'Re-construction' challenge no #1: Modern Gandhari

A forum for all topics related to constructed languages
Post Reply
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Granted, this is not the first 'reconstructo-lang' I have been working on -- a cursory browse on my posts will reveal many such languages. But Gandhari, like the languages I shall post under the 'label', poses a challenge, viz. the scantiness of resources. (In contrast, linguists have written 1000-page long volumes on the nature of the Tocharian verb alone, ). It is therefore my hope that this thread shall help organise my research on the historical Gandhari spoken a couple of millenia ago and my progress on the conlang -- as well as earn gainful feedbacks.

On Reconstructing Gandhari

The only grammar of Gandhari that I currently have recourse to is Baum's Outline of Gāndhārī Grammar (available on his website -- https://stefanbaums.com/). Over half of the grammar is a list of sound changes, which in the process of re-construction is more essential than morphological or syntactical details, on which the grammar doesn't cover too much in detail.

A good online lexicon of the language , which I have been accessing for quite some while, may be accessed from this link: https://gandhari.org/ (the marvels of technology..)

Since one can only know so much about historical Gandhari, comparison with related languages is inevitable. For one we have Sanskrit, roughly speaking the 'Latin' of all Indo-Aryan languages. Reconstruction of some forms would need to go back to Sanskrit. A comparison with modern Dardic languages, especially those spoken in the proximity of Peshawar, viz. the Kohistani dialects Torwali and Tirahi, will give us an idea of how the language ought to look like.


Modern Gandhari (Gandhari: ganhārī گنھاری or prĕğwarī پرږوری) is an Indo-Aryan language spoken (not always in the majority) in an area stretching approximately from Taxila to Jalalabad, mainly by Gandharans, though the Gandharans are not the only people to speak the language. Some Pashtuns migrating to the Valley of Peshawar have adopted Gandharan as a spoken language (in OTL it was the other way around), and most Hazaras settling in the valleys of the Hindu Kush in the 14th century have adopted the local language (and, to some extent, culture) spoken within, which, we presume in this scenario, is a 'mountain dialect' of Gandharan.

Gandhari, though may be held in proximity to the Dardic languages, also have features which binds it to the North-western Indo-Aryan languages. It can thus, llike Kashmiri, be characterised as a "transitional language" (correct term?) bridging Dardic and Lahnda.

(On Old Gandhari and Dardic: Cheung suggested that "that “Proto-Dardic” was actually Gandhari Prakrit, if not identical, then at least, very closely related to it." The first part of his assertion sounds less plausible than the second -- my (admittedly quite amateur-ish) research on Gandhari and Dardic leaves me with a hunch that not all Dardic languages may descend from old Gandhari, for there are a few features in Dardic that cannot be accounted for through old Gandhari, viz. the so-called "Dardic metathesis", to which attested Gandhari words show quite discongruent outcomes (more on that later, infinitives in -ik in some Dardic languages (also found in some neighbouring Pamiri languages [Iranic], therefore sprachbund / possible substrate influence?), and the non-Dardic 'lowland' features found in attested Gandhari and some Dardic languages such as Kashmir (which sporadically shows forms and features of the Lahnda and Pahari sort). Indeed, some scholars have questioned the existence of a single, united Dardic language family. The classification of Dardic languages, though, is outside the scope of the present discussion.)

Old Gandhari didn't paint the picture of a united literary language -- dialectal variations (i.e. non-diachronical ones) of forms mainly but not solely related to phonology may be detected in the documented inscriptions and texts. Modern Gandhari may be as much fragmented than her predecessor. One can class the dialects based on geographical delimitations (for which we have the Peshawari, Nanghari, Hazara Gandhari, and the one spoken in that valley northwest of Islamabad-Rawalpindi which includes Attock and Haripur (names, anybody?)) or linguistic features (most prominently being the "rhotic dialects" vs. "non-rhotic dialects" [which works not like counterparts in English -- and more on this later]. The prestige dialect is the one spoken in Peshawar -- which has been the cultural and intellectual capital of Gandhara for over 2000 years.

Gandhari is currently written using the Perso-Arabic script, mainly in Nasta'liq types (as opposed to the Naskhi used for Pashto). There has been some sporadic attempts to revive Kharosthi (popularly known by the first six letters of the abecedary: "Arap-canal" 𐨀𐨪𐨤𐨕𐨣𐨫)

(Phew! Now we're done with the introduction, bickering on which I have delayed a week or so posting it to the forum. You see, I don't write as eloquently as I used to do...)
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Phonology and Orthography

The orthography of Gandhari is meant to be a diasystem, a system that accounts for all accents of the language, which means that not all distinctions recorded in the orthography are realised in each of the spoken accents (more on Accents later).


p b m f* پ ب م ف
t d n ت د ن
ṭ ḍ ṇ ٹ ڈ ݨ
c dz څ ځ
ć j ń چ ج ڃ‎
y r l w ی ر ل و
ṛ ḷ ڑ لؕ
s z س ز
ś ź ش ژ
x́ ǵ ښ ږ
x* ğ* h** q* خ غ ہ ق

The above table, for the sake of economy, does not take into account:
a) orthographical aspiration, denoted with a trailing ھ, which could occur following any consonant except h /ɦ/ but including all of the sonorants (mhallē /mʱɐl:e ~ mɐ̀l:e:/ مھلے 'grand-father', whĕź /ʍəʒ ~ wə̀z/ وھژ 'greed'). The voicing of said aspiration matches the preceding consonant; a voiceless consonant with aspiration is realised aspirated, while voiced ones (including sonorants) are either realised with some sort of breathy voice, or the distinction is transferred to the following vowel -- devoiced and realised with a low tone (especially in the eastern dialects influenced by Lahnda).
b) 'tensed' consonants kk کۿ
gg گۿ
tt تۿ
ṭṭ ٹۿ
cc چۿ
which are either distinguished from their aspirated and non-aspirated co-articulants by tensing (cf. Korean), glotallisation, or merged into the aspirated or non-aspirated series of consonants (variation by dialect).

Supplementary remarks:
* found only in loan-words. /f/ and /q/ realised in vulgar parlance as wh (فائدہ /ʍəɪdɐ/ 'use') and k.
** always realised as voiced /ɦ/.


Code: Select all

a i u ĕ e o 
ā ī ū ē ō
ya wa
ai au

ā آ (word-initial) ا (word-medial & final)
ī, y, ya ی
ē ې (word-initial & medial) ے (word-final)
ai ئی
ō, ū, w, wa و
au ؤ

Nasalisation, which can fall onto any vowel, is indicated by the use of the letter ں (nūn bē-nuqta), which is never joined to any following letter: شېں‌چا śēnnća /ʃẽ:t͡ʃɑ:/ 'Saturday'

The short consonants may be denoted using diacritics (zabar, zer, pesh, etc.), though they are omitted almost most of the time. Since all Gandharan words and phrases included in this sketchpad are accompanied by transliterations, use of diacritics is deemed unnecessary.
Last edited by dva_arla on 27 Aug 2021 12:17, edited 1 time in total.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Months of the folk calendar, based on attested Old Gandhari forms. The folk calendar is mainly used to determine the dates to festivals, while the civil Gregorian calendar finds more currency in everyday use.

وښاز ماز wĕx́āz māz

چېتر ćetr Mar - Apr
ہېژاہ hēźāh Apr - May
ڃٹۿ ńeṭṭ May - Jun
آږاڑھ āǵārh Jun - Jul
ښاونے xāwne Jul - Aug
پروتۿہ proṭṭa Aug - Sept
شپیو śpayū Sept - Oct
کتۿی kattī Oct - Nov
منگشیر mengśīr Nov - Dec
? Dec - Jan
? Jan - Feb
پھگن phaggĕn Feb - Mar

ہفتیاز دیوز haftyāz dyōz / dōz

آچا āćā Sunday
چندا / ​سوما ćandā / somā Monday
منگرا mongrā Tuesday
بپا boppā Wednesday
بہسپا behspā Thursday
جمعہ* jummā Friday
شېں‌چا śēnnća Saturday

*irregular final ہ for ā instead of a.

(No comments yet? Now I don't mean to be an attention hog, but feel free to share your thoughts and suggestions in the comments!)
Last edited by dva_arla on 27 Aug 2021 13:18, edited 1 time in total.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Numbers 1 to 10:

۱ اک ekk
۲ دوی دوے doi / doē / dwē
۳ ترے trē
۴ چؤ ćau
۵ پنچ panć
۶ ښوں x́ōnn
۷ ستۿ satt
۸ اٹۿ aṭṭ
۹ نوں nōnn
۱۰ دژ daź
Zythros Jubi
Posts: 412
Joined: 24 Nov 2014 17:31

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by Zythros Jubi »

Maybe you can make a non-Muslim dialect as well, like the Kalashas' (Dardic) language, or a sociolect like Kashmiri Brahmins.
Lostlang plans: Oghur Turkic, Gallaecian Celtic, Palaeo-Balkanic
Posts: 112
Joined: 23 May 2016 00:40

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by IEPH »

Also, I would like to ask how the Gandharans would fare in the modern era? Are they Muslims today (and would this have any bearing on them if much of them live in Afghanistan say from 1996-2001 and from 2021 onward?)
Posts: 32
Joined: 16 Sep 2017 14:24
Location: Between a rock and a hard place.

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by Pāṇini »

Although I'm not an Indo-Aryanist by any means, it certainly reminds me of what I've seen of the Dardic languages. Excited to see how the morphology works out [:D]
/aɪ kænʔ r̼̊ ʌnəɹstʲænd r̼̊ jəɹ æksɪnt r̼̊/
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Figured out that I should first get the sound changes sorted out before proceeding any further; here's a sketch of the first leg of the changes -- gathered from Braum's Outline of Gandhari Grammar (2016) and other sources. Phonemic classes employed (represented by capitals) are hopefully self-evident -- feel free to inquire if elaboration is necessary.

Sanskrit (~2000 - 500 BC) > (Old) Gandhari (~300 BC - 100 AD)

1 Vowels

ai > ē, au > ō
aya > ē, ava > ō in certain morphological desinences

ya > yi
ay > ey
av > o
Ma > Mu "in some cases"

ṛ > i, u after labials

V: > V _# (original or the result of MIA cons. loss)
i > e, u > o _#
in closed syllables "especially when followed by retroflex or aspirated consonants" (Baums), but proof is sporadic
Edit: apparently, also before sibilants, cf. pustaka > postaka, postao

2 Consonants (could get this section sorted out better with a table)

n, ṇ > n

Lenition of single intervocalic consonants:
p, b > v V_V
c, j, k, g > y
[edit: k, g > v in the neighbourhood of labials]
t,d > ð
ṭ, ḍ > ṛ
ph, bh > vh, but h "ïn weak terminational position and adjacent to labial vowels"
kh, gh > h
th, dh > z
s, ś, ṣ > z, ź, ż

Weakening of final vowels
K, h > ø _# K, T = plosives
m, n > ø V:_# but ṽ (nasalised approximant) V_#

Assimilation of clusters
K1K2 > K: #first plosive in a pair assimilated to the second
NK > NG; NG > N: #mostly in Khotan (only occasionally in other texts)
Nv > N:
my > mm but ny, ṇy > ññ
l, r, h + C > C:
ṣṭ, sṭh > ṣṭ
sc > c: ? (Glass: 62)
sth, sp(h) > th, ph
sk, skh "unclear", but evidence points to > x (/x/)
sḱ or ḱ in a few texts both corresponds to Sanskrit sk, skh and is used to transcribe "a non-Indian guttural" in loanwords (Glass 2000: 52)

SN: > S (with nasalisation of following vowel) or N:
Sm, v > Sm > Sp
ṣy > ś: but sy > s:

Ks > Ts
kṣ > /tṣ/ (Baums) or /hṣ ~ ṣh/ (Cheung (2013: p. 624), on the basis of old Indic loanwords in Pashto; specifically, ṣ, kṣ > Pashto ṛ)
ks, ts, ps > ts, etc.

Kl > Kr
K,Ty > C
but ɉ (Kharosthi j with bar above) corresponding to Sanskrit dhy- ( > MIA jh-) might refer to a different sort of affricate than the former),
and Brahmi (i.e. MIA, note that jh is rare if extant at all in Sanskrit) jh is associated with Kharosthi z (Baums: 67, 110).
jñ > ññ

Km, v > Kv "in strong position"
> K: "in weak terminational position"
but ttv > pp

hN > N:
hy > ś ( > ź if word-medial)
iha > (ihya? >) iźa

Approximants y r l v undergo assimilation to the 'strength hierarchy' l > v > y > r: rv > vv, ly > ll etc.

Triconsonantal clusters:
CCr unchanged
C//CC second and third cons. assimilate separately
NCC assimilates together
CC//C third cons. dropped

1.3 and 3. Vc(r,y) > VC: ('Law of two Morae') > V:C (in some endings & lexical items)
4. Metathesis and intrusion of r : C > Cr _ ..... r
5. Assimilation / dissimilation of sibilants
ś > s _ ... ś, ć (anticipatory dissimilation), but
s > ś ś, ć ... _ (preservatory assimilation)
6. Methatesis: ar > ra #_ only in two word-families, viz. arhant > rahanta and araṇya > rañña

s > z in gen. sg. suffix -asya > -assa > *-āsa > -āza
therefore Law of two Morae > simplification of double cons. & compensatory lengthening > lenition
Last edited by dva_arla on 28 Oct 2021 06:44, edited 3 times in total.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

IEPH wrote: 31 Aug 2021 22:43 Also, I would like to ask how the Gandharans would fare in the modern era? Are they Muslims today (and would this have any bearing on them if much of them live in Afghanistan say from 1996-2001 and from 2021 onward?)
I picture a Muslim majority (~80%) with considerable minority communities (9% Buddhists, 7% Sikhs, 4% Hindus and others). Most of the speakers of Gandharan would be living in Pakistan since most of Gandhara lies to the east of the Durand Line -- that is if the latter does exist (i.e. everything in history stays as and when they were).
Zythros Jubi wrote: 29 Aug 2021 17:44 Maybe you can make a non-Muslim dialect as well, like the Kalashas' (Dardic) language, or a sociolect like Kashmiri Brahmins.
Great idea! I am not well versed in Indo-Pakistani sociolinguistics, and have so far considered of Gandharan chiefly geographical variation save for the Hazara dialect/language (they are bound to crop up here and there -- variations of a language are a good way to distribute choices from questions one finds hard to decide; besides, evidence points out to phonological variation even in 'Old' Gandharan).
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Old Gandhari (~400 AD) to Modern Gandhari (~1600 AD)

1. Outcome of certain Old Indo-Aryan consonants and consonant clusters

Outcome of kṣ

Modern Gandhari shall feature the two different outcomes posited by Baums and Cheung, reflecting dialectal inmixture:

kṣ > ṣh > x̌ (occasionally h, especially in the neighbourhood of a sibilant), or
> c̣h

Outcome of word-initial v-

The retention of word-initial v- in Kashmiri, the most spoken Dardic language, seems to suggest that the consonant would be retained in most languages of the family (and therefore in Gandhari); however, Dardic languages, upon closer inspection, show a variety of outcomes.

Compare the words for twenty (Sanskrit viṃśati) in the following Dardic languages (adapted from Turner, grouped by affinity):

K. Kashmiri (Kāśmīrī) wuh, wīh

Kho. Khowār (Dard.) bíšir
Kal. Kalasha (Kaláṣa — Dard.) bĭ̄ši

Mai. Maiyã̄ (Dard.) bīš
Tir. Tirāhī (Dard.) byeh
Tor. Tōrwālī (Dard.) bīš

Sh. Shina (Ṣiṇā — Dard.) bì(h)
Phal. Phalūṛa (Dard.) bhĭ̄š
Sv. Savi (Dard.) biš

Paš. Pashai (Pašaī — Dard.) wəst

Gaw. Gawar-Bati (Dard.) išīˊ
Shum. Shumashti (Šumāštī — Dard.) ísī
Niṅg. Niṅgalāmī (Dard.) isīˊ

Also compare:
Lahnda vīh
Punjabi vīh, bīh
Hindi bīs

In Tirahi and Torwali initial v- strengthens to b-, thus in Modern Gandhari v- > b-.

Outcome of word-initial y-

Torwali seems to retain initial y- in most words, (yāti 'to go' > y-, yavya 'stream' > yāb, yava 'barley' > yōu), while turning it into ź in some words (yantra 'machine' > žān 'mill').

Tirahi has zau for yava.

Kalami ('Bashkarik' in Turner), another Kohistani language, has for the same word yō, and for yātr̥ 'husbandś brother's wife' > yēl and yuga 'plough' yū̃, but ǰūī for yūkā 'louse'

Both y- and ź- will be reflected in Modern Gandhari, again reflecting dialectal mixing or interference from other lects.

Outcome of syllable-final r

Old Gandhari documents seem to show two sorts of developments: metathesis (to a preceding or following consonant) : dirgha > drigha, sarva > savra or assimilation: sarva > sava, though Baums seems to conclude that the latter is the ultimate development.

The outcome in most Dardic languages is metathesis where the preceding consonant is a stop, dirgha > Katarqala lig (dr > l), Phalura drigo, though Tirahi has dérega (i.e. epenthesis), though metathesis wouldn't be tenable where the preceding consonant is a sonorant or sibilant. Here instead the r remains before a stop in Khowar: vartana 'turning (part.) > bartun 'spindle disk' or gets assimilated in Pashai: vartayati 'to turn' > waṭṭ 'to spend time'. An r before a sonorant, on the other hand, gets assimilated in most Dardic languages: sarva > Bshk. Bashkarīk sō, sūo, Tor. Tōrwālī sōw.

In Gandhari metathesis to the preceding consonant tends to happen after d-, but note the following doublet: driggh 'high (of a mountain, building, etc) but deggh 'high (of a person). In any case the following consonant is doubled/lengthened. In other cases the 'trailing' r is assimilated to the following consonant: thus sarva > sau.

Outcome of nasal + voiceless stop clusters (nt, mp, ñc)

In Old Gandhari "Clusters of a nasal with a following homorganic plosive generally remain unchanged, but in the language of the Khotan Dharmapada (and occasionally in other texts) the nasal causes voicing of an unvoiced plosive and complete assimilation of a voiced plosive."

The stops tend to be voiced in most Dardic languages. Compare the following outcomes for Sanskrit pamca '5':

K. Kashmiri (Kāśmīrī) pānċ *

Kho. Khowār (Dard.) ponǰ
Kal. Kalasha (Kaláṣa — Dard.) pañ, poñ

Bshk. Bashkarīk (Dard.) panž
Mai. Maiyã̄ (Dard.) panz
Tir. Tirāhī (Dard.) panz
Tor. Tōrwālī (Dard.) painǰ
Woṭ. Woṭapūrī (language of Woṭapūr and Kaṭārqalā — Dard.) panj̈

Phal. Phalūṛa (Dard.) pānž
Paš. Pashai (Pašaī — Dard.) pāˊnǰa, pōnǰ, painǰ, paiñ, paẽ

Gaw. Gawar-Bati (Dard.) pōnċ
Shum. Shumashti (Šumāštī — Dard.) pōn
Niṅg. Niṅgalāmī (Dard.) paṇ

Also compare: Lahnda, Punjabi pañj, Hindustani pã̄c

*the voiceless affricate in Kashmiri is exceptional and seems to have been resulted from the influence of another IA language, most probably Urdu.

Modern Gandhari voices voiceless stops following homorganic plosives: NT > ND.

Outcome of nasal + voiced stop clusters (nd, ṇḍ, ndh, etc.)

Voiced consonants in such positions, on the other hand, see varying outcomes in Dardic languages: either their retention (ṇḍ > ṇḍ) or assimilation (ṇd > ṇṇ). , though most Kohistani languages (also Shina) seem to be in unison in preferred the latter outcome: píṇḍa 'lump' > Kalami pīṇ ʻ calf of leg ʼ; Torwali pīn ʻ heel ʼ, Maiya pīṇ ʻ leg ʼ.

A similar outcome can be seen in Modern Gandhari, though in the latter the plosive remains long and the preceding vowel therefore left without compensatory lengthening: píṇḍa > penn 'lump, pimple'. This outcome can also be seen in the language's endonym: Ganhārī.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Outcome of Sanskrit c /tʃ/ and j /dʒ/:

Retained, as in Tirāhi, Torwali, Burushaski, Khowar, and Pashai, as opposed to being de-alveolarised (/tʃ/ > /ts/ [> /s/], /dʒ/ > /z/) in Kashmiri, Maiya and Gauro among the Kohistani languages, and Pashto.

Outcome of Sanskrit ṣ and ẓ:

ṣ and ẓ are debuccalised into /x/ and /ɣ/, a shift also to be seen in Tirahi and Pashai among the Dardic languages, and the variety of Pashto spoken around Peshawar, viz. Northern Pashto.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

2. Other sound changes:
  • Development of fixed, penultimate stress [barring certain exceptions] (happened in most IA languages, perhaps dating back to Prakrit)
    Elision of weak, pre-stressed vowels
    Dissimilation, most notable of which being the change mn > ml:
manas > manaz > mnaz > mlaz

Loss of intervocalic /ð/ and /v/. This results in the hiatus of two vowels, in turn resolved by their contraction:

/əðə/ becomes /ə:/, a phoneme not merged with other long vowels, then heightened to /ɘ:/ sometime around the beginning of the second millenium. This phoneme in turn finds various realisations in the various accents: /e:/ in western ones and /o:/ in most eastern ones, though represented in writing by a letter of its own: أ
/iðə/, /ijə/, /uðə/, /uvə/ > ī, ū
/ivə/ > /eə/ > ē in the so-called 'dēź dialects', named after their pronunciation of the word for 'day' (< divaśa) (which includes the Peshawari-based prestige),
but /ivə/ > ī in the 'dīź dialects'
/o:ðə/ (common in the 3sg. of verbs) > /o:j/ (c.a. 1000 AD) > /e:/ in the dēź dialects but /u:j/ > /u:/ in the dīź dialects

/əjə/ > ē or ī and /əvə/ > ō
Final aka > aga > /ə(v)o/ > /əɔ/ > /ɔ:/. This was later raised (and shortened) to /a/ or /ə/ in the standard; the pronunciation /ɔ:/, becoming somewhat stigmatised, being retained in the country. The final -a(h)s from Persian loanwords was subsumed onto this final; both are represented in writing with the letter ه, being differentiated by a hamza atop the ha (ۀ) representing the so-called 'short ha' resulting from the (historical) lenition of other vowels: svasu > *śpaz > śpă śister'. Both sounds merge in some accents; in some others the latter is realised as a schwa or the front vowels /æ~ɛ/.

Weakening, merger, then apocope of word-final vowels: word-final i > e and u > o merges with a > o becoming a weak schwa a, though the former seems to have been retained longer when in hiatus.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

More on the outcome of syllabic and syllable-final -r.

Burrow: The Dialectical Position of the Niya Prakrit
Caillat: Connections between Ashokan and Niya Prakrit
Kogan: Once more on the language of the documents from Niya (East Turkestan) and its genetic position

The outcome of syllable-final -r is one of the contentious problems that surrounds a 'projection' of Modern Gandharan; due to the varying outcomes, sometimes conflicting, of the phoneme in question. The cases and its varying reflexes may be summed up thus:
  • Syllabic ṛ: becomes ri (ru in labial environments) or fully vocalised: i/u.
  • Syllable final r: either retained or gets elided/assimilated into the following consonant -- sometimes causing (and sometimes not) the cerebralisation (i.e. retroflexion) of a dental.
This variety may be accounted for the wide areal range in which Old Gandharan texts are found, or -- to place cause and effect in order -- the sheer amount of text subsumed under the general category Gandharan, North-Western Prakrit, etc. despite variety chronological and diachronic. Even so one may encounter variations even amongst texts from the same area , perhaps due to chronological reasons (more conservative features = older) or scribal practices.

I have hesitated to adopt into Modern Gandhari tendencies originating from those manuscripts labelled Ashokan, fearing that it might have in some way resulted from more easterly 'Bharati' influence, perhaps from scribes originating from outside Gandhara. But Caillat reassured me of the geographical affinities of Ashokan Prakrit when she subsumes Shahbazgarhi into Ashokan.

On the outcomes of the change in Niya and (Northwestern) Ashokan/Shahbazgarhi Prakrits Burrow has these to say:

1. On syllabic r: "r appears as ri (r) much more consistently in Niya than in Ashokan". In Ashokan, on the other hand, "the r is lost in the majority of forms", though the forms with the r's are not entirely absent — only much less frequent in occurence.

2. On the cluster rT: "In Ashoka, on the other hand, the assimilation of r + a following dental is almost the rule..."

3. "The group -rs- is preserved It is usually assimilated in Niya... it is usually assimilated in Ashoka.

4. ść > -c in Niya (distinguished from c with a bar above the glyph [Glass, 2000]). "In Ashoka no distinction is made" (though this might be as well a matter of orthography rather than pronunciation; Niya orthography distinguishes much more phonemes than does Ashokan).

5. ts is preserved in Niya, while assimilated to s (/ss/), though Burrow provides only one example for the change, viz. cikisa. This change I have so far yet to take into account.

6. ṃs > ṃts in Niya while > s in Ashoka; similarly ṃś > ṃc in Niya vs. > ś in Ashoka.

7. -ly- preserved in Niya while > l or > y in Ashoka.

8. -lp- preserved in Niya while > p in Ashoka.

9. sv > sp in Ashoka, śv in Niya

10. sm (of the loc. sg.) > sp or > s in Ashoka, while > ṁm in Niya

I have so far ascribed the conservativeness of Niya Prakrit due to its isolation from its source (cf. Modern French chaise /ʃɛz/ vs. Modern English /t͡ʃɛə(ɹ)/, the latter retaining the /t͡ʃ/ of its Old French source). Burns, though, suggested to me that the difference may rather find more geographical motivation:
"Obviously, we cannot derive the Niya Prakrit from the language of Ashokan, and the most natural conclusion to draw from the fact that phonetically it is better preserved is that its home is to be sought further to the west. Because it seems clear (then as now) that the more remote a language was in the direction of the North-West the less liable it was to phonetic decay." Some of the outcomes in Niya are reminiscent of those in Torwali (Kogan, 230) and though Kogan declines a close affinity between Torwali and Niya due to glottochronological findings, it may be certainly stated that Niya is closer in affinity to Dardic languages than Ashokan Prakrit, or (or to be more precise, dialects).

Kogan, on the other hand, presents the following comparison between Niya Prakrit and Dardic:

"The usual Niya reflex of PII syllabic *r... is ri" but Kogan notes "a number of cases" where it is, like in Ashokan, fully vocalised. This outcome is seemingly shared with Dardic, though according to Kogan Niya arrived to the same outcome by a different route. In Niya, r̥ > ri or ru without any intermediary, subsequently influencing the quality of the dental or causing the loss of another /r/ through distant assimilation and disassimilation respectively, while in Dardic languages r̥ > ir or ur and only assumes their attested outcomes after having undergone Dardic Metathesis.

Niya: prābhr̥ta > *prābhruda > *prābhuḍa > prahuḍa

Dardic: kṛṣṇa > *kirṣṇa > *kṛiżna > (kriżna, kruhun, etc.)

The intermediate stage in Dardic might be more plainly discerned from the lexemes where "Dardic metathesis" is not possible; i.e word-initially or where it would violate sonority constraints. Thus for r̥kṣa 'bear' we have Dardic forms without an initial ("Pashai ẽč, Shumashti, Gawar-Bati, Sawi, Shina ī̃č̣h, Kalasha ič̣, Phalura ĩč̣, Bashkarik ič̣h, Torwali īṣ, Indus Kohistani īčh ‘bear’"), which may be explained were we to assume an intermediary form *irc̣h. The outcome of the word is not attested in Gandhari, but compare with the forms with r- in non-Dardic languages.

One may therefore be convinced of Kogan's description the evolution of syllabic r̥ in Dardic; nevertheless, the same cannot be said of his proposal of the evolution thereof in Niya: one cannot say whether or the evolution thereof in Niya differs (i.e. without the intermediary *ir) from the former. Since the outcome of word-initial syllablic r̥ in Niya could not be discerned, one lacks a suiting comparanda, and thus whatever Kogan proposes for the development of r̥ in Niya may, at best, remain conjectural.

Both Burrow and Caillat mention the Ashokan form dhrama, dhramo, seemingly proving a shift in Ashokan akin to the well-known Dardic Metathesis, though like outcomes are scarcely attested elsewhere.

Kogan cites a Niya form paḍi corresponding to OIA prati, which might suggest the (rather 'Bharati', un-Gandharan) simplification of initial stops, though it would most likely seem that the loss of the /r/ is triggered by the following retroflex dental or trill (with the pronunciation of /ɖ/ or /ɻ/ for ḍ being as likely as the other, though it can be said that the latter gives more impetus for dissasimilation and thus elision).

prəti > /prədi/ > /prəɖi/ > /pəɖi/ or
" > " > /prəɻi/ > /pəɻi/

Another feature of Niya Prakrit, viz. the (scantily attested) development of nasal + voiced consonants (nd, mb, ndh, etc.), is connected to North-Western (non-Dardic) IA languages instead, i.e. "Sindhi, Lahnda, Punjabi, and West Pahari".
Last edited by dva_arla on 28 Oct 2021 13:29, edited 1 time in total.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

Note on the outcome of final -aka > -aga:

Inconsistencies in spelling in the Khotan Dhammapada seems to imply the value /əjə/ for this ubiquitous nominal and adjectival ending. Some inscriptions (from Hadda, etc.), on the other hand, directly attest to the ending -ao. I have dismissed the former development as one unique to Khotan (i.e. an extra-Gandharan one), though after becoming convinced of closer relations between 'Khotanese Prakrit' and its Gandhari counterparts through Kogan etc., begun to ponder of the possibility of ao perhaps being merely a scribal convention?

One may alternatively see both as equally valid developments : -ago (of the nom. sg; the ending is typical for Gandharan masculine nouns) > /əvɔ ~ əvə/ > ao, while -aka (of the pl. and as to be found elsewhere) > /əjə/ ( > Mod. Gandhari -ē); the differing outcomes of the lenition of /g/ being explained by the surrounding vowels.
Posts: 150
Joined: 25 Oct 2019 21:03
Location: Realm of Ideas

Re: 'Re-construction' challenge no #1: Modern Gandhari

Post by dva_arla »

It has almost been a month since I last posted on this thread. During the span of time I have been continuously facing the quandary of deciding outcomes to the phoneme /r/ in various environments, to which one may bestow the sobriquet 'r-dilemma' (in the same fashion my dilemma on final nasals (or nasalised vowels) in Khotanese may be called the 'n-dilemma'). In this respect attested outcomes show greatly varying outcome, even sometimes in texts found on the same site. After a few weeks spent in the dilemma, I decided to synthesise information from seven or so papers on different dialects of (Old) Gandhari, and by doing so managed to get a better picture of the variation to be found in those documents

I may, to my own relief, say that I have made up my mind on several points on:

syllabic ṛ (CṛC):
outcomes: CriC, CroC > CrəC
but CiT: (or CeT:?) where T is a dental or alveolar sound and Ṭ the corresponding retroflex.

The change might not have ever occurred in words such as hṛdaya , where the initial made vocalisation in the manner described above infeasible. Corresponding to the Sanskrit word only the form hidao, hidaya are found.

glide r proceeding a vowel (CrVC):
outcomes: CrVC retained
but CrVT > CVṬ:, where T is a dental or alveolar sound and Ṭ the corresponding retroflex.

Clusters with the glide are often preserved (in common with the Dardic languages but unlike the rest of the Indo-Aryan languages). The few cases where simplification of the cluster have occurred in the neighbourhood of a dental: krtya > kica; krta > kiḍa (with 'underscore' below the ḍ, perhaps for ḍḍ?), kiṭraṃ; prathama > paḍhama (the latter case is attested in a document unmistakably 'young' enough to merit it importance)
or, to a lesser extent elsewhere: gr̥hīta > giḍa (with 'underscore' below the ḍ, perhaps for ḍḍ?), gr̥hṇāti > griṇadi, giṇadi; kisamnae, etc.
Dental stops and, to a lesser extent, sibilants seems to act as an '/r/-magnet'...

syllable-final r (CVrC):
outcome: CVrC > CVC:

Some of the older documents display a phenomena reminiscent of the famed 'Dardic metathesis' (dharma > dhrama, dirgha > drigha). The feature, however, are only attested, besides in the older Ashokan edicts (c.a. 200 BC), in the translation of the Dhammapada; in Niya and the younger Gandharan documents (c.a. 200 AD) they are virtually absent. One may perhaps assume that this feature declined along with the dialects that had them.

A more proximal sort of metathesis may be found for trills following a semivowel: the word sarva may be occasionally encountered in the form savra. These however only occurs in the Ashokan Edicts, and in the texts where they do occur the form sava (consonant length being rarely indicated in Kharosthi) is also to be encountered. While the form sarva may indeed exist at some point in the evolution of the language (at least in some dialects), it eventually gave way to the latter version (with -rv- serving as intermediary to the shift -vr- > -vv-).

Miscellaneous features:

The distribution of dessinences of the declension seems to be geographically delimited (though not without exceptions): of the nominative one finds in the area west to the Indus mainly -e, and to its east mainly -o. Since the standard variety of modern Gandharan is based on the one spoken in Peshawar, which is deep to the east of the Indus.

Only in the Dhammapada one finds instances of v > m in the neighbourhood of nasals.
Post Reply