origins of Pali

Michael Witzel witzel at FAS.HARVARD.EDU
Sun Jan 16 17:25:24 EST 2000

I am afraid, we have to agree to disagree on many points.
I had no time due to foreign visitors, but today I took a brief look at the
books I brought home from my office and to gain some impression of the
'situation on the ground' from DC Sircar's Selcted Inscriptions.
(Unfortunately I forgot Pischel, Geiger, v. Hinueber, so again only from


Anyhow, even with just Bloch's Asoka inscriptions, with Mehendale's  Asokan
grammar, and DC Sircar's Selected Inscriptions, it quickly becomes clear
that the MIA 'mess' can be sorted out along broad lines. To reach safer
ground, however, instead of translating the Gita (etc.) once again, a team
of scholars should finally set about to edit and  index the ten thousand(s)
of Middle Indo-Aryan inscriptions from all over South Asia ...
As I mentioned last time,  we have to distinguish clearly between Vedic
(with some early MIA influences, such as jyotiS RV, nApita Satapatha Br.),
early *MIA,  Pali as we have it now, Asokan MIA and post-Asokan MIA, plus
the 'classical' Prakrits (and of course MIA influence on Epic, Buddh. Skt.,
 Since we have early and exhaustive materials only for Vedic, Pali, and
post-Asokan MIA, and since Asoka's inscr. take a special place in all of
this, some circumspection, as you also plead, is indeed advisable.

I leave out many of the details now and concentrate on the main points
since we do not seem to agree on them and further discussion is pointless
without such agreement.

The real problem is of course that of Asoka's language, the Magadhan
chancellery style, within the framework of the other really existing MIA
dialects at the time.  And the  transposition of Asoka's edicts into local
You maintain that
>> The point at issue is whether forms such as the nominative in -o are
>>geographically an exclusively western feature. And in fact they are not.
>>The inscription of Khaaravela does not have either the nominative
>>singular in -e nor the use of l for r. It is not alone in
Subsequent inscriptions from Western India in the century after A'soka do
not closely resemble the language of Girnaar. Conversely the inscriptions
from Bhaarhut and Saa~ncii have elements in common with Pali which are not
found in the inscriptions of A'soka. <<

But that puts the horses behind Asoka's chariot:
Rather, we have to look for local items in Asoka's texts which do NOT
conform to his Patna language, not the other way round.

*   1st, we have to describe what Asoka's language is like. Not too
difficult. And post-Asokan inscriptions in the Magadha area confirm his
Magadhi bias:
the famous nom. -e , l for r, etc. etc.  <note, below, and cf. the
Bangladesh inscr.>

* 2nd, we have to see how this language gets changed in other parts of the
subcontinent. The starting point should be Asoka's own language, then the
(partial) deviations from it, i.e. adaptation to local dialects in the NW,
Saurastra etc.,  and then, comparison with the actual local dialect as
evident from the slightly  later inscriptions.
As has long been observed, the Asoka's official Patna style is found from
the Yamuna to South India (summary in C. Caillat's paper in her Dialectes,
Paris 1989: 414 sqq.; cf. again in "Inside the texts", mentioned last time).

Colette Caillat, Dialectes dans les littratures indo- aryennes. Actes du
Colloque International ... 16-18 Septembre 1986.  Paris  (Coll´ge de
France, Institut de Civilisation Indienne) 1989

And, it has always been claimed that the NW, Saurastra
(and now Afghanistan) differ (Caillat p. 416: "negligence of Kalsi,
hybridization of Mansehra,  Sanskritization at Girnar").
It is, indeed, not very surprising that Asoka's Orissan or even the NW
Mansehra inscriptions have Magadhisms,  it rather is surprising how far
*local* elements have penetrated into the various versions of edicts in the
non-Patna style ("eastern") inscriptions. Thus, we should not  be
surprised, e.g. that Orissa so closely matches Patna, and then point out
that post-Asokan Orissa inscription do not...  NB: the descr. of 'real'
Oriya MIA is another point, still to be done, cf. below.

3rd. For Afghanistan (idiomatic Aramaic and Greek, not a literal
translation, see Caillat's summary, p. 415) these deviations are not
surprising. But they are, when we look into the transpositions made into
other MIA dialects, such as the NW (pre-Gandhari) one. (NB: the official
language of the Persian empire was Aramaic, so the NW MIA dialect does not
count as 'local admin. background' language).

When we progress in this fashion, the 'rough' dialect map basically stands
as those 'conservative' people, Pischel, v. Hinueber etc., have sketched
them for the past 100++ years.

*A NW dialect,  (e.g., nom. -e),  the later Gandhari
*A Central dialect, (nom -o) the later  Sauraseni
*An eastern dialect, (nom -e) the later Magadhi
*A Southwestern / Southern dialect, (nom.-o),  in part: the later Maharastri
(unfortunately I cannot quote from the latest summary here,v. Hinueber's 1986)

Some details ad 2:
Here, it is C. Caillat (her paper in: Dialectes dans les litteratures
indiennes, Paris 1989) who takes, at least in my view, a sensible middle
path: after reviewing the arguments pro/contra "eastern"/western dialect
features in Asoka, she discusses those which are peculiar for the NW,
Saurastra,  etc., i.e. where they diverge from Asoka's dialect.

It is only such *local* forms that should be compared to the early
post-Asokan inscriptions. (In general, we will have to observe, just as in
Asoka, the political motivations and other historical details behind many
of these inscriptions, note, e.g. Damsteegt's discussion of various types
of 'foreign' (= outside of local area) influences in many of the Mathura
and other inscriptions; Th. Damsteegt, Epigraphical Hybrid Sanskrit, Leiden

Unfortunately, there still is (in the following selection) no evidence of
what exactly happened in northern Bihar, exactly the area we really would
like to hear about (but at least, nearby, Kanishka's partly Sanskritized
Sarnath inscr., Sircar no. 37-39) has r forms)
Some details, to be used as evidence,  from Sircar's (post-Asokan goup of)
inscriptions  include:

1.  Nom. masc.   -o/-e:
Eastern UP:   Kosam (Allahabad) inscr., Sircar no. 10 A:  gahapattiko
Magadha area:  Patna Image, Sircar no. 8: nom -e (yakhe = yakSaH, etc.);
Orissa: -o/e : the exception  seto in Asoka's Dh is important as it  does
not conform with Magadhi but with later Orissa, Kharavela ++;
Kharavela's Hathigumpa inscr. poses many problems, as is well known.
Unfortunately,  Sircar has nothing but this one and one of his queen's (no.
92), plus one near Puri of the  end of the 1st c. BCE, which is almost too
short to be of use here. (but see -patino, below).  I would like to see
more of Or. MIA before deciding on Orissa (and cf. my suspicions in the
last email).
-e forms in Zeda, near Und, Rawalpindi Dst.:  Inscr. of Kanishka,  Sircar
no. 42  --  but   -o forms in Manikiala from the  same Dst., Sircar no. 43,
on the other hand,  -e also in Kurram inscr. Sircar  no. 47. As has long
been pointed out, even the NW gets influenced by the central dialect (cf.
Caillat on Asoka's Mansehra!)

The same / similar points could be made for:
2. forms in  -as, see list in Bloch p. 47sq.

3. or the  gen. -as > e/o.
For all these forms it is likely that the Central dialect, the continuant
of the Rgvedic Kuru dialect,  has split up the area of older -e forms which
remained in the NW and the E, and has introduced central -o forms.  (-e
once in the RV: sUre,  and of course in Asoka++, in Gandhara).
Post Asokan : Barhut, Besnagar (MP, Sircar p. 87sqq, no. 1-2) have:  -no
for -nas,
Parkham Image (Sircar no. 7) :  kRtaH > kato,   -rAjaH   = -raJo (ra~o),
Orissa:  Kharavela, but also no. 93: Manchapuri inscr. -patino cf. -siirino.
MP.  Gunji inscr., Raigarh, Chattisgarh (n. 93A) : ra(m)~o, (cf. also NW
-ch- : savachare!)

4. The r/l  distinction has a  similar distribution:
'asurya' -l- in the east: SatapathaBr. (he 'lavo);
cf. Mehendale  p. IX  and Bloch p. 46-47 on Central and S. India.
Note in Mauryan time: Nagarjuni Hill, Gaya, (Sircar p. 77 no. 41) with l
In many Sanchi inscr. immediately after Asoka, -r-.
Already Mehendale's position is  sensible: the Asokan Patna style text has
been preserved in many places, not the local pronunciation...

It must be underlined that the  exact boundary of r/l is not necessarily
the same one as that of  -e/-o  or cch /kh. Isogloss bundles of this type
do not always overlap in their combined area, a fact well known from
dialect studies.  There is an often broad 'fan' of intermediate forms
between dialect A and dialect B, where certain features of the other
dialect are found, but not all of them (i.e., a sort of Ardha-xyz).

5. Instructive: the vRkSa case, see Mehendale p. 3,  cf. H. Berger, Zwei
Probleme der mittelindischen Lautlehre, Munchen 1955 (p. 73), -kh- almost
eg. kalpavRkSa = kapa-rukhe in Kharavela,  and Pali  rukkha (in Pali only 2
times  (north)western -vaccha, see Berger, p. 73).

More information about the INDOLOGY mailing list