorities
on the subject, including many of those (e.g., Ojha, BPLM 31-7) who do not accept
a Semitic origin for Brahmi. R. B. Pandey (PIP 53-8) is virtually alone in completely
rejecting a Semitic origin for KharosthI in favor of the traditional view (as attested
in Chinese sources; see 2.3.5) that "it was invented by an Indian genius whose nick-
name was Kharostha, as the letters resembled ass-lips" (58). But Pandey's arguments
against the Semitic origin are not persuasive and do not present any significant chal-
lenge to it. More recently, C. D. Chatterjee171 and following him V. S. Pathak172 have
expressed doubts about the Aramaic origin of KharosthI, but their arguments are
equally unconvincing.
This is not to say, however, that the Aramaic derivation as formulated by Biihler
et al. is free of problems in all details. Biihler was severely criticized on method-
ological grounds by Halevy ("Un dernier mot," 380-1), who charged, not without
cause, that Biihler used Aramaic forms from widely differing periods and places for
his hypothetical prototypes of KharosthI characters.173 And among Biihler's specific
formulations, Halevy (op. cit., 383) rejected the derivation of KharosthI sa from
Aramaic het as an "idee bizarre"; while this, like many of Halevy's criticisms, seems
unduly harsh, it does legitimately point out a weakness in Biihler's formulation. Also
problematic on paleographic grounds is Biihler's derivation of-f tha as a secondary
form from5 ta. Dani's suggestion (DIP 258-9) to derive KharosthI tha directly from
Aramaic 6 tet (which, along with 'ain, is one of the two Aramaic letters for which
Biihler found no correspondent in KharosthI) seems preferable, though the equation
is paleographically uncertain.
Although such problems are not nearly weighty enough to cast serious doubts
on the Aramaic derivation as a whole, they do show that the final word has yet to

169. According to Biihler (OKA 61 = OIBA 109), the extra stroke of gha and similarly derived
aspirates represents a reduced form of the consonant 2 ha.
170. According to Halevy ("Essai sur 1'origine," 248-9), the KharosthT vowel diacritics for i and
u reflect the use of yod and waw as matres lectionis for these vowels in Aramaic, and the adoption of
a as the inherent vowel in KharosthI is due to the absence of any such graph for that vowel in the par-
ent script.
171. "The Aramaic Language and Its Problems in the Early History of Iran and Afghanistan," in
Samaresh Bandyopadhyay, ed., Acarya-vandana: D. R. Bhandarkar Birth Centenary Volume (Calcutta:
University of Calcutta, 1984), 205-26 (esp. 210-4).
172. JESI 13, 1986, 7-8.
173. Cf. Halevy's similar criticism of Biihler's Brahmi derivation, mentioned in section [Link].4.
54 Indian Epigraphy
be said on the subject, which, after all, has not been comprehensively examined
for nearly a hundred years.174 The subject needs to be reevaluated in light of new
data which was not available in Biihler's time, particularly the Mauryan Aramaic
inscriptions.175

2.3.7 Connections between Kharosthi and BrahmT

Certain common features of the Kharosthi and Brahml scripts, including both the
overall similarity in their graphic principles and several specific paleographic fea-
tures, suggest the possibility of some kind of historical connection between them.
For example, Biihler (OKA 62-3 = OIBA 110) and others have noted parallels in the
location of the postconsonantal vowel diacritics in the two scripts, with i, e, and o at
the top of the consonantal character and u at the bottom. Among the formal resem-
blances between consonantal characters which could indicate some historical rela-
tionship, kha, pa, and la are, with due allowance for the normal processes of inver-
sion and reversal, quite similar in the earliest forms of Brahml and KharosthT; but
these similarities could be attributed to a common Aramaic or other Semitic proto-
type. The apparent resemblance of the KharosthT and BrahmT signs for bha (fi> if) is
intriguing, but the internal derivation of this letter in both scripts is problematic, so
that it is difficult to draw any conclusion from it. Halevy ("Essai sur I'origine," 281-5),
as a proponent of a direct derivation of Brahml from KharosthT ([Link].2), made
much of other apparent resemblances such as BrahmT and KharosthT ha (P, A) and da
(</, ,J), but their significance is doubtful as they appear to be separate secondary de-
velopments within the two systems.
Although many of the specific relationships of individual characters remain prob-
lematic, it can be generally assumed that the primary direction of influence would
have been from KharosthT to BrahmT,176 since it is now fairly certain (see 2.1.2) that
the former is the older of the two scripts. Moreover, if the theory of an invention of
Brahml under Asoka, for which persuasive arguments have recently been proposed
by von Hiniiber and Falk, is correct, the common graphic system of both scripts (2.1.3)
must have originally been developed in Kharosthi and later adapted and refined in
Brahml.
Thus, while there are enough points of similarity to suggest at least a secondary
relationship between KharosthT and BrahmT, the overall differences between the two
render a direct linear development connection unlikely (cf. [Link].2). The paleo-
graphic dissimilarity of the great majority of the basic characters of the two scripts
suggests that KharosthT and BrahmT essentially developed separately from their pre-
sumptive Semitic prototype(s).

174. Das Gupta (DKS 284-90) criticizes several of Biihler's derivations on paleographic grounds
(e.g., ma from mem, 287), but since he does not propose alternatives, his criticisms do not represent
any significant advance.
175. Besides the example of sa mentioned earlier, other cases where these new inscriptions may
clarify the derivation of Kharosthi characters include da and ya (as noted by Das Gupta, DKS 285-6).
176. On apparent influences of BrahmT on Kharosthi in later times, see section 2.3.8.
 Writing and Scripts in India 55

2.3.8 The paleographic development of KharosthT

Unlike Brahmi and its derivatives, KharosthT script did not undergo extensive pa-
leographic changes in the course of its historical development. This is no doubt
mainly due to the fact that its history (at least as known to us from extant docu-
ments) covers only some five centuries and takes place within a relatively limited
geographical area, as opposed to Brahmi, which developed over two millennia
throughout India and in many other parts of Asia. KharosthT can be treated essen-
tially as one and the same script throughout its history, and, with the exception of
the central Asian variety, does not have clearly differentiated local variants. The
changes which the individual characters underwent were gradual and for the most
part minor, so that it is difficult to establish reliable criteria for the paleographic
dating of Kharosthl documents in anything other than broad terms such as "early,"
"middle," and "late." Even the central Asian form of the script is not essentially
different from the Indian; many of its peculiar characteristics, such as its generally
more cursive aspect, are attributable as much to the different physical character of
the documents, which were written in pen and ink rather than carved into stone or
metal, as to geographical variation.
Among the characters which do show clear chronological development, sa is often
cited as the most reliable "test letter." In the early forms of KharosthT of the Asokan
and Indo-Greek period, the head portion of sa is usually closed 0»). In the middle
period of the Indo-Scythian and Indo-Parthian era, the head is typically open, but
the vertical line still has a projection toward the top of the letter (/>). Finally, in the
later phase of the script during the Kusana period, the head is fully open (/>). This
character, along with a few other test letters such as ya and ca, can be used to date
Kharosthl documents in a broad fashion. But, as pointed out by Das Gupta (DKS,
102,146), the three types of sa actually overlap considerably throughout the history
of the script, so that to use it alone as a criterion for paleographic dating is an uncer-
tain method at best (see 5.4.1).
Several other characters underwent similar developments involving cursivization
and stroke reduction or reordering. The original form of 7i ka, for example, devel-
oped into a secondary one in which the extra stroke at the right is combined with the
top stroke (7?); this, in turn, led to a late form in which the right-hand portion curves
around to the left to meet the vertical (?). In certain cases paleographic changes re-
sulted in the development of ambiguous letters, the most important case being ya,
whose original angular form (/i) developed a horizontal line across the top (n) which
often makes it virtually identical with sa (n). In some cursive hands, ta and ra, to
and da, va and initial a, and various other pairs of letters are so similar as to cause
serious difficulties in interpretation.
The diacritic vowels were also subject to minor changes, notably postconsonantal
M, which originally was written as a short horizontal line to the left of the foot of the
consonant to which it is attached (e.g., > ku) but later developed into a loop (p>).
Similarly, the original mark foranusvara (e.g., £ gam), clearly recognizable as a min-
iature ma, later changed into a hook shape (<?) which disguises its origin. Certain com-
binations of consonants and vowel diacritics were subject to special cursive devel-
56 Indian Epigraphy

opments leading to ligatures in which the constituent parts are no longer readily ap-
parent; for example,,/ mu (with several other varieties) and £ de.
Several modified characters developed in KharosthI of the middle and later pe-
riods. For instance, { sa frequently appears in alternation with ordinary sa (?), of
which it is presumably a modification, perhaps derived from a reduction of the
conjunct j? sya.111 Several consonants, such as ga, da, and va, were often written
in intervocalic position with a diacritic mark consisting of a horizontal line added
at the lower right, which is usually thought to represent a fricative pronunciation;
for example, f, g 'a. Such diacritically marked consonants are particularly common
in the central Asian documents but also occur in some later Indian inscriptions,
notably in the Wardak vase inscription (CII 2.1, 165-70). Another diacritic sign,
again most common in central Asia but also found in India, is a horizontal line or
dot above the character, which usually indicates an abridged conjunct consonant;
thus f sa = sna, >j ca = sea, and so on.
Various further modifications arose in varieties of KharosthI used to write texts or
words of non-Gandhari origin. Most important, late KharosthI records sometimes
added a line or curve at the lower right of a consonant character to indicate a long vowel
(e.g., 7> = ka). A few sporadic and somewhat uncertain examples of this device are noted
in inscriptions from India,178 but it is much more common in the central Asian docu-
ments, especially those in Sanskrit (e.g., KIII nos. 511,523). In the latter we also find
signs for syllabic r, the vrddhi vowels ai and au (formed by combining the ordinary
diacritics for e and o with the long vowel sign), vowelless consonants (see KI III.297),
na, and visarga (the last two being perhaps borrowings from Brahml). In the central
Asian documents, non-Indie names and loan words were represented with the help of
new conjuncts such as *f pga, J, cma, and ^jhbo. Many of the diacritically modified
characters mentioned previously were also used for this purpose.

2.4 Numbers and Numerical Notation

BIP 76-87; 1C 11.683, 702-9; HEP 125-8; BPLM 103-29; S. L. Gokhale, Indian
Numerals; SAI 168-76.

2.4.1 Numerical notation in Brahmi and the derived scripts

[Link] The old additive/multiplicative system
Numerical figures are found in Indie inscriptions from the earliest times, especially
in the recording of dates but also in the details of financial or other transactions, the
numbering of verses in metrical inscriptions (see [Link]), and in various other con-
nections. Numbers in all periods have also been commonly expressed in words, either
directly or (in later centuries) by means of chronograms and other indirect expres-
sions (see [Link].2).

177. See Brough, GD 67-70; see also GD 75-7, for the similar problem of the characters ^ (ha
and tf t'ha.
178. See Konow, CII 2.1, cxx, and Salomon, StII 7, 1981, 14.
 Writing and Scripts in India 57

The original Brahml numerical system (see table 2.6), as first found in the Asokan
and other early inscriptions, is essentially additive/multiplicative in principle. It had a
full array of separate symbols, not only for the digits from 1 to 9 but also for the decades
from 10 to 90, as well as separate signs for 100 and 1,000; the system thus comprised an
array of twenty basic signs.179 The figures for 100 and 1,000 were treated multiplica-
tively; for example, 400 would be represented as a ligature of f 4 and 9 100, that is, •$.
But multiples of 2 and 3 were indicated by the addition of one or two horizontal ticks,
respectively, at the right of the sign for 100 or 1,000; thus 7- = 200 and •> = 300.

[Link] Origin and history of the old Brahml numerical system

Early specimens of Brahml numerals are found in the Asokan inscriptions (only 4,
6, 50, and 200), in the cave inscriptions of the first century B.C. from Nanaghat (SI
1.192-7) and of the first century A.D. from Nasik (El 8, 59-96), and in the inscrip-
tions of the Kusanas. A complete set of the figures is first found in the unique series
of dated coins issued by the Western Ksatrapas in the second to fourth centuries A.D.,
which provided important clues for the decipherment of the numbers. Besides the
limited body of early materials, the analysis of the old numerical system is further
complicated by the fact that the symbols were subject from the earliest times to con-
siderable graphic variability, which increased over the centuries and eventually re-
sulted in the notoriously diverse character of the numerical symbols in the different
BrahmT-derived scripts. Fortunately, in many inscriptions, for example at Nasik,
numbers are expressed in both words and numerical characters, and these instances,
together with valuable hints furnished by the dates of the Ksatrapa coins, were in-
strumental in the decipherment of the Brahml numerals.180
Although the system itself is nowadays fully understood, the historical origins
of the numerical characters remain controversial. Several theories on this question
are based on their apparent resemblance to certain letters of Brahml and related
scripts. The first of these was Prinsep's hypothesis181 that the number signs are
derived from the characters for the initial letter of the Sanskrit word for each num-
ber. This theory, though accepted by F. Woepcke,182 is obviously unsatisfactory.
In 1877, Bhagvanlal Indraji183 declared that "all of [the numerals] . .. except the
first three express letters or groups of letters"; thus the sign for 6 is interpreted as
phra or phra, 7 is gra or gra, and so on. In his postscript to BhagvanlaTs article
(47-8; see also BIP 77-81), Biihler accepted the theory, though he admitted that
the actual "origin" of these numerals remained obscure, since no rationale could
be discerned for the particular phonetic values attributed to the numeral signs.184
179. On the notation of fractions in the old system, see Dines Chandra Sircar, "Fractions in an
Early Inscription," Journal of the University of Gauhati 1, 1950, 133-6.
180. See section 6.2.3 for details.
181. JASB 7, 1838, 352-3 = Essays on Indian Antiquities 11.77.
182. "Memoire sur la propagation des chiffres indiens," JA, ser. 6, vol. 1, 1863, 27-79, 234—90,
and 442-529 (esp. 70-3).
183. "On the Ancient NagarT Numerals," IA 6, 42-8 (quoted 43).
184. As will shortly be shown, Biihler did not feel this theory to be incompatible with his belief
in an ultimate Egyptian origin of the BrahmT numerals, though he did not completely clarify his ideas
as to the relationship between the two factors.
 TABLE 2.6. Numerical Notation in Brahmi and Kharosthi

Value Brahmi Kharosthi

1
2
3
4
5
6
7
8
9
10
20
30
40
50
60
70
80
90
100
200
300
1,000

2,000 (not attested)

4,000 (not attested)

Representative Brahmi forms are shown, mostly from Western Ksatrapa coins; in practice,
number signs in BrahmT show much variation. The Kharosthi forms are from various inscrip-
tions, with alternate forms from central Asian documents.
 Writing and Scripts in India 59
For this and other reasons185 the idea that the numerical symbols are actually Brahml
letters is not widely accepted.185 While it is true that some of the numerical sym-
bols in Brahml and several Brahml-derived scripts do bear a striking resemblance
to phonetic characters, the correspondences are not regular enough to establish a
convincing connection. It is more likely that we are dealing with an originally sepa-
rate set of signs (whose origin remains to be determined) which tended to be adapted
to graphically similar phonetic signs of the script by the scribes of various times
and places. The relationship between the phonetic and numerical signs in Brahml
thus seems to be essentially secondary and superficial.
Other scholars have therefore attempted to explain the Brahml numerals as bor-
rowings from other scripts. Alexander Cunningham187 claims to have been the first
to note that the early Brahml figures for 5 through 9 resemble the "Ariano-Pali" (i.e.,
KharosthI) letters for the initial syllables of the words for each number (e.g., Brahml
f 5 resembles KharosthI/1 pa forpanca, etc.). This superficial resemblance was quite
rightly rejected by Edward Thomas already in 1855,188 as well as by J. Dowson189
and Biihler (OIBA 52-3 n. 1). It nevertheless reappeared in modified form in E. Clive
Bayley's essay "On the Genealogy of Modern Numerals,"190 in which he assigns to
the Brahml numerical system an "eclectic character" (346) arising from "a process
of mixed borrowing and adaptation" (360) from various sources including Phoenician,
"Bactrian" (i.e., KharosthI), Egyptian, and possibly cuneiform. Although Bayley's
hypothesis itself and the arguments used to support it are rather far-fetched, his ref-
erences to the Egyptian hieratic numerals (356-8) are of interest.191 For there are
striking similarities in principle, and to some extent in the specific symbols, between
the Egyptian and Indian numerals. The hieratic and demotic number system agrees
with that of BrahmT in having separate characters for the units, decades, 100, and
1,000, and its method of indicating multiples of the larger denominations is also simi-
lar, if not exactly identical, to the Indian method.
These factors induced Biihler (BIP 82) "to give up Bhagvanlal's hypothesis" (i.e.,
the phonetic theory) and to accept the view that "the Brahma numeral symbols are
derived from the Egyptian hieratic figures," which position he expounded further in
OIBA 115-9 (Appendix II: The Origin of the Ancient Brahma Numerals). While many
problems remain, it must be admitted that the similarities are quite striking. As was
the case with Biihler's derivation of the Brahml alphabet from a North Semitic pro-
totype (see [Link].4), some of the resemblances between the corresponding Egyp-
tian and Indie numeral signs are strong, but others much less so, so that establishing

185. Note, for example, such far-fetched identifications as that of 50 with "the anundsika . . . as
it occurs in the manuscripts of the Madhyandina Sakha of the White Yajurveda," and 80 and 90 with
the upadhmanlya and jihvamulTya, respectively (Bhagvanlal, op. cit., 47).
186. See the critical comments by Ojha (BPLM 106) and others (e.g., Gokhale, Indian Numer-
als, 48).
187. JASB 23, 1854, 703-4.
188. JASB 24, 558ff.
189. JRAS 20, 1863, 228-9 n. 1.
190. JRAS, n.s. 14, 1882, 335-76 (esp. 348-55).
191. A possible connection with Egyptian (demotic) numerals had already been briefly proposed
by Burnell in 1874 (SIP 65), although Bayley does not refer to him.
60 Indian Epigraphy
connections between them involves considerable manipulation. In a few cases, no-
tably the signs for 80 and 90, Buhler admits, "There is no actual resemblance be-
tween the Egyptian and Indian forms" (118). Another problem with the Egyptian
derivation is historical, since, as Buhler admits (OIBA 119), the evidence for con-
tact between India and Egypt in very early times is weak (Bayley's attempts [op. cit,
361-3] to establish such contacts notwithstanding), and in the absence of such cor-
roboration the derivation can hardly be considered authoritative.
Recently, Falk (SAI175-6) has proposed a possible influence from the early Chi-
nese system of numerical notation, but the similarities are in fact less striking than those
with Egyptian, and the historical and geographical arguments are even less cogent.
Thus some authors, unconvinced by the arguments for borrowing, hold out for an
indigenous origin of the Brahml numbers. Ojha, for example (BPLM 114), thinks that
they are "independent creations of the Indo-Aryans" (bharatiya aryom ke svatantra
nirmana). Sircar, too (HEP 126), says that they "appear to have developed out of certain
signs of the pre-historic writing of India"; but neither author offers any direct proof
for the presumed indigenous origin. A. H. Dani192 has recently presented a scheme
for an internal derivation of the Brahml numerical symbols from a (hypothetical)
basic sign for 10, but this purely formal hypothesis is not compelling.
In conclusion, the problem of the origin of the Brahml numerals is roughly analo-
gous to that of the BrahmT script itself. We have, on the one hand, a proposal by Biihler
et al. for a foreign origin, which is convincing to some extent but falls considerably short
of real proof, on both formal and historical grounds;193 and on the other hand, a camp
which rejects Blihler's and other theories of borrowing and holds out for an indigenous
origin, but without concrete evidence. However, it must be noted that the development
of numerical notation figures does not necessarily follow the same principles as linguis-
tic notation, nor is it necessarily linked to it. For whereas phonetic signs are normally
completely arbitrary in form, numerical signs can often be derived as cursive reduc-
tions of collocations of counting strokes, as in the case of the figures for 1, 2, and 3,
which resemble each other in many systems around the world, and, more significantly,
in the hieratic Egyptian figures for the decades which are clearly derived from com-
binatory groups of the old hieroglyphic sign for 10. Thus, it is possible for indepen-
dently derived systems of numerical notation to independently develop similar principles
and even similar forms, so that it may well be that the old Indian numerical system was
entirely indigenous in origin despite its apparent resemblance to the Egyptian system.
This numerical system could also have been in origin entirely separate from, and perhaps
older than, the BrahmT script with which it came to be associated,194 and this would ex-
plain the persistent failure of all attempts to interpret the numerical signs in terms of pho-
netic values derived from apparent resemblances to characters of BrahmT or other scripts.

192. Preface to the second edition (1986) of DIP, ix-xvii.

193. It should be kept in mind that the formal derivation of numeral signs does not necessarily follow
the same principles as those which apply to the derivation of phonetic characters. For while the latter are
(effectively) purely arbitrary symbols, the former are often derivable, at least as far as the lower numbers are
concerned, from combinations of simple counting strokes; so that the numbers from 1 to 3 or 4, and sometimes
higher ones as well, may often resemble each other in widely separated and unrelated systems. For this rea-
son, purely formal similarities between numerical symbols are less significant than those of phonetic signs.
194. Cf. G. R. Kaye, "The Old Indian Numerical Symbols," IA 40, 1911, 49-55 (esp. 50, 54).
 Writing and Scripts in India 61

[Link] The development of place-value notation

Despite outward changes in form, the old additive system of Brahml numerals remained
stable until about the seventh century A.D., at which point the modern system of deci-
mal place-value notation, which can represent any number of any size with only the
nine digit symbols plus a sign for zero, began to come into use. There are, however,
several problems involved in determining exactly what is the earliest epigraphic attes-
tation of the new, or "place-value,"195 system. First of all, there is the problem of spu-
rious inscriptions; several inscriptions which were previously thought to provide early
specimens of place-value numbers have proven to be spurious or at least of doubtful
authenticity.196 Most important among these is the MarikanI (formerly called Sankheda)
copper plate inscription (CII 4.1, 161-5), bearing in place-value characters the date
346, which is presumably attributable to the Kalacuri era and hence equivalent to A.D.
594/95 (if a current year) or A.D. 595/96 (if expired). Although this inscription is widely
cited both in specialized publications (e.g., BIP 83 and HEP 127) and in general works
on the history of mathematics197 as providing the earliest specimen of a place-value
date, cogent historical and contextual arguments have been presented by V. V. Mirashi198
to show that this is very likely a spurious record, and hence does not prove the use of
the new system in the sixth century A.D.
Second, there is the problem of the great variability and consequent difficulty in
interpreting the numerical figures. A case in point here is the Sakrai stone inscrip-
tion, dated, according to B. Ch. Chhabra (El 27, 27-33), in [Vikrama] 699 = A.D.
643, which would make it the earliest new-system date (disregarding the MarikanI/
Sankheda plates). But the reading of the date is controversial, and the first figure might
be 8 rather than 6 (op. cit., 30), so that this too cannot be definitely considered the
earliest place-value date.
Actually, the earliest unquestionable inscriptional dates in the new system are not
from India proper but from Southeast Asia. G. Coedes, in his important article "A propos
de 1'origine des chiffres arabes,"199 pointed out several early place-value dates in the
Saka era on indubitably authentic Southeast Asian stone inscriptions, including two
inscriptions from Indonesia and Cambodia dated in Saka 605 = A.D. 683. Coedes thus

195. In many publications on the subject of Indian numerals, the old and new styles are routinely
referred to as the "numerical" and "decimal" systems, respectively. But this terminology is inaccurate,
since in fact both systems are both numerical and decimal. I therefore prefer the terms "additive" and
"place-value," or simply "old" and "new."
196. This problem is not merely coincidental; for it would only be expected that forgers of copper plate
grants in the centuries following the transition from the old to the new notation system would predate
the forged documents in the new style to which they were accustomed, thus giving rise to a body of docu-
ments which purport to give dates in new-style numbers before they were actually in use (see also 5.3).
197. E.g., Georges Ifrah, From One to Zero: A Universal History of Numbers, tr. Lowell Bair (New
York: Viking, 1985), 437^tO. Ifrah, in accepting the evidence of the Sankheda (i.e., MarikanT) plates,
states that "[t]o the best of my knowledge, no serious reason for questioning the authenticity of the
Indian copperplate deeds has ever been stated" (440). But the objections raised by Mirashi, as cited
shortly, are indeed "serious" reasons for doubt.
198. CII 4.1, 161-3; see also "A Note on the Mankani Grant of Taralasvamin," Journal of the
Ganganatha Jha Research Institute 1, 1944, 389-94.
199. BSOAS 6, 1930-32, 323-8.
62 Indian Epigraphy

concludes that this system therefore must have been in use by the late seventh century,
and probably earlier, in Southeast Asia and presumably also in India.200
Returning to India proper, the earliest definite specimen of pure place-value nota-
tion (as opposed to mixed or partial place-value notation, discussed later) now seems
to be in the Siddhantam plates of the Eastern Gariga king Devendravarman (El 13,
212-6) dated (see Hultzsch, El 18,308) in the [Gariga] year 195, equivalent to approxi-
mately A.D. 693, or just ten years later than the Southeast Asian examples mentioned
earlier. Another specimen from the same period is the Sudava plates (El 26, 65-8) of
Anantavarman, son of the aforementioned Devendravarman, dated [Ganga] 204=A.D.
702. But undisputed cases of place-value notation continue to be somewhat scarce
through the eighth century A.D.201 Examples from the ninth century, such as the Torkhede
plates of [Saka] 735 = A.D. 812 (El 3, 53-8) and the Buckala stone inscription of
[Vikrama] 872 = A.D. 815 (El 9, 198-200; see esp. 199 n. 1) are more secure.
After this time, that is, from about the middle of the ninth century onward, epi-
graphic notation of dates by the place-value system becomes standard. But the old
system also continued to be used through the eighth century and sometimes even
into the ninth, for instance, in the Barah copper plate of [Vikrama] 893 = A.D. 836
(El 19, 15-19). An interesting illustration of the process of transition is provided by
the Ahar stone inscription (El 19, 52-62). This is a composite record often separate
documents of different dates; the two earliest documents (nos. II and I), dated in
[Harsa] 258 and 259 = A.D. 864 and 865, respectively, use the old additive notation
(cf. the editor's note, 58 n. 2), while all the later dates, from [Harsa] 261 = A.D. 867
(no. IX) on, are in the new place-value system. This document thus allows us to pre-
cisely date the transition from old to new style in the area concerned (Bulandshahr
Dist. in central northern India) to around A.D. 866.
We also find during the transitional period in and around the seventh century not
a few cases of a curious hybrid system which combines features of the old and new
systems. Several cases of this have been found, interestingly enough, among the
copper plate inscriptions of the Eastern Ganga kings in whose later records, as men-
tioned previously, were found the earliest definite cases of the place-value system.
Thus, as early as ca. A.D. 578 we find in the Urlam plates of Hastivarman (El 17,
330^1; corrected reading of the date in El 18, 308) the [Ganga] year 80 written with
the old-style sign for 80 plus the new-style sign for 0, and in the Purle plates of
Indravarman (El 14, 360-3; see also El 18, 308) the Ganga year 137 = ca. A.D. 635
written as 100-3-7 and the day 20 as 20-0.202 It thus appears that the transition from

200. Although both of the Southeast Asian inscriptions are in local languages, rather than San-
skrit, Coedes is no doubt correct in stating that the data do not justify the "etrange opinion" (324) of
G. R. Kaye ("Notes on Indian Mathematics—Arithmetical Notation," JASB, n.s. 3, 1907, 475-508)
that the symbolic number system originated in Southeast Asia and was imported thence into India, rather
than vice versa. The early Sanskrit inscriptions of Southeast Asia generally give the dates in chronogram
form, which presupposes the use of place-value numeration (see n. 203).
201. The Dhiniki inscription of Vikrama 794 = A.D. 738 (IA 12, 151-6) is now generally agreed to
be spurious, but the Samangad inscription (IA 11, 108-15) of the Rastrakuta king Dantidurga, dated Saka
675 = A.D. 753/4, is a more likely instance, though its authenticity too has been called into question by V.
S. Sukmankar in "Palaeographic Notes" (see n. 115), 313-7, mainly on paleographic grounds. The re-
marks of G. R. Kaye (op. cit., 484) on this point are misleading, as he quotes Fleet out of context.
202. For further references and examples see E. Hultzsch, El 18, 1925-26, 307-8; HEP 127;
G. S. Gai, "Two Epigraphic Notes," Journal of the Ganganatha Jha Research Institute 6, 1949, 306-7;
 Writing and Scripts in India 63

the old additive system to the new place-value system, at least as far as epigraphic
usage (which presumably mirrors popular usage) is concerned, was a gradual pro-
cess. The displacement of the old style evidently took place slowly over a period of
at least a century or so, and the new system itself underwent a process of gradual
evolution in the seventh century, during which it was sometimes used in combina-
tion with features of the old style.203 This matter has broader implications for the
important question of the origin of the zero sign and the now universal decimal place-
value system of numerical notation, which, according to some scholars (see, e.g.,
Ifrah, op. cit. [n. 197], 428ff.) is ultimately of Indian origin; but the literature on this
question is vast and goes far beyond the scope of the present book.

[Link] Later developments of the numerical notation system

Once the new system came into universal use in or around the ninth century A.D., the
subsequent history of numerical notation in the Brahml-derived scripts is essentially only
a matter of the developments of outward forms of the symbols. These developments,
however, are quite complex and not very well documented. Being a purely paleographic
matter, they will not be presented here in detail; for further information the reader is
referred to the standard works on paleography, especially BPLM, which presents this
material in detail (103-27 and charts 71-76 and 84).204 The later phases of development
of the numerals in the regional scripts in particular have not yet been adequately studied.

2.4.2 Numerical notation in Kharostht

The numerical notation of Kharosthi script (see table 2.6) follows an additive/multi-
plicative method essentially similar to that of the old Brahml system but consider-
ably different from it in detail. Unlike the latter, which has separate signs for each of
the units and decades, KharosthT has symbols only for 1,4 (a later development; see
next paragraph), 10, 20, 100, and 1,000. All other numbers are indicated additively,
somewhat as in Roman numerals; thus, for example, 76 would be written with the fig-

Subrata Kumar Acharya, "The Transition from the Numerical to the Decimal System in the Inscrip-
tions of Orissa," JESI 19, 1993, 52-62; and B. N. Mukherjee, "The Early Use of Decimal Notation in
Indian Epigraphs," JESI 19, 1993, 80-3.
203. Such is the picture to be derived from strictly epigraphic evidence; but if literary sources are
taken into account, it would seem that the actual origins of the decimal place-value notation system
must have been considerably earlier. Walter Eugene Clarke ("Hindu-Arabic Numerals," in Indian Studies
in Honor of Charles Rockwell Lanman [Cambridge, Mass.: Harvard University Press, 1929], 217—36
[esp. 225-8]) noted indications from literary and scientific texts that place-value notation must have
been in use well before the late seventh century; particularly noteworthy is the allusion to the siinyabindu,
or zero-sign, in Subandhu's Vasavadatta, which must be earlier than the seventh century (Clarke, 225).
Also significant is the attestation as early as Saka 526 = A.D. 604 of the chronogram or numerical word
system, which seems to presuppose the place-value system, in Sanskrit inscriptions from Southeast Asia
(Clarke, 226; Coedes, op. cit. [n. 199], 325), though such chronograms are not attested until consider-
ably later in India itself (Kaye, op. cit. [n. 200], 479-80; BIP 86).
204. Several of the standard paleographic studies treat numbers only briefly or not at all. Dani's
DIP, e.g., did not cover the numerical symbols (cf. Gokhale, Indian Numerals, preface), although he
has added some material on this subject in the preface to the second edition (see [Link]).
64 Indian Epigraphy

ures 20-20-20-10-4-1 -1 (or in the earlier script, 20-20-20-10-1 -1-1-1 -1 -1). The figures
for 100 and 1,000 were subject to multiplicative treatment, as in BrahmT; for example,
300 would be written as 3-100, and 2,000 as 2-1000.205
Like the KharosthI script itself, its numerical system did not change greatly in
the course of its history, though some significant developments did take place. The
sign for 4, shaped like a Roman x, was apparently a secondary development, since
in Asokan KharosthI 4 and 5 are written as l-l-l-l(-l), while in all subsequent
KharosthI texts the special sign for 4 is used. The sign for 1,000 to date is known
only from central Asian KharosthI documents, though it probably existed in India as
well. As shown by H. W. Bailey,206 it is derived from an Aramaic ligature 'LP (i.e.,
'alap, 'thousand'). Also in the central Asian documents, we find cursively written
combinations of the sign for 1 which have evidently developed into distinct charac-
ters for 2 (f) and 3 (y).
The origin of the KharosthT numerical system presents no difficulties, as its deri-
vation from Aramaic, like that of the script itself, is clear. Not only the system207 but
also all of the individual symbols208 can be easily associated with the Aramaic, as
represented, for instance, in the papyri from Elephantine (Egypt).209

2.5 Techniques of Epigraphic Writing

BIP 87-92; PIP 88-119; HEP 128-31; SIE 83-97.

2.5.1 General comments

The technical features and visual qualities of Indie inscriptions vary widely, ranging
from undecorated, strictly functional records to highly ornate and polished works of
art. Especially in the earlier centuries, many inscriptions exhibit a rather casual atti-
tude toward aesthetic quality and orthographic accuracy (see 3.3.8), which may be
attributed in part to the marginal value traditionally accorded to the written as op-
posed to the spoken word in India (2.1.1). It is true that in some cases, for example,
in Buddhist reliquary inscriptions that were meant to be permanently interred in
stupas, legibility and visual aesthetic appeal may be considered inherently insignifi-
cant. Nevertheless, it is not uncommon to find extremes of carelessness in the plan-

205. On the decipherment of the KharosthI numerical system, see section 6.3, p. 218.
206. "A Problem of the KharosthI Script," in Essays and Studies Presented to Stanley Arthur
Cook . . . , 121-3 (see n. 138).
207. The Aramaic system represented the digits up to 9 by simple combinations of the vertical
stroke representing 1; this confirms that the later KharosthI sign for 4 is a separate Indian develop-
ment.
208. The KharosthI symbol for 20 is often described (e.g., in HEP 125) as a doubling of the
KharosthI 10, but comparison with the Aramaic prototype indicates that it is rather a direct borrowing
of the Aramaic 20.
209. See Emil G. Kraeling, The Brooklyn Museum Aramaic Papyri: New Documents of the Fifth
Century K.C. from the Jewish Colony at Elephantine (London: Yale University Press for the Brooklyn
Museum, 1953). Papyrus 4 (168-9) has several numerical figures which are almost identical to the
corresponding ones in KharosthT.
 Writing and Scripts in India 65

ning and execution of inscriptions which were clearly meant to be seen and read;
and while this is most common in minor private records, it can also be observed in
royal inscriptions, as, for example, in the notorious case of the Erragudi rock edicts
of Asoka (see n. 82).210 There are, of course, many exceptions to this pattern, espe-
cially in the late medieval period in which the arts of writing and engraving attained
their highest development (see [Link]), but on the whole one must agree with Sircar
(SIE 86) that "[t]he number of inscriptions carelessly written and engraved is ... by
far larger than those written and engraved carefully."

2.5.2 The technical execution of inscriptions

[Link] Preparation of surfaces
As with the general quality of the inscriptions, the degree of care in the preparation of
the surfaces to be inscribed varies widely. In early records particularly, the inscrip-
tions might be written directly on the intended surface with little or no preparation
thereof; such is the case, for example, with many of the Asokan rock inscriptions.211
As time went on, however, more sophisticated techniques were developed, culminat-
ing in the very skillfully dressed and prepared polished stone slabs, often of large
dimensions, used for many inscriptions of the medieval period (cf. SIE 70). Similarly,
inscribed copper plates were frequently prepared with considerable skill and artistry,
especially in later centuries, though cruder examples are also by no means rare.

[Link] Execution of the inscription

After the preparation of the intended surface, the letters of the inscription itself would
usually be first written onto it in some temporary medium such as ink or chalk by a
writer or scribe, known as lipikdra, lekhaka, karana/karanika, kayastha, and so on
(see BIP 100-2 for further details). This preliminary step is discernible in occasional
cases wherein the ink or paint is still visible in the inscription (e.g., at Bandhogarh,
El 31, 168), or where the final step of carving the inscription was never carried out
(e.g., the Ginja hill inscription, ASIR 21,1883-85, 119-20 [cf. El 31, 168], and the
Kasia copper plate, ASIAR 1910-11, 73-7). The actual carving would then be car-
ried out by an engraver (sutradhdra, sildkuta, rupakdra, etc.; cf. SIE 85). Sometimes,
again mostly in the earlier period, the engravers seemed to be illiterate or minimally
skilled, but in later times stone engraving became a legitimate art unto itself, and its
expert practitioner came to be a person of some status (cf. IESIS 32-5) who would
often (especially in later stone prasastis) be mentioned by name at the end of the
inscription along with the composer and/or scribe.212 The verbs typically used to

210. Note also J. Ph. Vogel's comments (El 20, 1929-30), regarding the Nagarjunakonda inscrip-
tions, that "the careless manner in which they have been recorded is astonishing" (11).
211. The pillar edicts are written on exquisitely polished stone surfaces, but this preparation was
evidently an aspect of the construction of the pillars themselves as monumental works of art, rather
than as preparation for the inscriptions themselves.
212. For examples, see Appendix, no. 8, vv. 22-23, and no. 12, v. 13. For cases where all three
persons involved in the preparation of the record, viz. the composer, scribe, and engraver, are men-
tioned, see El I, 42, 11. 27-8; 48-9, 11. 21-2; and 80-1, 11. 25-6 (cf. BIP 101 n. 12).
66 Indian Epigraphy
denote the activities of composing, writing, and engraving are, respectively, ~^kr or
(v/-) Vrac; (ul- or vi-) ^Slikh;213 and ut-^kf or (ni-)^khan.
The normal method of writing was by incising or carving lines into the hard
surface with a stylus, chisel, or other suitable implement. Some early inscriptions,
mostly KharosthI relic dedications on metal plates, are written with a series of dots
(e.g., the Taxila silver scroll, CII 2.1,70-7). In the later medieval period we some-
times find Indie-language inscriptions composed of raised letters (e.g., the Peshawar
Museum inscription, El 10, 79-81) in imitation of the standard practice in Islamic
inscriptions.

[Link] Punctuation, numeration, and ornamentation

In general, punctuation is treated in Indie inscriptions (as in manuscripts) rather in-
formally, and seems to have been considered more or less optional. In early inscrip-
tions punctuation is often totally absent, or used very sporadically at best. A notable
exception is the Kalsl version of the Asokan rock edicts XI-XIII, where a single
vertical line of punctuation is used frequently, albeit promiscuously and inconsis-
tently.214 Punctuation marks began to be employed with any regularity only in the
Gupta era, when some metrical inscriptions have the ends of half verses and full verses
marked regularly by a single short horizontal line and a double vertical line, respec-
tively (e.g., the Mandasor pillar inscription of Yasodharman, CII 3,142-8). But even
in this period these marks are used sporadically, and many inscriptions are still en-
tirely unpunctuated.
In the inscriptions of the early medieval and succeeding periods, punctuation
marks begin to be used somewhat more systematically, particularly in verse records.
The system continues to consist principally of two marks, a half stop with a single
line and a full stop consisting of two lines, but these marks take various forms at
different times and places: straight or curved lines arranged vertically or horizon-
tally, lines with various hooks and other appendages, and so on.215 In general, the
single mark is used to mark the half verse and the double mark or full stop the end of
a verse, but usage in this respect is quite loose. Sometimes only the verse end is punc-
tuated, sometimes each pada is marked, and not infrequently the usage is inconsis-
tent even within an inscription.216
The verses are often numbered consecutively, either in conjunction with or in-
stead of the punctuation system mentioned previously. Verse numbering is first noted
in the Allahabad inscription of Samudragupta (CII 3, 1-17), and is common in in-
scriptions of the medieval period and later; but here, too, usage is inconsistent.
As a rule, Indie inscriptions are written continuously to the end of the line with-
out regard for grammatical or structural divisions. But in many early Sanskrit inscrip-
tions before and during the Gupta era the lines are planned so as to correspond to
metrical units, usually half verses (e.g., in the Mandasor pillar inscription, CII 3,
213. On the technical sense of this term, see Fleet, CII 3, 99 n. 3.
214. Note that these inscriptions are incompetently written in other respects as well; see section
2.2.4.
215. For details see BIP 88-9.
216. See, e.g., Appendix, no. 10.
 Writing and Scripts in India 67

142-8).217 This system was largely abandoned in later centuries218 (except in South-
east Asia, where it remained in use much longer),219 and thereafter all inscriptions,
verse and prose, are lineated at random without regard to content.
In general, especially in Sanskrit inscriptions, there is no separation or other
explicit word division. In some earlier inscriptions in Prakrit and "Epigraphical
Hybrid" dialect, however, words or phrases are occasionally separated by a space;
such is the case, for example, in several of the ASokan inscriptions (e.g., Rummindel;
Appendix, no. 1) and some of the Nasik cave inscriptions (e.g., Nasik no. 10, El 8,
78-81), as well as in some Kharosthi inscriptions (e.g., Sui Vihar, CII 2.1,139). Only
rarely do we find a mark consisting of two short vertical lines one above the other,
indicating a word divided between lines (equivalent to the modern hyphen; Nalanda
ins., Appendix, no. 12, end of lines 2,4, 5, etc.).
Abbreviations are commonly used in inscriptions of various periods, especially in
connection with dating formulae. The word samvatsara 'year' in particular is subject
to various abbreviations such as samvat, samva, sam, sa, and so on (BIP 91).220 Other
technical terms are frequently abbreviated, especially in copper plate inscriptions; for
instance, du for dutaka, bra for brahmana, or dm for dramma (see HEP 129).
It was customary from the earliest times to place various auspicious signs at the
beginning of inscriptions, and sometimes also at the end or even in the middle (e.g.,
MI §44). Symbols such as thesvastika and the triskele are used in some of the Asokan
inscriptions, for example, in separate rock edicts nos. 1 and2atJaugada(CII 1,111-8).
The Hathlgumpha inscription of Kharavela (SI 1.213-21) has srivatsa and svastika
at the beginning in the left upper margin, and the "tree-in-railing" symbol at the end.
Beginning in the early Christian era, it became customary to put the word siddham
'success(ful)' or its abbreviations and Prakrit equivalents (sdha, sidham, etc.) at the
beginning of an inscription or in the left margin. In the Gupta era, a symbol, origi-
nally consisting of a curved stroke open to the left, was substituted for the word
siddham, and this sign, in various developed forms, became the standard invocatory
mark thereafter. It appears to have later been sometimes understood to represent om,
but, as shown by Nalinikanta Bhattasali,221 it is probably more accurate to read it as
equivalent to siddham. Other words and phrases, such as siddhir astu 'May there be
success' or svasti 'well-being' (frequently used together with siddham) were also
often put at the beginning or end of inscriptions. A sign consisting of a circle with a
dot in the middle was sometimes written at the end of a record or section thereof;
this was later confused with and written as the syllables tha or cha. Similar symbols

217. See the comments of Liiders, El 24, 1937-38, 198, and n. 5; also Sircar, El 35, 1963-64, 17.
218. This presumably happened as a result of the increasing use of different meters of varying
length in ornate inscriptions, which makes it impossible to maintain lineation by verse, except by leav-
ing blank spaces at the end of shorter verses; but this, with the characteristic Indian horror vacuui, was
never done.
219. See, e.g., Appendix, no. 8.
220. For further examples of abbreviations in dating formulae, see section [Link].
221. El 17, 1923-24, 352; see also SIE 92-5. This interpretation has recently been further sup-
ported and refined by two studies published in G. Bhattacharya, ed., Deyadharma: Gustav Roth,
"Mangala-Symbols in Buddhist Sanskrit Manuscripts and Inscriptions" (239-44), and Lore Sander,
"Om or Siddham—Remarks on Openings of Buddhist Manuscripts and Inscriptions from Gilgit and
Central Asia" (251-61).
68 Indian Epigraphy
and marks, sometimes developed into elaborate floral motifs, are used at the end of
records, where they function as marks of both decoration and punctuation (see, e.g.,
El 31, plate facing page 127,1. 212).
Many inscriptions, especially in medieval and later times, are accompanied by
engraved illustrations of various sorts. These may take the form of auspicious sym-
bols such as Siva's bull Nandi, the Siva-linga (e.g., CII 6, pi. XC), the conch shell
(satikha), lotus, and so on; or they may be representations of dynastic symbols, such
as the Garuda on the copper plates of the Paramaras (CII 7.2, pi. I, IV, VI, etc.). Other
illustrations refer specifically to the content of the record itself, such as the represen-
tation of the sun and moon (e.g., CII 6, pi. XC), alluding to the traditional blessing
that a gift should last as long as the sun and moon; or of a donkey copulating with a
woman or a pig, representing the mother of a violator of the gift according to the
traditional "ass-curse" (e.g., IA 48,1919,47 and n. 33; El 40,1973-74, plate facing
page 88), which is also sometimes explicitly spelled out in the text of the inscrip-
tion.222 Sati and other types of memorial stone inscriptions (see 4.1.4) are usually
accompanied by representations, either visual or symbolic (typically in the form of
a hand- or footprint), of the person memorialized.
Corrections and omissions are indicated by various means. Sometimes, as in
Asokan rock edict XII at KalsT, 1. 31 (CII 1,42), an incorrectly written passage was
simply crossed out and the correct reading entered above the line. In copper plates,
however, the standard practice was to beat the incorrect passage flat and rewrite it,
or simply to write over it without erasure (see, e.g., Appendix, no. 10,1.36). Omitted
letters or words could be inserted by simply writing them into, above, or below the
appropriate place in the line. Longer omissions would typically be added in the
margins; their intended location was usually indicated by placing the beginning of
the added passage in the upper or lower margin at the appropriate vertical position,
and denoting the line where it belonged by a number indicating the number of lines
from the top or bottom. Numerous examples of corrections of these types are pointed
out in the editor's notes to the Asankhali plates of Narasimha II, El 31, 109-28. Al-
ternatively, the placement of a marginal addition could be indicated in the text by
the cross-shaped mark known as kakapada or hamsapada (see, e.g., El 3, 51 n. 9).

2.5.3 Calligraphic writing

IESIS 32-8; R. Salomon, "Calligraphy in Pre-Islamic India," in Asher and Gai,
IEBHA, 3-6.
Unlike in China or the Islamic world, in traditional India calligraphy did not achieve
the level of a major art, which is not surprising in view of what has been said earlier
(2.1.1, 2.5.1) about the relatively low status of the written word in India. Nonethe-
less, many specimens of calligraphic writing from various periods do survive in Indie
inscriptions, and some of them demonstrate a very high degree of skill and artistry,
although this aspect of Indian epigraphy and paleography has received little atten-
tion in scholarly writing.

222. E.g., CII 7.2, 207, yo na dadati tasya mata [sic] gardabho jabhati [sic], and pi. LVIII. See
also El 9, 1907-8, 164, jo anyatha karoti tasya pita gardabhah sukarT mata.
 Writing and Scripts in India 69

[Link] Calligraphic developments of the standard BrahmT-derived scripts

Even in relatively early times we sometimes find BrahmT223 records which are en-
graved with considerable care and skill—for example, in the Nasik cave inscriptions
and in several of the Mathura inscriptions (MI §18 and 118). But calligraphic writ-
ing proper, that is, writing with a systematically developed decorative component,
first begins to appear around the second and third centuries A.D., for example, in the
Bijayagadh Yaudheya inscription (CII3, pi. xxxvi.B) in the north and in the inscrip-
tions of the time of the Iksvaku kings at Nagarjunakonda, AmaravatT, Jaggayyapeta,
and so on, in the south. These early calligraphic scripts are principally characterized
by various forms and degrees of elongation and elaboration of the vowel diacritic
signs, which is to remain the principal technique for calligraphic elaboration through
the following centuries. The other major technique of decorative embellishment
consists of various creative treatments of the principal consonantal characters them-
selves, as, for example, in the "box-headed" scripts of central India (see [Link]) in
which all the letters are modeled into harmonious squared and angular forms. This
type of embellishment by stylization was also employed in some later calligraphi-
cally developed scripts such as Siddhamatrka as well as in the "special calligraphic
scripts" (see [Link]), often in conjunction with the technique of diacritic decoration.
The development of these techniques reached its climax in northern India in the
seventh and eighth centuries, from which period several inscriptions in Siddhamatrka
constitute masterpieces of calligraphic art; see, for example, the Lakkha Mandal
prasasti (Appendix, no. 8; fig. 18), displaying the characteristic features of artisti-
cally angularized consonants and highly elaborated vowel signs. This type of orna-
mental script attains extreme forms in some cases, particularly in signatures such as
those of the Emperor Harsa on the Banskhera plate (El 4, 208-11; fig. 1) or of one
Purnnaditya at the Rajlvalocana temple at Rajim (IEBHA, pi. 4). In the latter ex-
ample the decorative portions of the vowel diacritics are elaborated into floral and
animal-like motifs, a technique which is to be further developed in southern Indian
calligraphy of succeeding centuries.
The golden age of calligraphic writing in the south occurred in the tenth to twelfth
centuries. The calligraphic inscriptions of this period are characterized by extreme
elaboration of the vowel diacritics and subscript consonants of the first and last lines,
forming elaborate floral and zoological motifs as a decorative border at the top and
bottom of the inscription; see, for example, the Sravana-Belgola inscription of
Mallisena (El 3, plates between pages 194 and 195).

[Link] The special calligraphic scripts

Besides the previously mentioned calligraphic varieties of the standard scripts, there
also developed in the Gupta period and succeeding centuries calligraphic forms of
writing which, though ultimately derived from standard Brahml, became so stylized
and elaborate as to virtually constitute separate scripts. The first of these scripts, vari-

223. KharosthT, for the most part, remained essentially a "clerk's script" and did not attain any
significant degree of calligraphic development, though some specimens (e.g., the Mamane DherT ped-
estal inscription, CII 2.1, 171-2) show a skillful hand.
70 Indian Epigraphy

FIGURE 1. Signature of King Harsa on the Banskhera copper plate: svahasto mama mahara-
jadhijraja-sriharsasya ("This is the signature of me, the Great King of Kings, SrI-Harsa").
From El 4, plate facing page 210; copyright, Archaeological Survey of India.

ously referred to as "Ornamental Brahml" or (preferably) "Ornate Brahml," is charac-

terized by squat, squarish, highly stylized characters. Short inscriptions in this script
have been found in considerable numbers at various sites in northern, central, and upper
southern India (see IEBHA, 5). They presumably record personal names or pilgrims'
records (see 4.1.6), following the pattern, noted earlier, that the most elaborate calli-
graphic writings are primarily used for signatures. Because of the highly stylized forms
of the letters of the Ornate BrahmT inscriptions, they are extremely difficult to read,
and few, if any, of them can be considered to have been satisfactorily deciphered.224
The second of the "special" scripts is Saiikhalipi or "shell characters" (fig. 2).
Like Ornate BrahmT, this highly ornate script, also evidently used principally for sig-
natures, was in wide use over virtually all of India except the far south from approxi-
mately the fourth to the eighth centuries A.D. ; over six hundred examples of this script
are now known.225 The characters are highly stylized, tending to conform to a gen-
eral shape superficially resembling a conch shell or sarikha, whence the name ap-
plied to the script. Statistical and formal analysis of the characters indicates that they
are derived from the characters of standard Brahml, but the changes are so extensive
that it is still very difficult to authoritatively identify them.226 Several claims of de-
cipherment have been put forward, but they are not convincing.227

2.5.4 Biscript inscriptions

Biscript inscriptions, that is, those written in the same language repeated in different
scripts, are found occasionally in various periods, though they are much less com-
mon than bilingual inscriptions (see 3.6). In the northwest, inscriptions were occa-
sionally written in both BrahmT and KharosthT scripts, for instance, the Kanhiara and
Pathyar inscriptions (CII 2.1, 178-9).228 Another Brahml/KharosthI biscript has re-
cently been discovered in Mathura (El 40, 168-9). Inscribed seals often have leg-

224. See also R. Salomon, "Undeciphered Scripts of South Asia," in Jayanta Chakrabarty and
D. C. Bhattacharyya, eds., S. K. Saraswati Commemoration Volume: Aspects of Indian Art and Cul-
ture (Calcutta: Rddhi-India, 1983), 201-12 (esp. 203-5).
225. See R. Salomon, "New Sarikhalipi (Shell Character) Inscriptions," Studien lur Indologie und
Iranistik 11/12, 1986, 109-58, and references given there (esp. 109-10 n. 4).
226. Salomon, op. cit., 145-52.
227. See, for example, R. Salomon, "A Recent Claim to Decipherment of the 'Shell Script,'" JAOS
107, 1987, 313-5.
228. See also J. Ph. Vogel, "Two Brahmi and Kharoshthi Rock-Inscriptions in the Kangra Val-
ley," El 7, 1902-3, 116-9.
 Writing and Scripts in India 71

FIGURE 2. Shell character (Sarikhalipi) inscription from Siddha ki gupha, Deogarh (Lalitpur
Dist., UP). Photograph courtesy of Michael D. Willis.

ends in both Brahml and KharosthI (e.g., CII 2.1, 100, no. 2; 102, no. 11); one seal
in the British Museum has the name of the owner in these two scripts and in Greek as
well (JRAS 1905, 809-13).
Biscripts are also sometimes found among later inscriptions; these typically in-
volve a combination of a local form or derivative of Brahml and a more widely used
script such as Devanagarl. A well-known example is the Pattadakal pillar inscrip-
tion of the time of the Calukya king Klrttivarman II (A.D. 754; El 3,1-7; see [Link]),
with the full text given in both northern Indian Siddhamatrka and in the current local
southern (proto-Telugu-Kannada) script. In other cases, such as the Kangra Jvala-
mukhl pras'asti (El 1, 190-5), part of the inscription is written in the local script
(Sarada) and the rest in Devanagarl.

2.6 Undeciphered Scripts

Besides the well-known problem of the Indus script (see [Link]), there remains a
small residue of Indie scripts which are still substantially undeciphered.229 These
include the problematic extreme calligraphic "Ornate Brahml" and "shell characters"
discussed earlier ([Link]), and also some lesser-known scripts. Notable among these
are a script found in Afghanistan which seems to resemble KharosthI and which
Fussman230 suggests may have been used to record the KambojT language, and an-
other one, broadly resembling Brahml but of uncertain heritage, found on several
terra-cotta seals from various sites (notably Chandraketugarh and Tamluk) in West
Bengal and Bangladesh.231

229. See R. Salomon, "Undeciphered Scripts . . ." (see n. 224), 201-12.

230. BEFEO 61, 1974, 34.
231. See, e.g., D. K. Chakravorty, "Some Inscribed Terracotta Sealings from Chandraketugarh,"
JNSI 39, 1977, 128-34; and Salomon, "Undeciphered Scripts . . ." (see n. 224), 205-7, for further
references.
 3
The Languages of Indie Inscriptions

3.1 Middle Indo-Aryan ("Prakrit")

3.1.1 General remarks
In early Indian inscriptions the Prakrit, or, more precisely, the Middle Indo-Aryan
(MIA) languages predominated. Indeed, from about the third to the first century B.C.
these were the only languages in epigraphic use, and they continued to predominate
for at least two centuries more, and longer in some regions. It was only in the early
centuries of the Christian era that the "mixed dialect," or "Epigraphical Hybrid San-
skrit" (EHS), and Sanskrit began to appear, and for several centuries all three—MIA,
EHS, and Sanskrit—were in use simultaneously. But the latter two gradually gained
the upper hand over the Prakrits, and then Sanskrit gradually superseded both Prakrit
and EHS (see 3.3.1-3.3.7). Thus the Prakrits practically ceased to be epigraphic lan-
guages by the early fourth century A.D. in northern India, and about a century later in
the south.
The inscriptional Prakrits are diverse and largely unstandardized. Much more than
is the case with the literary Prakrits of later times,1 the morphology and especially
the orthography of the inscriptional dialects is unstandardized and inconsistent, to
the extent that it is not unusual to find the same word spelled several different ways
within the same inscription. The orthography can be characterized as approximative,
as contrasted to the precise orthography of Sanskrit inscriptions. Thus, for example,
geminate consonants are almost always (except in some late specimens) represented
by the single consonant, and the anusvdra is only sporadically used to indicate a
homorganic nasal before a stop consonant, the nasal often being left unindicated.
Among the Prakrit inscriptions in general, the earlier the inscription, the looser and
more colloquial the language and orthography are likely to be (although the Asokan
inscriptions constitute in some respects an exception to this rule). Thus some of the
earliest examples preserve a very informal, colloquial level of Prakrit, which may
present serious problems in interpretation, while the later specimens often tend to
approach Sanskritic usage in style, vocabulary, and orthography.

1. Literary Prakrits do appear occasionally in later inscriptions, but from a linguistic point of view
this is a separate matter; see section 3.1.5.

72
 The Languages of Indie Inscriptions 73

3.1.2 The Prakrits of the Asokan inscriptions

Grammatical descriptions: E. Hultzsch, Inscriptions ofAsoka, Ivi-cxxxi; J. Bloch,Les
Inscriptions d'Asoka, 43-88.
In general, the Prakrits of the Asokan inscriptions (see [Link]) represent an early
stage of development of MIA, in which, for example, the voicing and elision of
intervocalic consonants are only occasional and sporadic, and some archaic features
are preserved which are lost in later Prakrits, such as atmanepada verb forms of the
present tense (e.g., karate, Girnar IX. 1-3) and the present participle (vijinamane [sic],
Kalsi XIII.36). Moreover, the resolution of consonantal groups by assimilation or
epenthesis (svarabhakti) is less advanced, especially in the western and northwest-
ern dialects, than in later literary and inscriptional Prakrits. In these respects, then,
the Asokan Prakrits on the whole (and especially the western dialect of the Girnar
and Bombay-Sopara rock edicts) more closely resemble Pali than any other of the
literary forms of MIA.
Overall, the Asokan inscriptions can be said to comprise three main dialects:2
the eastern, the western, and the northwestern.

[Link] The eastern Asokan dialect

Among the three Asokan dialects the eastern dialect3 clearly predominates, as it was
the local language of the imperial capital at Pataliputra (whence it is sometimes re-
ferred to as the "Chancellery Language"). The original versions of the edicts were
drawn up in this dialect, to be translated later into the other dialects where deemed
necessary. The majority of the Asokan inscriptions, including all the pillar edicts and
minor rock edicts (MRE),4 as well as the rock edicts at Kalsi, Dhauli, Jaugada, and
Erragudi, are in this dialect, and in some of them it is obviously being used far be-
yond its original geographical range.
The distinctive features of the eastern dialect include:
1. The three sibilants of Sanskrit are represented by s.5
2. The dialect has only / for Sanskrit r and /; e.g., /a/a = raja.
3. The Sanskrit nasals n, n, and n all become n.
4. Sanskrit ks becomes kh.
5. Consonantal groups are subject to resolution by svarabhakti more frequently
than in the other dialects.

2. Some writers, e.g., Sukumar Sen (A Comparative Grammar of Middle Indo-Aryan [Poona: Lin-
guistic Society of India, 1960], 10-11), posit a fourth dialect, the "Middle Eastern," for some of the
rock edicts. But on the whole it is preferable to consider the languages of these texts, which show a
mixture of eastern and western characteristics, as essentially the eastern dialect with sporadic admix-
ture of westernisms, rather than as a distinct dialect.
3. For a specimen of this dialect see Appendix, no. 1 (Rummindel minor pillar edict).
4. Though some of the MREs, especially those in the south, show an admixture of western forms,
e.g., in preferring r to /; see n. 2.
5. Rock edicts X-XIII at KalsT show all three sibilants used indiscriminately (e.g., X.27 has both
yaso and yaso for Sanskrit yasas-). Though this peculiarity has never been fully explained, it is evi-
dently a matter of scribal confusion rather than a true dialect feature. The retroflex and palatal sibilants
also occur occasionally in the minor rock edicts, sometimes with etymological justification (e.g., Maski
1. 2, vasani = Skt. varsani).
74 Indian Epigraphy

6. Final -as of Sanskrit (as in the nominative sing. masc. of noun stems in a-)
becomes -e.
1. The locative sing, masc./neut. ends in -si (for -ssi).
8. The present participle atmanepada often ends in -triina (e.g., payamina,
Delhi-Topra V.8).
The main features of this dialect—the predominance of/, the nominative mascu-
line in -e, the loss of n and n, and the sibilant s6—accord on the whole with the later
literary Prakrit known as Ardha-MagadhI, whence it has been called by Liiders and
others7 "Old Ardha-MagadhI."

[Link] The western Asokan dialect

The western dialect appears only in the rock edicts of Girnar and the Bombay-Sopara
fragments. Its characteristics include:
1. The three Sanskrit sibilants become s.
2. Both r and / are retained as in Sanskrit.8
3. The Sanskrit nasals n, n, and n are retained.
4. Sanskrit ks generally becomes cch (but occasionally also kh).
5. Many consonantal clusters are retained, especially those involving the
semivowels (kr, tr, pr, vy, etc.).9 Sanskrit st and sth are sometimes reflected
by the anomalous combination st (written ts\ e.g., stita = sthitah, Girnar, VI.4).
6. Sanskrit -as becomes -o.
1. The locative sing, masc./neut. ends in -e or -mhi.
8. The gerund ends in -tpa (written -pta; see 2.2.4) = Sanskrit -tva.
In general, the western dialect is the one most similar to literary Pali and the later
inscriptional Prakrits. It contrasts with the other dialects (especially the eastern one)
in its generally archaic aspect, in both phonetics and morphology (e.g., the atmanepada
verb karate, cited earlier). The Girnar edicts also frequently differ from the other
dialect versions in vocabulary; for instance, Girnar hnspas(s)ati (-pasyati, 1.5) where
the others have dakhatildekhati, and pamthesu instead of mag(g)esu (II. 8; for other
examples see Bloch, Les Inscriptions d'Asoka, 83-4).

[Link] The northwestern Asokan dialect

The language of the rock edicts at Shahbazgarhl and Mansehra is essentially an early
form of the MIA dialect which is nowadays known as Gandharl (see [Link].1).10
Like all records in Gandharl, these inscriptions are written in the KharosthI script,

6. This is the principal feature which distinguishes the dialect from the "true" MagadhI of later
literature, which has only /.
7. See Oskar von Hiniiber, Das iiltere Mittelindisch im Uberblick (Osterreichische Akademie der
Wissenschaften, Phil.-hist. Klasse, Sitzungsberichte 467 = Veroffentlichungen der Kommission fur
Sprachen und Kulturen Siidasiens 20; Wien: Akademie der Wissenschaften, 1986), 27—8.
8. The Bombay-Sopara fragment of rock edict IX even has r in place of Sanskrit / (phara, mangara
= phala, mangala). It is not certain whether this is a true dialectal feature or the result of hypercorrection.
9. In the script, however, many of the conjuncts are reversed. E.g., the conjunctpr is actually written
as rp, with the sign for r above the p ([,). This is probably a graphic rather than a dialectal peculiarity;
see section 2.2.4.
10. On dialectal differences between Asokan and later "GandharT," see Burrow, "The Dialectal
Position of the Niya Prakrit," BSOS 8, 1935-37, 419-35 (esp. 419-22).
 The Languages of Indie Inscriptions 75

whose graphic peculiarities—particularly the absence of notation of vowel quantity—

tend to obscure some of the features of the language. The principal distinctive fea-
tures of the dialect include:
1. The three sibilants are maintained, generally as in Sanskrit (but see also item
5).
2. Both r and / are retained as in Sanskrit.
3. The Sanskrit nasals n, n, and n are retained.
4. Sanskrit ks is apparently retained in the form of a special character (Y; see
2.3.4), variously transliterated as ks, ch, ch', and so on, whose precise pho-
netic value is uncertain."
5. Many consonantal clusters are retained, especially those with r, but the lat-
ter are frequently subject to metathesis, for example, in dhrama = dharma
and krama = karma.u Special developments include sy > s (e.g., manusa =
manusya, Shahbazgarhl II.4; cf. item 8) and sv and sm (in the locative sing,
ending) > sp (e.g., spagram = svargam, Shahbazgarhl VI. 16).
6. Final -as in Sanskrit is represented by -o at Shahbazgarhl, but at Mansehra
generally by -e (see later remarks).
7. The locative sing, ends in -e or -spi (<* -smin; cf. item 5).
8. The future tense affix is -isa- (< -isya- [see item 5]; e.g., vadhisati = vardhisyati,
Shahbazgarhl IV.9).
The dialects of the two northwestern sets of rock edicts diverge in several respects,
notably in the treatment of final -as noted earlier (item 6). Thus as the equivalent of
Sanskrit rajnah (genitive), Shahbazgarhl has rano but Mansehra rajine, reminiscent
of the eastern form lajine. Some dialectal peculiarities of Mansehra have thus some-
times been dismissed as "Magadhisms," but this may not be the case; later Gandharl/
KharosthI inscriptions also vary between -o and -e (CH 2.1, cxii-cxiii), and recent
research into the historical grammar of Gandharl13 has shown that the notation of
final vowels is highly inconsistent in Gandharl generally.14
All in all, the Asokan inscriptions give a broad view of the dialect spectrum of
MIA vernaculars in the third century B.C. But it must also be understood that they do
not provide anything like a real dialect map of the time. For the geographical distri-
bution of the dialects—especially of the eastern dialect—can hardly correspond with
linguistic reality; the eastern dialect was obviously not the mother tongue of resi-
dents of the far north and the central south, though it was used for inscriptions (Kalsl,
Erragudi, etc.) in those regions. Moreover, the languages as they are presented in the
inscriptions are surely not exact renditions of the contemporary vernaculars. It has
been suggested,15 for example, that such features as the retention of consonantal clus-

11. H. W. Bailey (BSOAS 11, 1946, 774) suggests the pronunciations its I or Its'/.
12. This metathesis was formerly thought by some (notably Senart; IA 21, 1892, 148-50) to rep-
resent merely an orthographic peculiarity of Asokan KharosthI, as seems to be the case with the re-
versed conjuncts noted in the Girnar BrahmT inscriptions (see n. 9). But it has been pointed out by
Grierson (JRAS 1913, 682-3) and others since him (see discussion and references in von Hiniiber, Das
dltere Mittelindisch, 28-9) that such metatheses are characteristic of the Dardic languages of the north-
west to the present day, and that such forms in Asokan (and in later Gandharl/KharosthI) inscriptions
therefore must reflect actual dialect features.
13. Notably G. Fussman, "Gandharl ecrite, Gandharl parlee," 473 and 479-80.
14. See also the comments of Brough in GD 115.
15. See for example Senart, IA 21, 1892, 146-8.
76 Indian Epigraphy

ters reflect orthographic conservatism and the influence of Sanskrit, and that the actual
stage of phonetic development of MIA in Asoka's time could have been further ad-
vanced than it would seem from the inscriptions.

3.1.3 Other inscriptions of the Mauryan era

The scanty epigraphic remains of the Mauryan period other than the records of Asoka
(see [Link]) are all from the northeast, and most of them exhibit the typical Ardha-
Magadhl features of the "Chancellery Language" (-e for -as, I for r, s for all sibi-
lants, etc.). Some of these inscriptions, notably Mahasthan (SI 1.79-80) and Sohgaura
(SI 1.82-3), seem to reflect a more colloquial and/or archaic variety of this dialect.
The Jogimara cave inscription (ASIAR 1903-4, 128-31) is the only one from
any period preserving what seems to be pure Magadhi, showing the sibilant s instead
of s (e.g., sutanuka. . . devadasikyi - sutanuka devadasi) besides the aforementioned
characteristic features of Ardha-Magadhl.16

3.1.4 Later inscriptional Prakrits

[Link] Inscriptions in the west-midland dialect
After the Mauryan period there is a major shift in the linguistic features of the in-
scriptional Prakrits. The predominance of the eastern dialect of the Asokan and other
inscriptions of the Mauryan period ends abruptly; in fact, not a single inscriptional
record in eastern dialect has been found from the post-Mauryan era. The dominant
role in all regions except the northwest and Sri Lanka falls hereafter to a variety of
Prakrit which most resembles, among the Asokan dialects, the western dialect of the
Girnar rock edicts, and which among literary languages has the most in common with
Pali and archaic forms of Saurasenl. In other words, this dialect partakes of the typi-
cal characteristics of the western and central MIA languages: nominative singular
masculine in -o, retention of Sanskrit r and /, predominance of the sibilant s, and so
on.17 Like the Asokan Prakrits, this central-western epigraphic Prakrit is still rela-
tively archaic, with only occasional intervocalic voicing of unvoiced stops and eli-
sion of voiced stops. But unlike some of the Asokan inscriptions, consonant groups
from Sanskrit are nearly always assimilated.18
The causes of the abrupt dialectal shift from east to west undoubtedly lie in po-
litical and historical developments, that is, the decline of Magadha as the center of
power in northern India after the collapse of the Mauryan empire and the movement

16. The NagarjunI hill inscriptions of Asoka's grandson Dasaratha (SI 1.77-8) have in some
places the sibilant s, e.g., in the king's name dasalatha, and in others s. But this variation may not
be dialectally significant; cf. the similar cases in Kalsl and some other Asokan inscriptions noted
previously ([Link]. n. 5).
17. For a specimen of this common epigraphic Prakrit, see Appendix, no. 3 (Bharhut label
inscriptions).
18. An apparent exception is the Besnagar pillar inscription of Heliodoros (Appendix, no. 2), which
contains conjuncts such as tr (putrena, 1. A-3) and dr (bhagabhadrasa, 1. A-6). But these examples
would seem to fall into the category of early orthographic Sanskritisms (see [Link]) rather than of
phonetic archaisms.
 The Languages of Indie Inscriptions 77

of the center of political power in the following centuries toward the west and north-
west.19 Like the eastern dialect under Asoka, the central-western dialect of the post-
Mauryan era was used far beyond what must have been its original homeland. Thus
we find inscriptions in this standard epigraphic Prakrit as far afield as Orissa in the
east, for instance, in the Hathlgumpha inscription (SI 1.213-21), while in the south it
is abundantly attested in inscriptions from such sites as Nagarjunakonda and AmaravatT.
This central-western MIA dialect was, in fact, virtually the sole language in epigraphic
use in the period in question, and therefore seems, like Pali, to have developed into
something like a northern Indian lingua franca, at least for epigraphic purposes, in
the last two centuries B.C.
This is not to say that the inscriptions in this dialect, which Senart called "Monu-
mental Prakrit,"20 are totally devoid of local variations. On the contrary, we do oc-
casionally find, in the southern inscriptions, for example, instances of the influence
of the local (Dravidian) language, as in ayira-haghana = arya-sanghanam at
Nagarjunakonda (AyaA:a-pillar ins. C-2; SI 1,232,1.10). There are also considerable
variations in orthography, with inscriptions from the later part of the period in ques-
tion, that is, those of about the third and fourth centuries A.D., showing an increasing
precision, most notably in indicating geminate consonants which had hitherto almost
always been represented by the corresponding single consonant (see examples cited
in SIE 43-4). Stylistically too, there is a considerable range of variation, from the
brief and unadorned style of the donative inscriptions from Buddhist stupa sites such
as Sand and Bharhut to the quasi-literary compositions of the Hathlgumpha inscrip-
tion of King Kharavela or the Nasik inscription of the time of Vasisthiputra Pulumavi
(SI 1.203-7). But all in all, the standard epigraphic or "Monumental" Prakrit can be
treated as essentially a single language whose use spread far beyond its place of ori-
gin, and which should not be taken to represent the local vernacular of every region
and period where it appears.21

[Link] Inscriptions in other MIA dialects

[Link].1 Gandhan
Grammatical descriptions: S. Konow, CII 2.1, xcv-cxv; J. Brough, GD 55-118;
G. Fussman, "GandharT ecrite, Gandhan parlee."
In the northwest and adjoining regions, inscriptions of the post-Mauryan era (see 4.3.2)
continued to be written in essentially the same language as the Asokan rock edicts
from this area, that is, the MIA dialect which used to be referred to as "Northwestern

19. See the comments of Mehendale, HGIP, xxxvii.

20. IA 21, 1892, 258.
21. For this reason, the methods of analysis followed by Mehendale in HGIP are misleading. His
classification of Prakrit inscriptions of the post-Mauryan epoch into geographical groups (western,
southern, central, and eastern) is valid only with reference to their findspots; but to trace separate lin-
guistic developments within and influences among the various groups, as Mehendale does, is based on
the misconception that each of them constitutes a distinct linguistic entity. As Senart showed in Les
Inscriptions de Piyadassi, this is not the case, and most of the conclusions drawn by Mehendale con-
cerning the trends and directions of phonetic and morphological change cannot be accepted. A defini-
tive historical linguistic treatment of the post-Asokan epigraphic Prakrit, bearing in mind the compli-
78 Indian Epigraphy

Prakrit" but which is nowadays more commonly called "Gandharl," following

H. W. Bailey's suggestion.22 The inscriptions in this language are invariably written
in KharosthI script and are virtually all Buddhist in content. They range in date from
approximately the second century B.C. to the third or possibly fourth century A.D. (see
2.3.2), and their findspots extend from Afghanistan in the northwest to Mathura in
the southeast, with stray finds farther to the east and north (2.3.1).23
Inscriptional Gandharl preserves most of the characteristic features already men-
tioned in connection with its early form seen in the northwestern Asokan dialect
([Link]), such as the retention of the three Sanskrit sibilants and the tendency to
preserve certain consonant groups, especially those with r. It also exhibits several
further developments, notably the following:
1. Fricative pronunciations of certain intervocalic consonants, sometimes
marked by a diacritic sign (cf. 2.3.8) resembling the subscript r, as in the
Wardak inscription (CII 2.1, 170,1. 1), bhag(r)avada = bhagavatah.
2. A voiced sibilant (Izl) derived from intervocalic s, dha, etc., indicated by
various spellings such as majhe < mase (Manikiala stone ins., CII 2.1, 149,
1. 1) and bosisatva < bodhisattva (Taxila silver scroll, CII 2.1, 77,1. 3).
3. In general, a more advanced stage of development with respect to the voic-
ing and elision of intervocalic consonants, as compared to the contemporary
BrahmT inscriptions (cf. [Link]); see examples cited by Fussman, "GandharT
ecrite, Gandharl parlee," 457.
4. A weakening of the distinction between Sanskrit n and n, reflected in ortho-
graphic inconsistency in the use of the two characters (CII 2.1, cii-civ).
5. A weakening of the distinction between aspirate and nonaspirate consonants,
perhaps attributable to the influence of the neighboring non-Indie languages
(CII2.1, ci-cii).
6. Extreme inconsistency in the notation of final vowels, which according to Fuss-
man (473) indicates that they were largely neutralized in pronunciation. This is
especially notable in the nominative singular masculine, for which the endings
-o, -e, and -a are attested in the same phrase (examples in Fussman, 459-60).24
7. The occasional use (not surprising in view of the geographical position of
the dialect) of loanwords from non-Indie languages such as Greek (e.g.,

cations introduced by loose standardization on the one hand and the vagaries of scribal usage and local
influences on the other, remains to be written.
22. BSOAS 11, 1943-46, 764. Although Bailey's suggested name has been accepted and come
into general use, it should be remembered that it is a modern creation and has no attestation in tradi-
tional materials (see 2.2.2, on the traditional names of scripts).
23. Until recently, besides the many inscriptions and central Asian secular documents in Gandharl,
only one literary specimen of the language, namely the Dharmapada ms. from Khotan, definitively
published by Brough (GD), was known. But a large collection of fragmentary GandharT mss. on birch
bark scrolls has recently been discovered, confirming, as had already been suspected on the basis of
traces of Gandharl words in Buddhist texts translated into Chinese (see Fussman, "GandharT ecrite,
Gandharl parlee," 442 n. 20), that Gandharl was a major Buddhist literary language. See R. Salomon,
"A Preliminary Survey of Some Early Buddhist Manuscripts Recently Acquired by the British Library,"
JAOS 117, 1997, 353-8, and Ancient Buddhist Scrolls from Gandhara (London/Seattle: British Li-
brary and University of Washington Press, forthcoming).
24. Cf. the earlier comments ([Link]) concerning -o and -e in northwest Asokan. Konow (cxiii)
divided the post-Asokan inscriptions into "an eastern o-dialect and a western e-dialect," but the recent
study of Fussman, based on a much larger corpus, has shown that this distinction does not hold. The
 The Languages of Indie Inscriptions 79

meridarkha = (leptSdpXTic,; Swat ins., CII 2.1, 4) and Saka (erjhuna =

alysanai/e'ysanai; Takht-i-bahl ins., 1. 5, CII 2.1, 62 and xcvii).
The KharosthI documents from central Asia (see [Link]) provide specimens of
a separate dialect of GandharT used far beyond its native region. These documents
show that Gandharl became the official language of the Indianized kingdom of
Shan-shan or Kroraina in the southeastern Tarim basin in about the third century A. D.
(cf. 2.3.1 and 2.3.2). This "Niya Prakrit" or "Kroraina Prakrit," which was analyzed
in detail by T. Burrow in The Language of the Kharosthi Documents from Chinese
Turkestan, agrees on the whole fairly closely with the language of the post- Asokan
GandharT inscriptions from India,25 but also differs from Indian Gandharl in several
significant particulars, such as the following:
1 . Certain phonetic tendencies, such as a confusion between voiced and unvoiced
consonants (more pronounced than in Indian GandharT), and particularly a
tendency to deaspirate Indie aspirates, apparently reflecting the phonetic struc-
ture of the (unattested) local dialect (Burrow, Language, 5 and 9).
2. The loss of distinction between endings for the nominative and accusative end-
ings. The ending is generally -a, in both the singular (<-e according to Burrow,
"The Dialectal Position," 421 and 424) and the plural (presumably for -a).
3. The use of an ending -tu for the second person singular of all tenses of the
verb; for example, labhisatu 'you will get'. According to Burrow (Language,
43), this is "probably taken from the 2nd person of the pronoun."
4. The development of a periphrastic construction for the preterite active, con-
sisting of the past participle plus the personal endings of the present, except
for the third-person singular where the ending is omitted; for example, trithemi
= *drstah + -mi 'I saw'; tadita = *taditah 'he beat' (Language, 50-3).
5. The frequent use of the infinitive in -amnae (also occasionally found in Indian
GandharT; Burrow, Language, 49); for example, karamnae 'to do'.
6. The frequent use of postpositions not found in other MIA dialects, such s&prace
'concerning' (<*pratyayam) and vamti 'to, with' (<upante; Language, 42).
7 . The extensive use of loanwords from various non-Indie languages, especially
from Iranian languages (Language, vii-viii), from the local "Krorainic" lan-
guage which apparently was closely related to Tocharian (Language, viii-ix),
and occasionally from Greek (e.g., milima = |ie5i|ivo<;; Language, 111).

[Link].2 Sinhalese Prakrit

Grammatical descriptions: S. Paranavitana, Inscriptions of Ceylon, [Link]-xlv; Wilhelm
Geiger and D. B. Jayatilaka in Jayatilaka et al., eds., A Dictionary of the Sinhalese
Language (Colombo: The Royal Asiatic Society, Ceylon Branch, 1935), I.I, xxiv-xxvii;
Saddhamangala Karunaratne, EZ 7, 39-44.
The language of the Old Sinhalese or Sinhalese Prakrit inscriptions of about the third
century B.C. to the third or fourth century A.D. (see 4.3.7. 1 ) is characterized by several

entire question of the subdialects of Gandhari remains unanswered, due at least in part to the problems
of the casual orthography which is characteristic of the Kharosthi/Gandhari inscriptions.
25. See also Burrow, "The Dialectal Position of the Niya Prakrit" (see n. 10), 419-35 (esp. 423-
5); and recently, Colette Caillat, "Connections Between Asokan (Shahbazgarhi) and Niya Prakrit?" IIJ
35, 1992, 109-19.
80 Indian Epigraphy
unusual phonological features. Palatal s generally predominates over the other sibi-
lants, but Sanskrit sibilants also are frequently reflected by h. Sanskrit aspirates are
generally represented by the corresponding nonaspirates (e.g., dama = dharma), but
jh regularly replaces./' (e.g., rajha = raja; Paranavitana, xxxi; Karunaratne, 9). Other-
wise, consonants in intervocalic position are generally preserved as in Sanskrit, though
they are often subjected to voicing. Long vowels are commonly shortened, though
according to some authorities this is merely a graphic device (Geiger and Jayatilaka,
xxv-xxvi; contra, Paranavitana, xxviii).
The morphology of Sinhalese Prakrit is on the whole not much different from
that of the other old inscriptional Prakrits. The nominative singular ends in -e and
the genitive in -sa, -sa, or -ha.26 In these respects, as in others such as the predomi-
nance of the palatal sibilants, the dialect appears to have been influenced by the
MagadhI of the Buddhist missionaries sent to Sri Lanka from northeastern India in
the Mauryan period (Geiger and Jayatilaka, op. cit., xx).

3.1.5 Literary Middle Indo-Atyan in inscriptions

[Link] The literary Prakrits
The standardized literary Prakrits (mainly Maharastrl, sometimes also SaurasenI and
MagadhI) were occasionally employed in inscriptions of the medieval era,27 either
in independent Prakrit inscriptions or within Sanskrit inscriptions. The outstanding
example of an independent literary Prakrit inscription is an inscription of the elev-
enth century A.D. at Dhar consisting of two "centuries" (actually 109 verses each) of
Maharastrl verses in praise of the tortoise avatara (avanikurmmasatam) (El 8,241-60;
cf. 4.1.8). Not surprisingly, Prakrit was also sometimes used for Jaina inscriptions,
most notably the Ghatayala inscription (JRAS 1895, 513-21), which is composed
almost entirely of Arya verses in Maharastrl.
The literary Prakrits also occur in the inscriptional dramas (see 4.1.8 and [Link].2),
where they play their usual role as the language spoken by the lesser characters; thus
we find specimens of Maharastrl and SaurasenI in the Parijatamanjan-natika at
Dhar and the Harakeli-nataka at Ajmer, and of these dialects plus MagadhI in the
Lalitavigraharaja-nataka, also from Ajmer. Occasionally a Prakrit verse is also in-
cluded in an ordinary Sanskrit inscription, for instance, verse 12 (in Maharastrl) of
the VaranasI copper plate inscription of Kama (SI II.340).28

[Link] Pali
Inscriptions in canonical Pali29 from India proper are relatively rare. Those examples
which do occur are usually Buddhist inscriptions of cultic content (cf. 4.1.7), such as

26. Due to the brevity and formulaic character of the Sinhalese Prakrit inscriptions, there is al-
most no material for the study of finite verb morphology; such few verb forms as do appear are nearly
all participial.
27. This is of course a purely literary phenomenon, historically unrelated to the use of the MIA
vernaculars in early inscriptions.
28. For other examples of literary Prakrit together with other languages, see section 3.6.
29. It should be noted that in some early (and even some more recent) epigraphic publications the
term "Pali" has been inaccurately used to refer to various other MIA dialects.
 The Languages of Indie Inscriptions 81

the two Sarnath stone inscriptions (El 9, 291-3) recording the "Four Noble Truths"
and the "Buddhist creed." Other Indian inscriptions in Pali or closely related dialects
include the Pratltya-samutpada formulae on the Devnlmori casket and in a Ratnagiri
inscription.30
Unlike in India, Pali inscriptions are found in large numbers in some Buddhist
countries of Southeast Asia, especially Burma and Thailand, and also in smaller
numbers in Sri Lanka and Cambodia (see [Link] and [Link]-[Link]).

3.2 Mixed or "Hybrid" Dialects

3.2.1 The character of "Epigraphical Hybrid Sanskrit" (EHS)
From about the first to the fourth century of the Christian era, a large number of
inscriptions from northern, central, and occasionally also southern India were written
in a peculiar language which is neither fully Sanskrit nor fully Prakrit but partakes of
the characteristics of both. This language has accordingly been referred to by such terms
as "mixed dialect," "Sanskrit influenced by Prakrit," or "Prakrit influenced by San-
skrit." In 1978, Th. Damsteegt coined the term "Epigraphical Hybrid Sanskrit" (EHS)
in his book of the same name, on the analogy of Franklin Edgerton's "Buddhist Hybrid
Sanskrit"; and although, like its prototype, the term is potentially misleading in some
cases, it deserves to be, and largely has been, adopted for general use.
EHS texts are typically more or less Sanskritic in orthography but Prakritic in
morphology and syntax. Thus a typical EHS word occurring frequently in the in-
scriptions is bhiksusya 'of (the) monk', corresponding to Sanskrit bhiksoh and MIA
bhikkhus(s)a. Here the stem is spelled as in Sanskrit, while the inflection is a pseudo-
Sanskrit rendition of the Middle Indie -ssa. Similarly, the standard EHS dating for-
mula is etaye purvaye or variants thereof, equivalent to Sanskrit etasyam purvayam
'on the aforementioned [date (tithau)]'.
An example of a typical EHS text,31 in both content and form, is the Kankall Tlla
(Mathura) torana inscription (Liiders, MI §20 [Liiders' tr.]):
bhadata-jayasenasya amtevasinlye dhamaghosaye dan[o] pasado
"The temple (is) the gift of Dhamaghosa (Dharmaghosa), the female disciple of
the venerable Jayasena."
Notable here, among other characteristic features, is the juxtaposition of the San-
skrit genitive masculine ending -sya with the Prakrit feminines in -dye and -rye; the
semi-Sanskritized stem of the proper name dharmaghosa- (according to Liiders'
interpretation) contrasting with the un-Sanskritized stem of pasado (Skt. prasada-);
and the spellings, influenced by MIA orthography, of bhadata- (for bhadanta) and
amtevasinlye (for amte-).
It must be emphasized that EHS, like the epigraphic Prakrits, is by no means a
standardized or a unified language. The orthography and morphology of EHS in-

30. See O. von Hiniiber, "Epigraphical Varieties of Continental Pali from Devnimori and Ratnagiri,"
in Buddhism and Its Relation to Other Religions [for full reference see Index of Inscriptions Cited s.v.
DevnTmorT], 185-200 for these and for references to other Pali inscriptions.
31. For a longer specimen of EHS, see Appendix, no. 5 (Sarnath umbrella shaft ins.).
82 Indian Epigraphy
scriptions vary widely and unpredictably, and they display various grades or degrees
of hybridism. In general, one may think in terms of more Prakritic varieties ("Prakrit
influenced by Sanskrit," in the terminology used by Sircar in SI I) versus more highly
Sanskritized varieties ("Sanskrit influenced by Prakrit").32 But it must be understood
that these divisions, like any categorization of EHS, are inevitably somewhat arbi-
trary. In actual practice, the boundaries between the two main types of EHS, and
indeed between EHS as a whole and MIA on the one side and Sanskrit on the other,
are far from clear.33 It is probably impossible to establish fully objective criteria for
EHS, particularly for the more Sanskritized varieties where there is no clear divid-
ing line between Sanskritic EHS and informal epigraphic varieties of standard San-
skrit.34 It may therefore be more appropriate to think of EHS in terms of a broad
spectrum of partial Sanskritization, verging into pure MIA at one end and standard
Sanskrit at the other.

3.2.2 Geographical and chronological distribution of EHS

The city of Mathura and its environs are the source of the earliest and largest number
of EHS inscriptions. Liiders' Mathura Inscriptions (MI) thus constitutes the largest
and most convenient corpus for the study of these documents. But from around the
late first century A.D., many other inscriptions in EHS have been discovered over a
wide range of India, particularly from the west, at sites such as Andhau in Gujarat
and Nasik in Maharashtra; from north central and northeastern India, at Sarnath,
Kosam, and so on; from central India, at Sand; and even from the south, at Nagar-
junakonda.35 Throughout the period in question, EHS coexisted as an epigraphic
language with Prakrit and/or Sanskrit, but it was definitely the predominant language
overall for the first three centuries of the Christian era in northern and central India.
Although individual records of similar date may vary widely in their degree of
hybridism, EHS does follow an overall pattern of development toward greater
Sanskritization over the first four centuries of the Christian era. Thus from about the
third century A.D. on, the standard epigraphic dialect ranges from moderately to highly
Sanskritized EHS, with some specimens in completely or nearly standard Sanskrit.
This variety is characteristic, for instance, of the later inscriptions of the Western
Ksatrapas and related dynasties of the third century.
Overall, the pattern of distribution of EHS inscriptions gives the impression that
they radiate out from Mathura toward the northeast and southwest. Thus the origins
and development of EHS seem to parallel the foundation and spread of the kingdoms
of the Scythian and Kusana rulers in the heartland of India (see [Link]).

32. Damsteegt (EHS 143) similarly divides EHS inscriptions into "a) those basically in MIA, but
with Sanskrit features; [and] b) those basically in Sanskrit, but with MIA features."
33. Note that Damsteegt devotes an entire section of his study (2.1, "Criteria") to a consideration
of marginal cases of EHS.
34. See, e.g., MI §67; and for a highly Prakritic specimen, §137.
35. Damsteegt (156) also includes certain southern Indian copper plate inscriptions such as
Penugonda, Mattepad, and Basim as examples of EHS; I would not agree with this classification, and
treat them (3.3.4) as bilingual (Prakrit and Sanskrit) records.
 The Languages of Indie Inscriptions 83

3.2.3 The linguistic nature ofEHS

[Link] Relation to Buddhist Hybrid Sanskrit
The epigraphic mixed dialects bear a clear and close linguistic relation to the idio-
syncratic Sanskrit of Buddhist texts generally and particularly to the true "Buddhist
Hybrid Sanskrit" (BHS) of texts such as the Mahavastu. Although the epigraphic
and literary hybrids do differ in particulars, this is at least in part due to the different
character and content of the documents preserved in them. In any case, the variabil-
ity of both of these unstandardized languages presents major obstacles to a detailed
comparison. But despite these complications, there can be no doubt that the two rep-
resent different manifestations of the same basic linguistic phenomenon, namely, the
gradual Sanskritization of MIA vernaculars. It is also certain that the hybrid language
used in the first three centuries of the Christian era was neither an exclusively Bud-
dhist nor an exclusively literary construction, as seems to be assumed by certain schol-
ars (notably Edgerton) who did not sufficiently consider the epigraphic evidence.
For among the hybrid inscriptions of this period, we have documents of not only
Buddhist but also Brahmanical and Jaina content which are original compositions in
the hybrid dialect.
Thus the epigraphic evidence contradicts the notion espoused by Edgerton and
others that the specimens of hybrid dialects that we have are merely incomplete or
imperfect translations into Sanskrit of texts originally written in MIA dialects; in other
words, that the hybrid language is merely a literary artifact and not a real language.
The inscriptions prove that hybrid Sanskrit was, or at least eventually became, a liv-
ing linguistic entity, and thus support the position of Lamotte (HBI 638) and others
that some Buddhist texts were actually composed in BHS.

[Link] The origins ofEHS

This still leaves unanswered the question of the historical origins of the epigraphic
(and Buddhist) hybrid dialect. In general, opinions on the controversial question of
the motivations and mechanics of the semi-Sanskritization of MIA fall into two camps.
The first is the school which holds that the hybrid is little more than a bad imitation
of Sanskrit, that is, that its incomplete and imperfect Sanskritization reflects incom-
petence and ignorance on the part of the composers of the documents. Thus T. Bur-
row says that "such a mongrel language was actually employed by those who wished
to employ the superior Sanskrit language but were not able to master its grammar."36
K. R. Norman elaborates this view: "They [the donors] would presumably dictate in
Pkt, and once it had become fashionable to write inscriptions in Skt, the scribes would
'translate', to the best of their ability, into that language. The correctness of the Skt
depended, therefore, on the scribes' competence."37

36. The Sanskrit Language, The Great Languages (London: Faber and Faber, 1955), 61.
37. Review of Damsteegt, EHS, in Lingua 48, 1979, 293. A similar position was maintained by
Senart in his important but by now largely outdated discussion in Les Inscriptions de Piyadassi of "Mixed
Sanskrit," which he saw as essentially an orthographic phenomenon with no real linguistic basis: "Mixed
Sanskrit is only a manner of writing Prakrit, consisting in going as near as possible to the orthography
and the etymological forms known to the religious language [i.e., Sanskrit]" (IA 21, 1892, 275).
84 Indian Epigraphy
The other school of thought tends to view the degree of hybridism in a given
document as a matter not so much of knowledge or ability as of taste; thus Lamotte
describes hybrid Sanskrit as "une langue litteraire ou le dosage du prakrit et du Sanskrit
etait laisse a leur appreciation personelle" (HBI 642; see also 638 and 645). The
epigraphic evidence in particular tends to support the views of this second school.
Especially worthy of note are cases such as the Mora well inscription (MI § 113; see
3.3.2) where a single inscription comprises portions both in hybrid and in more or
less standard Sanskrit, distributed according to function; here, as is usually the case,
the prose portion is in hybrid language while the verse—which is composed for the
occasion, not quoted from another source—is in good Sanskrit. Such cases indicate
that Lamotte was correct in supposing that the use and degree of hybridism were
essentially controlled by the taste and judgment of the composers of the texts, and
that at least some of them could write in standard Sanskrit when they saw fit.
Certain grammatical features of the epigraphic language also support this view.
Particularly interesting in this regard is the sporadic application of Sanskrit sandhi
rules. The "wrong" application of sandhi is one of the most characteristic features of
the epigraphic hybrid, and is often what most clearly distinguishes inscriptions in
highly Sanskritized hybrid from fully standard Sanskrit; a characteristic example is
the label inscription on the Kaniska portrait statue from Mat (Mathura; MI §97), which
reads mahardja rajatiraja devaputro kanisko. Here the MIA nominative masculine
ending -o is used instead of Sanskrit visarga, not only in pausa (kanisko) but also
before k (devaputro k-). This usage is very common in EHS, and of course actually
reflects not "wrong sandhi" but the retention of the old MIA ending in the otherwise
(i.e., orthographically) Sanskritized word. Contrast with this the peculiar sandhi forms
found in later Mathura inscriptions of the posthybrid era, such as sakyabhiksunyar
jayabhattayaryad... in the Katra pedestal inscription (MI §8) and visnusyah gomin-
draputtrasyah ha[ku]datta-p[au]ttrasyah (Mathura Naga statue ins., MI § 161). Both
of these inscriptions date from the Gupta period, when Sanskrit had become the stan-
dard epigraphic language, and in them we clearly do have incompetent attempts to
write Sanskrit; the incorrect sandhi does not reflect underlying MIA forms, but sim-
ply the misapplication of Sanskrit rules improperly learned or understood by the
composer.
Thus the combined evidence of grammar and usage speaks against the assump-
tion that the hybrid language simply represented failed attempts to write Sanskrit and
supports the notion that hybrid was a coherent (though not rigidly standardized) lan-
guage in and of itself, and that those who wrote it did so intentionally. This is not to
say, however, that the hybrids were consciously formulated and developed. Rather,
the available evidence suggests that hybrid Sanskrit arose in the course of a gradual
Sanskritizing movement which had its origins in the late centuries B.C., expanded in
the early centuries of the Christian era, and culminated in the final triumph of clas-
sical Sanskrit in the Gupta era. Early tendencies toward Sanskritization, in the form
of sporadic semi-Sanskritized orthography, appear in some Prakrit inscriptions of
the pre-Christian era. Such "Prakrit influenced by Sanskrit" is seen, for example, in
the Besnagar pillar inscription (Appendix, no. 2) in such spellings as bhagabhadrasa
tratarasa (1. A-6; cf. n. 18), and in the Pabhosa cave inscription (SI 1.96), raj no
gopaliputrasa (1. 1). The hybrids would seem to reflect the extension and regulariza-
 The Languages of Indie Inscriptions 85
tion of these early tendencies toward Sanskritization. In other words, it seems that
the semi-Sanskritized Prakrits of the early Christian era gradually attained a de facto
status of their own as semiliterary languages; in Jean Filliozat's words, it was not
merely a matter of "seulement sanskritisation de textes, mais encore sanskritisation
de dialectes."38
Thus it is more likely that scribes and authors wrote in such a dialect not because
they were unable to write in "pure" Sanskrit (or "pure" Prakrit) but simply because
this had become the prevalent and preferred style of their time and place. The lin-
guistic borders, in other words, between "Sanskrit" and "Prakrit" were probably not
as strictly fixed in practical usage as they appear to be when the matter is viewed (as
it usually is) from the point of view of formal literature. From the less formal Bud-
dhist texts, and especially from the epigraphic remains of the period in question, we
instead get the impression that a variety of dialects, or perhaps rather of dialectal styles,
covering a broad spectrum from pure MIA to pure standard Sanskrit were available
for varying purposes and contexts. The choice of a given dialect by a given writer
was, to be sure, governed to some extent by his knowledge and level of education;
but to an equal, and perhaps greater, extent, it was the content and nature of the docu-
ment he was writing which would determine the appropriate level of Sanskritization.
Due to the limitations of the materials, it is not possible to specify with any pre-
cision the MIA dialect or dialects underlying the hybrid language. But both the geo-
graphical concentrations of the inscriptions in the Mathura area and its predominant
grammatical features such as the nominative masculine in -o indicate that the epi-
graphic dialect, like the Buddhist literary hybrid, reflects an underlying midland dialect
(HBI645).
Attempts to explain the motivations and linguistic forces shaping the development
of the hybrids tend to focus on two main concepts. First, there is the idea, promoted by
Joseph Mansion39 and Filliozat,40 that they essentially arose as a lingua franca enabling
Buddhist monks from various regions of India to converse easily at a period in history
when their local MIA dialects were beginning to diverge to the point where they were
no longer easily mutually intelligible. This assumes, first of all, that the hybrid remains
reflect an actual spoken language,41 which would be difficult to prove. Other elements
of the theory too are less than totally convincing. For one thing, it is questionable whether
the MIA dialects of the time were really so different; from the available literary and
inscriptional data, it would appear that they were not yet so widely divergent as to present
major difficulties of communication. Moreover, while it is true that the literary remains
of the hybrid language are virtually all Buddhist, the inscriptional data, as we have seen,
include hybrid documents of all three major religious traditions, that is, Buddhist,
Brahmanical, and Jaina; the statistical predominance of Buddhist records may simply
reflect the predominance of Buddhism itself at the time in question. The development
of hybrid Sanskrit may well have been influenced and promoted by the Buddhists, but
the epigraphic evidence shows that hybridization was actually part of an overall lin-
38. Review of F. Edgerton, Buddhist Hybrid Sanskrit Grammar/Dictionary, T'oung Pao 43, 1954,
147-71 (quoted 168).
39. Esquisse d'une histoire de la langue sanscrite (Paris: Paul Geuthner, 1931), 109.
40. Op. cit., 164-6.
41. Cf. Filliozat, op. cit., 166, describing the hybrid language as "un langage reellement parle."
86 Indian Epigraphy
guistic trend which transcended sectarian divisions. In the words of Lamotte, "le Sanskrit
mixte n'est pas un phenomene strictement bouddhique, mais s'insere dans 1'evolution
generate de la langue indienne" (HBI638).
The second theory of hybridization places more emphasis on the status value of
Sanskrit as the traditional language of the learned elite. According to this view, the
early adoption of Sanskritic spellings and the ever-increasing Sanskritization of the
epigraphic language, and eventually the adoption of classical Sanskrit itself, all re-
flect the irresistible influence of Sanskrit as the language of learning par excellence.
In this connection it has been pointed out by several authors, from Sylvain Levi42 to
Damsteegt (EHS 207-9), that the Sanskritizing trend was accelerated by the influ-
ence of the Scythian and Kusana rulers of northern and western India in the period
concerned. As foreigners, they were evidently inclined to patronize the elite language
in an effort to legitimize their rule and emphasize their own Indianization. Thus it is
no accident that the Sanskritized hybrid came into extensive use at precisely the time
and place that the Kusanas and Saka Ksatrapas were ruling; just as it is no coinci-
dence that the first major literary Sanskrit inscription, Rudradaman's Junagadh rock
inscription (see 3.3.3), was written on behalf of a Saka king who, like his predeces-
sors in Mathura, was trying to appear "more Indian than the Indians."
All in all, the theories based on the status of Sanskrit seem to provide the stron-
gest explanation for the gradual hybridization of the inscriptional Prakrits, and for
the ultimate Sanskritization of Indian epigraphy. The status of Sanskrit was so deeply
ingrained in traditional India that it slowly but surely eroded the precedent estab-
lished by Asoka (presumably for Buddhistic motivations) of using Prakrit for epi-
graphic purposes.

3.3 Sanskrit
3.3.1 The earliest Sanskrit inscriptions
It is, in the words of Louis Renou, "le grand paradoxe linguistique de 1'Inde"43 that
Sanskrit, the linguistic parent of MIA, first appeared in inscriptions much later than
its own descendants. For Sanskrit began to come into epigraphic use only in the first
century B.C., according to the now generally accepted dating (mainly on paleographic
grounds) for the oldest Sanskrit inscriptions, namely, the Ayodhya (SI 1.94-5) and
the Ghosundl (SI 1.90-1) and Hathlbada (El 22, 198-205) stone inscriptions.
The Ayodhya inscription records in two lines of essentially standard Sanskrit the
foundation by the "righteous king" (dharmarajha) Dhana[*deva?J of a structure
(ketanam) in memory of his father Phalgudeva.44 The formation of the compound
dharmarajnd instead of the theoretically correct dharmarajena is of no great conse-

42. "Sur quelques termes employes dans les inscriptions des Ksatrapas," JA, ser. 9, vol. 19, 1902,
95-125 (esp. 117-9).
43. Histoire de la langue sanskrite, Les Langues du Monde 10 (Lyon: IAC, 1956), 84.
44. The paleographic estimate of a date in the first century B.C. is corroborated by the statement
that the donor is the "sixth [in descent] of [i.e., from; cf. SI 1.95 n. 3] General Pusyamitra" (senapateh
pusyamitrasya sasthena), since the Pusyamitra referred to is presumably the founder of the Suiiga dynasty
in ca. 187 B.C.
 The Languages of Indie Inscriptions 87

quence, since compounds of this type are found frequently in less formal registers of
Sanskrit. Also worthy of note is the apparent use of the genitive instead of the abla-
tive inpusyamitrasya sasthena, assuming that the usual interpretation45 as "sixth [in
descent] of [i.e., 'from'] .. ." is correct. But here, too, the intrusion of the stronger
genitive case into the domain of the weaker ablative is not uncommon in informal
Sanskrit usage.
The three inscriptions from HathTbada and one from Ghosundl are all separate ren-
derings of the same text, recording the construction of a structure46 for the worship of
the deities Samkarsana and Vasudeva. The text, which except for the beginning of the
first line can be reconstructed from the four extant fragmentary versions, is again in
essentially "correct" Sanskrit with a few possible exceptions. For instance, the word
accompanying the names samkarsana-vasudevabhyam is spelled bhagavabhyam in-
stead of the expected bhagavadbhyam in both versions (Hathlbada A and Ghosundl)
in which it is preserved; but this may be no more than a scribal error (cf. El 22,201).
Also, it has been suggested (SI 1.91 n. 1) that the donor's personal name, given as
sarvatata, "may be actually Sarvatrata." But this is uncertain, as the king in question is
otherwise unknown, and even if the inscription did record a Prakritic form of the ruler's
name this would be of no great linguistic import, since personal names are frequently
recorded in Prakritic forms in Sanskrit inscriptions.
In conclusion, then, the language of the Hathlbada-GhosundT inscriptions, like
that of Ayodhya, is essentially standard Sanskrit, though with some marginal indica-
tions of informal usage and style.

3.3.2 Early Sanskrit inscriptions from Mathura

Except for these very few examples from the last century before the Christian era,
the earliest Sanskrit inscriptions are found in Mathura, which has yielded several
records of the first and second centuries A.D., that is, the time of the Saka Ksatrapas
and the early Kusanas, which are written in Sanskrit or a dialect very closely approach-
ing it.47 The earliest of the Sanskrit inscriptions from Mathura are probably those of
the time of the Ksatrapa Sodasa, who is dated with reasonable certainty to the early
years of the first century A.D. The most important of these is the Mora well inscrip-
tion discussed earlier ([Link]; MI §113), which seems to record the dedication of a
shrine to five epic heroes. The opening portion of this fragmentary inscription, re-
cording its date, was written in the hybrid dialect (mahaksatrapasa rajuvulasaputrasa
svami...), while the remaining three lines are in Sanskrit, evidently including a verse
in the Bhujangavijrmbhita meter (El 24, 198). But for minor orthographic liberties
of a type often seen in Sanskrit inscriptions (cf. [Link]), such as arcadesam sailam
instead of arcadesan sailan (accusative plural), the latter part of the inscription is
virtually standard Sanskrit.
A similar linguistic situation is observed in another fragmentary Mathura stone
slab inscription of Sodasa's time (MI § 178) whose first two lines appear to be in hybrid
45. Cf. SI 1.95 n. 3.
46. [Link] narayanavataka; the exact sense of the phrase is uncertain.
47. Here, as in the case of the hybrid dialects discussed earlier, it is not always absolutely clear
whether or not a given text should be labeled "Sanskrit."
88 Indian Epigraphy
dialect ("not quite correct Sanskrit," in Liiders' words) as indicated by the nonstand-
ard sandhi mparvatoprasado, 1. 2, but whose third and fourth lines seem to contain
an upajati verse in standard Sanskrit. A third inscription of Sodasa's time (MI § 115)
on a doorjamb from Mora is in standard Sanskrit except for the typical hybrid form
svami(syd) (I. 10). Other undated Sanskrit inscriptions attributable to the same gen-
eral period include a fragment of a stone inscription from Karikall Tlla (possibly a
Jaina inscription) with text in sardulavikridita meter (MI §21) and a fragmentary
Mathura coping-stone inscription in "pre-Kusan characters and composed in pure
Sanskrit" (MI §162), which is probably a Brahmanical dedication.
The Sanskrit inscriptions from the earliest phase at Mathura show certain inter-
esting patterns. First, they are mostly Brahmanical in affiliation. While the Karikall
Tlla inscription may be Jaina (it is too fragmentary to be identified), in any case there
are no Buddhist inscriptions among them. Second, some of them comprise two dia-
lects, with the practical portion (containing the date, etc.) in the hybrid language and
a eulogistic portion in standard Sanskrit verse. Others, such as the Mora doorjamb
inscription, are in standard Sanskrit but show occasional hybrid tinges in morphol-
ogy or orthography.
Moving on to the period of the Great Kusanas (i.e., Kaniska and his successors,
provisionally assumed here for purposes of discussion to have begun ruling in A.D.
78; cf. [Link]), we now find more Mathura inscriptions in reasonably standard San-
skrit, including for the first time some of Buddhistic content. For instance, several of
the pillar base donations from the Jamalpur mound (e.g., MI §47-9 and 53) are in
standard Sanskrit or very close to it; for instance, §48, aya[m kum]bh[ako da]nam
sanghapmkrtan[a]m Bh[ad]raghosa-pramukha(nam). A longer inscription is the
Isapur yupa (MI §94) of the year 24 (= ca. A.D. 102?) of Vasiska, commemorating
the performance of a dvadasaratra sacrifice in proper Sanskrit but for some minor
orthographic variations (grsma- for grisma- [1. 3]; hiatus in -cchandogena istva
[1. 6]). A few decades younger is a pedestal inscription from Tokri Tlla (MI §99)
recording donations, apparently Brahmanical, during the reign of Huviska (the date
is lost) in more or less correct Sanskrit but with some peculiarities such as the appar-
ent gerund drsya (1. 3) and the unclear word nanayat (I. 1).
Overall, a fair number (but still only a fraction of the total corpus) of the inscrip-
tions of Mathura from the early first to about the middle of the second century A.D. are
written in fairly standard Sanskrit. Most of the Sanskrit inscriptions are Brahmanical
in affiliation, and the Sanskrit of Buddhist inscriptions is more prone to hybridisms,
though these are also not totally absent from the Brahmanical ones. In conclusion,
although the hybrid dialect is still predominant, Sanskrit is beginning to establish
itself as an epigraphic language in this era, especially in Brahmanical circles, con-
tinuing the trend which began in the first century B.C.

3.3.3 Sanskrit inscriptions from western India

in the Ksatrapa period
The inscriptions of the earlier house of Western Ksatrapa kings, namely, the Ksaharata
line of Nahapana (middle of the first century A.D.?), are mostly in Prakrit or in hybrid-
ized Prakrit, with the important exception of Nasik inscription no. 10 of Nahapana's
 The Languages of Indie Inscriptions 89
son-in-law Usavadata (SI 1.167-70). The opening portion of this inscription (11. 1-
3), eulogizing Usavadata, is written in a fair approximation of standard Sanskrit with
some hybrid features such as frequent sandhi hiatus (e.g., dharmatmana idam, 1. 3)
and hybrid morphology such as bhojdpayitra (1. 1). The remainder of the inscrip-
tion, recording the actual donations, is in a somewhat more hybridized style. Senart
(El 8, 79) pointed out that about the first half of the eulogistic portion is virtually a
Sanskrit rendition (or, as Senart puts it, a "reproduction in Sanskrit orthography") of
the description of Usavadata in the Prakrit Karle cave inscription of the time of
Nahapana (SI 1.171-2). The linguistic innovation in this inscription presumably re-
flects contemporary developments in Mathura and adjoining regions, though it is not
clear why Sanskrit was used for this inscription only; apparently, this is an instance
of the orthographic options alluded to previously ([Link]).
The Junagadh rock inscription of Rudradaman (SI 1.175-80), the greatest king
of the second Western Ksatrapa line of Castana, was written shortly after A.D. 150
and represents a turning point in the history of epigraphic Sanskrit. This is the first
long inscription recorded entirely in more or less standard Sanskrit, as well as the
first extensive record in the poetic style. Although further specimens of such poetic
prasastis in Sanskrit are not found until the Gupta era, from a stylistic point of view
Rudradaman's inscription is clearly their prototype. But as noted by Kielhorn48 and
others,49 the language of the Junagadh inscription is not pure classical Sanskrit in the
strictest sense of the term. The orthography is inconsistent in the use of anusvara
and visarga and in the notation of double consonants, and the nonclassical retroflex
/ occurs several times (e.g.,pdll, 1.1). Local dialect features are probably reflected in
the lengthened vowels in riirvyajam avajityavajltya (1. 12). VTsad- instead of stan-
dard vimsati- (1. 7) also reflects local dialect pronunciation of the epic variant vimsat-
(cf. slha for simha in the Gunda ins., discussed in the next paragraph). Patina in-
stead of patya (1. 11) in the sense of 'lord' is likewise a common epicism, though
technically incorrect. Other questionable uses from the classical point of view in-
clude anyatra samgramesu instead o f . . . samgramebhyah (1. 10; cf. Kielhorn, El 8,
40 n. 2) and the redundant a garbhat prabhrti (1. 9). The language of the Junagadh
inscription is thus a close approach to high classical Sanskrit, but, like early literary
Sanskrit generally, it shows the influence of the less formal epic-vernacular style in
which some of the grammatical niceties of Paninian/classical Sanskrit were not ob-
served. But the nonstandard features in question are in general not of the same order
as those which characterize the hybrid and hybrid-influenced inscriptions of Mathura.
Thus although we can suspect that the inspiration for using Sanskrit for epigraphic
purposes emanated from Mathura (there are clear historical connections between the
Western Ksatrapas and the Scythian dynasties of Mathura), the source of the San-
skrit of the Junagadh inscription was evidently the preclassical literary style current
among the literati of the day rather than the Sanskritized hybrid of Mathura and ad-
joining regions.
This literary style of Sanskrit was not, however, employed in the inscriptions of
the time of the Western Ksatrapa rulers who succeeded Rudradaman. These inscrip-

48. El 8, 39-40.
49. E.g., Renou, Histoire de la langue sanskrite, 96, refers to "des epismes linguistiques."
90 Indian Epigraphy
tions, which are unofficial records, reflect a less formal style which retains some
characteristics of the hybrid language. A typical example is the Gunda inscription
(SI 1.181-2) of the time of Rudrasimha I (Saka 103 = A.D. 181), which contains
such nonstandard forms as (tri)yuttarasate for tryuttara- and rudrasihasya for
rudrasimhasya (cf. visad- forvimsat- in the Junagadh inscription), as well as the hybrid
sandhi rajno ksatrapasya. Inscriptions of contemporary dynasties in western India
show similar linguistic characteristics. A Satavahana inscription of the time of
VasisthTputra SrI-Satakarni (second century A.D.) from Kanherl (Biihler, IA 12,1883,
272) is in hybridized Sanskrit (e.g., srT-sata[karn]I[s]ya, 1. 1). From a slightly later
period, the Kanakhera inscription (SI 1.186-7) of A.D. 279 is in mostly standard San-
skrit but still shows some features reminiscent of hybrid, such as the causative par-
ticiple khdnapito (1. 6).
Also from the third century A.D., we have several Sanskrit inscriptions on yupas
from Badva (SI 1.91-2) and Barnala (El 26, 118-23) in Rajasthan. Their language
still shows significant hybrid characteristics, most strikingly in the dating formula
krtehi or kritehi for Sanskrit krtaih, i.e., "in [lit. 'by'] Krta [- Vikrama] years." This
follows the familiar pattern of inscriptions from the early centuries of the Christian
era (e.g., the Mora well inscription discussed earlier), with the portions concerning
the date and other mundane information in more Prakritic language. This suggests
that everyday documents were still being written in MIA or mixed dialects at this
time, so that people would habitually employ set phrases like krtehi in recording dates,
even at the head of documents which were to be composed in Sanskrit.50

3.3.4 Early Sanskrit inscriptions from the

Deccan and southern India
In general, Sanskrit began to appear in southern Indian inscriptions somewhat later
than in the north, and also lagged behind in its gradual adoption there as the primary
epigraphic language. The first significant body51 of southern Sanskrit inscriptions is
from Nagarjunakonda, where, in addition to many Prakrit inscriptions, a few in San-
skrit have also been found. Most of these date from the reign of the later Iksvaku
king Ehavala Cantamula, who probably reigned in the late third and/or early fourth
centuries A.D. The earliest dated Sanskrit inscription from Nagarjunakonda, of the
year 11 of king Ehavalasri (= Ehavala Cantamula; El 33, 147-9), records a Brah-

50. The Devnlmorl stone casket inscription (SI 1.519), dated in the year 127 of the "Kathika kings"
(kathikanrpanam), is written entirely in correct classical style. The date of the inscription, however, is
controversial; Sircar (ibid.) interprets it as a year of the Saka era, equivalent to A.D. 205. Others, how-
ever, such as P. Srinivasan (El 37, 67-8), judge the inscription to be considerably later on paleographic
grounds and think that the otherwise unknown "Kathika" era may be equivalent to the Kalacuri-Cedi
era, in which case the date of the inscription would be ca. A.D. 376. The latter opinion is preferable on
linguistic grounds, as the classical style would be quite typical of the early Gupta era but unusual for
the early third century (see also [Link]).
51. An anomalously early inscription in what appears to be hybrid Sanskrit (cf. EHS 144) is the
fragmentary AmaravatT slab inscription of the Satavahana ruler GautamTputra SrT-Yajna Satakarni,
of about the late second century A.D. (JAIH 4,7-8). But here (as also in the case of the Kanherl Satavahana
inscription discussed earlier) the unusually early Sanskritization is probably due to Satavahana contacts
with the Western Ksatrapas.
 The Languages of Indie Inscriptions 91
manical (Saiva) donation in good classical verse (anustubh and sragdhard meters).
A pillar inscription of the sixteenth regnal year of the same king (El 34,17-20), also
of Saiva content, is in Sanskrit prose but with numerous hybrid characteristics such
as the frequent absence of sandhi (e.g.,naptryah mahatalavarasya, 1. 6) and the sup-
pression of the dative in the introductory invocation (namo bhagavate mahadevasya
puppabhaddrasvaminah, 1. 1). A Buddhist inscription of Ehavala's year 24 in San-
skrit prose (El 35,11-13) shows similar linguistic features, including hybrid inflec-
tions such as bharyyaya sresthiriiya. A fragmentary fourth Sanskrit inscription on a
pillar from the same site (El 35,17-18), probably of the same period, records a Bud-
dhist donation in good classical verse.
Thus we have at Nagarjunakonda examples of both standard and hybridized
Sanskrit in both Buddhist and Brahmanical records, and all from a period when Prakrit
inscriptions were also still being written. The determining factor in the linguistic
choice seems to be neither sectarian nor chronological but verse versus prose: stan-
dard or near-standard Sanskrit is used in versified inscriptions, while hybridized
Sanskrit appears in the prose texts. This distinction is reminiscent of similar patterns
in earlier inscriptions from the north, notably the Mora well inscription.
Several of the early specimens of epigraphic Sanskrit from other southern Indian
sites occur in bilingual Sanskrit and Prakrit records. A typical example is the Basim
copper plates of the Vakataka ruler Vindhyasakti II (SI 1.430-5), who ruled around
the middle to late fourth century A.D. In this inscription the introductory genealogi-
cal portion (11. 1-5) is in Sanskrit, while the remainder, that is, the functional portion
of the grant, is in Prakrit.52 Here once again the situation is comparable to that of
some northern inscriptions of an earlier period, such as Nasik no. 10 (see 3.3.3).
Similar patterns emerge in this period in the far south. For example, the older
copper plate inscriptions of the early Pallavas (Mayidavolu and Hirahadagalli, SI
1.457-61 and 461-6), datable to about the fourth century A.D., are in Prakrit, but on
some of them the king's name on the seal is given in Sanskrit. Some slightly later
records, such as the Gunapadeya copper plate of the time of Skandavarman (SI
1.467-9) have imprecatory verses (bahubhir vasudha datta ... , etc.) at the end in
Sanskrit. A further step toward the adoption of Sanskrit is illustrated by the Mattepad
copper plates of King Damodaravarman (El 17,327-30). These come from approxi-
mately the same region and period (late fourth century A.D.) as the Gunapadeya plates
but are written primarily in Sanskrit, with only the portion enumerating the donees
and their shares in Prakrit; and even the Prakrit shows the influence of Sanskritic
orthography, for example, in kasyapanandijjasya amso 1 (1. 13).
This shift from Prakrit to Sanskrit around the latter part of the fourth century A.D. in
southern India is also attested in the inscriptions of several other dynasties (see SIE
44—5 for further references). A particularly clear case appears in two copper plate in-
scriptions of the Salankayanas from Kanukollu (El 31,1-10). The earlier set, dated in
the year 14 of Nandivarman, is Prakrit except for the imprecatory verses at the end (as
in the Gunapadeya inscription mentioned previously), while the later set, issued in the
first year of Nandivarman's grandson Skandavarman, is in Sanskrit. According to Sircar
(SIE 44), these inscriptions are datable to the fourth and fifth centuries, respectively.

52. See Sircar's comments in SI 1.430 n. 2.

92 Indian Epigraphy

Finally, after this transitional period in the fourth and early fifth centuries A.D.,
Prakrit fell out of use completely in southern Indian inscriptions. For the next few
centuries Sanskrit was the sole epigraphic language, until the regional Dravidian
languages began to come into use around the seventh century (see 3.5.1).

3.3.5 Early Sanskrit inscriptions from other regions

With the exception of the very early Ayodhya inscription discussed earlier (3.3.1), there
are few early Sanskrit inscriptions from eastern and northeastern India. Some brief
records of the "Magha" kings of KausambI of the second century A.D. are in Sanskrit
or in a highly Sanskritized hybrid: for example, the Bandhogarh inscriptions nos. 18
and 19 of Vaisravana (and apparently also the fragmentary no. 14 of Sivamagha; El
31,184-6) and the Kosam pillar inscription (El 24,146-8) of Vaisravana. Apart from
such scattered and marginal examples, the earliest true Sanskrit inscription from the
northeast (besides Ayodhya) is probably the Susuniya (West Bengal) rock inscription
(SI 1.351-2), datable to about the middle of the fourth century.
From the far north, a notable set of relatively early Sanskrit inscriptions in more
or less correct classical style are the Jagatpur (Dehradun Dist., UP) asvamedha brick
inscriptions of King SHavarman (SI 1.98-9), dated paleographically to about the third
century A.D.
Outside of India proper, the oldest Sanskrit inscription is probably the V6-canh
(Vietnam) stele inscription, composed in standard Sanskrit prose and verse (vasanta-
tilaka and sardulavikrldita).53 Its date is highly controversial, but some time in the
third century A.D. is the most likely (see [Link]); if this date is correct, this record
would be at least as advanced linguistically and stylistically as the contemporary
epigraphs within India.54 From about the fifth century A.D., Sanskrit inscriptions
become common in various countries outside of India, particularly Cambodia, Viet-
nam, and Nepal.

3.3.6 The emergence of Sanskrit in the Gupta period

It was during the reign of the early Gupta emperors in the fourth century A.D. that
Sanskrit was finally established as the epigraphic language par excellence of
the Indian world. The turning point appears in the inscriptions of Samudragupta
(middle to late fourth century), especially the Allahabad pillar inscription (SI
1.262-8), which, despite a few trivial orthographic irregularities, is often held up
as a model of high classical literary style of the mixed prose and verse (campu)
class. From this point on, all the inscriptions of the Guptas and their neighbors and
feudatories in northern India were written in correct classical Sanskrit; and, as we
have already seen (3.3.4), similar developments followed soon after in southern
India and elsewhere.

53. See Claude Jacques, BEFEO 55, 1969, 117-24.

54. It is interesting to note that, somewhat surprisingly, Prakrit was not used as an epigraphic lan-
guage in Southeast Asia; see Jean Filliozat, BEFEO 55, 1969, 112.
 The Languages of Indie Inscriptions 93
Thus by about the end of the fourth and beginning of the fifth centuries A.D.,
Sanskrit had at last established itself as virtually the sole language for epigraphic use
throughout India. Prakrit, from this time onward, virtually fell out of epigraphic use,
with occasional exceptions for literary effect or sectarian considerations (see 3.1.5).
Sanskrit continued to enjoy its privileged position in the north for many centuries,
until regional NIA and "Islamic" languages (3.4 and 3.5.2) began to appear in in-
scriptions of the medieval period; even then Sanskrit was never completely supplanted,
and has continued to be used sporadically up to modern times. In the south, the re-
gional (i.e., Dravidian) languages made their appearance earlier and more prominently
(3.5.1), but there too not entirely at the cost of Sanskrit, which continued to be used
as an alternative to or in bilingual combination with the Dravidian languages through-
out the ancient and medieval periods.

3.3.7 Summary: Historical and cultural factors in the

development of Sanskrit as an epigraphic language
Near the end of the pre-Christian era, we find a smattering of inscriptions of Brah-
manical content recording religious donations and foundations in standard or nearly
standard Sanskrit, and we may assume that these are isolated survivals of what must
have been an increasingly common practice in this period. About the beginning of
the Christian era, we begin to find more examples of epigraphic Sanskrit among
the abundant inscriptions from Mathura and surrounding regions, and it appears to
be more than coincidence that this development appeared at precisely the time when
this area of northern India came under the domination of the Scythian "Ksatrapa"
rulers (see [Link]). This suspicion is confirmed by the appearance, in the next cen-
tury, of Sanskrit inscriptions in the domains of the early Western Ksatrapas in
Maharashtra and Gujarat, culminating in Rudradaman's Junagadh rock inscription,
the first long epigraphic text in virtually classical language and style.
Thus it appears that the use of Sanskrit for inscriptions was promoted, though not
originated, by the Scythian rulers of northern and western India in the first two centu-
ries of the Christian era. Their motivation in promoting Sanskrit was presumably a desire
to establish themselves as legitimate Indian or at least Indianized rulers and to curry
the favor of the educated Brahmanical elite. In other words, the forces and motivations
behind epigraphic Sanskritization were evidently the same as those which promoted
the development and spread of the hybrid language at the cost of MIA, and indeed these
two developments must have taken place more or less simultaneously. As discussed
earlier, for several centuries there was available to the composers and scribes of in-
scriptions a range of linguistic choices comprising MIA, hybrid, and Sanskrit, from
which the appropriate dialect could be chosen according to such factors as the purpose
and contents of the record, its sectarian affiliations, or simply personal preferences and
abilities. But the direction of movement along this spectrum was consistently toward
Sanskrit, promoted by the previously mentioned legitimizing motivations of the non-
Indian rulers, as well as by the inherent status of this elite language. Thus eventually,
and inevitably, Sanskrit completely supplanted Prakrit and the mixed dialects.
The spread of epigraphic Sanskrit to the south in subsequent centuries can also
be attributed to the influence, direct or indirect, of the Western Ksatrapas. In this
94 Indian Epigraphy
connection it is significant that the earliest southern Indian Sanskrit inscriptions come
from Nagarjunakonda (see 3.3.4), since other inscriptions from the same site attest
to the connections of the Ksatrapas and other western Indian rulers with it; for in-
stance, a Nagarjunakonda memorial pillar inscription of the time of King Rudra-
purusadatta (El 34,20-2) attests to a marital alliance between the Western Ksatrapas
and the Iksvaku rulers of Nagarjunakonda.
The movement toward Sanskrit was thus already well entrenched by the early
years of the Gupta empire, when Sanskrit was adopted as the sole administrative
language for epigraphic and (presumably) other purposes, and when the high classi-
cal style became the standard mode. The Guptas thus merely brought to its logical
conclusion a gradual process which had been going on for the previous four centu-
ries or so. The adoption of Sanskrit by the Guptas is sometimes thought to represent
a Brahmanical revival under their auspices; and while there may be something to
this, it would be a serious oversimplification to picture the triumph of Sanskrit merely
as a victory of the Brahmanical language over the MIA and hybrid dialects associ-
ated with the Buddhists and Jains. It is certainly true that, on the whole, Sanskrit was
first and most frequently employed epigraphically in Brahmanical circles (as in
Ayodhya, Hathibada/Ghosundl, etc.), and that many of the earliest and best speci-
mens of Sanskrit from subsequent sites such as Mathura and Nagarjunakonda are in
Brahmanical records. But several other early Sanskrit inscriptions from these sites
are Buddhist, and possibly also Jaina (Karikall Tlla). Particularly interesting is the
situation at Nagarjunakonda, where the Sanskrit inscriptions seem to be distributed
equally, in terms of number and style, between Buddhist and Brahmanical records.
Thus while the Brahmans and their clients may have led the way in the Sanskritization
of epigraphic language, the Buddhists did not lag far behind and were no doubt also
influential in the process.
These patterns should not, however, be uncritically extrapolated to nonepigraphic
contexts. It should be kept in mind that, in the words of Burrow, "[T]he inscriptional
evidence gives a very one-sided picture of the contemporary linguistic conditions.
. . . Sanskrit was always, even when the use of Prakrit was most flourishing, the pri-
mary literary language of India."55 In other words, the limited and sporadic use of
Sanskrit in inscriptions prior to the Gupta era does not mean that Sanskrit as a lan-
guage of literature, culture, and ritual was in abeyance but simply that inscriptions
were not yet felt to be literary documents worthy of its use. The gradual Sanskritization
of inscriptions reflects, on the one hand, the formalization of inscriptions into a mode
of literary expression, and on the other, the spread of Sanskrit into the administrative
realm, which was fully accomplished under the Guptas and their contemporaries.

3.3.8 Linguistic characteristics of inscriptional Sanskrit

The style and quality of inscriptional Sanskrit varies over a wide range from the fin-
est classical to the near-illiterate. In broad terms, one may discern three levels.
The first level comprises more or less standard classical Sanskrit, found most fre-
quently in inscriptions from the Gupta era down to the early medieval period, particu-

55. The Sanskrit Language (see n. 36 for reference), 59.

 The Languages of Indie Inscriptions 95

larly in records of tiieprasasti class. Examples of the high classical style are very nu-
merous; among the outstanding ones, in terms of both grammatical correctness and
literary polish, may be mentioned the Allahabad inscription, the Aihole inscription of
the time of Pulakesin II (SI 11.443-50), and the Deopara inscription of Vijayasena (SI
II. 115-22).56 But even in inscriptions of this class, it is not uncommon to find occa-
sional lapses from the strictest standards of classical grammar and orthography (cf.
[Link]). Thus in the Allahabad inscription we find / for d in -vyalulitena (1. 8) and
prithivyam for prthivyam (1. 24), and in Deopara mansa for mamsa (1. 8). Some in-
scriptions written in what may broadly be called classical Sanskrit show stronger ten-
dencies toward a colloquial style, as in the Junagadh inscription, whose epicisms and
other nonclassical features have been discussed earlier (3.3.3).
The second level comprises a looser, more vernacular style, characteristic of what
might be termed "functional Sanskrit,"57 frequently found in Sanskrit inscriptions,
particularly of the medieval era. Such documents partake in substantial but varying
degrees of vernacularisms in orthography, vocabulary, syntax, and so on (see
[Link]-[Link] for details). The first part of the STyadonI inscription (El I, 162-79)
is a good example of this variety of epigraphic Sanskrit, whose author(s), in Kielhorn's
words, "were evidently influenced by, and have freely employed words, phrases, and
constructions of, their vernacular" (163). A typical example of this style is the fol-
lowing version of a formulaic malediction (1.6): yo kopipurusahparipanthanakhasra
karoti utpadayati sa pancamahapatakai lipyati ("Any person who causes or insti-
gates obstruction [or] damage58 is guilty of the five great sins").
The third level comprises semiliterate Sanskrit, found in some inscriptions,
typically later copper plate charters, often from eastern India. An example is the
Madras Museum plates (originally from an unknown site in Orissa) of the time of
Narendradhavala (El 28,44-50), whose language, in the words of the editor, "is only
seemingly Sanskrit and is greatly influenced by the local dialect" (45). Here, for
example, the malediction is rendered (11.21-2) assadatam vaparadatam va/yo hareti
vasundhara / visthayam krmi bhuta pitrbhi saha pacyate.59 A comparison of this with
the example cited in the preceding paragraph (the inscriptions are roughly contem-
porary, both dating from around the tenth century) shows that we are indeed dealing
with a different level of linguistic competence; while the author of the SlyadonT in-
scription evidently had a somewhat limited command of Sanskrit, the text of the
Madras plates is, as Sircar states, merely a poor imitation of Sanskrit.
The following summary of the typical peculiarities of inscriptional Sanskrit there-
fore does not take the third class of inscriptions into account, since from the linguis-
tic point of view they are of more interest for the study of the underlying vernacular

56. See also section [Link].

57. Similar types of Sanskrit used for practical functions are also attested in some relatively late
nonepigraphic sources, including manuals for letter writing such as the Lekhapaddhati or for instruc-
tion in spoken Sanskrit such as the Uktivyaktiprakarana.
58. For khasra Kielhorn suggests "compare the Hindi khasar 'damages, loss, injury, fraud'" (165).
59. The intended verse is the familiar svadattam paradattam va yo hareta vasundharam / svavi-
stayam krmir bhutva pitrbhis saha pacyate ("He who would steal land given by himself or by someone
else becomes a worm in his own excrement and rots along with his ancestors"). Cf. section [Link](1),
and see SIE 196 for references and variants.
96 Indian Epigraphy
than of Sanskrit. The features described are characteristic mainly of the second class
of inscriptions, that is, those in informal or "practical" Sanskrit, although, as already
noted, many of them are also to be found in the generally more formal and correct
inscriptions of the first group. This fact is linguistically significant because it indi-
cates an overall consistency within epigraphic Sanskrit; the features enumerated here
are not a mere random accumulation of "errors," but rather establish consistent pat-
terns of usage of nonformal Sanskrit which not rarely slip over into the more formal
register. In other words, epigraphic usage in Sanskrit has a grammar of its own,60 a
sketchy outline of which is given in the following sections; a more detailed study of
the subject remains to be done.

[Link] Orthography and sandhi

R is often written as ri, and sometimes also vice versa. See the example of ri for r in
the Allahabad inscription, noted earlier, and examples of the converse in the Ramtek
stone inscription of the time of Ramacandra (El 25, 7-20), pryatama forpriyatama
(1. 19) and trbhuvana- for tribhuvana- (1. 31).
L is sometimes written for d, mainly in early Sanskrit inscriptions. See the ex-
ample noted earlier (Allahabad).61
Notation of anusvara and nasals is often in violation of strict orthographic rules.
Anusvara is commonly used inpausa in place of in. Nornis, often written form before
sibilants and h, for example, vansa- for vamsa- (Bhitarl ins., SI 1.323,1. 13); mansa
(Deopara ins., cited in 3.3.8). Final -n before consonants is sometimes written asm, for
example, pancedram sthapayitva (Kahaum pillar ins., Appendix, no. 7,1. 11).62
Notation of doubled consonants is often inconsistent. Words such as sattva,
ujjvala, and sattra are very commonly written satva, ujvala, satra, and so on. The
doubling of intervocalic ch is not always observed. The optional doubling of conso-
nants in conjunction with rory (e.g., karttd or karta) is frequently but by no means
regularly applied, even within the same inscription.
In many later Sanskrit inscriptions, especially from northern and eastern India,
the distinction between v and b is not consistently maintained, and often the two are
represented by the same character and not distinguished at all. Similarly, in many
such inscriptions s and s, and sometimes also s, are frequently interchanged. The
distinction between n and n is not always correctly maintained.
Besides these common orthographic irregularities, Sanskrit inscriptions also
occasionally display spellings which reveal nonstandard or vernacularized pronun-
ciations, such as asvoja for dsvayuja in the Mandasor stone inscription of the time of
Naravarman (SI 1.397-8,1. 3).

60. This form of Sanskrit is by no means exclusive to the inscriptional language. Most of its char-
acteristics can also be found in, or at least resemble, those of the less formal literary varieties of the
language, particularly those of the epic and of some strata of Buddhist literature; see R. Salomon,
"Linguistic Variability in Post-Vedic Sanskrit," in C. Caillat, ed., Dialectes dans les litteratures indo-
aryennes, 275-94 (esp. 282-4, "Epigraphical Sanskrit").
61. Further examples given in H. Liiders, "Zur Geschichte des / im Altindischen" in Antidoron:
Festschrift Jacob Wackernagel (Gottingen: Vandenhoeck & Ruprecht, 1923), 294-308 = Philologica
Indica, 546-61.
62. Sec also examples in the Mora well ins., cited earlier (3.3.2).
 The Languages of Indie Inscriptions 97

Sandhi is left unapplied not only in prose (see 3.3.2) but also, not infrequently,
in verse, especially at the juncture between padas.

[Link] Morphology
The formation of causative verbs and their derivatives with the -p- affix where not
called for by Paninian rules is particularly characteristic of epigraphic Sanskrit. This
occurs in both early Sanskrit inscriptions, as in bandhapitas in the Gunda stone ins.
(A.D. 181; SI 1.181-2,1. 5), and later, for example, karsapayato (= karsayato) in the
Dhulev plate (El 30, 4, 1.3). The late (A.D. 1264) Veraval inscription (SI II.402-8),
which presents a good example of the second class of "functional" epigraphic San-
skrit described earlier, contains numerous forms of this type, including several
derivatives of^vrt(varttapanartham, varttapariiyam, varttapayatam, 11. 21, 34, 35),
and lopapayati (1. 42).
Also typical of epigraphic Sanskrit, particularly of the second class mentioned
previously, are nonstandard gerunds formed contrary to the rule of distribution of
the suffixes -tva and -(t)ya to simple and prefixed roots, respectively. A notable case
is the Jodhpur stone inscription of Baiika (SI 11.236-41) with three gerunds of this
type: prahatva, tyajya, and stambhya (11. 15, 17).
A pleonastic -ka suffix is frequently applied to nominal and participial forms,
for example, in -[u]tpannakotpadyamanaka- and karitaka- in the Khoh copper plate
of Sarvanatha (CII 3, 135-9,11. 9,11).63
Other miscellaneous morphological peculiarities are far too numerous to list here.
One example is the frequent formation of compounds ofmahant- with mahat- or other
stem forms, rather than maha-, as the prior member; for example, mahaddyuti in the
Sakralinscription (El 27,27-33,1.6) andmahaddharmma- (also mahantadharmma-\)
in the Slyadoni inscription (El 1, 162-79,11. 8, 20, 29, etc.).

[Link] Syntax
A notable peculiarity of epigraphic syntax is the mixing of active and passive con-
structions, as in prapittracaryena... krtavan in the Parasuramesvara temple inscrip-
tion (El 26,126-7,11.1-2) andra casaca... talidatavya in Slyadoni (El 1,174,1.9).
Like many of the features enumerated in this section, this peculiarity has parallels in
inscriptional Prakrit, for instance, in MI § 150, kutubiniye... pratithapeti, and in some
of the Nagarjunakonda inscriptions, for example, B 5 mahadevi rudradharabhatarika
imam selakhambam.. . patithapitam (SI 1.231,1. 4).

[Link] Vocabulary
Sanskrit inscriptions (like those in other languages) contain a large number of words
which are rarely or, quite often, never attested in nonepigraphic Sanskrit. The ma-
jority of this vocabulary consists of technical terms, typically connected with such
matters as agriculture, weights and measures, coinage and currency, revenues and
taxation, local and territorial administration, and various other official terms and

63. See also the comments of Fleet in CII 3, 69, and of Buhler in El 1, 74 n. 28.
98 Indian Epigraphy
64
titles. Also included in this category are abbreviations of technical and other terms
(see [Link]). The definitive collection of epigraphic terminology, particularly of
technical vocabulary but also including some of the other types mentioned subse-
quently, is D. C. Sircar's Indian Epigraphical Glossary (IEG).
There is also a stock of semitechnical words which are exclusive to or character-
istic of epigraphic Sanskrit. This class includes terms such as purvd in the special
sense of 'the preceding [date]' (cf. 3.2.1) or 'the preceding [prasasti]', and satka
'belonging to' (El 1, 164).65
Inscriptions not infrequently also provide examples of rare or otherwise unattested
words of a poetic or literary rather than technical character; see, for example, the
"rare Sanskrit words" (asviya 'a number of horses', anandathu 'joy', etc.) cited by
Hultzsch from the Motupalli inscription (El 12, 188). Inscriptions thus also consti-
tute an important source for the lexical study of classical Sanskrit, for which see S.
P. Tewari's Contributions of Sanskrit Inscriptions to Lexicography.66
It has traditionally been the practice in epigraphic studies to regard orthographic
and grammatical peculiarities of the type noted previously as mere errors, and to
correct them either in the text itself (usually by adding the "correct" form in paren-
theses) or in the notes. This often leads to a situation where inscriptional texts, particu-
larly those written in the less formal modes, are burdened with an inconvenient number
of notes correcting often trivial variants. Moreover, as pointed out by Ramesh in his
essay "Indian Epigraphy and the Language Medium" (RIE 44-8), it is not only im-
practical but also misleading to indulge in such overcorrection of informal epigraphic
Sanskrit; in his words, such "departures from Paninian rules of grammar, which are
dubbed as inaccuracies by the epigraphists, need not necessarily have appeared un-
grammatical to the composers and contemporaneous readers of those inscriptions and,
on the other hand, may have been accepted as legitimate usages" (45). This position
is supported by the fact that many inscriptions contain what Ramesh (44) calls a "for-
mal" portion, containing genealogical and eulogistic passages written in the high lit-
erary style, as well as an "operative" or technical portion presenting the legal details
of the document formulated in the less formal vernacularized epigraphic Sanskrit
described earlier (3.3.8). Thus Ramesh is no doubt correct in his claim that inscrip-
tions prove that "there was, in the early and medieval periods, a Sanskrit for the com-
mon man, a living Sanskrit as against the literary or classical or Paninian Sanskrit"
(46). The peculiar features of epigraphic Sanskrit, whether one prefers to look upon
them as errors or as legitimate linguistic or stylistic variants, are an important source
of linguistic data which throws light both on the history of the Sanskrit language67 as
a means of functional and official communication (as opposed to a purely literary
vehicle) and on its relations with the various vernaculars with which it coexisted.

64. Not surprisingly, such vocabulary includes many loanwords, borrowed directly or in a pseudo-
Sanskrit form, from the local languages, including MIA, NIA, Dravidian, and others; see, e.g., the case
mentioned in n. 58.
65. See Renou's Histoire de la langue sanskrite (ref. in n. 43), 100 n. 1, for a brief sampling of
such typical epigraphic vocabulary.
66. See also K. Bhattacharya, "Recherches sur le vocabulaire des inscriptions sanskrites du
Cam bodge."
67. E.g., the frequent orthographic interchange of r and ri alluded to earlier ([Link]) establishes
that the modern northern Indian pronunciation of syllabic r as ri goes back at least as far as Gupta times.
 The Languages of Indie Inscriptions 99

3.4 The New Indo-Aryan (NIA) Languages

D. B. Diskalkar, "Inscriptions in Sanskritic Provincial Languages," JOI 6, 1956-57,

129-39; SIE 53-60.
The study of inscriptions in the New Indo-Aryan languages could well be described
as the stepchild of Indian epigraphy. Compared with the attention which has been
paid to Old and Middle Indo-Aryan inscriptions on the one hand, and to Dravidian
on the other, NIA epigraphy is virtually an untouched field. Reliable collections and
studies are lacking for most of the NIA languages (Tulpule' s Pracina Marathi Koriva
Lekha is one of the few important exceptions), and many of the collections and edi-
tions of individual inscriptions are in local publications and/or in the various regional
languages concerned, rendering them difficult of access to the scholarly community
as a whole. Although editions of NIA inscriptions are occasionally published in El
and other major journals, the majority have been merely reported in ARIE, often with
only the vague designation of "local dialect," and never properly published elsewhere.
Because of these problems, it is difficult to provide a comprehensive and reliable
survey of the subject; what is presented here is based on the available published
materials but can hardly be considered complete or authoritative.
The neglect of the NIA inscriptions is attributable to various factors (see Diskalkar,
129). For one thing, none of them can claim any great antiquity. Moreover, they have
on the whole less historical importance than those in the other Indie and Islamic lan-
guages (i.e., Arabic and Persian; see 3.5.2), the majority of them being private records
of a rather humble character. Nevertheless, they are worthy of much more attention
than has until now been accorded to them, as they are potentially excellent sources
of data for such subjects as social history, religion, and historical linguistics.
The earliest clear-cut specimens of NIA languages (notably Marathi and Oriya;
see 3.4.1 and 3.4.2) appear in inscriptions of about the eleventh century. It is often
difficult, however, to specify a precise date for the earliest epigraphic attestation of
a given NIA language, as their emergence is gradual (not unlike the situation with
Sanskrit at a much earlier time). Occasional traces of NIA have been observed within
Sanskrit records somewhat before the eleventh century, and the earliest definite speci-
mens are typically mixed in varying degrees with Sanskrit. Even in later centuries, it
is very common for NIA inscriptions to be bi- or multilingual with Sanskrit, Dravidian,
other NIA languages, or Islamic languages.
The dates of earliest epigraphic attestation for the various NIA languages vary
widely, as do the degrees to which they are developed as epigraphic languages. These
discrepancies are due to various historical and geographical factors, most importantly
the date and extent to which Islamic governments were established in the different
regions (see SIE 53). In those portions of northern and central India in which Islamic
rule was firmly entrenched at an early date, the Islamic languages directly supplanted
Sanskrit as the principal epigraphic medium, whereas the NIA languages developed
in this function in areas where Islamic rule was established only later in the medi-
eval period. This accounts, for example, for the discrepancy between Oriya, which
became one of the most important NIA epigraphic languages, and neighboring Ben-
gali, which was almost negligible in this respect; for Bengal was one of the first parts
of India to come under Islamic rule, while Orissa was among the last.
100 Indian Epigraphy
In some regions—again, most notably in Maharashtra and Orissa—the regional
NIA vernaculars were even elevated to the rank of imperial languages. In such cases
one may suspect the influence of neighboring Dravidian-speaking regions, in which
epigraphic use of the local vernacular, instead of or in addition to Sanskrit, was a
long-established tradition.68 In these regions we find the vernaculars regularly used
for official records of the usual sorts. Elsewhere, NIA inscriptions are predominantly
private in character; especially common are memorial records of various types.

3.4.1 Marat hi
Sam. Go. Tujpule, PracTna Marathi Korlva Lekha ["Old Marathi Inscriptions"];
A. Master, "Some Marathi Inscriptions, A.D. 1060-1300," BSOAS 20, 1957, 417-35;
A. V. Naik, "Inscriptions of the Deccan," 55-8, etc.; V. V. Mirashi, Inscriptions of the
$ilaharas (CII 6); D. B. Diskalkar, JOI 6, 132-7; SIE 53-5.
The epigraphic material in Marathi is the most abundant and best-documented
among the NIA languages. The principal reference sources are Master's brief an-
thology of sixteen inscriptions and Tulpule's larger though by no means exhaus-
tive compilation of seventy-six texts. Diskalkar (132) estimates a total of some three
hundred inscriptions in Marathi, of which about two hundred are from Maharashtra
proper and the rest from neighboring territories, mainly Karnataka and Andhra
Pradesh. Marathi first began to appear in inscriptions around the eleventh century
A.D.,69 and Marathi inscriptions became especially common in the twelfth and thir-
teenth centuries when the language was widely used by the Yadavas of Devagiri
and the Silaharas of northern Konkan (see CII 6) for their inscriptions. Marathi
continued to be used epigraphically in the later medieval period, for instance in
the records of the Adil Shahis, and even into the European period; an interesting
late (A.D. 1803) Marathi inscription is the long history of the Tanjore Marathas
inscribed on the wall of the Brhadlsvara temple in Tanjavur (V. Srinivasachari and
S. Gopalan, Bhomsle Vamsa Caritrd).
Most Marathi inscriptions are of the usual types of the period concerned, that is,
donative or memorial records on stone, and copper plate charters; according to
Diskalkar (132), the former type constitutes about three-quarters of the total. The
majority are written in Devanagarl, but some are in the Modi and Kannada scripts.
As with other NIA language inscriptions, many of the Marathi inscriptions are bilin-
gual. Very common are bilinguals of various types with Sanskrit, for example, San-
skrit copper plate charters with some of the "functional" portions written in Marathi
and exhibiting varying degrees of linguistic admixture between Sanskrit and Marathi
(see Master, 417-8). Many later Marathi inscriptions are bilingual with other lan-
guages such as Kannada, Telugu, or Persian. There are, of course, also monolingual
Marathi inscriptions; the Dive Agar plate, for example (Tulpule, 10-14; Master,
422-3; El 28, 121-4), is an early copper plate grant composed entirely (except for
the date in Sanskrit) in Old Marathi.

68. Cf. the comments of Kunjabihari Tripathi in The Evolution ofOriya Language and Script, v.
69. Occasional Marathi words and usages have also been noted in some inscriptions of the tenth
century, e.g., the Marmuri copper plates (Journal of the Bombay Historical Society 2, 213-4).
 The Languages of Indie Inscriptions 101

3.4.2 Oriya
K. Tripathi, The Evolution of Oriya Language and Script; idem, Pracma Oria
Abhilekha; R. Subrahmanyam, Inscriptions of the Suryavamsi Gajapatis ofOrissa.
After Marathi, Oriya is the most abundantly attested and important among the in-
scriptional NIA languages; Diskalkar (JOI 6, 129) estimates the number of Oriya
inscriptions at 150, and Tripathi presents a selection of 71 Oriya inscriptions from
A.D. 1051 to 1568 in The Evolution of Oriya Language and Script (222ff.).70 These
records are found in various districts of Orissa as well as in neighboring districts of
Andhra Pradesh, Madhya Pradesh, and West Bengal. The early manifestations
of Oriya in inscriptions follow a pattern similar to that of Marathi. Characteristics of
the Oriya language can first be discerned in inscriptional Sanskrit of the tenth cen-
tury A.D., for instance, in the Madras Museum Plates (El 28,44-50).71 Oriya proper
began to appear sporadically in inscriptions of the eleventh and twelfth centuries,
the earliest specimen being dated A.D. 1051 (Tripathi, The Evolution, 222-4), and
became common in the thirteenth and following centuries. Oriya was used frequently
in the records of the Eastern Gariga kings and of their successors, the Suryavamsi
Gajapatis. Many later inscriptions of local dynasties as well as private records con-
tinued to be written in Oriya into the eighteenth, nineteenth, and even the early twen-
tieth centuries (e.g., ARIE 1951-2, no. A.20).
Most Oriya inscriptions are the usual donative stone records and copper plate
charters. The early ones were written in "GaudI" or "Proto-Bengali" script, which
gradually developed into a distinct Oriya script from the fourteenth century onward.
Oriya inscriptions were also sometimes written in the Devanagarl (e.g., Tripathi,
Evolution, 248-9) or Telugu (ibid., 229-31) scripts. Like the Marathi inscriptions,
many of them are bilingual (or trilingual), but the proportion of inscriptions in Oriya
alone is larger than for Marathi. The multilingual records include inscriptions in Oriya
together with Sanskrit and/or Telugu, Tamil (e.g., Bhubaneswar stone ins., El 32,
229-38), and Hindi (e.g., Baripada Museum stone ins., OHRJ 2, 94-8).

3.4.3 Gujarati
D. B. Diskalkar, Inscriptions ofKathiawad.
Inscriptions in Gujarati from various parts of the modern state of Gujarat, especially
the Kathiawar and Kacch regions, number well in the hundreds; for examples see
Diskalkar, nos. 79,82-3,85,87-8,90, and so on, and ARIE 1969-70, nos. B.22-138.
Gujarati linguistic features appear in some Sanskrit inscriptions of the fourteenth
century (e.g., Diskalkar nos. 36-7, etc.; New Indian Antiquary 1, 587-8), and in-
scriptions written entirely or nearly entirely in Gujarati date from the second half of
the fifteenth century on. The latest Gujarati inscription cited by Diskalkar (no. 193)
is dated Vikrama Samvat 1935 = A.D. 1879.

70. Several Oriya inscriptions are also transcribed (in Devanagan) in SH 5 (nos. 1006, 1119, 1121,
1132-3, 1152, 1161) and 6 (nos. 654, 697, 700-3, 748-9, 778, 895, 903, 908-9, 927, 1089, 1145-65).
71. See [Link] and SIE 58.
102 Indian Epigraphy

The majority of Gujarat! inscriptions are private dedicatory or memorial records.

The former typically record land grants, the foundation or repair of temples or mosques,
or the digging of wells. Memorial inscriptions are very frequently seen on memorial
pillars orpaliyas, typically recording deaths in battle ("hero-stones") or by sahagamana
("raff-stones"). Pilgrims' records are also common (e.g., ARIE1969-70, no. B.24, etc.).
Many of the Gujarati inscriptions are informally or incompetently written.
Gujarati inscriptions are generally written in DevanagarT script or its local vari-
ant, called Boriya, but some are inscribed in script forms similar to modern Gujarati
(NIA 1, 588). Bilingual (and sometimes trilingual) inscriptions, usually with San-
skrit or Persian, are found in some numbers, but the majority of Gujarati inscriptions
are monolingual.

3.4.4 Hindi and related languages and dialects

Due to the linguistic complexities involved and the generally poor documentation of
the materials, the study of inscriptions in Hindi and related languages is particularly
problematic. As a matter of convenience the term "Hindi" is used here in a very loose
sense, as in the other literature on the subject. The documents concerned actually
cover a wide geographical and dialectal range, but since no comprehensive linguis-
tic study of them has been attempted it is in most cases not possible to specify the
dialects concerned, and it is regrettably necessary to resort to vague terms such as
"Hindi" or "Rajasthanl."
Inscriptions in various dialects of Hindi in the relatively strict sense of the term
are found mainly in central and eastern Madhya Pradesh.72 (NIA inscriptions from
Uttar Pradesh are rare, presumably because of the early establishment of Islamic
dynasties there; see Diskalkar, JOI6, 138, and section 3.4.) There seems to be a par-
ticular concentration of Hindi inscriptions in the territory of the former Gwalior State
and adjoining regions; these inscriptions are catalogued and briefly described in
H. N. Dvivedl's Gvaliyara Rajya ke Abhilekha.13 They consist largely of sail- and
other memorial stones, image inscriptions, and other private records from the thir-
teenth century on.74
Hindi inscriptions from the eastern regions come mainly from relatively remote
districts such as Damoh and Bastar. Here we find official stone inscriptions and cop-
per plate charters of various local dynasties dating from the sixteenth and following
centuries, for example, the Dantewara bilingual75 (Sanskrit and Hindi) inscription
(El 12, 242-50).
From Rajasthan and adjoining areas of western Madhya Pradesh we have numer-
ous inscriptions of the fourteenth and following centuries in the local NIA dialects,

72. For inscriptions in Hindi (and Punjabi) from Azerbaijan, see section [Link].
73. The concentration of Hindi inscriptions in this region of central India may, however, be only
apparent, as a result of their being relatively well documented in this book.
74. A Jaina image inscription from Radeb tentatively dated Vikrama (10)78 (= A.D. 1022) and de-
scribed as being in Hindi was noted in the Annual Report of the Archaeological Department of Gwalior
State for Vikrama 1992 (35, ins. no. 39), but this has apparently never been published in detail and
hence cannot be taken as firm evidence of the epigraphic use of Hindi in the eleventh century (see SIE
55 and n. 3).
75. Like several of the NIA bilingual inscriptions, this is a "true" bilingual (see 3.6), with (ap-
proximately) the same text repeated in Sanskrit and Hindi.
 The Languages of Indie Inscriptions 103

which, however (for reasons stated earlier), cannot be specified with any linguistic
precision on the basis of currently available published materials. Most of the mate-
rial concerned is published only in the form of brief summaries in the various num-
bers of ARIE (e.g., 1962-63, nos. B.849-954, passim), wherein the language is usu-
ally cited only as "local dialect" or occasionally as "Rajasthanl."76 The vernacular
inscriptions of Rajasthan are mainly memorial and saff-stones77 and other stone
records concerning the usual matters, that is, grants and donations, temple founda-
tions and renovations, construction of wells, pilgrims' records, and so on. A few of
these records have been published in detail in various journals; see, for example,
several inscriptions in Sanskrit and "Rajasthanl Bhasa" published by L. P. Tessitori
inJPASB,n.s. 12, 1916,92-116.
Of an entirely different character is a unique inscription from Dhar containing a
poetic composition entitled Raiila-vela by a poet named Roda (Bharatiya Vidya 17,
130-46; 19, 116-28), which is datable on paleographic grounds to approximately
the eleventh century. The precise identification of the language of this composition
is problematic, as it appears to imitate characteristics of various contemporary dia-
lects by way of a pastiche. But the underlying dialect seems to be a transitional stage
between late Apabhramsa and early western NIA,78 and hence this curious inscrip-
tion can be considered as an early epigraphic specimen of Hindi in the broad sense
of the term.
Vernacular languages were also used epigraphically in some of the hill-states of
the Himalayan foothills in medieval times. Among these languages (again, loosely
classed here under "Hindi"), CambyalT is the most important as an inscriptional
medium. This language, which shows affiliations with eastern dialects of Punjabi,79
was used in combination with Sanskrit in the copper plate charters of the local kings
of Chamba from the fourteenth to the nineteenth centuries. These inscriptions were
published in a scholarly edition in B. Ch. Chhabra's Antiquities of Chamba State,
Part II, which is thus one of the few authoritative collections of NIA inscriptions.
Unlike most of the "Hindi" inscriptions described in this section, which are written
in the several varieties of Devanagarl script, the Chamba records are in Devasesa, a
local- script intermediate between Sarada and TakarT (Chhabra, 2-3). Some late
(nineteenth-century) inscriptions from Garhwal in the "local dialect" have also been
reported (ARIE 1948-49, nos. A.7-10; SIE 55 n. 8).
As is generally the case with NIA, the inscriptions in Hindi and affiliated lan-
guages are frequently bilingual, especially with Sanskrit as already noted, but also
occasionally with other languages such as Oriya (3.4.2). Also common are inscrip-
tions from central India written in Hindi or Rajasthanl together with Persian or
Urdu.80
76. For further references see Mangllal Vyas' Rajasthana ke Abhilekha and Marvara ke Abhilekha.
77. On these see B. D. Chattopadhyay, "Early Memorial Stones of Rajasthan: A Preliminary
Analysis of Their Inscriptions," in S. Settar and G. D. Sontheimer, eds., Memorial Stones, 139-49.
78. See R. S. McGregor, Hindi Literature from Its Beginnings to the Nineteenth Century, A His-
tory of Indian Literature, vol. 8, fasc. 6 (Wiesbaden: Otto Harrassowitz, 1984), 7-8; cf. Bhayani in
Bharatiya Vidya 17, 132.
79. See B. Ch. Chhabra, Antiquities of Chamba State, Part II, 13.
80. Urdu, though strictly speaking an Indo-Aryan language, is traditionally treated in epigraphic
contexts along with Persian and Arabic as a part of Islamic epigraphy, and hence will be discussed in
that connection in section 3.5.2.
104 Indian Epigraphy
3,4.5 Bengali and other eastern NIA languages
As noted earlier (3.4), the eastern NIA languages, except for Oriya, on the whole
have not been widely used for epigraphic purposes. In Bengali we have several dedi-
catory temple inscriptions,81 mostly from the eighteenth and nineteenth centuries but
dating back in one case to the fifteenth.82 Some of these are bilingual with or mixed
with Sanskrit. There are also a few other stone inscriptions in Bengali (e.g., ARIE
1975-76, no. B.45/D.272, bilingual with Arabic), and some late copperplate records
of the kings of Tripura in the seventeenth and eighteenth centuries (ARIE 1951-52,
nos. A.13-19) are written in Bengali and Sanskrit.
Assamese inscriptions are similarly late and few in number, at least as far as
they are reported in published sources. Some copper plates of the Ahom kings of
the eighteenth century in Sanskrit and Assamese have been noted (ARIE 1957-58,
nos. A.3-5), as well as a few stone inscriptions (ibid., nos. B.65, 74-5).
A seventeenth-century inscription in Maithili has been published in R. K. Choud-
hary's Select Inscriptions of Bihar, 138 (cf. SIE 60).

3.4.6 Nepali
Nepali did not become an important inscriptional language until relatively recently,
the principal epigraphic languages of Nepal in earlier times being Sanskrit and Newari
(see [Link] and [Link]). It is only during the time of the Shah kings in the eigh-
teenth and nineteenth centuries that donative and dedicatory stone inscriptions in
Nepali (often together with Sanskrit) were written in good numbers.83

3.4.7 Sinhalese
Due no doubt to its special geographical and cultural setting, the pattern of the de-
velopment of Sinhalese as an epigraphic language is entirely different from that of
the NIA languages of India proper. Unlike other NIA languages, Sinhalese and its
direct linguistic predecessors were the main languages of inscriptions from early times
in Sri Lanka. According to Geiger and Jayatilaka's formulation,84 the period of the
Sinhalese Prakrit or Old Sinhalese inscriptions (see [Link].2) ended about the third
or fourth century A.D. and was followed by a "Proto-Sinhalese" period from the fourth
to the eighth century, represented by the relatively sparse (see [Link]) epigraphic
remains of that period. The more abundant inscriptions of the medieval period, es-
pecially of the ninth to the thirteenth century, represent the "Medieval Sinhalese"
phase, at which stage, in Geiger's words, "the language has now become a modern
Indian idiom and . . . the principal and most characteristic features of the modern
language are recognisable" (op. cit., xxix).

81. See A. K. Bhattacharyya, A Corpus of Dedicatory Inscriptions from Temples of West Ben-
gal, nos. 56, 58, 62, 74, 96, 106, 115, 116, etc.
82. Bhattacharyya, op. cit., no. S-l, A.D. 1490.
83. See section [Link] for references.
84. In A Dictionary of the Sinhalese Language (see [Link].2), xxiv-xxxi.
 The Languages of Indie Inscriptions 105
Most of the medieval and later Sinhalese inscriptions, some of which are of great
length, are on stone. They generally record foundations, grants, regulations, and other
matters pertaining to Buddhist monastic establishments.85

3.5 Other (Non-Indo-Aryan) Languages in Indian Inscriptions

Although the primary subject matter of this book (see the preface) is restricted to
epigraphic material in the Indo-Aryan languages, it may be useful to briefly summa-
rize the other inscriptional material found in India. Needless to say, the following is
meant only as a very general introduction and has no pretensions to completeness.

3.5.1 Dravidian languages

The epigraphy of the Dravidian languages constitutes an enormous field of study in
itself, which can be presented here only in the broadest outlines, although ideally
Dravidian epigraphy cannot be completely separated from Indo-Aryan epigraphy,
there being a great deal of geographical, historical, linguistic, and stylistic overlap-
ping between the two fields. Although for the most part Dravidian inscriptions are
less ancient than the Indo-Aryan ones, their numbers overall, and especially in the
medieval period and in the Tamil and Kannada languages, are very great, perhaps
greater than those of the Indo-Aryan languages.
Major publication and reference sources for Dravidian inscriptions are mentioned
in section 8.1.2 (esp. South Indian Inscriptions andEpigraphia Carnatica). The com-
plete literature of the field is vast, but there is unfortunately still no single authorita-
tive study of the subject as a whole.86

[Link] Tamil
Inscriptions in what is now generally agreed to be an early form of Tamil date from
approximately the last centuries before and/or the first centuries after the beginning
of the Christian era (see [Link]). There are also a few Tamil inscriptions datable on
paleographic grounds to the second through sixth centuries A.o.,87 but Tamil's emer-
gence as a full-fledged epigraphic language actually began during the reign of the
Pallavas, some of whose copper plate inscriptions from the seventh century onward
were bilingual in Tamil and Sanskrit. This practice was continued by succeeding
dynasties of the Tamil country such as the Colas and Pandyas, but some of their copper
plates (especially in the later centuries) were in Tamil only. The majority by far of
Tamil inscriptions are donative and other stone records, especially on temple walls,
mostly dating from the Cola era and later; these number in the many thousands. As

85. See also section [Link] for further information.

86. Brief surveys of Dravidian inscriptions are given by D. C. Sircar in SIE 46-52 and by D. B.
Diskalkar in JAHRS 21, 1950-52, 163-8.
87. See I. Mahadevan, "Tamil-Brahmi Inscriptions of the Sangam Age" (ref. in [Link] n. 103),
84-5.
106 Indian Epigraphy

with other Dravidian languages, Tamil generally predominates in stone inscriptions

and private records, while Sanskrit tends to be retained for imperial grants on copper
plates.

[Link] Kannada
Next to Tamil, Kannada is the earliest and most important of the Dravidian epigraphic
languages. The earliest Kannada epigraphs, such as the Halmidi (S. Srlkantha Sastri,
Sources ofKarnataka History [for reference see Index of Inscriptions Cited], 20) and
Badami Vaisnava cave (IA 10, 59-60) inscriptions, date from around the late sixth
or early seventh century A.D. From this time onward, inscriptions in Kannada, mainly
private donative records on stone but also some royal copper plate charters, are ex-
tremely common.88 Many of these are bilingual with Sanskrit or occasionally other
languages.

[Link] Telugu
Like Kannada, Telugu first appeared in inscriptions at about the end of the sixth
century A.D. ; the earliest Telugu inscriptions are usually considered to be the Kalamalla
and Erragudipadu inscriptions of the early Telugu Cola dynasty (El 27,220-8). The
subsequent development of Telugu as an epigraphic language also follows similar
lines to those of Kannada, with private dedicatory stone records predominating.
Though numerous, Telugu records are not as abundant as those in Kannada.

[Link] Malayalam
Compared with the other three major Dravidian literary languages, Malayalam is of
much less epigraphic significance, coming into use only around the fifteenth cen-
tury. The number of stone and copper plate inscriptions in Malayalam is thus far
smaller than in the other Dravidian languages.
Among other Dravidian languages, a few late inscriptions in Tulu have also been
reported; for references see SIE 52.

3.5.2 Islamic languages (Arabic, Persian, Urdu)

Z. A. Desai, "Arabic and Persian Inscriptions," AI 9, 1953, 224-32; SIE, 31-8; V. S.
Bendrey, A Study of Muslim Inscriptions.
Inscriptions in the Islamic languages89 first appeared in India in the last decade of
the twelfth century A.D. (Desai 226) and gradually became more numerous, especially
in the sixteenth and seventeenth centuries, and continued into the nineteenth. In the
earlier centuries Arabic was the preferred language, but from the fourteenth century
on Persian became more prevalent, while Urdu came into use only from the mid-

88. Diskalkar (JAHRS 21, 167) estimates that there are twenty-six thousand Kannada inscriptions.
89. "Islamic" is used here as a term of historical and cultural convenience, rather than as a lin-
guistic category; cf. n. 80.
 The Languages of Indie Inscriptions 107

eighteenth century on. Bilinguals among these languages, especially Arabic and Per-
sian, are common, as are bilinguals with Indie languages (e.g., Persian and Sanskrit
or Kannada). In general, Islamic inscriptions are most numerous in northern India,
but they are found in virtually all parts of the subcontinent.
The majority of the inscriptions in Islamic languages are dedicatory in nature,
recording the construction of mosques or other religious buildings, or of secular works
such as wells, gates, forts, and so on. Other common types are endowments, admin-
istrative records, and memorials (tombstones, etc.).
Unlike Indo-Aryan and Dravidian inscriptions, the Islamic inscriptions do not
constitute a major primary source of historical information, not only because they
are of no great antiquity but, more importantly, because (unlike earlier times) exten-
sive historical chronicles are available for the period concerned. Nonetheless, Indo-
Muslim epigraphy is important as a source of corroborative and supplementary his-
torical and cultural data.90

3.5.3 Other non-Indie languages

D. B. Diskalkar, "Inscriptions of Foreign Settlers in India," JIH 34, 1956, 39-52.

[Link] (Non-Islamic) Semitic, Iranian, and other Near Eastern languages

Aramaic and a mixed "Aramaeo-Iranian" are represented in six inscriptions of the
Mauryan era from Pakistan (Taxila) and Afghanistan (see [Link]). Several hundred
brief inscriptions in Sogdian have been found in the Shatial region on the upper
Indus,91 as well as a few others from Ladakh.92 Some inscriptions in Parthian and
Bactrian were also found on the upper Indus.93 Inscriptions in Pahlavi have been found
in western and southern India; the most important examples are the Pahlavi pilgrims'
records in the Kanheri caves (IA 9,265-8 = ICTWI62-6) dating from the early elev-
enth century, and the Christian inscriptions in the churches at Madras and Kottayam
(IA 3, 308-16; West, El 4, 1896-97, 174-6), some of which may be as early as the
seventh or eighth century. A few Christian inscriptions in Syriac have also been found
at Kottayam and other places in South India (e.g., IA 3, 314).
Some medieval inscriptions in Hebrew have been found in southern India, for
example, a tombstone of A.D. 1269 from Chennamangalam (IA 59,134-5).94 A similar
Hebrew inscription of A.D. 1251, together with two others in Himyaritic, were found
in Bhuj (El 19, 300-2), but these are thought to have been brought to India from
South Arabia (301). There are also a few Hebrew graffiti from near Chilas among

90. For examples see Desai, 229-32.

91. Helmut Humbach, "The Sogdian Inscriptions of Thot-Shatinl," Journal of Central Asia 8,1985,
51-7; Nicholas Sims-Williams, "The Sogdian Inscriptions of the Upper Indus: A Preliminary Report,"
in K. Jettmar, ed., Antiquities of Northern Pakistan, 1.131-7; Sims-Williams, Sogdian and Other Ira-
nian Inscriptions of the Upper Indus.
92. N. Sims-Williams, "The Sogdian Inscriptions of Ladakh," in K. Jettmar, Antiquities of North-
ern Pakistan, II. 151-63.
93. Humbach, op. cit., 57. Some other inscriptions of the Kusana period in Bactrian, notably the
Surkh-Kotal (Afghanistan) inscription of Kaniska (JA 246, 345-440), are found outside of India proper.
94. For other examples and references, see Diskalkar, 43-4.
108 Indian Epigraphy
the newly discovered inscriptions on the upper Indus (K. Jettmar, Orientalla losephi
Tucci Memoriae Dicata [for reference see Index of Inscriptions Cited], 11.667-70).
Armenian tombstone inscriptions, mostly of the seventeenth and eighteenth centu-
ries, are found in fair numbers in various parts of India, especially in Bengal, Ma-
dras, and the west coast, for example, the epitaph at the Little Mount, Madras
(El 6, 89).95

[Link] European languages

Although Greek was extensively used in coin legends in northern India in the "for-
eign period," and although from the neighboring region of Afghanistan we have
Asokan and other Greek inscriptions (see [Link]), in India proper only a very few
minor epigraphic specimens of it have been discovered in the form of inscribed pot-
sherds from the Swat region (e.g., Birkot-Ghundai; Journal of Central Asia 7,49-53).
There are also a few tombstone inscriptions from the colonial period in modern Greek
(ARSIE 1912, 86).
Several other European languages are represented in inscriptions of the modern
period, that is, from the sixteenth century onward. The earliest of these are the
numerous Portuguese inscriptions from Goa, Daman and Diu, Bassein, and other
Portuguese settlements.96 Some of the inscriptions of the Portuguese and other
European settlers are composed in Latin. Tombstones and other inscriptions in Dutch
dating from the seventeenth and eighteenth centuries are found in the Dutch settle-
ments in western and southern India (e.g., Chingleput; El 24, 123-6).97 There are
similar though less numerous records in French, mostly from the eighteenth and nine-
teenth centuries. Finally, inscriptions in English are very common all over India from
the seventeenth century on; these include inscriptions on tombs and other memorial
records, administrative orders, foundations, and the like.

[Link] Sino-Tibetan languages and Chinese

Only a few late inscriptions in Sino-Tibetan languages of the border regions of India
have been published. Inscriptions in Manipuri have been compiled by M. Bahadur
and P. G. Singh in Epigraphical Records of Manipur. A few late copper plate in-
scriptions in Ahom are noted in SIE 60. Stone and copper plate inscriptions in Newari
of the Malla period, usually bilingual with Sanskrit, are regularly found in Nepal from
the late fourteenth century A.D. and become very common in the sixteenth to eigh-
teenth centuries (see [Link] for references). Also quite common are ritual and
historical inscriptions in Tibetan found in many parts of the Himalayan regions of
India; see, for example, IA 35, 1906, 237-41, 325-33 (A. H. Francke).
Inscriptions in Sino-Tibetan languages of East and Southeast Asia are also occa-
sionally found in India, generally in the form of pilgrims' records or Buddhist dona-

95. For other examples, see Diskalkar, 45-6, and ARIE 1962-63, nos. D.98-9, and 1966-67, no.
B.192.
96. Sec J. H. da Cunha Rivara, Inscripfdes Lapidares da India Portugeuza (Lisboa: Imprensa
Nacional, 1894). For other references see Diskalkar, 48-9.
97. Sec also ARSIE 1909, 121-4; 1911, 90-1; and 1912, 85-6.
 The Languages of Indie Inscriptions 109

tions. At Bodh-Gaya, for example, inscriptions have been found in Burmese (El 11,
118-20) and Chinese (JRAS, n.s. 13,552-72). Some Chinese inscriptions have also
been discovered in other parts of India, for example, several recently found at Chilas,
Hunza-Haldeikish, and other sites in northern Pakistan.98

3.6. Bilingual and Multilingual Inscriptions

As will be clear from the several examples already noted in this chapter (and as would
have been expected in any case in such a linguistically complex cultural area as
India), inscriptions in two or more languages are common. The majority of these in-
volve Sanskrit and one or more other languages; for example, Sanskrit and Prakrit
(examples cited in 3.3.4) in inscriptions of the transitional period of the fourth and
fifth centuries A.D. Beginning around the sixth century, inscriptions in Sanskrit and
one of the Dravidian languages become common. In the medieval period, Sanskrit is
often combined with one of the Islamic languages, especially Persian, or with one of
the NIA languages.
These Sanskrit bilinguals are for the most part not "true" bilinguals, that is, the
same text repeated in full in two languages, but rather contain a single text divided
on functional grounds between the languages concerned. Typically, the invocatory,
genealogical, and concluding portions will be in Sanskrit, while the "functional"
portions recording the specific details of the gift, transaction, and the like, will be in
the other language (3.3.4). In the medieval period we also find non-Sanskrit bilinguals,
involving two NIA languages or an NIA and a Dravidian language (e.g., the Bhu-
baneswar Oriya-Tamil inscription, cited in 3.4.2). Among this group we do find some
true bilinguals, for example, the Baripara Museum Oriya-Hindi inscription (3.4.2).
Trilingual inscriptions are also not rare. Examples include the Kurgod inscrip-
tion (El 14,265-78) in Sanskrit, Prakrit, and Kannada, and the Ciruvroli copperplates
(El 34, 177-88) in Sanskrit, Telugu, and Oriya.

98. Ma Yong in Jettmar, Antiquities of Northern Pakistan, 1.139-57; Thomas O. Hollman in ibid.,
H.61-75.