The Buddha Spoke Kosalan

by Lennart Lopin

Reflections based on Stefan Karpik's argument (see below)

Re: [Pali] Stefan Karpik article

Just 2 weeks ago while reading some material on Mahinda's
mission to Sri Lanka and re-thinking this whole topic for the 100th time, I
could not help myself as to write down some of the argument on both sides
for myself - but especially those arguments in favor of Pali being the
Buddha's language. In that little resume of mine I came to similar rational
(not philological) conclusion as you. This was not always the case. When I
learnt Pali I was under the impression it was Buddha's language. Then, like
most, I learnt about the scholar's objections and had no way to argue
against it...however, over the last decade my doubts grew and now I am back
to my former position. Here some of my thoughts on this matter:

The first time I ever came across someone making a case for Pali as the
language of the Buddha, was Wiliam Stede (Pali English
made the convincing argument, that Kosala, and not Magadha, was the
center of affairs at the time of the Buddha. Yes, Kosala's dominant power
was about to fade during the next few centuries, but we should never
interpret history based on future events. Buddha was raised in Kosala, lived
in Kosala, spent most of his adult life in Kosala. Most rain seasons start
with Sāvatthiyam and not Nalanda or Rajagaha. Yes, it may have been that
Magadha was "hip" and "cool" as the Aryan conquest had pushed further and
further eastward and the power-centers of the Veda time shifted further and
further east. But still, the culture and learning was in the West and so the
Buddha's Kosalan dialect may have well been influenced by
Magadha-dialectism. What would that result in? Right, a nice mix. A prakrit
with some heavy vedic/sanskrit voila, "Pali" is born.

Now, fast forward to the time of Ashoka. Kosala is gone, swallowed up by
Magadha. It is the heartland of Magadha now as well. The Buddha is seen as a
son of the Magadhan empire - at least from the perspective of the Sinhala!!
They of course take Mahinda's transmission to be "Magadhan". Even Mahinda
may have felt that way.

Also: (But maybe some of the very knowledgeable philologists can help me
understanding) why does Buddhist Hybrid Sanskrit read like someone trying to
translate Pali into Sanskrit...and not just any Prakrit - it reads just like
a Sanskritized pali text. The amount of Pali vocabulary and forms are so
staggering (esp. compared to other brahmanic Sanskrit texts) that I would
not look at a few examples which time and place may have changed, but at the
overall me it seems so blatant that Brahmanic Buddhist monks
pushed the usage of that "new" and "hype" Sanskrit, but because most of
their terminology was so influenced by the Buddha (in Pali?!) the best way
to move Buddhist philosophy into a Sanskrit language (as a medium to express
new philosophic ideas because Pali became to sacred to touch) was by simply
basing it on Pali and then formulate the new ideas in Sanskrit.

There is that other argument you brought up, Buddhaghosa's interpretation
that the Buddha did not want a translation of his teaching...I am ambivalent
about that, as from a teaching standpoint the Buddha clearly advocates that
we should not grasp at nirutti - but, at the same time, we see the monks and
lay people at the life time of the Buddha take extreme care (!) not to
misrepresent the Buddha's words (unlike today, unfortunately)...

Now, if for a moment I acknowledge however, that it was understood to mean
that you are supposed to learn the Buddha's word in your own language, then
why did Mahinda not tell that the Sinhalese people? Why did they
"misunderstand" Mahinda in such a big way that they were supposedly the only
ones learning Mahinda's "dialect" by heart and not keeping the canon in
their own Sinhalese dialect. This does not make any sense at all! It only
makes sense, if we acknowledge that Pali was indeed Buddha's language and
they were instructed by Mahinda to keep it that way.

So any way, Geiger, Walleser, Stede seem to be - at least on the grounds of
reason - much closer to the truth than the idea that there has been this
"mysterious" Pali coming out of nowhere - at the same time while the
Buddhist tradition was able to transmit one of the best and most complete
literary canons in the world through 2500 years... I just guess when in
doubt one would favor the simplest solution and that would just be that Pali
was the Kosalan dialect of the Buddha, influenced by his tours through
Magadha and probably filled with altenative forms as the centuries passed by
before it was "solidified" in writing.


For a full discussion see:

Below Stefan's article:


The Buddha Spoke Pāli

by Stefan Karpik

The Problem

Pāli means “text” and the Theravadin commentarial tradition tells us that the actual

language of the texts is Māgadhī, which is what the Buddha spoke.1 However, this is

dismissed by most Western scholars and Theravadin Buddhist writers have deferred to

their opinion. Even among Thai monastics, the opinion that the Buddha spoke Māgadhī,

and not Pāli, is common. I have been impressed by the confidence of the claims of

Western scholars and rather surprised. For implied in these claims are (a) that the

Buddha’s words were translated into Pāli and (b) that, once translated, the original words

were lost. This process is fairly typical of a written tradition, such as the Christian New

Testament and Classical authors, but it would be, so far as I know, unique, if it were true,

in an oral tradition. This is therefore a large claim to suggest that without the benefit of

writing materials, a translation equivalent to thousands of pages of scripture was made. It

is the more remarkable because, in the same culture and geographical region, other

canons, such as the Vedas and the Jain texts, have been transmitted orally over millennia.

Prominent in claims that the Buddha spoke something other than Pāli are K.R. Norman2

and Richard Gombrich3. Both are former Presidents of the Pāli Text Society (PTS), but I

trust that neither would wish me to accept their opinions without examining some of the

evidence. Their argument runs as follows:

a) Pāli contains features which appear to come from several dialects and includes

incorrect backformations from Sanskrit. It therefore bears the hallmarks of an

invented or literary language translated from other authentic dialects or languages.

b) Furthermore, the Buddha specifically allowed translation of his words.

c) Māgadhī, the language of Māgadha (now Bihar) was clearly different from Pali.

However, now I have examined the evidence, I am far from convinced.

An Out of Date View

Before arguing against Norman and Gombrich, I will explain what my position is.

Essentially. I follow Geiger4, whose introduction to the reissue of his grammar for the

PTS was replaced by Norman in favour of a new one by Gombrich:

Pali forms are derived from Vedic, the language of the Vedas, not from Sanskrit,

the language of the Upanishads, and stand beside Sanskrit forms as later

formations from Vedic. Numerous double forms indicate that Pali is a mixed

dialect, but it had its origin in a particular dialect.5 The Theravada tradition states

that Pāli is Māgadhī; however, distinguishing features of Māgadhī, such as no r or

1 Vibhanga Commentary 388 sammāsambuddho pi tepit_akam_ buddhavacanam_ tantim_ āropento

Māgadhibhāsāya eva āropesi.

2 K.R. Norman Pali Literature Otto Harrassowitz: Weisbaden 1983

K.R. Norman The Value of the Pali Tradition Collected papers Vol 3 PTS Oxford 1992

K.R. Norman A Philological Approach to Buddhism SOAS 1997

3 R. F. Gombrich What is Pāli? A Pāli Grammar W. Geiger, tr.B. Ghosh, Ed K.R. Norman PTS 1994

4 Wilhelm Geiger. Pāli Literature & Language. Oriental Books Reprint Corporation, New Delhi 1978

5 Much, I suppose, as BBC English has its origins in the South East of England, but slips into

Americanisms, such as “prehaps, nucelar, did you do that yet, like he said”.


s and ending nominative masculine and neuter -a stems in -e are absent in Pāli.

Even so, Geiger regards Pāli as a form of Māgadhī spoken by the Buddha:

Ārsa, the language of the Jain suttas, is called Ardha Māgadhī, “half-Māgadhī”,

but the special features of Māgadhī are also absent. Therefore, Pāli may be

considered as a kind of Ardha Māgadhī.

Pāli is a language of the higher and cultured classes which had been brought into

being in pre-Buddhist times through the needs of intercommunication in India

(following Rhys Davids, Buddhist India p140ff)

Pāli is a lingua franca free of the most obtrusive dialectical characteristics6, but

flavoured by the dialect of the speaker. The Buddha, though not a Māgadhan, was

active in Māgadha and this could have influenced his language.7

The Pāli canon was not translated. Its features can be explained by a gradual

integration of elements from various parts of India, a long oral tradition and the

texts being written down in a different country.

Cullavagga V.33.1 of the Vināya (anujānāmi bhikkhave sakāya niruttiyā

buddhavacanam pariyāpun_itum_) is mistranslated as “I allow you, oh brethren, to

learn the words of the Buddha, each in his own dialect” by Rhys Davids and

Oldenberg. Two monks were complaining that monks were spoiling the Buddha’s

words with their own dialects and suggested putting them into Sanskrit verse.

Geiger follows Buddhaghosa, the most prominent commentator, in concluding the

Buddha refused any translation both negatively by refusing verse and positively

by requiring the Buddha’s words to be learnt in the Buddha’s language. The

correct translation should be “I ordain the words of the Buddha to be learnt in his

own language (in Māgadhī, the language used by the Buddha himself)”.This is

more in keeping with Indian tradition and the context in which the monks were

complaining that the Buddha’s words were being corrupted.

Against the Modern View

I will now attempt to argue against the Norman and Gombrich view. Please bear in mind

that I am in no position to contradict their learning, but I am questioning their deductions:

a) Pāli is not an invented language.

A mixture of dialects is a feature of most living languages. Regularity is the exception

rather than the rule and belongs to artificial languages such as Esperanto and classical

Sanskrit. The fact that a language shows mixed features does not imply that it has been

translated. It would be absurd for future 45th century professors of Middle Americanoid

languages to suppose that a transcript from a current BBC programme had been

translated from an earlier language because it shows ancient features e.g. “oxen” instead

of “oxes”, “mice” instead of “mices” or includes Americanisms or has regional variations

like “chimdey” for “chimney”.

Actually, the mixing of dialects was not great. Even Norman (1997:44-45) states: “We

can see that in the canon as a whole there are very few non-Pāli characteristics and most

6 Similar, I imagine, to received pronunciation/ Queen’s English/ BBC English. M iii 235 states that a

monk should not insist on local language and should not override normal usage: janapadaniruttim_

nābhinveseyya samaam_ nātidhāveyyāti.

7 Much, I suppose as a northerner might change “bath” to “bāth” when moving to the South of England.


of these are due to a consistent introduction, at a later date, of Sanskritisms, which are

restricted in number, the most obvious being the absolutive in tvā.”

The evidence of artificiality from supposed Sanskritisation can be looked at differently:

Geiger believed that Pali developed from Vedic in parallel with Sanskrit. Thus Pāli could

have developed its own forms without any backformation. Backformations are in any

case a normal feature of languages of living languages; notice the debate on referenda or

referendums, phenomena or phenomenons; I once met a professor of a Canadian

university who argued that the plural of pizza was pizzae.

b) The Buddha specifically disallowed translation of his words.

As noted above, Geiger and Buddhaghosa disagree with this interpretation of the Pāli8. I

would like to add that I don’t see how translation could be practicable in an oral tradition

equivalent to over 5,000 pages of Vinaya and Suttas when written down in the 1st century

B.C.E9. The labour involved in producing a standard consistent translation capable of

recitation among a group of monks would have been enormous and without parallel in

world history so far as I know. Not only that, but those disagreeing with Buddhaghosa are

implying this heroic effort was repeated several times over in different regions of India.

In fact, the word “translation” is misleading, though arguably academically correct.

Normally we think of translation from one different language to another, but the

boundaries between language and dialect are not always clear. A speaker of Norwegian

can chat with a speaker of Swedish and understand the other, though speaking different a

language. A speaker of Mandarin Chinese and a speaker of Cantonese Chinese cannot do

this, though speaking different dialects. Some linguists therefore claim that the difference

between a language and dialect is political.10 With regard to the different states of

Northern India in the Buddha’s time, it would therefore be academically correct, though

misleading to laypeople, to talk of several languages being spoken. In common parlance,

however, the languages of North India could be said to represent different dialects of the

same language. Ashokan inscriptions, dating 150 years after the Buddha’s death in c. 400

B.C.E., are available in up to six dialects, which look very alike and whose differences

would probably not cause much difficulty to native speakers. My point is that there was

no need for “translation”, which is why the Buddha should not be interpreted as allowing


Gombrich (1994: xxvii) implies that I am on the right lines: “Before the texts were ever

written down, it is not likely that their dialect was ever completely fixed, or even that the

differences between the dialects were clearly conceptualised; it must have been a matter

of reciting in what appeared like “regional accents.”

If the Buddha did not allow translation and translation was unnecessary, what was the

nature of the corruption of the teachings that was complained of? Perhaps the monks

were using such different accents that they could not chant together harmoniously. The

8 Cullavagga V.33.1 anujānāmi bhikkhave sakāya niruttiyā buddhavacanam pariyāpun_itum_.

9 This is when palm leaves and ink became available to replace rock and clay tablets.

10 This argument is aired in The Story of Language by Mario Pei, Allen & Unwin 1952.


PTS dictionary gives for nirutti, the word for what was spoiling the Buddha’s speech,

“pronunciation, dialect, way of speaking, expression”. Norman (1980) thinks it was the

glosses of the Buddha being altered and the passage requires monks to use the Buddha’s

own glosses. I could similarly guess that “term” is a suitable translation for what was

altered, i.e. some monks were imprecisely saying, for example, citta for nāma, kāya for

rūpa; however, I admit I don’t know, for the Pāli is too open to interpretation.

The reader may wonder how Norman can stretch the dictionary meaning and how I can

presume to guess it; the reason is that academic work in this area is often speculative.

Here is Norman (1997:59):”We very commonly find in books and articles about early

Buddhism such statements as: “The Buddha preached in the Prakrits, the language of the

common people, and resisted the suggestions of some of his ex-Brahman followers to

translate his sermons into Sanskrit”. There is frequently no hint that these statements are

anything other than accepted fact, but readers need to be very wary, because such

statements are frequently not fact, and are anything but accepted by all scholars working

in the field. To tell you the truth, there is a great deal of the Bellman principle in the

academic world. You all know about the Bellman in Lewis Carroll’s The Hunting of the

Snark, who maintained that: “what I tell you three times is true”. I am as guilty in this

respect as anyone else, I fear. I may have an idea about something, and so I incorporate

my idea, as a suggestion, in an article I am writing, and wait for someone to reject or

disprove it. No one does, and I repeat the idea, still as a suggestion, in another article.

again, no one rejects it. I do this a third time, and if there is still no reaction, it becomes

fact in my mind – I have said it three times, so it must be true, and I consequently refer to

it in future publications as an established fact. The thought that no one ever reads my

articles and so no one has ever seen my suggestion, and so no one had had any desire to

reject it, or the alternative explanation, that those who read my first article thought that

the idea was so preposterous that it was not worth wasting paper and ink refuting it, so

that the second and third repetitions were dismissed as: “I see that Norman is still pushing

that stupid idea of his”, does not enter my head.”

c: Pāli and Māgadhī are not different

There is no clear reason to doubt the Theravadin tradition:

1. Norman and Gombrich object to the notion that the Buddha spoke Pāli, which

is not the language of the area called Māgadha. The earliest examples of

anything that could be called Māgadhī are Ashokan inscriptions. Gombrich11

gives the gap between the Buddha’s death and the ascension of Ashoka to the

throne as 136 years; the inscriptions are later. There was therefore time for the

dialects of Māgadha to change from something similar to Pāli to something

similar to the Māgadhī of the Ashokan inscriptions.

2. Alternatively Pāli was a Western dialect, used across other parts of India,

which went out of fashion in Eastern India by the time of Ashoka. Perhaps the

prestigious dialect of Māgadha’s court became a lingua franca and the basis of

the eastern dialect of the Ashokan inscriptions.

11 R.F. Gombrich “Dating the Buddha: A Red herring Revealed” The Dating of the Historical Buddha Part

2 ed H. Bechert, Gottingen 1992


3. In an early work Norman12 comes to some conclusions which are compatible

with Geiger’s: the occasional use of the -e form in the nominative case in Pāli

suggests that it was spoken in Eastern Māgadha; Māgadhī was considered the

root language of all languages and in the usage of the time of introducing

Buddhism to Ceylon, it was correct to call Pāli Māgadhi; the Buddha spoke

Māgadhī and Pāli was spoken in Māgadha; it is broadly speaking correct to

call the language of the Theravada canon ‘Māgadhī’.

Some Evidence

My belief is that the differences between the North Indian dialects 150 years after the

Buddha’s death would not be significant for native speakers. Below I give the six

Ashokan inscriptions13 of Rock Edict I, which exemplifying local dialects of c.250

B.C.E. The first three are considered Western dialects and the last three are Eastern;

Girnār is considered by some to be most like Pāli. Unfortunately, there are no parallels of

Girnār inscriptions with any from Māgadha, but Māgadhan inscriptions resemble those of

Dhaulī and Jaugada. I include translations into English, Pāli and Sanskrit:

Location: Śāhbāzgāhrī

(Pakistan, NW Frontier)

Location: Mansehrā

(Pakistan, near Rawalpindi)

Location: Girnār (India:

Gujarāt State, Girnār Hills)

Ayam dhramadipi devana

priasa …. rao


hida no kici jive arabhitu


Ayi dhramadipi devana

priyena Priyadraśina rajina


hida no kici jive arabhitu


Iyam dhammalipī devānam

priyena Priyadasinā rāā


idha na kimci jīvam

ārabhitpā prajūhitavyam

Location: Kālsi (India:Uttar

Pradesh, near Mussourie)

Location: Dhaulī (India:

Orissa, near Cuttack)

Location: Jaugad_a (India:

Orissa, near Lake Chilka)

Iyam dhammalipi devānam

piyenā Piyadasinā


hidā nā kichi jive ālabhitu


Iyam …devānam

piyena … lājinā


…. kichi jīvam ālabhitu


Iyam dhammalipī devānam

piyena Piyadasinā lājinā


hida no kichi jīvam ālabhitu


Translation by writer Pāli composed by writer Sanskrit by JamesWhelan14

This law edict was ordered to

be inscribed by King

Piyadasi15, Beloved of Gods.

Here no living being is to be

slaughtered and sacrificed.

Ayam dhammalipi devānam

piyena Piyadasinā raā


idha na kici jīvam ārabhitvā


iyam dharmalipir devānām

priyena Priyadarśinā rājā


iha na kicij jīvam ārabhya


12 K.R. Norman The dialects in which the Buddha preached. In “The Language of the earliest Buddhist

Tradition” Ed Bechert Gottingen 1980

13 A.C. Woolner Asoka:Text & Glossary Low Price Publications Delhi 1993

14 My thanks to James Whelan of the Totnes Pāli study group for the translation into Sanskrit.

15 “Piyadasi” is the name Ashoka gives himself.

16 The Pāli word pāhuneyyam_ has exactly the same root as the other examples. However, Buddhism

changed its meaning from “should be sacrificed” to “should be offered gifts”. In an early example of “spindoctoring”,

the meaning of “sacrifice” was changed to “gift to monk”. Similarly the meaning of “Brahmin”

was changed from “one of good birth” to “one of good conduct”.



There is no clear evidence to suggest that the language in which the Buddha taught was

different from Pāli. Pāli shows the features of a normal, non-literary dialect and has a

resemblance to real dialects of Northern India 150 years later. Pāli does not represent a

translation, as that would be neither practicable nor necessary. Pāli is acknowledged to

contain features of the Eastern Ashokan dialect. Therefore it could well have been spoken

in Māgadha 150 years before the Ashokan inscriptions and there is no reason to distrust

the tradition that it was indeed spoken there.

This is not to claim that the Pāli canon is a pristine representation of the Buddha’s words.

There is certainly fabrication17 and possibly omissions. One would expect any real

language to change subtly despite the checks imposed by communal recitation; for

example, bhikkhavo (monks!) is less common than the bhikkhave, which is more in

keeping with the Ashokan Eastern dialect . Possibly, there was a shift towards the

Ashokan Eastern dialect which was not much noticed by monks who regarded the forms

as equivalent.

With these caveats, I have come to the opinion that, if today the Buddha heard Sri Lankan

monks (who have the best pronunciation) reciting Pāli, he would recognise at least some

of his own words in a slightly different accent. That the Buddha’s actual words can

resound to us after 2,400 years is perhaps a romantic notion, but that does not prevent it

being the truth.

Stefan Karpik

Totnes, England

[email protected]

December 2004

Suggestions to improve this article are welcome.

17 There is an anachronism in Miii 253 where Mahāpajāpatī Gotamī, the first Buddhist nun, is still a

laywoman and is nonetheless referring to existing Buddhist nuns.