As mentioned in the Introduction and Context section, due to the complexity and subtlety involved in the investigation of the pronoun forms of Kannada and other Dravidian languages, this document has become unavaoidably detailed and long. So, for the ease of access, publishing, collaboration, revision tracking as well as for some technical reasons, this document has been split into two parts.
- Introduction and Context - A brief introduction to the topic and the context for this investigation, such as a listing of the various pronoun forms in Kannada and other Dravidian languages and highlights of some relevant existing scholarly work can be seen here.
- Provenance of Pronouns - The current document. The actual investigation of pronoun roots of Kannada and other Dravidian languages. This section refers to the relevant parts from the Introduction and Context section as required during the investigation of the pronoun roots of Kannada and other Dravidian languages.
ಈ ಲೇಖನವು ಮೊದಲು ಕನ್ನಡದಲ್ಲಿ ಪ್ರಕಾಶಿತವಾಗಿದ್ದು, ಆ ಕನ್ನಡದ ಆವೃತ್ತಿಯನ್ನು ಇಲ್ಲಿ ನೋಡಬಹುದು.
This document was originally published in Kannada and that original Kannada version can be seen here.
Kannada and other Dravidian words and word fragments are transliterated in this document using ISO 15919 wherever possible and is unambiguous. Where ISO 15919 transliteration is unclear or ambiguous, IPA transliteration has also been given alongside the ISO 15919 transliteration.
For example, the anusvāra is transliterated as ṁ
in ISO 15919 but the pronunciation may not be clear to anyone not familiar with Indian languages.
But this is workable for Kannada words because the pronunciation is fairly clear to most Kannada speakers.
Nasalisation is similarly transliterated as m̐
in ISO 15919 but the pronunciation may not be clear to many Kannada speakers because the nasalisation has now been mostly lost in many dialects.
So, to make it clear at least in the transliteration, IPA transliteration (̃
) has been given alongside the ISO 15919 transliteration (e.g. am̐
/ə̃
, ǣm̐
/æ̃ː
etc.)
- Provenance of Pronouns
- Background
- ಅನುವಾದದ ಕುರಿತು / About Translation
- About Transliteration
- Content
- License
- ಬರೆಹಗಾರರು / Authors
- Third-person Pronominal Roots
- Dravidian Languages
- Kannada, Tamil
- The euphonic infix v in the masculine and feminine singular and rational plural forms
- Neuter Gender Forms
- Proximal demonstrative ŭ in Tulu
- Dative Presentative ŭṁda of Tulu
- Origins of the form and semantics of the the root vowel u
- Neuter gender singular suffixes of finite verb forms in Havyaka Kannada
- Neuter gender singular suffixes of finte verb forms in Kannada
- The relationship between the anta of Tamil and anta, atta and antu of Kannada
- Semantics of and the processes behind aṁta/iṁta/eṁta
- Are the third-person demonstrative pronominal roots nasalised?
- Third-person demonstrative pronouns in Kannada and their relationship to case suffixes
- The medial/presentative udu and the aggregating suffix uṁ
- Indo-European Languages
- Dravidian Languages
- Second-person Pronominal Roots
- Dravidian Languages
- Indo-European Languages
- First-person Pronominal Roots
- Dravidian Languages
- Caldwell's arguments against nān/nā being the first-person pronominal roots
- Caldwell's arguments for nān/nā being the first-person pronominal roots
- Analysis of Caldwell's arguments
- The vowel that is intermediate to ā and ē suggested by Masti Venkatesha Iyengar
- An afterward on the first, second and third person pronominal roots
- Indo-European Languages
- Dravidian Languages
- Interrogative Pronominal Roots
- Dravidian Languages
- Features of the interrogative pronominal forms
- Interrogative pronominal root for all the variations starting with ā, e/ē, yā, ō and vā
- The difference between the self-referential æ and the interrogative æ
- The processes behind the interrogative/speculative/skeptical suffixes ē, ā, ō and the emphasising suffix ē in Kannada
- Producing all the various interrogative forms
- Indo-European Languages
- Dravidian Languages
- Some Loose-ends
- Summary
- Revisions
It is not surprising that there is a kind of consensus among scholars regarding the third-person pronominal origins in Dravidian languages.
It is easy to see how all the third-person demonstrative pronominal forms can be produced from the
demonstrative single vowels, distal a
,
proximal i
and the medial u
.
The simplicity, beauty and the logical consistency of this system may not surprise the native speakers of Dravidian languages
only because they have naturally inherited it as part of their mother tongue.
All the different third-person demonstrative pronominal forms are produced when grammatical gender and number suffixes are appended to these third-person demonstrative pronominal roots.
an (Masculine Singular) |
aḷ (Feminine Singular) |
du (Neuter Singular) |
ar (Rational Plural) |
vu (Neuter Plural) |
Elongated Vowel (Demonstrative Determiner) | |
---|---|---|---|---|---|---|
a | avan | avaḷ | adu | avar | avu | ā |
i | ivan | ivaḷ | idu | ivar | ivu | ī |
u | uvan | uvaḷ | udu | uvar | uvu | ū |
Only the Old Kannada forms have been given here and Tamil and Modern Kannada forms have been omitted because they vary very little from these forms.
The appearance of the infix v
in the masculine and feminine singular and the rational plural forms (avan
, ivaḷ
, uvar
etc.) needs an explanation.
This is because, normally, when a suffix starting with a vowel (an
, aḷ
, ar
etc.) is appended to a word fragment having an ending vowel,
other euphonic processes might operate to produce forms different from the ones listed above.
For example,
a + an => an
- The ending vowel of the preceding fragment drops according to lōpasaṁdhi.i + aḷ => iyaḷ
- A euphonic infixy
appears between the two fragments according to yakārāgama saṁdhi.u + ar => uvar
- A euphonic infixv
appears between the two fragments according to vakārāgama saṁdhi.
Of these, only the form uvar
appears in Kannada and Tamil; the other forms don't.
So, some other kind of process is required to explain the appearance of the v
infix in these forms (avan
, ivaḷ
etc.).
A previous research article "A Nose for Nasalisation" proposes such a set of processes.
In these processes, nasalisation appears to mediate whenever suffixes are appended to roots.
This nasalisation then further transforms into many different forms according to a related set of processes.
Turning into m
or v
is among such transforms that the mediating nasalisation enables.
I.e.,
In ISO 15919,
a + m̐ + an => am̐ + an => avan
i + m̐ + aḷ => im̐ + aḷ => ivaḷ
In IPA,
ə + ̃ + ən̪ => ə̃ + ən̪ => əʋən̪
ɪ + ̃ + əɭ => ɪ̃ + əɭ => ɪʋəɭ
I have proposed in "A Nose for Nasalisation" that the ending n
in the masculine singular suffix an
is also originally just a nasalisation.
I.e., an
/ən̪
is originally derived from am̐
/ə̃
.
In light of this hypothesis, it is easy to see how not just the Old Kannada form of avan
/əʋən̪
but also avam̐
/əʋə̃
of Havyaka Kannada as well as ava
/əʋə
(without the trailing n
) found in some regional spoken dialects of Kannada are produced.
In ISO 15919,
a + m̐ + am̐ => am̐ + am̐ => avam̐
- The first nasalisation turns intov
.=> ava
- The trailing nasalisation disappears.=> avan
- The trailing nasalisation turns inton
.
In IPA,
ə + ̃ + ə̃ => ə̃ + ə̃ => əʋə̃
- The first nasalisation turns intoʋ
.=> əʋə
- The training nasalisation disappears.=> əʋən̪
- The trailing nasalisation turns inton̪
.
When a mediating nasalisation can (among other things) turn into v
or drop off, it wouldn't be incorrect, for the sake of logical consistency,
to invoke the mediating nasalisation even in the forms involving the medial demonstrative pronominal root u
as well.
I.e.,
In ISO 15919,
u + m̐ + ar => um̐ + ar => uvar
- It can be considered that either the nasalisation has disappeared (and the euphonic infixv
has appeared betweenu
anda
according to vakārāgama saṁdhi) or the nasalisation itself has turned intov
.
In IPA,
u + ̃ + əɾ => ũ + əɾ => uʋəɾ
- It can be considered that either the nasalisation has disappeared (and the euphonic infixʋ
has appeared betweenu
andə
according to vakārāgama saṁdhi) or the nasalisation itself has turned intoʋ
.
The following discussion about neuter gender forms and Tulu third-person pronouns seems to support the case for the mediating nasalisation processes.
There is quite a bit of supporting evidence for mediating nasalisation after word fragments ending in the vowel a
, in Kannada.
The Tulu inflexed forms like maroṁkŭ
(dative), maroṁtŭ
/marattŭ
(ablative), maroṁta
/maratta
(genitive), maroṁṭŭ
/maraṭṭŭ
(locative) etc. show that this is true in Tulu as well.
Here, the alternative geminated forms like marattŭ
and maraṭṭŭ
are clearly derived from the corresponding nasalised forms (maroṁtŭ
and maroṁṭŭ
).
In the case of word-fragments ending with other vowels, the traditional wisdom is that the nominative suffix (really, just a nasalisation) disappears when other suffixes apply. Some Tulu inflexed forms for word fragments with other vowel endings seem to reveal the influence of the nasalisation that is more than just disappearing.
For example,
naḍuṭŭ
/naḍuṭṭŭ
-naḍu
ends inu
.puḍeṭŭ
/puḍeṭṭŭ
-puḍe
ends ine
.bariṭŭ
/bariṭṭŭ
-bari
ends ini
.
Though no nasalisation exists in these examples, the optional gemination seems to betray an ancestry from an original nasalisation.
Among the neuter gender singular pronominal forms in Kannada, forms ending with du
(adu
, idu
, udu
etc.) are most common.
Therefore, du
has been mentioned above as the neuter gender singular suffix.
But there are reasons to think that this is not entirely correct.
For example, the verse 158 of kēśirāja's
śabdamaṇidarpaṇaṁ, mentions the forms ninatu
, enatu
, tanatu
and ninattu
, enattu
, tanattu
for the common forms ninnadu
, ennudu
, tannadu
.
These alternate forms are no longer found in Modern Kannada.
The suffixes in these forms are atu
, attu
and not adu
.
I.e.,
nin + atu => ninatu
en + attu => enattu
This suggests the possibility of tu
being the original neuter gender singular suffix,
because t
transforming into d
is a natural process in Kannada and Dravidian languages.
The reason for the optional gemination between atu
and attu
forms can be found in "A Nose for Nasalisation",
where the nasalisation mediating the appending of suffixes to the third-person pronominal roots can explain the the optional gemination between atu
and attu
.
In ISP 15919,
a + m̐ + tu => am̐tu
=> atu
- The nasalisation disappears.=> aṁtu
- Nasalisation turns into anusvāra which is pronounced asn
because of thet
that follows.=> attu
- The anusvāra which was pronounced asn
turns into a gemination of the following consonantt
.
In IPA,
ə + ̃ + t̪u => ə̃t̪u
=> ət̪u
- The nasalisation disappears.=> ən̪t̪u
- Nasalisation turns into anusvāra which is pronounced asn̪
because of thet̪
that follows.=> ət̪t̪u
- The anusvāra which was pronounced asn̪
turns into a gemination of the following consonantt̪
.
More details about these processes can be seen in "A Nose for Nasalisation".
The fact that other such occurrences of optional gemination of words ending with tu
survive in Modern Kannada (haḷ + atu/attu => haḷatu/haḷattu
, hos + atu/attu => hosatu/hosattu
etc.) strengthens this argument.
So, it can be supposed that the neuter gender singular suffix du
has been derived from the original form tu
.
I.e.
atu => adu
But the forms like ninadu
, enadu
, tanadu
do not appear in Kannada; only forms like ninnadu
, ennadu
, tannadu
are found.
The reason for this is another process of "dvitvasaṁdhi" which is natural to Dravidian languages, especially to Kannada.
It can be seen that, in the word fragments nin
, en
, tan
that appear before adu
, before the trailing consonant n
is a single light syllable (ni
, e
, ta
).
Gemination of the trailing consonant (in this case n
) in such a word fragment, when it is appended with a suffix starting with a vowel,
is a natural process in Kannada which is called dvitvasaṁdhi.
I.e.,
nin + adu => ninnadu - DUMdada (DUM - heavy syllable, da - light syllable)
I have argued in a previous research article "vakārada vibhakti"
(please note that the document is in Kannada and is yet to be translated to English)
that such gemination is one of a set of processes used in Kannada (and other Dravidian languages too to varying extent)
to avoid the daDUM
rhythm which is unnatural in Kannada (even more so in Modern Kanada) at the beginning of words
(especially, at the beginning of metrical units and sentences).
But daDUM
rhythm cannot be seen in the ungeminated form ninadu
either.
So, the reason for the geminated form ninnadu
seems to be the nasalisation hidden (but subsequently disappeared) inside adu
.
That nasalisation, in some production sequences, is capable of producing dada
or daDUM
rhythms both of which are unnatural to Kannada at the beginning of words and metrical units.
So, the trailing consonant n
gets geminated (ninnadu
, ennadu
, tannadu
) in Kannada to avoid such dada
or daDUM
rhythms.
The production sequences could be as follows.
In ISO 15919,
-
nin + a + m̐ + tu => (nin + a + m̐) + tu => ninnam̐ + tu
- Becauseninam̐
would be in the (undesirable)dada
rhythm.=> ninnam̐tu => ninnatu
- The nasalisation disappears.=> ninnadu
-t
turns intod
.
-
nin + a + m̐ + tu => nin + (a + m̐ + tu) => nin + am̐tu => nin + aṁtu
-aṁtu
is found in Kannada but with different semantics.=> nin + aṁdu
-t
turns intod
.aṁdu
is also found in Kannada but with different semantics.=> ninnaṁdu
- Becauseninaṁdu
(daDUMda
) would start with the (undesirable)daDUM
rhythm.=> ninnadu
- The anusvāra, which is pronounced asn
here, disappears.
In IPA,
-
n̪ɪn̪ + ə + ̃ + t̪u => (n̪ɪn̪ + ə + ̃) + t̪u => n̪ɪn̪n̪ə̃ + t̪u
- Becausen̪ɪn̪ə̃
would be in the (undesirable)dada
rhythm.=> n̪ɪn̪n̪ə̃t̪u => n̪ɪn̪n̪ət̪u
- The nasalisation disappears.=> n̪ɪn̪n̪əd̪u
-t̪
turns intod̪
.
-
n̪ɪn̪ + ə + ̃ + t̪u => n̪ɪn̪ + (ə + ̃ + t̪u) => n̪ɪn̪ + ə̃t̪u => n̪ɪn̪ + ən̪t̪u
-ən̪t̪u
is found in Kannada but with different semantics.=> n̪ɪn̪ + ən̪d̪u
-ən̪d̪u
-t̪
turns intod̪
.ən̪d̪u
is also found in Kannada but with different semantics.=> n̪ɪn̪n̪ən̪d̪u
- Becausen̪ɪn̪ən̪d̪u
(daDUMda
) would start with the (undesirable)daDUM
rhythm.=> n̪ɪn̪n̪əd̪u
- The anusvāra, which was pronounced asn̪
here, disappears.
More details about these processes can be seen in "vakārada vibhakti".
The existance of ninattu
, enattu
, tanattu
forms which start with the (undesirable) daDum
rhythm, might seem to contradict
the above mentioned processes.
But like the exceptions listed in the same "vakārada vibhakti", these exceptional forms are found only in Old Kannada
and have disappeared from Modern Kannada.
This supports the view that the processes related to dvitvasaṁdhi have dominated more in Modern Kannada when compared to Old Kannada.
The argument that there is a nasalisation hidden inside the third-person pronouns finds strong support in the Tulu proximal demonstrative pronouns (ŭṁbye
, ŭṁbaḷŭ
, ŭṁdu
etc.).
Here, the nasalisation appears as full-blown anusvāra which is pronounced as the euphonic nasal consonant of the succeeding consonant.
The b
that appears in the masculine, feminine singular and the rational plural forms (ŭṁbye
, ŭṁbaḷŭ
, ŭṁberŭ
etc.) seems to be
a latter phenomenon that was not present in the root forms.
I.e.,
In ISO 15919,
ŭ + m̐ + e
-e
seems to be the root masculine singular suffix in Tulu.=> ŭṁ + ye
- The nasalisation turns into anusvāra which is pronounced here asm
and an euphonicy
infixes beforee
.=> ŭṁye
- The anusvāra is pronounced here asm
,ŭmye
.=> ŭṁbye
- Pronouncedŭmbye
.
ŭ + m̐ + aḷŭ
- As in other Dravidian languages,aḷŭ
is a feminine singular suffix in Tulu too.=> ŭṁ + aḷŭ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ŭṁaḷŭ
- The anusvāra is pronounced here asm
,ŭmaḷŭ
.=> ŭṁbaḷŭ
- Pronouncedŭmbaḷŭ
.
ŭ + m̐ + du
- As in other Dravidian languages,du
is the neuter gender singular suffix in Tulu too.=> ŭṁdu
- The nasalisation turns into anusvāra which is pronounced here asn
due to thed
that follows,ŭndu
.
ŭ + m̐ + erŭ
-erŭ
seems to be the root rational plural suffix in Tulu.=> ŭṁ + erŭ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ŭmerŭ
- The anusvāra is pronounced here asm
.=> ŭṁberŭ
- Pronouncedŭmberŭ
.
ŭ + m̐ + kŭḷŭ
-kŭḷŭ
(likegaḷu
in Kannada) seems to be a plural suffix in Tulu.=> ŭṁ + kŭḷŭ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ŭṁkŭḷŭ
- The anusvāra is pronounced here asm
,ŭmkŭḷŭ
.=> ŭṁbŭkŭḷŭ => ŭṁbakŭḷŭ
- The anusvāra is pronounced here asm
,=> ŭmbŭkŭḷŭ => ŭmbakŭḷŭ
.
In IPA,
ɯ + ̃ + e
-e
seems to be the root masculine singular suffix in Tulu.=> ɯm + je
- The nasalisation turns into anusvāra which is pronounced here asm
and an euphonicj
infixes beforee
.=> ɯmje
- The anusvāra is pronounced here asm
.=> ɯmbje
- The anusvāra is pronounced here asm
.
ɯ + ̃ + əɭɯ
- As in other Dravidian languages,əɭɯ
is a feminine singular suffix in Tulu too.=> ɯm + əɭɯ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ɯməɭɯ
- The anusvāra is pronounced here asm
.=> ɯmbəɭɯ
- The anusvāra is pronounced here asm
.
ɯ + ̃ + d̪u
- As in other Dravidian languages,d̪u
is the neuter gender singular suffix in Tulu too.=> ɯn̪d̪u
- The nasalisation turns into anusvāra which is pronounced here asn̪
due to thed̪
that follows.
ɯ + ̃ + eɾɯ
-eɾɯ
seems to be the root rational plural suffix in Tulu.=> ɯm + eɾɯ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ɯmeɾɯ
- The anusvāra is pronounced here asm
.=> ɯmbeɾɯ
- The anusvāra is pronounced here asm
.
ɯ + ̃ + kɯɭɯ
-kɯɭɯ
(likegəɭu
in Kannada) seems to be a plural suffix in Tulu.=> ɯm + kɯɭɯ
- The nasalisation turns into anusvāra which is pronounced here asm
.=> ɯmkɯɭɯ
- The anusvāra is pronounced here asm
.=> ɯmbɯkɯɭɯ => ɯmbakɯɭɯ
- The anusvāra is pronounced here asm
.
The appearance of an euphonic b
as an infix after the anusvāra which is pronounced as m
is similar to the linguistic processes
in other languages, where an euphonic consonant comes as infix after a nasal consonant.
For example, the name Henry
has an alternate form Hendry
; similarly, Thomson
has the alternate form Thompson
.
So, it would not be wrong to say that ŭṁbye
, ŭṁbaḷŭ
, ŭṁberŭ
etc. are derived from ŭmye
, ŭmaḷŭ
, ŭmerŭ
etc., respectively.
Furthermore, the alternate forms of these pronouns which start with i
instead of ŭ
(iṁbye
, iṁbaḷŭ
etc.) must have been derived
by the ŭ
turning into i
and the alternate forms which start with the consonant m
instead of a vowel (mekḷŭ
, mokḷŭ
, merŭ
, mōḷŭ
etc.) must have been derived
from the forms without the euphonic b
infix (ŭmkŭḷŭ
, ŭmerŭ
etc.) by losing the initial ŭ
.
From this discussion, it becomes clear that nasalisation is indeed hidden inside the third-person demonstrative pronouns of Dravidian languages.
Also, the close relation between the euphonic infixes v
, found in ivanu
, ivaḷu
, ivaru
of Kannada
and b
preceded by the nasal anusvāra/m
, found in ŭṁbye
, ŭṁbaḷŭ
, ŭṁberŭ
of Tulu is notable.
Clearly, the euphonic v
, anusvāra/m
, b
are derived from the original nasalisation.
The similarities in pronunciation of v
, m
and b
is also noteworthy.
In most of the other Dravidian languages, i
and u
are used as the proximal and medial demonstrative pronominal roots, respectively.
But in Tulu, ŭ
is used as the proximal demonstrative pronominal root
and the medial demonstrative pronominal forms do not exist (at least in the modern dialects).
Let's proceed to investigate the reasons behind this peculiarity next.
Tulu has a dative presentative ŭṁda
,
where the anusvāra is pronounced as n
due to the d
that follows (ŭnda
).
Here ŭ
is used in both the proximal demonstrative as well the presentative senses.
Kannada has the equivalent ikō
/igō
which are clearly derived from the proximal demonstrative i
.
Also, the anusvāra/n
here is clearly derived from the original nasalisation.
The derivation of the Kannada equivalents for the Tulu ŭṁda
, viz. ikō
/igō
, akō
/agō
can be seen here.
Tulu using ŭ
in the the proximal demonstrative pronouns and the dative presentative form ŭṁda
(the Kannada equivalent being ikō
/igō
),
suggests that the original form and semantics of the root vowel u
might be subtler than the commonly accepted medial demonstrative sense.
Regarding the origins of the form, a possibility is that u
might be originally derived from ŭ
.
This is supported by the fact that Kannada and Telugu use u
wherever Tulu uses ŭ
and that wherever it doesn't affect semantics, alternate forms for the words with ŭ
can be seen even in Tulu where ŭ
turns into u
.
But it is difficult to conclude that u
is derived from ŭ
without further supporting arguments and evidence.
The medial demonstrative pronominal forms with the root u
are quite rare in most Dravidian languages (especially so in the modern spoken dialects).
They are almost non-existant as independent pronominal forms in Modern Kannada.
But the root is still found in fragments and suffixes.
For example,
naḍe + udu => naḍevudu
naḍe + u => naḍeyu
naḍeyu + udu => naḍeyuvudu
tiḷi + udu => tiḷivudu
tiḷi + u + udu => tiḷiyuvudu
ā + da + udu => ādudu
ā + udu => āvudu
yā + udu => yāvudu
ā + uvu => āvuvu
yā + uvu => yāvuvu
The u
found in these forms turns into a
in the northern dialects of Kannada.
For example, aṁbikātanayadatta in his famous poem "hr̥daya samudra" says miṁcubaḷaga tereteregaḷāgi aleyuvadu
puṭṭapūrā.
Havyaka Kannada also shows the same process of u
turning into a
.
For example, naḍevadu
, tiḷivadu
etc.
Words ending in a consonant being suffixed with a euphonic u
(kaṇ => kaṇṇu
, kāl => kālu
) is a well known phenomenon in Modern Kannada.
This is also very common in Telugu.
Spoken dialects of Tulu, Tamil and Malayalam usually pronounce the consonant ending with a suffixed vowel ŭ
(kaṇ => kaṇṇŭ
, kāl => kālŭ
etc.).
Also, in Modern Kannada, when such words where the original ŭ
has turned into u
are suffixed with anything starting with a vowel, the trailing u
of the preceding word usually disappears following the process of lōpasaṁdhi (kaṇṇu + iḍu => kaṇṇiḍu
etc.).
It is also common for u
to disappear in the spoken dialects of Modern Kannada even in the middle of words.
For example, ādudariṁda => āddariṁda
.
It is tempting to speculate that u
being originally derived from ŭ
might be the reason behind u
sometimes dropping off or being replaced by a
,
because the pronunciation of ŭ
seems fit to turning into a
, u
and disappearing.
But it is hard to conclude based only on these speculations without further supporting arguments and evidence.
So, I leave the discussion on the origins of the form of the demonstrative root vowel u
here, incomplete.
However, the above discussion sheds more light on the semantics of the demonstrative root vowel u
too.
As already mentioned above, most scholars have considered the vowel u
to be the medial demonstrative pronominal root and rightly so.
But that may not be its only semantics.
It seems appropriate to suppose that, in addition to the medial demonstrative semantics, it also has presentative semantics.
Presentative as in here you go
or how a lawyer might present evidence saying Exhibit A
;
but here, I propose that Dravidian languages use the root vowel u
in the presentative sense much more systematically
(as they do the other pronominal roots) than in English.
As seen above, the semantics of the Tulu ŭṁda
is clearly presentative.
Even in Kannada, the suffix udu
in māḍuvudu
, baredudu
etc. adds the semantics of presence or relevance to the
action indicated by the verb root; to say it adds the semantics of medial distance to the action makes no sense.
Since the presentative semantics has clear overtones of both metaphorical and physical medial (and proximal) distance,
it seems appropriate to even say that the medial demonstrative semantics of u
is derived from the original presentative semantics.
Considering that the presentative semantics also indicates logical and/or physical proximity too,
it seems appropriate that Tulu uses ŭ
as the proximal demonstrative pronominal root.
In Kannada, masters of ceremony use the proximal demonstrative rational plural ivaru
to introduce / present important guests to the audience.
This also indicates a close relationship between not only the root vowels i
and u
but also the proximal, medial demonstrative and presentative
semantics.
Because the presentative pronominal forms derived from u
have now disappeared from use in Modern Kannada,
the semantically related proximal demonstrative pronominal form ivaru
seems to have been pressed into service in their stead.
If the presentative pronouns derived from u
were still in use in Kannada, it might have been more appropriate to use uvaru
on such occasions.
If it is unavoidable to lose one of either of the distal a
, the proximal i
or the presentative u
, it might be more logical to lose the proximal i
like in Tulu
than to lose the presentative u
like in Kannada.
But languages do not evolve based on logical analyses, of course.
But that languages largely show logical consistency is undeniable
and it is fortunate that these three root vowels (a
, i
and u
) have been used in such simple, logical and beautiful regularity
in the Proto-Dravidian language.
The fact that this simple and logical regularity has survived to a great extent (though partially) even in the modern Dravidian languages,
leaves significant clues that aid linguists in the search of pronominal roots.
The fact that the presentative semantics of u
metaphorically and/or physically indicates not only a medial (u
) distance but also proximity (i
),
will come in handy in the investigation of the second-person pronominal roots later.
As seen above, the third-person distal demonstrative neuter gender singular adu
must have been derived from the original atu
/attu
.
The neuter gender singular suffixes of finite verb forms in Havyaka Kannada seem to support this argument further.
For example,
hō + attu => hōdattu (with d infix)
- In Kannada,hōyitu
tin + attu => tiṁdattu (with d infix)
- In Kannada,tiṁditu
Similarly, that Old Kannada must have had the original forms utu
/uttu
from which the third-person medial demonstrative neuter gender singular udu
must have derived, is also clear from the neuter gender suffixes of finite verb forms in Havyaka Kannada.
For example,
hō + uttu => hōvuttu (with v infix)
- In Kannada,hōguvudu
hēḷ + uttu => hēḷuttu
- In Kannada,hēḷuvudu
As already seen above, though the third-person presentative/medial demonstrative
neuter gender singular pronoun udu
of Old Kannada might have disappeared in Modern Kannada, it has survived as a neuter gender singular suffix
of finite verb forms.
Similarly, the apparently third-person proximal demonstrative neuter gender singular itu
found in anitu
/initu
/enitu
,
can be seen as a neuter gender singular suffix of finite verb forms.
For example,
hō + itu => hōyitu (with the euphonic y infix)
tin + itu => tiṁditu (with d infix)
hēḷ + itu => hēḷitu
But even in Old Kannada (and in many modern regional dialects), there are alternate forms without the vowel i
.
For example, hōtu
, tintu
, hēḷtu
, etc.
So, proximal i
or presentative/medial u
might sometimes appear at the beginning of neuter gender suffixes of finite verb forms.
As already mentioned above, the neuter gender singular suffix du
must have been originally derived from tu
.
It was also seen that nasalisation mediates
whenever a suffix is appended to pronominal roots and leads to many different forms based on a set of related processes.
Of these forms of the third-person distal demonstrative pronoun, atu
, aṁtu
, attu
, adu
, aṁdu
are significant.
I.e.,
In ISO 15919,
a + m̐ + tu => am̐tu
=> atu
- The nasalisation disappears.=> aṁtu
- The nasalisation turns into anusvāra which is pronounced here asn
due to thet
that follows,antu
.=> attu
- The anusvāra turns into a gemination of the following consonantt
.
=> am̐du
-t
turns intod
.=> adu
- The nasalisation disappears.=> aṁdu
- The nasalisation turns into anusvāra which is pronounced asn
due to thed
that follows,andu
.
In IPA,
ə + ̃ + t̪u => ə̃t̪u
=> ət̪u
- The nasalisation disappear.=> ən̪t̪u
- The nasalisation turns into anusvāra which is pronounced here asn̪
due to thet̪
that follows.=> ət̪t̪u
- The anusvāra turns into a gemination of the following consonantt̪
.
=> ə̃d̪u
-t̪
turns intod̪
.=> əd̪u
- The nasalisation disappears.=> ən̪d̪u
- The nasalisation turns into anusvāra which is pronounced asn̪
due to thed̪
that follows.
Of these forms, atu
, attu
, adu
have the same semantics in Old Kannada as adu
in Modern Kannada.
But aṁtu
and aṁdu
have different semantics.
The same set of processes derive the proximal demonstrative forms itu
, iṁtu
, ittu
, idu
, iṁdu
and the interrogative forms etu
, eṁtu
, ettu
, edu
, eṁdu
as well.
Of these forms too, iṁtu
, iṁdu
, eṁtu
, eṁdu
have semantics different from their related standard pronominal forms.
Some forms (for example, edu
) are in use only in some regional dialects,
while others (for example, etu
, ettu
) are not at all found amongst the pronominal forms in Kannada.
Some closely related forms to the above mentioned pronominal forms are also notable here.
For example, aṁta
, atta
, anitu
.
Their corresponding proximal iṁta
, itta
, initu
and interrogative eṁta
, etta
, enitu
also exist.
Of these too, some forms (for example, eṁta
) are in use only in some regional dialects.
For the sake of argument, let us first keep the semantic differences aside and analyse only the similarities in the processes of derivation.
The similarities in the processes of derivation of
atu
/itu
/etu
, aṁtu
/iṁtu
/eṁtu
, attu
/ittu
/ettu
, adu
/idu
/edu
, aṁdu
/iṁdu
/eṁdu
are already shown above
and the processes that derive the forms aṁta
/iṁta
/eṁta
, atta
/itta
/etta
, anitu
/initu
/enitu
are not hard to see.
I.e.,
aṁtu + a => aṁta
ittu + a => itta
The derivation of anitu
/initu
/enitu
becomes clear by the set of processes explained in "A Nose for Nasalisation".
I.e., nasalisation mediates when the root vowels, distal a
, proximal i
, interrogative e
, are suffixed with the proximal demonstrative neuter gender singular pronoun itu
and the nasalisation then turns into n
, resulting in anitu
/initu
/enitu
.
I.e.,
In ISO 15919,
e + m̐ + itu => enitu
- The nasalisation turns inton
.
In IPA,
e + ̃ + ɪt̪u => en̪ɪt̪u
- The nasalisation turns inton̪
.
The Kannada dictionary published by
the Kannada Sahitya Parishat
gives the alternative forms anittu
/inittu
/enittu
for anitu
/initu
/enitu
.
This suggests that itu
/ittu
alternate forms also existed in Old Kannada
similar to the atu
/attu
alternate forms as can be seen in ninatu
/enatu
/tanatu
and ninattu
/enattu
/tanattu
as already mentioned above.
So, the existance of the forms and the processes of derivations of atu
/attu
and itu
/ittu
cannot be mere speculation.
The same dictionary gives the forms anibar
/inibar
/enibar
as the rational plural counterparts of anitu
/initu
/enitu
.
This confirms that atu
/itu
, attu
/ittu
forms had the neuter gender singular semantics in Old Kannada,
because it is easy to see that the anibar
/inibar
/enibar
forms are produced by suffixing the proximal demonstrative rational plural pronoun ivar
to the root vowels, the distal a
, the proximal i
and the interrogative e
.
I.e.,
In ISO 15919,
a/i/e + m̐ + ivar => anivar/inivar/enivar
- The nasalisation turns inton
.=> anibar/inibar/enibar
-v
turning intob
is a common process in Kannada.
In IPA,
ə/ɪ/e + ̃ + ɪʋəɾ => ən̪ɪʋəɾ/ɪn̪ɪʋəɾ/en̪ɪʋəɾ
- The nasalisation turns inton
.=> ən̪ɪbəɾ/ɪn̪ɪbəɾ/en̪ɪbəɾ
-ʋ
turning intob
is a common process in Kannada.
So, this confirms not only that anitu
/initu
/enitu
contain the suffix itu
,
but also that tu
is the original neuter gender singular suffix from which du
is derived,
because, while t
turning into d
is a natural process in Kannada, d
turning into t
is not so common.
Having analysed the similarities in the processes of derivation of these related pronominal forms in isolation, let us now turn towards the differences in their semantics.
As seen above, though not used as independent pronouns, atu
/attu
, itu
/ittu
were used in Old Kannada as suffixes
in the third-person demonstrative neuter gender singular pronominal semantics as seen in ninatu
/ ninattu
, anitu
/anittu
etc.
But aṁtu
/iṁtu
/eṁtu
, aṁta
/iṁta
/eṁta
, atta
/itta
/etta
, anitu
/initu
/enitu
have slightly different,
though still closely related to the neuter gender singular, semantics.
So, it is natural to question if it is possible to decide on the roots and processes of derivation of these forms
based only on the similiarities in the pronominal root and processes of derivation.
These forms (or forms close to these) are found in other Dravidian languages. But in some cases their semantics are close to and in other cases very different from their counter-parts in Kannada.
Please note that I am not very familiar with the transliteration of Tamil and Malayalam in ISO 15919. So, I might have made mistakes in the transliteration of Tamil and Malayalam here and elsewhere in this document. I would be grateful if knowledgeable readers point out the mistakes and help improve the document.
For example, in Tamil,
aṁdu
- (writtenantu
),in that way
aṁṟu
-that day
aṁda
- (writtenaṁta
),there
,that
iṁda
- (writteniṁta
),here
,this
iṁṟu
-this day
eṁdu
- (writteneṁtu
),how
eṁda
-which
,where
?enna
-what
ennŭkkŭ
-why
In Telugu,
aṁdi
-that day
aṁdu
-there
aṁta
-that much
aṁtu
-that much
?atti
-to that side
iṁta
-this much
iṁdu
-here
eṁdi
-which day
eṁdu
-where
eṁduku
-why
eṁta
-how much
enni
-how many things
In Malayalam
In Tulu,
In the different Dravidian languages, these words seem to have been produced from similar roots and by similar processes of derivation. But not only are the semantic differences obvious, there also seems to be difference in the semantic differences. I.e., there is no commonality even in the semantic difference. In other words, from language to language the ways by which semantics change also seem to be different. So, it doesn't seem appropriate to suppose that these words acquired different semantics because they were produced from different roots following different processes of derivation. Because if it was so, there wouldn't have been so much similarity in the word forms. Instead, it seems more appropriate to suppose that these words were produced from the same Proto-Dravidian roots, undergoing the same set of processes of derivation, leading to the original third-person neuter gender singular semantics, which later in due course of time in the various branches of the language family, came to acquire different semantics, though still closely related to the original third-person neuter gender singular semantics.
The way the semantics of these words changed in Kannada can be supposed as follows.
-
aṁdu
- The original distal (in the sense of space) demonstrative neuter gender singular pronoun (equivalent ofadu
) becomes distal in the sense of time, leading to the current Kannada semantics ofthat day
,that time
. If this is one possibility, another is that Kannadaaṁdu
, Tamilaṁṟu
and Malayalamannu
all derive from a common ancestral form. -
iṁdu
- The original proximal (in the sense of space) demonstrative neuter gender singular pronoun (equivalent ofidu
) becomes proximal in the sense of time, leading to the current Kannada semantics ofthis day
,this time
. If this is one possibility, another is that Kannadaiṁdu
, Tamiliṁṟu
and Malayalaminnu
all derive from a common ancestral form. -
eṁdu
- Thoughedu
now survives in the interrogative sense (what
) only in some regional dialects of Kannada, there are reasons to suppose thatedu
might be the original interrogative neuter gender pronominal form from whichyāvudu
is derived. More details about this can be seen here. So, the original interrogative (in the sense of space) neuter gender singular pronoun (equivalent ofedu
) becomes interrogative in the sense of time, leading to the current Kannada semantics ofwhich day
,which/what time
,when
. If this is one possibility, another is that Kannadaeṁdu
, Tamileṁṟu
and Malayalamennu
all derive from a common ancestral form. -
aṁtu
- The original distal (in the sense of space) demonstrative neuter gender singular pronoun (equivalent ofatu
) becomes distal in the sense of way or procedure, leading to the current Kannada semantics ofin that way
. -
iṁtu
- The original proximal (in the sense of space) demonstrative neuter gender singular pronoun (equivalent ofitu
) becomes proximal in the sense of way or procedure, leading to the current Kannada semantics ofin this way
,thus
. -
eṁtu
- The original interrogative (in the sense of space) neuter gender singular pronoun (equivalent ofetu
/edu
) becomes interrogative in the sense of way or procedure, leading to the current Kannada semantics ofhow
,in which way
. -
atta
- When the original distal (in the sense of space) demonstrative pronominal formattu
(equivalent ofatu
/adu
) is suffixed witha
(which, among other things, has genitive semantics), (i.e.,attu + a => atta
), it naturally acquires a distal directional semantics, leading to the current Kannada semantics ofon/to that side
,in that direction
. -
itta
- When the original proximal (in the sense of space) demonstrative pronominal formittu
(equivalent ofitu
/idu
) is suffixed witha
(which, among other things, has genitive semantics), (i.e.,ittu + a => itta
), it naturally acquires a proximal directional semantics, leading to the current Kannada semantics ofon/to this side
,in this direction
. -
etta
- When the original interrogative (in the sense of space) pronominal formettu
(equivalent ofedu
) is suffixed witha
(which, among other things, has genitive semantics), (i.e.,ettu + a => etta
), it naturally acquires an interrogative directional semantics, leading to the current Kannada semantics ofon/to which side
,in which direction
. -
aṁta
- When the original distal (in the sense of space) demonstrative pronominal formaṁtu
(equivalent ofatu
/adu
) is suffixed witha
(which, among other things, has genitive semantics), (i.e.,aṁtu + a => aṁta
), it naturally acquires a distal directional semantics, which includes the semantics of a distal demonstrative determiner like the much more widely usedā
. This argument is supported by the fact that the relatedaṁda
of Tamil (written asaṁta
) is indeed used as a distal demonstrative determiner. Butaṁta
is rarely used in the distal demonstrative determiner in Kannada. Perhaps it can be found as a suffix in the words likehāgaṁta
,ēnaṁta
, though, it is possible, that has a different derivation and semantics. -
iṁta
/iṁda
- When the original proximal (in the sense of space) demonstrative pronominal formsiṁtu
/iṁdu
(equivalents ofitu
/idu
) is suffixed witha
(which, among other things, has genitive semantics), (i.e.,iṁtu/iṁdu + a => iṁta/iṁda
), they naturally acquire a proximal directional semantics, which includes the semantics of a proximal demonstrative determiner like the much more widely usedī
. This argument is supported by the fact that the relatediṁda
of Tamil (written asiṁta
) is indeed used as a proximal demonstrative determiner. Butiṁta
/iṁda
also are rarely used in the proximal demonstrative determiner in Kannada. Butiṁta
can still be seen as a suffix in forms likebānigiṁta
,beṭṭakkiṁta
etc. where it signifies a direction physically or metaphorically different from the semantics of the preceding word fragment. Likewise,iṁda
can be seen as a suffix in the ostensibly instrumental forms likebāniṁda
,maradiṁda
etc. More detailed discussion of these suffixes can be seen here. -
eṁta
- When the original interrogative (in the sense of space) pronominal formeṁtu
(equivalent ofedu
) is suffixed witha
, (which, among other things, has genitive semantics), (i.e.,eṁtu + a => eṁta
), it naturally acquires an interrogative directional semantics, which includes the semantics of an interrogative determiner likeyāva
orwhich
. Since the specific (for example,which
) and the general (for eample,what
) interrogative determiner forms are closely related in many languages, for example, in English,Which delicacy is this?
vs.What delicay is this?
, in Kannada,yāva tiṁḍiyidu?
vs.ēnu tiṁḍiyidu?
, it is not surprising thateṁta
is used in some dialects of Kannada (like Havyaka Kannada, Mangaluru Kannada etc.) in both the specific (which
) and the general (what
) sense. For example,nīnu eṁta (eṁta = ēnu) hēḷiddu?
(what did you say?
),adu eṁta (eṁta = yāva) mara?
(which/what kind of tree is that?
). -
anitu
/initu
/enitu
,anibar
/inibar
/enibar
- When the original demonstrative root vowels, the distala
, the proximali
and the interrogativee
, are suffixed with the proximal demonstrative pronouns, the neuter gender singularitu
and the rational pluralivar
, the formsanitu
/initu
/enitu
andanibar
/inibar
/enibar
are produced. Here, it seems that, the proximal demonstrative pronounsitu
andivar
seem to have been used in the sense of quantity, leading to the semantics ofthis/that/how much/many
.
Thus, when variant forms are produced by a related set of processes acting on the same roots and suffixes, the variant forms naturally develop variations in semantics over time and in time, if the original language diverges into dialects and languages, different branches of the divergence are likely to diverge in the variations of the semantics too, which is what is seen in these sets of related words. So, just because words have different semantics in the modern dialects of languages, it is not necessary to conclude that they must have been produced by different roots, suffixes and processes.
Of the pronominal forms discussed above, it is a commonly held view that aṁta
/iṁta
/eṁta
have formed
from aṁtaha
/iṁtaha
/eṁtaha
by losing the trailing ha
.
Alternatively, it is also said that aṁta
/iṁta
/eṁta
have formed from aṁtha
/iṁtha
/eṁtha
(having the same semantics as aṁtaha
/iṁtaha
/eṁtaha
), by the aspirated th
turning into its unaspirated counter-part t
.
For example, the Kannada Wiktionary seems to consider iṁta
as an alternate form of iṁtha
.
But native Kannada (Dravidian, in fact) words do not have mahāprāṇas/aspirated consonants
and it is well known in linguistics that pronominal forms are usually native to the languages and rarely borrowed.
So, it is clear that aṁtha
/iṁtha
/eṁtha
have formed from aṁtaha
/iṁtaha
/eṁtaha
by fusing of the traliing ta
and ha
into the aspirated form tha
.
This means that aṁta
/iṁta
/eṁta
cannot have been derived from aṁtha
/iṁtha
/eṁtha
.
And it is not completely correct to say that aṁta
/iṁta
/eṁta
have formed from aṁtaha
/iṁtaha
/eṁtaha
either,
because eṁta
, which is widely used in the western dialects (Mangaluru and Havyaka Kannada etc.),
clearly has semantics different from eṁtaha
in those dialects.
aṁtaha
/iṁtaha
/eṁtaha
of the Middle and Modern Kannada have derived by suffixing aṁtu
/iṁtu
/eṁtu
with aha
,
which itself is derived from the appa
of Old Kannada.
I.e.
aṁtu/iṁtu/eṁtu + appa => aṁtappa/iṁtappa/eṁtappa => aṁtaha/iṁtaha/eṁtaha
Naturally, semantics of these forms are something like that kind of/this kind of/what kind of
.
While eṁta
form is indeed used in many dialects with these semantics, is almost never used with these semantics
in the western dialects of Kannada (Mangaluru, Havyaka Kannada etc.).
For example,
īga eṁta māḍudu?
- meaning,what to do now?
eṁta hēḷtā iddi?
- meaning,what are you saying?
niṁge illi eṁta kelasa?
- meaning,what work/business do you have here?
adu eṁta mara?
- meaning,what/which tree is that?
. It might seem the semantics ofwhat kind of
is also appropriate here, but seeing the above examples and the close relationship between the specific and the general interrogative determiners, the semantics ofwhat
/which
seems to be more appropriate here.nīnu eṁtakke illi baṁdaddu?
- meaning,why did you come here?
. It is notable that Tamil also uses the equivalent formennŭkkŭ
with the same semantics.
From these examples, it seems more appropriate to say that eṁta
, as used in the western dialects (Mangaluru, Havyaka Kannada etc.),
is produced when the interrogative neuter gender singular pronominal form eṁtu
(equivalent of etu
/edu
) is suffixed with a
,
which, among other things, has genitive semantics, rather than to say that it is derived from aṁtaha
/iṁtaha
/eṁtaha
.
If there were no possbile set of processes to derive aṁta
/iṁta
/eṁta
from anything other than aṁtaha
/iṁtaha
/eṁtaha
,
we might have had to be content to conclude that that is indeed the derivation and the change in semantics in the western dialects
might have happened over a period of divergence.
But since process of producing aṁta
/iṁta
/eṁta
by suffixing a
to aṁtu
/iṁtu
/eṁtu
is clear, as already shown above,
it is appropriate to say that eṁta
is also derived from eṁtu
and hence has the interrogative neuter gender singular pronominal semantics too.
Given this set of arguments, it seems unfortunate that the Kannada Dictionary of Karnataka Sahitya Parishat gives only the
aṁtappa/iṁtappa => aṁta/iṁta
derivation and does not even have an independent entry for eṁta
.
In all the processes described above, nasalisation mediates the appending of suffixes to the pronominal roots. Alternatively, the same phenomenon could be described as the pronominal roots are originally nasalised themselves. There doesn't seem to be much qualitative difference between these two descriptions of the same underlying phenomenon. Both the descriptions lead to the same set of processes and pronominal forms.
As already argued above, the forms like iṁta
, iṁda
are derived from the original proximal (in the sense of space) demonstrative neuter gender singular pronominal forms iṁtu
/iṁdu
(equivalents of itu
/idu
),
by suffixing a
, which, among other things, has genitive semantics, i.e., iṁtu/iṁdu + a => iṁta/iṁda
.
But they are not used as independent pronouns, but only used as suffixes in bānigiṁta
, bāniṁda
etc.
to indicate a different physical or metaphorical direction from the semantics of the preceding word fragment.
In Sanskrit, a single ablative form ākāśāt
has the semantics of both the example words bānigiṁta
and bāniṁda
.
Sediyapu Krishna Bhat, in his pathbreaking research article "paṁcamī vibhakti"
(this article has been included by Dr. Padekallu Vishnu Bhat
in the book "vicāraprapaṁca" which is a collection of writings of Sediyapu Krishna Bhat),
has argued against the modern scholarly consensus that Kannada doesn't have an ablative suffix and for ablative semantics too it use the
instrumental suffix.
He goes on to argue that the ostensibly instrumental suffixes in Kannada (iṁ
, iṁda
etc.) are really ablative suffixes
and that for the instrumental case, Kannada relies mostly on the locative suffix
(for example, kattiyalli kaḍi
etc.).
In support of this, he argues that a kind of (physical or metaphorical) separation is key to the ablative case.
I.e., in maradiṁda haṇṇu bittu
, maradiṁda
is the ablative form because of the separation of the fruit from the tree.
Similarly, in kattiyalli kaḍi
, kattiyalli
is the instrumental form because katti
is instrumental in the action kaḍi
,
though alli
is generally considered to be the locative suffix.
However, in the passive voice construction (which is unnatural and rare in Kannada)
rāmaniṁda ahalyeya śāpavimōcaneyāyitu
, rāmaniṁda
is the instrumental form because rāma
is instrumental in the lifting of ahalya
's curse.
So, iṁda
is indeed used as an instrumental suffix in Kannada in the passive voice constructions.
But given that passive voice is unnatural and rare in Kannada, Sediyapu Krishna Bhat's thesis that the sufixes iṁ
, iṁda
etc. are
mainly ablative suffixes in Kannada seems sound, overall.
It is notable that iṁta
suffix is only ever used in the ablative sense and never in the instrumental sense.
To recall again, iṁta
, iṁda
are derived from the original proximal (in the sense of space) demonstrative neuter gender singular pronominal forms
iṁtu
/iṁdu
(equivalents of itu
/idu
), by suffixing a
, which, among other things, has genitive semantics, giving them a kind of proximal directional semantics.
When such a proximal demonstrative/directional suffix is appended to any word fragment,
it suggests a physical or metaphorical proximal direction from the semantics of the preceding word fragment; in other words, separation,
which is crucial to the ablative semantics.
So, it is natural that they are used as ablative suffixes in Kannada.
Taken together, this validates Sediyapu Krishna Bhat's argument that iṁ
, iṁda
etc. are ablative suffixes in Kannada,
not only from the perspective of semantics and usage, but also from the Proto-Dravidian third-person proximal demonstrative pronominal root
and the linguistic and grammatical processes that act on (or make use of) it.
This inexorably leads to the conclusion that the ablative/instrumental suffix iṁ
in Kannada
and the Proto-Dravidian third-person proximal demonstrative pronominal root im̐
/ɪ̃
are one and the same.
I.e., Kannada has put the third-person proximal demonstrative pronominal root (im̐
/ɪ̃
) to use as the ablative/instrumental suffix too.
But Old Kannada scholars have insisted that iṁ
/iṁda
/iṁde
are instrumental suffixes and attaṇiṁ
is the ablative suffix
(for example, śabdamaṇidarpaṇaṁ, verse 103).
Some Modern Kannada scholars have argued that iṁ
/iṁda
/iṁde
are indeed only instrumental suffixes,
whereas deseyiṁda
is the ablative suffix, but an unnatural one, especially, in the spoken dialects of Modern Kannada
(for example , "kannaḍa kaipiḍi", pages 413, 414 and 406).
As already seen above, iṁ
/iṁda
/iṁde
are not exclusively instrumental suffixes, but in fact, they are more naturally ablative suffixes
(more details about the derivation of iṁde
form, iṁda + e => iṁde
, can be seen here; please note that the document is in Kannada and is yet to be translated to English).
However, deseyiṁda
is indeed unnatural to Kannada and rare outside works on linguistics and grammar.
Certainly, it did not catch on in speech.
It is relevant here to note that though deseyiṁda
has dese
in the beginning which gives it the directional semantics,
the ablative semantics is achieved only when it is suffixed with iṁda
, the ostensibly instrumental suffix.
Without the iṁda
suffix, only dese
is not fit to be used as an ablative suffix.
Now let's investigate the case for attaṇiṁ
as an ablative suffix in Old Kannada.
It is clear that attaṇiṁ
is formed from the fragments attu
, aṇ
and iṁ
.
It is well known in Old Kannada grammar that the case declensions of directional words get an aṇ
infix
(śabdamaṇidarpaṇaṁ, verse 120).
But I have argued in a previous research article that the aṇ
infix is in fact derived from the accusative
suffix aṁ
/am̐
/ə̃
mediating the case declension of such directional words (even in cases other than the accusative).
I.e., including the process of derivation of attu
, the process for deriving attaṇiṁ
could be as follows.
In ISO 15919,
a + m̐ + tu + a + m̐ + i + m̐ => (a + m̐ + tu) + (a + m̐) + (i + m̐)
=> (am̐tu) + am̐ + (im̐)
=> (aṁtu) + am̐ + iṁ
- The nasalisation turns into anusvāra which is pronounced asn
because of thet
that follows.=> (attu + am̐) + iṁ
- The anusvāra turns into a gemination of the following consonantt
.=> attam̐ + iṁ
- The trailingu
disappears according to lōpasaṁdhi.=> attaṇiṁ
- The nasalisation turned intoṇ
.
In IPA,
ə + ̃ + t̪u + ə + ̃ + ɪ + ̃ => (ə + ̃ + t̪u) + (ə + ̃) + (ɪ + ̃)
=> (ə̃t̪u) + ə̃ + (ɪ̃)
=> (ən̪t̪u) + ə̃ + ɪm
- The nasalisation turns into anusvāra which is pronounced asn̪
because of thet̪
that follows.=> (ət̪t̪u + ə̃) + ɪm
- The anusvāra turns into a gemination of the following consonantt̪
.=> ət̪t̪ə̃ + ɪm
- The trailingu
disappears according to lōpasaṁdhi.=> ət̪t̪əɳɪm
- The nasalisation turns intoɳ
.
So, the Old Kannada grammarians, having decided that iṁ
/iṁda
are exclusively instrumental suffixes, seem to have settled on using
the Proto-Dravidian third-person distal demonstrative pronominal form attu
as the base for the ablative suffix
because it has the appropriate distal directional semantics.
But having realised that this doesn't achieve the required ablative semantics fully, they seem to have resorted to appending the very same
iṁ
suffix (which they had decided was not the ablative suffix) before finally declaring that the required ablative semantics is achieved.
The key reason for this futile exercise seems to be the confusion that iṁ
/iṁda
suffixes are exclusively instrumental
and do not have the directional/separational semantics required for the ablative case.
As already shown above, iṁ
/iṁda
not only have the proximal demonstrative/directional semantics,
but also, when appended as a suffix to any word fragment,
they suggest a physical or metaphorical proximal direction (and hence, separation) from the semantics of the preceding word fragment,
which is all that is required for the ablative case.
So, even though iṁ
/iṁda
, have proximal demonstrative pronominal semantics when seen as independent words,
impart ablative semantics when used as suffixes.
Not noticing this subtlety seems to be at the heart of the confusion that iṁ
/iṁda
are not ablative suffixes.
Even if the attaṇiṁ
form was constructed to escape from the trap caused by this confusion, that construction too ended up having to use
the very same iṁ
, which was discarded in the first place as not being an ablative suffix, suffixed to attu
,
a form generated by a related set of processes working on the distal equivalent root of the very same iṁ
again.
To paraphrase a famous movie dialogue, grammar gods (i.e., Goddess Saraswati),
are not without a sense of irony.
It is lucky that ordinary Kannadigas did not fall into this trap as they have been using iṁ
/iṁda
as ablative suffixes freely.
The constructions attaṇiṁ
and deseyiṁda
never really caught on.
That even the grammarians who declared the motto "prayōgaśaraṇāḥ vaiyākaraṇāḥ" stumbled in this way, is a reminder to us all
to think a hundred times before declaring any usage in language as incorrect or ungrammatical.
Now that it is clear that the third-person proximal demonstrative pronominal root im̐
/ɪ̃
is also used in Kannada as the instrumental/ablative suffix,
it is not hard to see that its distal counterpart, the third-person distal demonstrative pronominal root am̐
/ə̃
,
is also used in Kannada as the accusative suffix.
This is because the accusative case indicates the target or object of the action indicated by the verb.
For this, a directional sense is need to be added to the root word's semantics to indicate the target or object of the action
and the distal demonstrative/directional am̐
/ə̃
when added as a suffix is perfectly suited to add that directional semantics.
I.e., appending the distal demonstrative/directional am̐
/ə̃
, makes the base word fragment the target or object of the action;
thus, achieving the accusative semantics.
This supports the argument that the distal demonstrative pronominal root am̐
/ə̃
is also used in Kannada as the accusative suffix,
not just by comparison with the proximal counterpart im̐
/ɪ̃
being used in Kannada as the instrumental/ablative suffix,
but also by the processes of derivation and their implication on semantics.
In Old Kannada, uṁ
is used as an aggregating suffix.
This is mentioned in śabdamaṇidarpaṇaṁ, verse 133.
amumeṁdusirva samuccaya[-] dumuvidhi nirutaṁ dvitīyeyoḷ pokkirkuṁ | samanisaveṁduṁ ṣaṣṭhiyo[-] ḷamumavu mikkina vibhaktigaḷa mēlirkuṁ || vr̥tti - am eṁduṁ um eṁdu pēḻva samuccayadoḷ uṁ eṁbudu dvitīyeya naḍuve pokkirkuṁ; ṣaṣṭhiyoḷ samuccayaṁ pattadu; am umgaḷ mikka vibhaktigaḷa mēle yathārūpamāgi pattuguṁ. prayōgaṁ - dvitīyeyoḷagaṇumuviṁge: - avarumanaṁtakanormeye savinōḍal... (1) oṁdeḍeyoḷ kaṭṭuvude puliyumaṁ kavileyumaṁ (2) mikka vibhaktigaḷa mēgaṇa samuccayakke: naranuṁ naranaṁdananuṁ guruvuṁ gurusutanuṁ (3) kāliṁdeyuṁ kaiyiṁdeyuṁ ghaṭṭisidaṁ. (4) caturthiyoḷ vikalpaṁ: drōṇaṁgaṁ ninagaṁ bil[-] jāṇikeyadu sakkasamane... (5) vaṁcisuvudavaṁgeyuṁ ninageyuṁ sahajaṁ biḍisalke barkumē. (6) jādiyattaṇiṁdeyuṁ kēdageyattaṇiṁdeyuṁ kaṁpu baṁdudu. (7) saptamige am pattuvudu: caladoḷamācāradoḷaṁ kalitanadoḷaṁ... (8) allayeṁbudaṟa mēle uṁ pattuguṁ: alliyumilliyumelliyaṁ naraṁ. (9)
The aggregating suffixes aṁ
/uṁ
do not apply in the genitive case; in the accusative case they apply before the accusative suffix;
in the other cases they apply after the case suffixes.
In Modern Kannada (perhaps in Middle Kannada too), uṁ
has turned into ū
.
For example,
rāmanū lakṣmaṇanū sīteyū kāḍige hōdaru.
- As in a typical childern's story title,
siṁhavū iliyū
.
As already discussed above, original semantics of the
medial demonstrative pronominal root um̐
/ũ
is presentative.
In light of this, it becomes clear that the aggregating suffixes, uṁ
in Old Kannada and ū
in Modern Kannada,
are originally derived from the presentative pronominal root um̐
/ũ
,
because the role of the aggregating suffixes uṁ
/ū
is to present more one or more things as context,
so that the rest of the sentence can continue on about them.
The derivation process could be as follows.
In ISO 15919,
um̐
=> uṁ
- The nasalisation turns into anusvāra which is pronounced asm
.=> ū
- The nasalisation elongates the vowel that was nasalised. More on this process can be seen here.
In IPA,
ũ
=> um
- The nasalisation turns into anusvāra which is pronounced asm
.=> uː
The nasalisation elongates the vowel that was nasalised. More on this process can be seen here.
The presentative semantics of ū
in siṁhavū iliyū
is easy to see.
Likewise, in rāmanū lakṣmaṇanū sīteyū kāḍige hōdaru
, the usage of ū
in the part rāmanū lakṣmaṇanū sīteyū
presents rāma
, lakṣmaṇa
and sītā
so that the kāḍige hōdaru
can be said only once but still apply to all of them together.
Similarly, in Old Kannada, naranuṁ naranaṁdananaṁ
, kāliṁdeyuṁ kaiyiṁdeyuṁ
, drōṇaṁgaṁ ninagaṁ
, jādiyattaṇiṁdeyuṁ kēdageyattaṇiṁdeyuṁ
,
caladoḷamācāradoḷaṁ
, alliyumilliyumelliyaṁ
etc., uṁ
presents one or more things for the rest of the sentence.
Though this analysis applies to the accusative case as well, the variation in Old Kannada of the aggregating suffix uṁ
applying before the accusative suffix (aṁ
/am̐
/ə̃
) seems peculiar.
I.e., in the above example from śabdamaṇidarpaṇaṁ, avarumaṁ aṁtakanormeye
, i.e., avar + uṁ + aṁ => avarumaṁ
.
But in Modern Kannada (perhaps in Middle Kannada too), the aggregating suffix ū
applies after the accusative suffix
just like in the other casess.
For example, the Modern Kannada "translation" of the above Old Kannada example would be avarannū aṁtakanommeye
,
i.e., avar + annu + ū => avarannū
.
The difference in the processes in Old Kannada and Modern Kannada in the accusative case can be reconciled as follows.
As already mentioned above,
the accusative case indicates the target or object of the action indicated by the verb.
In a crude comparison to English grammar, it can be said that if the nominative case indicates the
subject
in a sentence, the accusative case indicates the
object
.
In light of this, both orders of applying of aggregating suffix and the accusative suffix can be considered logically sound.
-
Modern Kannada, like in other cases, seems to prefer to "objectify" the root word first by appending the accusative suffix
aṁ
/am̐
/ə̃
and then present it for the rest of the sentence by appending the aggregating suffixū
. -
Old Kannada, in the accusative case, seems to prefer to present the root word first by appending the aggregating suffix
um
and then "objectify" it by appending the accusative suffixaṁ
/am̐
/ə̃
.
Since such "targetting/objectification" semantics is not relevant in the other cases, there is no difference between the aggregating processes in Old and Modern Kannada.
This investigation of the aggregating suffixes, uṁ
in Old Kannada and ū
in Modern Kannada,
not only shows that these aggregating suffixes are originally derived from the presentative pronominal root um̐
/ũ
,
but also clarifies the variation in the aggregating processes between Old and Modern Kannada in the accusative case
and the lack of such variation in the other cases.
This gives a glimpse of the depth of the semantics of the pronominal roots and the wide-ranging breadth of use Dravidian languages have put them into.
Though the Dravidian third-person demonstrative pronominal root vowels, the distal a
, the proximal i
and the medial u
,
are found in Indo-European languages too,
their form and derivation doesn't seem to be as systematic as in the Dravidian languages,
as can be seen in the following sample (please note that this is not a complete list by any means;
it is presented here only for the purpose of comparison).
It is interesting to note that the vowel i
appears most often in the proximal demonstrative pronominal forms
and the vowel a
or u
appear most often in the distal demonstrative pronominal forms, even in the Indo-European languages.
Language | Proximal with i |
Proximal with u |
Proximal with other vowels | Distal with a |
Distal with u |
Distal with other vowels | Single word for both distal and proximal |
---|---|---|---|---|---|---|---|
Sanskrit | इदम् (idaṁ), इदानीम् (idānīṁ), इयम् (iyaṁ), इतः (itaḥ), इति (iti), इत्थम् (itthaṁ) एषः (ēṣaḥ)?, एषा (ēṣā)?, एतत् (ētat)?, एतावता (ētāvat) | अत्र (atra), अयम् (ayaṁ), अधुना (adhunā) | अतः (ataḥ), अदः (adaḥ), सः (saḥ), सा (sā), तत् (tat), तत्र (tatra), तदा (tadā), तदानीम् (tadānīṁ) | यावत् (yāvat) | |||
Hindi | यह (yah), इस (is), यहाँ (yahām̐), ऐसे (aisē)?, इतना (itnā), इधर (idhar) | अब (ab) | तब (tab) | वह (vah), उस (us), वहाँ (vahām̐), वैसे (vaisē), उतना (utnā), उधर (udhar) | |||
Gujarati | આ (ā), અહીં (ahīṁ), આટલું (āṭluṁ) | તે (tē), ત્યાં (tyāṁ) | |||||
Punjabi | ਇਹ (iha), ਇਥੇ (ithē), इतना (inā) | ਹੁਣ (huṇa) | ਉਹ (uha), ਉੱਥੇ (uthē), ਉਂਜ (uṁja), उतना (utnā) | ਜਿੰਨਾ (jinā) | |||
Bhojapuri | ई (ī), ईहाँ (īhām̐), इतना (ētnā) | अब (ab) | तब (tab) | ऊ (ū), ऊहाँ (uūhām̐), ओइसे (ōisē), जोतना (jōtnā) | |||
Bengali | এই (ēi), এখানে (ēkhānē), এতটুকু (ētaṭuku)?, এখন (ēkhāna) | সে (śē), যে (jē), সেখানে (śēkhānē), যতটুকু (yatatuku) | |||||
English | it, this, these, here | now | that, then?, there? | those | he, she, thus | ||
German | dies, hier, jetzt? | dass, dann | dort | er, sie, es |
Though it might be more appropriate to investigate the first and second person pronominal roots together, because Caldwell has used his arguments regarding the second-person pronominal roots as supporting evidence/arguments for his arguments regarding the first-person pronominal roots, it is necessary (even at the cost of lucidity) to discuss the second-person pronouns first. But it must be noted that the discussion of the second-person pronouns in this section and the discussion of the first-person pronouns in the subsequent section are two parts of the same discussion.
As already seen in the "Introduction and Context" part, there is some consensus among scholars that nīn
is the second-person pronominal root in Kannada.
Caldwell goes further and asserts that nīn
is the second-person pronominal root in all Dravidian languages.
He also documents some evidences/arguments against this assertion.
But overall, he concludes that nīn
(with the vowel i
infixed between two consonant n
s) must be the second-person pronominal root
and used it as a part of the argument to support his assertion that nān
is the Proto-Dravidian first-person pronominal root.
Since the depth and breadth of Caldwell's analysis of the second-person pronominal roots is not seen in the other scholars mentioned
in the "Introduction and Context" section (if other scholars not mentioned here have done deeper analysis, the ignorance is entirely mine
and I will be grateful if knowledgeable readers point me towards such works),
Caldwell's arguments will be analysed first before continuing the investigation of the second-person pronominal roots.
First let's analyse the arguments Caldwell has mentioned for and against the case that nīn
is the second-person pronominal root.
-
In the personal terminations of the verb, this pronoun is much changed in all the Drâvidian dialects. It not only loses its initial 'n,' like the pronoun of the first person; but its final 'n' also disappears. Generally nothing remains in the verbal inflexions but the included vowel, and that also is more or less modified by use. In the Canarese verb it appears as 'i,' 'î,' 'îye,' and 'e:' in the ancient dialect of the Canarese it is 'âÿ,' as in Tamil. [From pages 306 and 307]
... In Malayâlam the nominative of this pronoun is 'nî;' but 'nin' is used, as in classical Tail, in the oblique cases. The dative has 'nan,' instead of nin; e.g., 'nanikka,' to thee - as if from a nominative in 'nan,' with 'a' as the included vowel. This use of 'a' is in accordance with the colloquial Tamil personal termination of the web, which is 'âÿ,' instead of 'îÿ.' [From page 308]
In Tulu... The personal terminations of the verb in the second person singular is represented by 'a.' [From page 308]
-
... In the colloquial obliques the initial 'n' entirely disappears, and does not leave even a 'y' behind it, as the initial 'n' of the first personal pronoun generally does. When the initial is discarded, the included vowel changes from 'i' to 'u.' 'u,' however, constitutes the included vowel of this pronoun, not only when the initial 'n' is lost, but sometimes, in the higher dialect, even when it is retained. 'nin,' 'nun,' and 'un' are severally used as the bases of the oblique cases. [From page 307]
-
The root of the verb is regularly used in Tamil as the second person singular of the imperative, without any pronominal suffix, and even without any euphonic addition: but the second person plural of the imperative in the colloquial dialect is formed by the addition of 'um' (the ordinary plural base of the oblique cases); which 'um' is derived from a singular form in 'un,' one of the bases of the oblique cases already referred to. ... [From page 307]
-
The Tulu nominative 'î' illustrates the fact already stated that each of the nasals of 'nîn' (both the radical initial and the formative final) has sometimes been worn off. [From page 308]
-
In the higher dialect of Telugu, 'îvu,' from an old nominative 'î,' which is identical with the Tulu, is occasionally used instead of 'nîvu.' [From page 309]
-
The nominative forms and their oblique cases have the root form that starts with the consonant
n
.-
In Canarese the nominative of this pronoun is 'nîn' or 'nîn-u;' [From page 306]
-
In Tamil 'nî,' which is properly the crude base, is invariably used as the isolated nominative... That 'nîn' originally constituted the nominative even in Tamil, appears from this that the oblique cases in the higher dialect agree in using 'nin' (shortened by rule from 'nîn') as the base to which the case suffixes are attached. [From page 307]
-
... In Malayâlam the nominative of this pronoun is 'nî;' but 'nin' is used, as in classical Tail, in the oblique cases. From page 308]
-
In Tulu the nominative is 'î;' but the oblique cases are formed upon the base of 'ni' or 'nin.' From page 308]
-
... The Telugu nominative is 'nîvu,' expanded from 'nî' by the addition of the euphonic particle 'vu:' 'nîvu,' Tel., thou ... In the oblique cases ... uses 'nî' as its inflexional base, and also as its possessive. The objective alone follows the example of the other dialects in abbreviating the included vowel, and appending a final nasal. That case is 'ni-nu,' 'nin-u,' or 'nin-nu,' and is evidently formed from a nominative 'nîn-u.' [From page 309]
-
-
The compound sound of 'ñj' or 'ny,' in the Malayâla 'ñjân' or 'nyân' is a middle point between 'n' of 'nân,' and the 'y' of 'yân.' It is a softened and nasalised form of 'n,' from which the change to 'y' is easily made. In like manner, 'nîn,' the original form of the pronoun of the second person singular in all the Drâvidian dialects, has become in Tamil, first 'nîÿ,' then 'nî;' and in the verbal terminations 'aiy,' 'i,' and 'ei.' [From page 294]
-
The verbal inflexions of the Telugu, use only the final syllable of the nominative of each of the pronouns, viz., 'nu' (from 'nênu,' I), 'vu' (from 'nîvu,', thou), and 'ḍu' (from 'vâḍu,' he). The most important and essential part of each pronoun has thus been omitted; and the fragments which have been retained are merely formatives, or at most signs of gender and number. [From page 295]
-
In the oblique cases the Telugu rejects the euphonic addition of 'vu,' and uses 'nî' as its inflexional base, and also as its possessive. ... In the personal terminations of the verb, the Telugu rejects every portion of the pronominal root, and employs only the euphonic addition 'vu' or 'vi.' [From page 309]
-
-
... On a comparison of the various Drâvidian dialects we arrive at the conclusion that the primitive form of this pronouns was 'nî,' 'nû,' or 'na;' - most probably the first. The only essential part of the pronoun appears to be the initial consonant 'n;' just as in the Indo-European languages 't' is the only essential part of the corresponding pronoun. In each family the vowel by the help of which the initial consonant is enunciated varies considerably, but evinces, on the whole, a preference for 'i' in the Dravidian languages, for 'u; in the Indo-European. [From page 309]
-
Supposing 'nî' to be the primitive form of the Drâvidian pronoun of the second person, and comparing it with 'nâ,' which we have seen to be the primitive form of the Drâvidian pronoun of the first person, it is deserving of notice that the only difference between the two is the difference between the two included vowels, 'a' and 'i.' ... The method adopted by the Dravidian languages of expressing the difference between the first person and the second by means of of the vowel 'a' and 'i,' used as auxiliaries, does not appear to be the result of accident. It is probably founded on some ultimate principle; Though it may be difficult or impossible now to discover what that principle is. [From pages 309 and 310]
-
If 'a' and 'i' be considered as identical with the demonstratives, an idea which would suit the signification, and which is corroborated by the circumstance that 'u' is also a demonstrative, we are met by the apparently insurmountable difficulty that in all the Drâvidian tongues, and (as far as the use of these demonstrative vowels extends) in all the tongues of the Indo-European family, 'a' is not the proximate, but the remote, demonstrative; and 'i' is not the remote, but the proximate; whilst 'u' is used in Tamil as an intermediate between these two. [From page 310]]
-
Is any weight to be attributed to the circumstances that 'a' has naturally the first place in all lists of vowels, and 'i' the second? [From page 310]]
-
If, as we have seen, 'nâ' is to be regarded as the primitive form of the Drâvidian pronoun of the first person, and the final 'n' as merely a sign of number, or as an euphonic formative, it might appear extraordinary, that in the pronominal terminations of the verb, the initial 'n,' the primitive sign of the personality has invariably and altogether disappeared; whilst the first person singular is represented by the final 'n' alone. We might almost be led to suppose the initial 'n' to be a formative prefix, and the succeeding vowel to be the real pronominal base. Formative and definitive pre-fixes, however, are utterly unknown to the Dravidian languages; ... [From page 296 and 297]]
The main thrust of the the above arguments (both for
and against) on nīn
/nin
/nī
/ni
being the second-person pronominal roots
seems to be that the starting consonant n
and the vowel i
that animates it are both essential for the second-person semantics.
Now let's go through the arguments one by one.
For lucidity, the arguments are picked for analysis in an order different from those mentioned above or in the "Introduction and Context" part.
I apologise if this causes confusion.
-
In the personal terminations of the verb, this pronoun is much changed in all the Drâvidian dialects. It not only loses its initial 'n,' like the pronoun of the first person; but its final 'n' also disappears. Generally nothing remains in the verbal inflexions but the included vowel, and that also is more or less modified by use. In the Canarese verb it appears as 'i,' 'î,' 'îye,' and 'e:' in the ancient dialect of the Canarese it is 'âÿ,' as in Tamil. [From pages 306 and 307.
... In Malayâlam the nominative of this pronoun is 'nî;' but 'nin' is used, as in classical Tail, in the oblique cases. The dative has 'nan,' instead of nin; e.g., 'nanikka,' to thee - as if from a nominative in 'nan,' with 'a' as the included vowel. This use of 'a' is in accordance with the colloquial Tamil personal termination of the web, which is 'âÿ,' instead of 'îÿ.' [From page 308]
In Tulu... The personal terminations of the verb in the second person singular is represented by 'a.' [From page 308]
This argument strongly opposes the case for the starting consonant n
in nī
/ni
being essential to the second-person semantics,
because the relationship between pronouns and the grammatical personal suffixes of finite verb forms seems to be an ancient one
and it is likely that the grammatical personal suffixes of finite verb forms are closer to the corresponding pronominal roots.
From this perspective, the lack of the starting consonant n
in the grammatical personal suffixes of finite verb forms (āy
, i
, ī
, īye
, e
, ai
),
strongly opposes the case for the starting consonant n
in the second-person pronominal root.
Since it is possible that a
, the grammatical personal suffixes of finite verb form in Malayalam and Tulu, is related to the Tamil āy
,
the fact of this suffix not having the vowel i
does not undermine this argument against the starting consonant n
.
-
The verbal inflexions of the Telugu, use only the final syllable of the nominative of each of the pronouns, viz., 'nu' (from 'nênu,' I), 'vu' (from 'nîvu,', thou), and 'ḍu' (from 'vâḍu,' he). The most important and essential part of each pronoun has thus been omitted; and the fragments which have been retained are merely formatives, or at most signs of gender and number. [From page 295]
-
In the oblique cases the Telugu rejects the euphonic addition of 'vu,' and uses 'nî' as its inflexional base, and also as its possessive. ... In the personal terminations of the verb, the Telugu rejects every portion of the pronominal root, and employs only the euphonic addition 'vu' or 'vi.' [From page 309]
-
Caldwell seems to have used this argument to weaken the strength of the argument by grammatical personal suffixes of finite verb forms
against the starting consonant n
.
The point he makes about the second-person suffixes of finite verb forms in Telugu is true enough,
but the same phenomenon of not using the vowel from the pronominal form in the corresponding grammatical personal suffixes of finite verb forms
is not seen in the other Dravidian languages.
So, it does not seem appropriate to generalise a feature that seems specific only to Telugu to all Dravidian languages
and therefore, this argument doesn't carry much weight.
-
The Tulu nominative 'î' illustrates the fact already stated that each of the nasals of 'nîn' (both the radical initial and the formative final) has sometimes been worn off. [From page 308]
To the contrary, this is, in fact, a strong argument against the starting consonant n
in the second-person pronominal root
and for the argument that the vowel i
/ī
is the second-person pronominal root.
It seems equally possible, and indeed more appropriate, to say that like in Tulu and literary Telugu,
the vowel i
/ī
is the second-person pronominal root than to say that nīn
/nin
is the second-person pronominal root
and the trailing consonants (n
) has been lost in the
grammatical personal suffixes of finite verb forms.
The reason for this can be seen in more detail here.
-
In the higher dialect of Telugu, 'îvu,' from an old nominative 'î,' which is identical with the Tulu, is occasionally used instead of 'nîvu.' [From page 309]
Like in the case of Tulu second-person pronoun, this is a strong argument against the Caldwell's case for nīn
/nin
being the second-person pronominal root
and for the vowel i
/ī
being the pronominal root.
It is notable that Tulu and Telugu have branched from the Proto-Dravidian earlier than Kannada, Tamil and Malayalam,
because it is often seen that earlier branches retain some features of the common ancestor that change or disappear in later branches.
For example, monotremes like the platypus
and the echidna are well known to be an earlier branch of mammals.
The fact that these monotremes are oviparous is strong argument for the case that
the common ancestor of all mammals was also oviparous,
even though the vast majority of mammal species today are viviparous.
However, it is important to remember that the surviving species of the earlier branches have evolved as much as those from the later branches.
For example, just because platypuses (or is it platypi?) have a duck-like bill,
it would be wrong to conclude that the common ancestor of all mammals also had a duck-like bill.
Which aspect is inherited from a common ancestor and which is a change or evolution in a later branch is to be discerned by careful consideration
of various arguments and evidences.
Back in the case of the second-person pronominal root, it seems more appropriate to say that the Tulu and Telugu pronominal form i
/ī
is the original second-person pronominal form with some processes leading to the incorporation of the leading and trailing consonant n
than to say that nīn
/nin
is the pronominal root and that the leading and trailing consonant n
has disappeared independently in two separate
(Tulu and Telugu) earlier branches.
- The nominative forms and their oblique cases have the root form that starts with the consonant
n
.-
In Canarese the nominative of this pronoun is 'nîn' or 'nîn-u;' [From page 306]
-
In Tamil 'nî,' which is properly the crude base, is invariably used as the isolated nominative... That 'nîn' originally constituted the nominative even in Tamil, appears from this that the oblique cases in the higher dialect agree in using 'nin' (shortened by rule from 'nîn') as the base to which the case suffixes are attached. [From page 307]
-
... In Malayâlam the nominative of this pronoun is 'nî;' but 'nin' is used, as in classical Tail, in the oblique cases. From page 308]
-
In Tulu the nominative is 'î;' but the oblique cases are formed upon the base of 'ni' or 'nin.' From page 308]
-
... The Telugu nominative is 'nîvu,' expanded from 'nî' by the addition of the euphonic particle 'vu:' 'nîvu,' Tel., thou... In the oblique cases ... uses 'nî' as its inflexional base, and also as its possessive. The objective alone follows the example of the other dialects in abbreviating the included vowel, and appending a final nasal. That case is 'ni-nu,' 'nin-u,' or 'nin-nu,' and is evidently formed from a nominative 'nîn-u.' [From page 309]
-
This is the main argument in support of nīn
/nin
being the second-person pronominal root in Dravidian Languages
and against only the vowel i
/ī
being the pronominal root.
There is no disputing that the second person pronouns in most of the Dravidian languages start with the consonant n
.
But Caldwell's examples for second-person pronominal forms in Tulu are inaccurate and incomplete.
It is indeed true that ni
/nin
are the inflexional prefixes in some regional dialects of Tulu;
but in other regional dialects i
/iṁ
/in
are the inflexional prefixes, as can be seen in the
Tulu second-person pronouns section of "Introduction and Context".
For example, inna
, iṁkŭḷŭ
etc.
Also, Caldwell has mentioned the inflextional prefixes starting with or including the vowel u
in Tamil.
-
... In the colloquial obliques the initial 'n' entirely disappears, and does not leave even a 'y' behind it, as the initial 'n' of the first personal pronoun generally does. When the initial is discarded, the included vowel changes from 'i' to 'u.' 'u,' however, constitutes the included vowel of this pronoun, not only when the initial 'n' is lost, but sometimes, in the higher dialect, even when it is retained. 'nin,' 'nun,' and 'un' are severally used as the bases of the oblique cases. [From page 307]
-
The root of the verb is regularly used in Tamil as the second person singular of the imperative, without any pronominal suffix, and even without any euphonic addition: but the second person plural of the imperative in the colloquial dialect is formed by the addition of 'um' (the ordinary plural base of the oblique cases); which 'um' is derived from a singular form in 'un,' one of the bases of the oblique cases already referred to. ... [From page 307]
Tulu and literary Telugu having second-person pronominal forms starting with the vowel i
/ī
,
spoken Tamil having the pronominal forms starting with the vowel u
instead of i
/ī
and the literary Tamil having pronominal forms starting with the consonant n
but including the same vowel u
,
work against the case that nīn
/nin
is the second-person pronominal root.
-
The compound sound of 'ñj' or 'ny,' in the Malayâla 'ñjân' or 'nyân' is a middle point between 'n' of 'nân,' and the 'y' of 'yân.' It is a softened and nasalised form of 'n,' from which the change to 'y' is easily made. In like manner, 'nîn,' the original form of the pronoun of the second person singular in all the Drâvidian dialects, has become in Tamil, first 'nîÿ,' then 'nî;' and in the verbal terminations 'aiy,' 'i,' and 'ei.' [From page 294]
Even though this argument might seem plausible at first sight, it does not survive careful scrutiny (at least in the case of the second-person pronoun).
As in the mentioned first-person pronouns, there are no other examples of the ñ
(ostensibly, middle point between n
and y
) turning into y
in Malayalam, Tamil or Tulu.
In fact, there are no second-person pronominal forms like ñī
/ñi
or yī
/yi
in any of these languages.
Besides, the nasal consonants ñ
and ṅ
are found in Malayalam much more than in other Dravidian languages.
So, it seems inappropriate to include such a process among the processes involved in producing the second-person pronominal forms
without considering the obvious possibility that this might be a process specific to Malayalam.
Likewise, the argument that nīn
has turned first into nīy
and then into nī
doesn't hold water,
because the trailing n
in nīn
is originally just a nasalisation; i.e., nīm̐
/n̪ĩː
; that nasalisation turns into n
to form nīn
or disappears to form nī
.
More details on these processes involving nasalisation can be seen here.
In fact, the process of n
turning into ñ
, then into y
and eventually disappearing is inaccurate not just in the case of the
second-person pronouns but also in the case the first-person as will be discussed later.
-
... On a comparison of the various Drâvidian dialects we arrive at the conclusion that the primitive form of this pronouns was 'nî,' 'nû,' or 'na;' - most probably the first. The only essential part of the pronoun appears to be the initial consonant 'n;' just as in the Indo-European languages 't' is the only essential part of the corresponding pronoun. In each family the vowel by the help of which the initial consonant is enunciated varies considerably, but evinces, on the whole, a preference for 'i' in the Dravidian languages, for 'u; in the Indo-European. [From page 309]
Caldwell seems to have included nu
and na
in the list of possible second-person pronominal roots
based on the Tamil second-person pronominal forms containing the vowel u
and the Tulu, Malayalam grammatical personal suffix a
of finite verb forms, correspondingly.
But his conclusion that Dravidian languages lean towards the vowel i
and the Indo-European languages towards the vowel u
seems correct based on even just a few examples like tvaṁ
, yūyaṁ
in Sanskrit, thou
, you
in English etc.
But on closer inspection, it becomes clear that Caldwell's position has changed from the starting consonant n
in nīn
/nin
being crucial for the second-person pronominal semantics
to the included vowel (i
, u
or a
) being crucial.
It is notable that this position has already been arrived at in the discussion of the second-person pronominal forms in Tulu and literary Telugu above.
-
Supposing 'nî' to be the primitive form of the Drâvidian pronoun of the second person, and comparing it with 'nâ,' which we have seen to be the primitive form of the Drâvidian pronoun of the first person, it is deserving of notice that the only difference between the two is the difference between the two included vowels, 'a' and 'i.' ... The method adopted by the Dravidian languages of expressing the difference between the first person and the second by means of of the vowel 'a' and 'i,' used as auxiliaries, does not appear to be the result of accident. It is probably founded on some ultimate principle; Though it may be difficult or impossible now to discover what that principle is. [From pages 309 and 310]
-
If 'a' and 'i' be considered as identical with the demonstratives, an idea which would suit the signification, and which is corroborated by the circumstance that 'u' is also a demonstrative, we are met by the apparently insurmountable difficulty that in all the Drâvidian tongues, and (as far as the use of these demonstrative vowels extends) in all the tongues of the Indo-European family, 'a' is not the proximate, but the remote, demonstrative; and 'i' is not the remote, but the proximate; whilst 'u' is used in Tamil as an intermediate between these two. [From page 310]]
By extending the argument that the included vowel (mainly, i
) carries the second-person pronominal semantics to the first-person pronominal roots as well,
Caldwell conjectures that the included vowel a
/ā
in nān
/nan
carries the first-person pronominal semantics
and compares the first-personal a
, the second-personal i
and the occasional second-personal u
of Tamil
with the third-person demonstrative pronominal root vowels, the distal a
, the proximal i
and the medial u
.
This is a great insight.
But in that comparison, Caldwell notes that equating the first-personal a
and the third-person demonstrative distal a
,
the second-personal i
and u
with the third-person demonstrative proximal i
and medial u
is not logically sound.
But Caldwell is mistaken in saying that the demonstrative proximal i
and the medial u
do not fit the second-person pronominal semantics.
To the contrary, the demonstrative proximal i
is very well suited to carry the second-person pronominal semantics
because the entity addressed in the second-person pronoun is always (physically or metaphorically) close-by.
It might seem right to say that the demonstrative medial u
does not fit the second-person pronominal semantics,
but on considering the already discussed presentative semantics of the
demonstrative vowel u
, it becomes clear that the presentative demonstrative u
is quite suited to carry the second-person pronominal semantics,
because the entity addressed in the second-person pronouns is obviously (physically or metaphorically) present.
So, the pronominal root vowels i
and u
being used in the distal and presentative semantics in the third-person demonstrative pronouns
as well as in the second-person semantics seems quite appropriate.
But Caldwell's objection to equating the third-person demonstrative distal a
with the first-person pronominal a
is indeed correct.
But it is not unsolvable like he supposed, because a
is not the first-person pronominal root vowel like he concluded.
This is discussed in more detail in the section on the first-person pronominal roots.
-
Is any weight to be attributed to the circumstances that 'a' has naturally the first place in all lists of vowels, and 'i' the second? [From page 310]]
It is difficult to decide about this speculation of Caldwell's.
But another possible reason for the vowels a
, i
and u
being used in this way in the pronominal forms
will be discussed later in this document.
What remains to investigate in the second-person pronominal forms is the crucial process by which the starting and trailing consonant n
are produced from the root vowel i
, which will be picked next.
The clue for the process for producing nīn
/nin
from the second-person pronominal root vowel ī
/i
can be
found in a previous research article "anusvārada anusāra",
which investigates the phenomenon of nasalisation mediating
wherever suffixes are appended to the root words, eventually producing various different forms.
Also, it is seen in the previous document on nasalisation
and also briefly above in this document,
that saying nasalisation mediates when suffixes are appended to the roots is equivalent to saying that the roots themselves are nasalised
and it does not make much difference to the processes and the forms that are produced.
It is seen while examining the variations in the nominative forms
like ava
/rāma
, avam̐
/rāmam̐
, avan
/rāman
etc., that the nasalisation originally existed in the nominative case too.
With this background, it seems appropriate to suppose that the second-person pronominal root ī
/i
is also nasalised in the nominative case
and Caldwell has already noted the phenomenon of the included vowel in the pronouns being elongated in the nominative forms.
So, the nominative and the inflexional prefix forms of the second-person pronominal root would be as follows respectively.
In ISO 15919,
īm̐
- Nominativeim̐
- Inflexional prefix
In IPA,
ĩː
- Nominativeɪ̃
- Inflexional prefix
It is easy to see how the Tulu forms like iṁkŭḷŭ
, literary Telugu forms like īvu
and the Tulu (and old literary Telugu) forms like ī
are produced from these nasalised root vowels.
I.e.,
In ISO 15919,
īm̐ => ī
- The nasalisation disappears in Tulu and old literary Telugu.īm̐ + u => īvu
- The nasalisation turns intov
in literary Telugu.im̐ + kŭḷŭ => iṁkŭḷŭ
- The nasalisation turns into anusvāra in Tulu, which is pronounced asṅ
because of thek
that follows.
In IPA,
ĩː => iː
- The nasalisation has disappeared in Tulu and old literary Telugu.ĩː + u => iːʋu
- The nasalisation turns intoʋ
in literary Telugu.ɪ̃ + kɯɭɯ => ɪŋkɯɭɯ
- The nasalisation turns into anusvāra in Tulu, which is pronounced asŋ
because of thek
that follows.
It is no surprise then that the anusvāra in iṁkŭḷŭ
is many times pronounced as a mere nasalisation rather than ṅ
/ŋ
,
i.e., as im̐kŭḷŭ
/ɪ̃kɯɭɯ
.
Similarly, since nasalisation turning into n
is also a known process, the trailing consonant n
in the nīn
/nin
can be easily explained.
I.e.,
In ISO 15919,
īm̐ => īn
- The nasalisation turns inton
.im̐ => in
- The nasalisation turns inton
In IPA,
ĩː => iːn̪
- The nasalisation turns inton̪
.ɪ̃ => ɪn̪
- The nasalisation turns inton̪
.
But the īn
, in
forms are not found in Dravidian languages.
Only the forms including the additional starting consonant n
are found.
However, Caldwell mentions the un
forms without the leading n
in Tamil.
I.e.,
um̐ => un
- The nasalisation turns inton
in the inflexional prefix of non-nominative cases of the second-person pronoun in spoken dialects of Tamil.
More details about these related set of processes of nasalisation disappearing, turning into anusvāra, n
, v
etc. can be seen here.
Now what remains is the process of producing the leading consonant n
in nīn
/nin
.
Caldwell has argued (rightly so) that prefixes are not natural to and rarely found in Dravidian languages
and has used this as a strong argument for the leading consonant n
in nīn
/nin
being crucial to the second-person pronominal roots.
-
If, as we have seen, 'nâ' is to be regarded as the primitive form of the Drâvidian pronoun of the first person, and the final 'n' as merely a sign of number, or as an euphonic formative, it might appear extraordinary, that in the pronominal terminations of the verb, the initial 'n,' the primitive sign of the personality has invariably and altogether disappeared; whilst the first person singular is represented by the final 'n' alone. We might almost be led to suppose the initial 'n' to be a formative prefix, and the succeeding vowel to be the real pronominal base. Formative and definitive pre-fixes, however, are utterly unknown to the Dravidian languages; ... [From page 296 and 297]]
Here, Caldwell seems to retreat from the threshold of solving the problem of the second person pronominal root.
He is indeed largely right that prefixes are unnatural to Dravidian languages
but if he had access to some regional dialects of Tulu, I have no doubt that he might have seen the possibility of
producing the leading consonant n
which doesn't involve prefixes.
Different regional dialects of Tulu have two different forms for the second-person rational plural pronouns, iṁkḷŭ
and niṁkḷŭ
(pronounced variously as iṅkḷŭ
/ɪŋkɭɯ
, niṅkḷŭ
/n̪ɪŋkɭɯ
, im̐kḷŭ
/ɪ̃kɭɯ
, nim̐kḷŭ
/n̪ɪ̃kɭɯ
etc.).
It might seem that niṁkḷŭ
is produced by prefixing n
to iṁkḷŭ
.
I.e.,
In ISO 15919,
n + im̐kḷŭ => nim̐kḷŭ
n + iṁkḷŭ => niṁkḷŭ
n + iṅkḷŭ => niṅkḷŭ
In IPA,
n̪ + ɪ̃kɭɯ => n̪ɪ̃kɭɯ
n̪ + ɪŋkɭu => n̪ɪŋkɭu
Though Caldwell is indeed right in arguing that Dravidian languages do not use prefixes,
it is important to note that n
is not just any old prefix
but a nasal consonant which makes a crucial difference.
It can be seen that while trying to pronounce the nasalisation or the nasal consonant clearly in iṁkḷŭ
in im̐kḷŭ
/ɪ̃kɭɯ
, iṅkḷŭ
/ɪŋkɭɯ
respectively,
the nasality can be heard strongly, almost before the starting vowel i
.
This process becomes even more clear while trying to pronounce just the nasalised vowel i
.
I.e.,
In ISO 15919,
im̐ => m̐im̐
- The clear pronunciation of nasalisation can be heard almost before the voweli
.
In IPA,
ɪ̃ => ̃ɪ̃
- The clear pronunciation of nasalisation can be heard almost before the vowelɪ
.
Such a clear nasalisation is capable of producing the transformations mediated by nasalisation
both before and after the vowel i
and the process of nasalisation turning into n
is already seen.
I.e.,
In ISO 15919,
m̐im̐ => nin
- The nasalisation of the voweli
turns inton
both before and after the vowel.
In IPA,
̃ɪ̃ => n̪ɪn̪
- The nasalisation of the vowelɪ
turns inton̪
both before and after the vowel.
In this process, there are no prefixes involved as objected to by Caldwell;
just the clear nasalisation of the vowel i
has turned into n
both before and after the vowel.
Seen together, the processes for producing nīn
, nī
of Old Kannada and Tamil from the second-person nominative pronominal root form īm̐
/ĩː
and inflexional prefix nin
of Old Kannada, Tamil, Malayalam and Telugu from the second-person pronominal root of the rest of the cases
could be as follows.
In ISO 15919,
īm̐
- The second-person nominative root form.=> m̐īm̐
- The clear pronunciation of nasalisation can be heard almost before the vowelī
.=> nīm̐
- The nasalisation heard before the vowelī
turns inton
.=> nīn
- The nasalisation heard after the vowelī
turns inton
as well.=> nī
- The nasalisation heard after the vowelī
disappears.
im̐
- The root form of the second-person inflexional prefix.=> m̐im̐
- The clear pronunciation of nasalisation can be heard almost before the voweli
.=> nin
- The nasalisation heard on both sides of the voweli
turns inton
.
In IPA,
ĩː
- The second-person nominative root form.=> ̃ĩː
- The clear pronunciation of nasalisation can be heard almost before the voweliː
.=> n̪ĩː
- The nasalisation heard before the voweliː
turns inton̪
.=> n̪iːn̪
- The nasalisation heard after the voweliː
turns inton̪
as well.=> n̪iː
- The nasalisation heard after the voweliː
disappears.
ɪ̃
- The root form of the second-person inflexional prefix.=> ̃ɪ̃
- The clear pronunciation of nasalisation can be heard almost before the vowelɪ
.=> n̪ɪn̪
- The nasalisation heard on both sides of the vowelɪ
turns inton̪
.
In comparison with this set of processes involving nasalisation, the opposite process of losing the nasal consonant n
(i.e., na => ña => ya => a
) suggested by Caldwell seems less plausible, not the least because the intermediate forms
are rarely found in any one Dravidian language
which is why Caldwell had to invoke examples for each form from a different Dravidian language.
Similarly, the processes producing the second-person pronominal forms starting with or including the vowel u
in the spoken and literary Tamil could be as follows.
In ISO 15919,
um̐
- The presentative and second-person pronominal root form in the non-nominative cases in Tamil.=> un
- The nasalisation turns inton
in spoken dialects of Tamil.=> um
- The nasalisation turns intom
in the plural form of grammatical imperative mood of verbs in Tamil.=> m̐um̐
- The clear pronunciation of the nasalisation can be heard almost before the vowelu
.=> nun
- The nasalisation heard on both sides of the vowelu
turns inton
.
In IPA,
ũ
- The presentative and second-person pronominal root form in the non-nominative cases in Tamil.=> un̪
- The nasalisation turns inton̪
in spoken dialects of Tamil.=> um
- The nasalisation turns intom
in the plural form of grammatical imperative mood of verbs in Tamil.=> ̃ũ
- The clear pronunciation of the nasalisation can be heard almost before the vowelu
.=> n̪un̪
- The nasalisation heard on both sides of the vowelu
turns inton̪
.
So, without invoking any processes involving prefixes, it is possible to produce all the second-person pronominal forms
found in the old and modern variants of Dravidian languages by a set of related processes where the clear pronunciation of
nasalisation of the proximal demonstrative and second-person pronominal root vowel, im̐
/ɪ̃
(and īm̐
/ĩː
) and
the presentative and second-person pronominal root vowel, um̐
/ũ
, is naturally heard on both sides of the vowel which then
proceed to undergo the usual set of related processes already discussed in "A Nose for Nasalisation".
With this, the last of the objections, due to which Caldwell abandoned pursuing the possibility of
the proximate demonstrative vowel i
/ī
being also the second-person pronominal root, is also rendered ineffective.
Taking all these arguments together, it seems appropriate to say that the nasalised third-person demonstrative pronominal root vowels,
the proximal im̐
/ɪ̃
(and īm̐
/ĩː
) and the presentative um̐
/ũ
, are also the second-person pronominal roots.
This brings the part of Caldwell's abandoned hypothesis equating the third-person demonstrative and the second-person pronominal root vowels
i
and u
to a positive logical conclusion.
The sole remaining objection to the rest of Caldwell's abandoned hypothesis that the third-person demonstrative distal pronominal root vowel a
is not suited semantically to be the first-person pronominal root will be taken up in the section on the first-person pronominal roots.
I have used IPA in the image because I could not get the ISO 15919 character ಁ
to render properly in the image.
As noted by Caldwell, the Dravidian languages favour the proximal demonstrative vowel i
for the second-person pronominal forms,
whereas the Indo-European languages favour the presentative vowel u
,
which can be seen in the following sample of second-person pronouns of some of the Indo-European languages
(please note that this is not a complete list by any means; it is presented here only for the purpose of comparison).
Language | Second-person Singular | Second-person Plural |
---|---|---|
Sanskrit | त्वम् (tvaṁ) | यूयम् (yūyaṁ) |
Hindi | तू (tū) | तुम (tuṁ) |
Gujarati | તું (tuṁ) | તમે (tamē) |
Punjabi | ਤੂੰ (tū) | ਤੁਸੀਂ (tusīṁ) |
Bhojpuri | তুমি (tumi) | তোমরা (tōmrā) |
English | thou | you |
German | du | sie |
Greek | σύ (sý) | εσείς (eseís) |
Latin | tu | voi |
As already mentioned in the section on second-person pronominal roots, it is to be noted that the discussion of the first-personal pronominal forms is but a continuation of the discussion of the second-person pronominal forms.
As already mentioned in the "Introduction and Context" part of this document, the kind of apparent (but incorrect) consensus that was seen in scholarly opinion on the second-person pronominal root forms is somewhat lacking in the case of the first-person pronominal root forms.
kēśirāja in śabdamaṇidarpaṇaṁ, verse 157, mentions only
the form starting with the vowel ā
(ān
), but does not clarify if the forms starting with the consonant n
are produced from the forms starting with the vowel ā
or by some other independent mechanism.
Caldwell asserts that the form starting with the consonanant n
(nān
) must be the first-person pronominal root in all Dravidian languages.
He also documents some arguments against this assertion,
but concludes that, overall, the form nān
with the leading and trailing consonant n
and the infixed vowel ā
must be the
Dravidian first-person pronominal root.
Masti Venkatesha Iyengar seems to have noticed this argument of Caldwell's,
but briefly mentions a new idea about the included vowel in the first personal pronominal forms which is not seen Caldwell's discussion.
So, let's first analyse Caldwell's position on the matter before seeing how Masti Venkatesha Iyengar's idea about the included vowel
helps in solving the mystery of the first-person pronominal root.
First, let us analyse the arguments Caldwell has mentioned for and against the case that nān
is the first-person pronominal root.
-
In Tamil the form which is used in the colloquial dialect is 'nân,' the inflexion of which is not 'nan,' as might have been expected, but 'en;' and this inflexion 'en' indicates the original existence of a nominative in 'ên.' Though 'ên' is no longer found in a separate shape, it survives in the inflexions of verbs; in which the sign of the first person singular is 'ên,' sometimes poetically shortened to 'en.' [From page 292]
-
In the higher dialects of Tamil 'yân' is more commonly used than 'nân,' the inflexion of which is not 'yan' but 'en,' as in the colloquial dialect.* [From page 293]
-
From the examples which have been adduced above, it would appear that there are three forms of the pronoun of the first person singular recognised in Tamil, viz., 'nân,' 'yân' and 'ên.' [From page 293]
-
It is not easy to determine whether the included vowel of this pronoun was originally 'â' or 'ê.' [From page 293]
-
A comparison of the corresponding plurals 'nâm,' 'yâm,' and 'em' (the inflexion of 'êm'), and of the plural terminations of the verb 'ôm,' 'âm,' 'am' and 'em,' leads to the conclusion that 'â' was most probably the original vowel. [From page 293]
-
It might appear, indeed, at first sight that 'yân' was an older form than 'nân;' ... [From page 294]
-
In Malayâlam, the nominative is 'ñjân' ('ny,' 'jñ,' or 'ñj,' the nasal of the palatal 'varga,' is to be pronounced as one letter, like the 'ni' of onion); but the oblique form, or inflexion, is 'en' as in Tamil, ... [From page 294]
-
The ordinary Malayâla verb is destitute of personal endings: but in the poetry an inflected form of the verb is occasionally used, in which the pronominal termination of the first person singular is 'ên,' precisely as in Tamil. [From page 294]
-
In Canarese, ... The ancient dialect uses 'ân,' the inflexion of which is 'en' - identical with that of the Tamil. [From page 294]
-
The pronominal terminations of the first person singular of the Canarese verb are 'en' in the ancient dialect, and 'êne,' 'ênu,' and 'enu' in the modern. [From page 294]
-
The Tulu nominative is 'yân,' the inflexion 'yân,' the pronominal ending of the verb 'e,' which is probably softened from 'en.' [From page 294]
-
In Telugu the nominative of this pronouns is 'nên-u;' in the higher dialect 'ênu' (answering to 'ên,' the Tamil-Canarese pronominal ending of the verb and 'en,' the Tamil and Ancient Canarese inflexion); ... [From page 295]
-
... this preference of 'ê' to 'â' appears also in the plural, which is 'mêm-u,' and in the higher dialect 'êmu.' [From page 295]
-
'nê' may be used at pleasure instead of 'nên-u,' like 'nâ' in Canarese; and in the higher dialect 'ên-u' is sometimes represented by 'ê' alone. [From page 295]
-
From the examples which have been adduced above, it would appear that there are three forms of the pronoun of the first person singular recognised in Tamil, viz., 'nân,' 'yân' and 'ên.' The first of these forms, though the most common, was probably the primitive one: its initial 'n' was first, I think, softened to 'y,' and finally abandoned. [From page 293]
-
A comparison of the corresponding plurals 'nâm,' 'yâm,' and 'em' (the inflexion of 'êm'), and of the plural terminations of the verb 'ôm,' 'âm,' 'am' and 'em,' leads to the conclusion that 'â' was most probably the original vowel. [From page 293]
-
... The final 'ôm' ... could not well have been corrupted from 'êm,' but would spring naturally enough from 'âm;' and of this we have a proof in the circumstance that 'âm' (from 'âgum,' it is, yes) is also sometimes converted into 'ôm.' Moreover, whilst there are many instances of the change of 'a' into 'e' of 'ei,' there is not any of the converse. ... [From page 293]
-
... I conclude, therefore, that 'â' was originally the included vowel of the Tamil pronoun of the first person, and that 'nân,' the ordinary colloquial form of the pronoun, is the most faithful representative of the primitive Tamil I. [From page 294]
-
It might appear, indeed, at first sight that 'yân' was an older form than 'nân;' but before our investigation is concluded, we shall be convinced, I think, that the 'n' is radical. 'n' is known to change into 'y;' but 'y' evinces no tendency to be changed into 'n.' [From page 294]
-
The compound sound of 'ñj' or 'ny,' in the Malayâla 'ñjân' or 'nyân' is a middle point between 'n' of 'nân,' and the 'y' of 'yân.' It is a softened and nasalised form of 'n,' from which the change to 'y' is easily made. In like manner, 'nîn,' the original form of the pronoun of the second person singular in all the Drâvidian dialects, has become in Tamil, first 'nîÿ,' then 'nî;' and in the verbal terminations 'aiy,' 'i,' and 'ei.' [From page 294]
-
In Canarese, ... 'ân' is evidently softened from the Tamil 'yân,' as 'yân' from 'nyân,' or 'ñjân,' and that from 'nân'; and the same softening is apparent in the Canarese plural 'âm' (instead of 'yâm' or 'nâm'), we. [From page 294]
-
The crude form of this pronoun ('nâ') is sometimes used in Canarese as a nominative, instead of 'nânu;' e.g., 'nâ bandenu,' I came; ... [From page 294]
-
'nênu,' I, takes 'nâ' for its inflexion or oblique form; and this shows that 'â' and not 'ê' was originally the included vowel in Telugu, as well as in Tamil and Canarese. This view is corroborated by the accusative of this pronoun in Telugu, which is 'nanu' or 'nannu,' me, (compare the Canarese accusative 'nanna' or 'nannu'), and which has evidently been derived from a nominative, 'nân' or 'nâ.' [From page 295]
-
From this comparison the weight of evidence appears to be in favour of our reading 'nân,' the Tamil nominative, as the best existing representative of the old Drâvidian nominative this pronoun, and 'nâ,' the crude form of the Canarese, as the primitive, unmodified root. This conclusion will be found to gain strength from the investigation of the pronoun of the second person, the root of which will appear to be not 'î' or 'yî,' but 'nî.' [From pages 295 and 296]
-
Each consonant of 'nân' evinces a tendency to be softened away. The initial 'n,' though the more essential of the two, has been softened into 'dñj' or 'ny,' then into 'y,' and finally has disappeared; and in none of the dialects has it, or any relic of it, been retained in the personal terminations of the verb. [From page 296]
-
The final 'n,' though not a part of the root, has shown itself more persistent, especially in the verbal terminations; but in the Telugu and Ku inflexion 'nâ,' in the Canarese crude nominative 'nâ,' and in the corresponding Telugu 'nê,' it has disappeared altogether. [From page 296]
-
The pronoun of the second person singular in Telugu is 'nîvu,' thou, from 'nî,' the radical base, and 'vu' an euphonic addition. This 'vu' is of so little importance that it totally disappears in all the oblique cases. Nevertheless, it forms the regular termination of the second person singular of the Telugu verb; and it has acquired this use simply through the accident of position, seeing that it has not even a sign of number, much less of personality, but is merely an euphonisation. [From page 297]
-
If, as we have seen, 'nâ' is to be regarded as the primitive form of the Drâvidian pronoun of the first person, and the final 'n' as merely a sign of number, or as an euphonic formative, it might appear extraordinary, that in the pronominal terminations of the verb, the initial 'n,' the primitive sign of the personality has invariably and altogether disappeared; whilst the first person singular is represented by the final 'n' alone. We might almost be led to suppose the initial 'n' to be a formative prefix, and the succeeding vowel to be the real pronominal base. Formative and definitive pre-fixes, however, are utterly unknown to the Dravidian languages; [From pages 296 and 297]
-
Supposing 'nî' to be the primitive form of the Drâvidian pronoun of the second person, and comparing it with 'nâ,' which we have seen to be the primitive form of the Drâvidian pronoun of the first person, it is deserving of notice that the only difference between the two is the difference between the two included vowels, 'a' and 'i.' ... The method adopted by the Dravidian languages of expressing the difference between the first person and the second by means of of the vowel 'a' and 'i,' used as auxiliaries, does not appear to be the result of accident. It is probably founded on some ultimate principle; Though it may be difficult or impossible now to discover what that principle is. [From pages 309 and 310]
-
If 'a' and 'i' be considered as identical with the demonstratives, an idea which would suit the signification, and which is corroborated by the circumstance that 'u' is also a demonstrative, we are met by the apparently insurmountable difficulty that in all the Drâvidian tongues, and (as far as the use of these demonstrative vowels extends) in all the tongues of the Indo-European family, 'a' is not the proximate, but the remote, demonstrative; and 'i' is not the remote, but the proximate; whilst 'u' is used in Tamil as an intermediate between these two. [From page 310]]
There is consensus among scholars that the first-personal pronominal root consists of three components, namely, the leading consonant, the vowel that animates it and the trailing nasal consonant n
or anusvāra.
Since this seems largely correct, the main task is to analyse how these three components were in the first-person pronominal root.
But before analysing Caldwell's arguments (both for
and against) on nān
being the first-person pronominal root,
a few inaccuracies in those arguments need to be addressed first.
-
The Tulu nominative is 'yân,' the inflexion 'yân,' the pronominal ending of the verb 'e,' which is probably softened from 'en.' [From page 294]
As can be seen in the "Introduction and Context", Tulu has both yān
and ēn
as the first-personal pronominal nominative forms in different dialects.
The inflexional prefix form for both these pronouns in the non-nominative cases is not yān
as Caldwell asserts,
but en
instead, just like in Old Kannada and Tamil.
Also, the first-person singular suffix of the verb finite forms is not e
as Caldwell asserts,
but æ
instead.
More details about the distinction between æ
and e
can be seen in the "Introduction and Conetxt"
as well as in another research article "drāviḍabhāṣegaḷa viśēṣasvaragaḷu"
(please note that the document is in Kannada and is yet to be translated to English).
-
'nênu,' I, takes 'nâ' for its inflexion or oblique form; and this shows that 'â' and not 'ê' was originally the included vowel in Telugu, as well as in Tamil and Canarese. This view is corroborated by the accusative of this pronoun in Telugu, which is 'nanu' or 'nannu,' me, (compare the Canarese accusative 'nanna' or 'nannu'), and which has evidently been derived from a nominative, 'nân' or 'nâ.' [From page 295]
The first-person singular accusative forms in Kannada are "nanna
or nannannu
" and not "nanna
or nannu
".
This is probably an error while typesetting/printing.
Of the three components that make up the first-person pronominal root, it is easier to decide about the trailing consonant n
or anusvāra.
Caldwell has insisted that this must be the consonant n
and Masti Venkatesha Iyengar also seems agree with this.
Though kēśirāja specifies anusvāra for the trailing consonants n
or m
in the verses of śabdamaṇidarpaṇaṁ
(for example, verse 103),
the nān
form with the trailing consonant n
does appear in some explanations and examples for the verses,
as can be seen in the "Introduction and Context".
But since forms without the trailing consonant n
also exist (for example, nā
of Kannada, nē
of Telugu),
the trailing component must be originally just a nasalisation, as explained in "A Nose for Nasalisation".
I.e.,
In ISO 15919,
nām̐
=> nān
- The nasalisation turns inton
in Old Kannada and Tamil.nān + u
=> nānu
- Kannada appends an euphonic vowelu
to the consonant ending.
=> nā
- The nasalisation disappears in Kannada.
ām̐
=> ān
- The nasalisation turns inton
in Old Kannada.ān + u
=> ānu
- Middle and Havyaka Kannada append an euphonic vowelu
to the consonant ending.
nēm̐
=> nēn
- The nasalisation turns inton
in Telugu.nēn + u
=> nēnu
- Telugu appends an euphonic vowelu
to the consonant ending.
=> nē
- The nasalisation disappears in Telugu.
ēm̐
=> ēn
- The nasalisation turns inton
in Tulu and literary Telugu.ēn + ŭ
=> ēnŭ
- Tulu appends an euphonic vowelŭ
to the consonant ending.
ēn + u
=> ēnu
- Telugu appends an euphonic vowelu
to the consonant ending.
=> ē
- The nasalisation disappears in literary Telugu.
In IPA,
n̪ɑ̃ː
=> n̪ɑːn̪
- The nasalisation turns inton̪
in Old Kannada and Tamil.n̪ɑːn̪ + u
=> n̪ɑːn̪u
- Kannada appends an euphonic vowelu
to the consonant ending.
=> n̪ɑː
- The nasalisation disappears in Kannada.
ɑ̃ː
=> ɑːn̪
- The nasalisation turns inton̪
in Old Kannada.ɑːn̪ + u
=> ɑːn̪u
- Middle and Havyaka Kannada append an euphonic vowelu
to the consonant ending.
n̪ẽː
=> n̪eːn̪
- The nasalisation turns inton̪
in Telugu.n̪eːn̪ + u
=> n̪eːn̪u
- Telugu appends an euphonic vowelu
to the consonant ending.
=> n̪eː
- The nasalisation disappears in Telugu.
ẽː
=> eːn̪
- The nasalisation turns inton̪
in Tulu and literary Telugu.eːn̪ + ɯ
=> ēnŭ
- Tulu appends an euphonic vowelɯ
to the consonant ending.
eːn̪ + u
=> eːn̪u
- Telugu appends an euphonic vowelu
to the consonant ending.
=> eː
- The nasalisation disappears in literary Telugu.
Caldwell mentions the first-personal suffix of verb finite forms only with a trailing consonant n
,
but single vowel e
without the trailing consonant n
is widely used in spoken dialects and even in poetry of Modern Kannada
(for example, Dasa and Vachana literature),
for example, nānu baṁde
etc.
These variations of the first-personal suffix of verb finite forms as well as the inflexional prefixes is also explained by the nasalisation-related processes.
I.e.,
In ISO 15919,
nam̐
=> nan
- The nasalisation turns inton
in the inflexional prefix of Kannada and Telugu.
em̐
=> en
- The nasalisation turns inton
in the inflexional prefix and the first-personal suffix of verb finite forms in Old Kannada, Tamil and Tulu.=> e
- The nasalisation disappears in he first-personal suffix of verb finite forms in Tulu and spoken dialects of Kannada.
In IPA,
n̪ə̃
=> n̪ən̪
- The nasalisation turns inton̪
in the inflexional prefix of Kannada and Telugu.
ẽ
=> en̪
- The nasalisation turns inton̪
in the inflexional prefix and the first-personal suffix of verb finite forms in Old Kannada, Tamil and Tulu.=> e
- The nasalisation disappears in he first-personal suffix of verb finite forms in Tulu and spoken dialects of Kannada.
So, seeing that the nasalisation-related set of processes produce all the variations of the first-person pronominal nominative, inflexional prefix and grammatical person/gender suffix of verb finite forms, it is safe to says that the trailing component in the first-person pronominal root is a nasalisation (or that the trailing component is not present in the root but the nasalisation mediates the appending of suffixes to it).
Let us examine the leading consonant of the first-person pronominal root next.
As already seen in the "Introduction and Context", kēśirāja mentions mainly the first-person pronominal root starting with a vowel (and not any consonant).
But first-person pronominal forms starting with the consonant n
are indeed found not just in Middle and Modern but also Old Kannada.
Caldwell clearly supports the starting consonant n
by saying that nān
is the first-person pronominal root.
He uses the nān
/nānu
, nān
, nēnu
forms of Kannada, Tamil, Telugu respectively, all of which start with the consonant n
,
as a strong argument in support of the starting consonant n
.
The Malayalam equivalent ñān
being closely related to nān
supports a starting consonant n
being present in the first-person pronominal root.
But he has recorded that the yān
of Tamil and Tulu as well as the ēnu
of literary Telugu show some variations in the starting component
of the first-person pronominal root.
However, the inflexional prefix en
of Old Kannada, Tamil and Tulu argues against the starting consonant n
.
Even more than this, the first-personal suffix of verb finite forms ēn
, en
, ēne
, ēnu
, enu
of Kannada, Tamil and Tulu,
all starting with a vowel, strongly opposes the case for the starting consonant n
,
because, as mentioned in the context of the second-person pronominal roots,
the relationship between pronouns and the grammatical person/gender suffixes of verb finite forms is an ancient one
and it is likely that the grammatical person/gender suffixes of verb finite forms are close to their corresponding pronominal roots.
Caldwell acknowledges and documents this in the context of the second-person pronoun.
This casts strong doubts about not only the starting component of the first-person pronominal root being the consonant n
,
but also the starting vowel being a
.
The first-personal suffix of verb finite forms suggests that the starting vowel of the first-person pronominal root might be e
instead of a
.
Though this suggestion might be close the truth, it is not entirely accurate, the reasons for which will be discussed later.
As seen in the context of the second-person pronoun, Caldwell seems to weaken the opposition
to the starting consonant n
by grammatical person/gender suffixes of verb finite forms, by citing the example of the first-personal vu
suffix
in Telugu verb finite forms which is really just the euphonic suffix and does not have any part of the original first-person pronominal root.
Caldwell uses this example to conclude that it is not just that the pronominal roots may be much modified
in the grammatical person/gender suffixes of verb finite form, but also,
as in the case of Telugu first-personal suffix of verb finite forms, nothing of the corresponding pronominal root might be retained.
But as already seen in the context of the second-person pronoun, while Caldwell is right that the first-personal suffix of verb finite forms in Telugu
do not retain any component of the first-person pronominal root, such a phenomenon is not seen in the other Dravidian languages.
So, hoisting a process specific to Telugu on all the rest of the Dravidian languages seems unsound.
Caldwell tries to explain the variations in the starting component of the first-person pronominal root by proposing the process of n
turning into
ñ
(as in Malayalam), then into y
(as in Tamil and Tulu) and then finally disappearing (as in Tulu and literary Telugu).
While this might seem plausible at first sight, it does not survive closer scrutiny.
-
While the consonant
ñ
is found in Malayalam,y
in Tamil and Tulu and the consonant disappearing (starting with vowele
) in Tulu and literary Telugu, other examples ofn
turning intoñ
,y
or disappearing are rare in these languages. -
Though it might seem that the consonant
ñ
is suitable for turning intoy
, in reality such a transformation is not found in Malayalam and Tulu where the consonantñ
abounds in comparison with other Dravidian languages. The reason for this sequence of transformations being rare and the misconception that the transformations are plausible seems to be the same, which is that, in most languages,ñ
is rarely used as independent consonant outside of consonant clusters with trailing palatal consonant, i.e.,ñ
is rarely used with a vowel directly following it. So, the speakers of languages whereñ
is not found immediately before a vowel, being unfamiliar with the clear pronunciation of the palatal nasal consonantñ
, pronounce it variously asy
(for example,jñāna => gyāna
), orn
(for example,jñāna => gnāna
), or as a consonant cluster ofn
followed byy
(ny
). But in reality, there is no hint of the pronunciation ofn
ory
in the pronunciation of the palatal nasal consonantñ
. Native speakers of languages where the palatal nasal consonantñ
is often followed by a vowel (like in Malayalam or Tulu), pronounce the palatal nasal consonant without any hint of the pronunciation ofn
ory
. This must be whyñ
rarely turns intoy
in Malayalam and Tulu and the misconception that palatal nasal consonantñ
is suitable to turn intoy
seems to be due to the unfamiliarity of the pronunciation of the palatal nasal consonantñ
. -
The consonant
y
disappearing (or turning into vowels likea
ore
) is also rare and is not a natural process. -
Malayalam does indeed have a lot of examples for the consonant
n
turning intoñ
. But the nasal consonantsñ
,ṅ
abound in Malayalam in comparison with the other Dravidian languages. So, hoisting the processes involving the consonantñ
on all the Dravidian languages without examining the obvious possibility that it might be specific to Malayalam seems unsound. -
The third-personal singular suffix of verb finite forms in Tulu,
e
is probably the transformed form of the third-person demonstrative distal pronominal rootam̐
/ə̃
(am̐ => an => anu
orə̃ => ən̪ => ən̪u
in Kannada). More details on this can be seen here. But even this process is rare in Tulu and unsuitable to be extended into processes in the Proto-Dravidian that transform the vowela
intoe
.
If there were no other viable processes that can produce the variations in the starting component of the first-person pronominal root,
accepting the argument of the consonant n
turning into ñ
, then into y
and finally disappearing might have seemed reasonable.
But as seen in the context of the second-person pronoun,
given the process of the clear pronunciation of the nasalisation of the starting vowel being heard on both the sides of the vowel
and turning into the consonant n
on both sides of the vowel exists, it seems unreasonable.
However, while the processes involving nasalisation can explain the starting consonant n
and n
turning into ñ
is natural in Malayalam,
the starting consonant y
or vowel e
are not readily explained by the processes related to nasalisation.
But as can be seen later,
it is indeed possible to explain the starting consonant y
or vowel e
with some additional processes
which complement the processes related to nasalisation.
Taking all these arguments together, given that the links in the case for the na => ña => ya => a => e
processes are either weak
or just plain non-existent, it is safe to reject this argument.
Caldwell notes that it may not be easy to decide if the starting vowel of the first-person pronominal root is a
or e
.
-
It is not easy to determine whether the included vowel of this pronoun was originally 'â' or 'ê.' [From page 293]
But he concludes that the starting vowel in the first-personal pronominal root must be a
after considering
the first-person plural suffixes of verb finite forms in Tamil.
-
A comparison of the corresponding plurals 'nâm,' 'yâm,' and 'em' (the inflexion of 'êm'), and of the plural terminations of the verb 'ôm,' 'âm,' 'am' and 'em,' leads to the conclusion that 'â' was most probably the original vowel. [From page 293]
-
... The final 'ôm' ... could not well have been corrupted from 'êm,' but would spring naturally enough from 'âm;' and of this we have a proof in the circumstance that 'âm' (from 'âgum,' it is, yes) is also sometimes converted into 'ôm.' Moreover, whilst there are many instances of the change of 'a' into 'e' of 'ei,' there is not any of the converse. ... [From page 293]
The argument that a
can turn into o
, e
and ai
but it is rare for ai
or e
to turn into a
seems correct.
So, Caldwell's case for the vowel in the first-person pronominal root being a
might seem convincing.
But the idea about the vowel in the first-person pronominal root from Masti Venkatesha Iyengar in "namma nuḍi" enables
an even more convincing argument.
As seen in the "Introduction and Context", Masti Venkatesha Iyengar proposes a new idea that is not seen in Caldwell's discussion regarding the vowel in the first-person pronominal root.
-
... tamiḷinalli ān, nān eṁdu idda śabda teluginalli ēnu, nēnu eṁdāyitu eṁdu hēḷuvudu vāḍike. iveraḍakkū mūladalli oṁdu madhyasthavāda svara iddirabēku. adu oṁdu bhāṣege ā eṁdu innoṁdakke ē eṁdu iḷiyitu eṁdu hēḷuvudu heccu sariyāgabahudu. [From page 52]
This is a key insight in unlocking the mystery of the first-person pronominal root.
Masti Venkatesha Iyengar has not elaborated about what the intermediate vowel between a
and e
might be
and if there are any modern Dravidian languages where it might still be surviving.
But we need not search too much for this intermediate vowel.
As can be seen in "drāviḍabhāṣegaḷa viśēṣasvaragaḷu"
(please note that the document is in Kannada and is yet to be translated to English),
Tulu not only has the vowel æ
(as well as and distinct from the vowel e
), but uses it as the first-personal singular suffix
of verb finite forms (for example, ēnŭ battæ
= nānu baṁde
).
On careful observation, even in Kannada where æ
is generally considered to be absent (it is certainly absent in the script),
the e
in the first-personal suffix of verb finite forms enu
, while not identical to æ
, is not fully e
either.
This becomes even more clear in the the alternate form e
of the first-personal singular suffix of verb finite forms in the spoken dialects of Kannada
(for example, nānu baṁde
).
The fact that Tulu uses æ
as the first-personal suffix and e
as the third-personal suffix of verb finite forms (āye batte = avanu baṁdanu
) might be a reason why
Tulu has retained the distinction between æ
and e
.
Because, pronouncing the first-personal suffix of verb finite forms as e
instead of æ
does not make too much of a difference to the semantics
and with the script anyway lacking the distinction between æ
and e
, the distinction between these vowels might have or be continuing to be blurred in Kannada.
However, if æ
is pronounced as e
in such contexts in Kannada, it might sound like toddler-speech or produce some comic effect.
So, the distinction between æ
and e
might still remain in Kannada, but not as much as in Tulu.
More details about this can be seen here.
This vowel æ
is clearly suitable to turn into a
on the one side and e
on the other, as supposed by Masti Venkatesha Iyengar.
I.e.,
æ
=> a
-æ
turns intoa
.=> e
-æ
turns intoe
.
Similarly, the elongated vowel ǣ
is suitable to turn into ā
and ē
in the first-person singular nominative case.
I.e.,
ǣ
=> ā
-ǣ
turns intoā
.=> ē
-ǣ
turns intoē
.
This avoids the objection of Caldwell that, while there are many examples of a
turning into e
or ai
,
not enough examples exist of e
or ai
turning into a
, because here æ
turns into a
or e
.
In addition to turning into a
or e
, æ
is suitable for producing y
as well,
because in Kannada, where there are no symbols for æ
and ǣ
in script,
it is a common practice to transliterate English words having the vowel æ
or ǣ
as a
or ā
but with an additional y
.
For example, apple
as ಆ್ಯಪಲ್
or bank
as ಬ್ಯಾಂಕ್
/byāṁk
and calendar
as ಕ್ಯಾಲೆಂಡರ್
/kyāleṁḍar
I.e.,
ǣ
=> yā
-ǣ
turns intoyā
This process of æ
turning into ya
seems more convincing than the process of na => ña => ya => a => e
argued for by Caldwell,
because it is a common process found in many occasions in Kannada and even in other Dravidian languages.
As already seen in the case of the second-person pronominal roots,
nasalisation is either present in the pronominal roots or mediates when suffixes are appended to such pronominal roots
and when the nasalisation of the vowel æ
is pronounced clearly, it is heard almost before the vowel.
I.e.,
In ISO 15919,
æm̐ => m̐æm̐
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.
In IPA,
æ̃ => ̃æ̃
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.
Such a clear pronunciation is suitable for undergoing the processes related to nasalisation on both the sides of the vowel
and the nasalisation turning into the consonant n
is already seen above and elsewhere.
I.e.,
In ISO 15919,
m̐æm̐
=> æn
- The nasalisation disappears before the vowel and turns inton
after the vowel.=> næn
- The nasalisation turns inton
on both the sides of the vowel.
In IPA,
̃æ̃
=> æn̪
- The nasalisation disappears before the vowel and turns inton̪
after the vowel.=> n̪æn̪
- The nasalisation turns inton̪
on both the sides of the vowel.
As seen in the case of the second-person pronouns, these processes are not subject to Caldwell's objection that
prefixes are rare and unnatural in Dravidian languages, because n
is not just any old prefix but a nasal consonant
which is formed when the clear pronunciation of the nasalision of the vowel ǣ
is heard on both sides of the vowel
which then turns naturally into the nasal consonant n
on both the sides of the root vowel.
The short vowel æ
seen in these processes becomes ǣ
in the nominative case which produces the variations in the
first-person pronominal singular nominative forms ǣ
, ǣn
, nǣ
, nǣn
.
In ISO 15919,
ǣm̐
=> m̐ǣm̐
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> nǣm̐
- The nasalisation turns in ton
before the vowel.=> nǣn
- The nasalisation turns in ton
after the vowel too.=> nǣ
- The nasalisation disappears after the vowel.
=> ǣn
- The nasalisation turns inton
after the vowel.=> ǣ
- The nasalisation disappears.
In IPA,
æ̃ː
=> ̃æ̃ː
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> n̪æ̃ː
- The nasalisation turns in ton̪
before the vowel.=> n̪æːn̪
- The nasalisation turns in ton̪
after the vowel too.=> n̪ǣ
- The nasalisation disappears after the vowel.
=> æːn̪
- The nasalisation turns inton̪
after the vowel.=> æː
- The nasalisation disappears.
Similarly, the first-person inflexional prefixes and suffixes of verb finite forms, en
, e
, nēn
are also formed.
I.e.,
In ISO 15919,
æm̐
,=> m̐æm̐
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> næm̐
- The nasalisation turns inton
before the vowel.=> næn
- The nasalisation turns inton
after the vowel too.
=> æn
- The nasalisation turns inton
.=> æ
- The nasalisation disappears.
In IPA,
æ̃
=> ̃æ̃
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> n̪æ̃
- The nasalisation turns inton̪
before the vowel.=> n̪æn̪
- The nasalisation turns inton̪
after the vowel too.
=> æn̪
- The nasalisation turns inton̪
.=> æ
- The nasalisation disappears.
Taken together, these processes can be seen to produce all the variations of the first-person pronominal forms
from just the nasalised vowels æ
/ǣ
.
I.e.,
In ISO 15919,
ǣm̐
=> m̐ǣm̐
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> nǣm̐
- The nasalisation turns in ton
before the vowel.=> nǣn
- The nasalisation turns in ton
after the vowel too.=> nān
-ǣ
turns intoā
in Old Kannada and Tamil.nān + u
=> nānu
- Kannada appends an euphonicu
to the consonant ending.
=> nēn
-ǣ
turns intoē
.nēn + u
=> nēnu
- Telugu appends an euphonicu
to the consonant ending.
=> nǣ
- The nasalisation disappears after the vowel.=> nē
-ǣ
turns intoē
in Telugu.
=> ǣn
- The nasalisation turns inton
.=> ān
-ǣ
turns intoā
in Old Kannada.ān + u
=> ānu
- Old Kannada and Havyaka Kannada (also in Telugu first-personal suffix of verb finite forms) append an euphonicu
to the consonant ending.
=> ēn
-ǣ
turns intoē
.ēn + ŭ
=> ēnŭ
- Tulu appends an euphonicŭ
to the consonant ending.
ēn + u
=> ēnu
- Literary Telugu appends an euphonicu
to the consonant ending.
=> yān
-ǣ
turns intoyā
in Tulu and literary Tamil.yān + ŭ
=> yānŭ
- Tulu appends an euphonicŭ
to the consonant ending.
=> ǣ
- The nasalisation disappears.=> ē
-ǣ
turns intoē
in literary Telugu.
æm̐
=> m̐æm̐
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> næm̐
- The nasalisation turns inton
before the vowel.=> næn
- The nasalisation turns inton
after the vowel too.=> nan
-æ
turns intoa
in the inflexional prefix in Kannada and Telugu.
=> æn
- The nasalisation turns inton
in the first-personal suffix of verb finite forms in Kannada.=> en
-æ
turns intoe
in the inflectional prefixes of Old Kannada, Tamil, Tulu and in the first-personal suffixes of verb finite forms in Kannada.en + u
=> enu
- Kannada appends an euphonicu
to the consonant ending.
=> æ
- The nasalisation disappears in the first-personal suffixes of verb finite forms in Tulu and Kannada.
In IPA,
æ̃ː
=> ̃æ̃ː
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> n̪æ̃ː
- The nasalisation turns in ton̪
before the vowel.=> n̪æːn̪
- The nasalisation turns in ton̪
after the vowel too.=> n̪ɑːn̪
-æː
turns intoɑː
in Old Kannada and Tamil.n̪ɑːn̪ + u
=> n̪ɑːn̪u
- Kannada appends an euphonicu
to the consonant ending.
=> n̪eːn̪
-æː
turns intoeː
.n̪eːn̪ + u
=> n̪eːn̪u
- Telugu appends an euphonicu
to the consonant ending.
=> n̪æː
- The nasalisation disappears after the vowel.=> n̪eː
-æː
turns intoeː
in Telugu.
=> æːn̪
- The nasalisation turns inton̪
.=> ɑːn̪
-æː
turns intoɑː
in Old Kannada.ɑːn̪ + u
=> ɑːn̪u
- Old Kannada and Havyaka Kannada (also in Telugu first-personal suffix of verb finite forms) append an euphonicu
to the consonant ending.
=> eːn̪
-æː
turns intoeː
.eːn̪ + ɯ
=> eːn̪ɯ
- Tulu appends an euphonicɯ
to the consonant ending.
eːn̪ + u
=> eːn̪u
- Literary Telugu appends an euphonicu
to the consonant ending.
=> jɑːn̪
-æː
turns intojɑː
in Tulu and literary Tamil.jɑːn̪ + ɯ
=> jɑːn̪ɯ
- Tulu appends an euphonicɯ
to the consonant ending.
=> æː
- The nasalisation disappears.=> eː
-æː
turns intoeː
in literary Telugu.
æ̃
=> ̃æ̃
- The clear pronunciation of the nasalisation can be heard on both sides of the vowel.=> n̪æ̃
- The nasalisation turns inton̪
before the vowel.=> n̪æn̪
- The nasalisation turns inton̪
after the vowel too.=> n̪ən̪
-æ
turns intoə
in the inflexional prefix in Kannada and Telugu.
=> æn̪
- The nasalisation turns inton̪
in the first-personal suffix of verb finite forms in Kannada.=> en̪
-æ
turns intoe
in the inflectional prefixes of Old Kannada, Tamil, Tulu and in the first-personal suffixes of verb finite forms in Kannada.en̪ + u
=> en̪u
- Kannada appends an euphonicu
to the consonant ending.
=> æ
- The nasalisation disappears in the first-personal suffixes of verb finite forms in Tulu and Kannada.
As mentioned above, n
turning into ñ
is a natural process in Malayalam, which produces the ñān
form.
I.e.,
nān => ñān
- The startingn
turns intoñ
in Malayalam.
Of the remaining first-person plural suffixes of verb finite forms ōm
, ām
, am
, em
, it is easy to see that ām
can be produced from ǣ
and am
, em
from æ
, which leaves only the process for producing ōm
to be explained.
The process of words ending with the nasalised vowel a
(even non-native, loan-words) being pronounced with the nasalised vowel o
is already discussed in the previous research document on the processes related to nasalisation.
For example, aparaṁ
is often pronounced as aprom̐
in spoken dialects of Tamil.
So, ām
can turn into ōm
by this natural process in Tamil, which has also been recorded and acknowledged by Caldwell.
I.e.,
ām => ōm
- The endinga
, when nasalised or followed by anusvāra orm
, turns intoo
in Tamil.
Taken together, the processes of producing all the first-personal suffixes ōm
, ām
, am
, em
in Tamil from ǣ
/æ
could be as follows.
In ISO 15919,
ǣm̐
=> ām̐
-ǣ
turns intoā
in the first-personal suffix of verb finite forms in Tamil.=> ām
- The nasalisation turns intom
in the plural.=> ōm̐
- The endingā
, when nasalised or followed by anusvāra orm
, turns intoō
in Tamil.=> ōm
- The nasalisation turns intom
in the plural.
æm̐
=> æm
- The nasalisation turns intom
in the first-person plural suffix of verb finite forms in Tamil.=> am
-æ
turns intoa
.=> em
-æ
turns intoe
.
In IPA,
æ̃ː
=> ɑ̃ː
-æː
turns intoɑː
in the first-personal suffix of verb finite forms in Tamil.=> ɑːm
- The nasalisation turns intom
in the plural.=> õː
- The endingɑː
, when nasalised or followed by anusvāra orm
, turns intooː
in Tamil.=> oːm
- The nasalisation turns intom
in the plural.
æ̃
=> æm
- The nasalisation turns intom
in the first-person plural suffix of verb finite forms in Tamil.=> əm
-æ
turns intoə
.=> em
-æ
turns intoe
.
Thus, all the variations of the first-person pronominal nominative forms, their inflextional prefixes and the corresponding
grammatical person/gender suffixes of verb finite forms are produced by these set of related processes acting on the nasalised vowel æm̐
/æ̃
(and ǣm̐
/æ̃ː
).
This leads to the inevitable conclusion that the nasalised vowel æm̐
/æ̃
(and ǣm̐
/æ̃ː
), and not nān
,
is the first-person pronominal root in all the Dravidian languages.
A cursory look at the the first-person pronoun forms in the Indo-European
(especially, in the reconstructed Proto-Indo-European)
seems to further support this conclusion.
I have used IPA in the image because I could not get the ISO 15919 character ಁ
to render properly in the image.
The depth and breadth of Caldwell's insight in comparing the third-person pronominal roots with the first and second person
pronominal roots becomes apparent in light of the conclusion, as seen above, that the distal vowel a
is the third-person demonstrative distal,
the proximal and medial/presentative vowels i
and u
are, correspondingly third-person demonstrative proximal and medial/presentative
as well as second-person and the self-referential vowel æ
is the first-person, pronominal roots in all the Dravidian languages.
-
Supposing 'nî' to be the primitive form of the Drâvidian pronoun of the second person, and comparing it with 'nâ,' which we have seen to be the primitive form of the Drâvidian pronoun of the first person, it is deserving of notice that the only difference between the two is the difference between the two included vowels, 'a' and 'i.' ... The method adopted by the Dravidian languages of expressing the difference between the first person and the second by means of of the vowel 'a' and 'i,' used as auxiliaries, does not appear to be the result of accident. It is probably founded on some ultimate principle; Though it may be difficult or impossible now to discover what that principle is. [From pages 309 and 310]
-
If 'a' and 'i' be considered as identical with the demonstratives, an idea which would suit the signification, and which is corroborated by the circumstance that 'u' is also a demonstrative, we are met by the apparently insurmountable difficulty that in all the Drâvidian tongues, and (as far as the use of these demonstrative vowels extends) in all the tongues of the Indo-European family, 'a' is not the proximate, but the remote, demonstrative; and 'i' is not the remote, but the proximate; whilst 'u' is used in Tamil as an intermediate between these two. [From page 310]]
That the proximal vowel i
and the medial/presentative vowel u
are, respectively, the third-person demonstrative proximal and medial/presentative
as well as the second-person pronominal roots was shown in the section on the second-person pronominal root.
Now in the discussion of the first-person pronominal root, the last remaining objection that the distal vowel a
is not suited for the
first-person semantics is also eliminated by showing that the self-referential vowel æ
, and indeed not the distal vowel a
,
is the first-person pronominal root.
So, it turns out that, while it might have been difficult, it is not impossible to discover the ultimate principle behind the root vowels of
all the first, second and third person pronominal roots in the Dravidian languages.
With a small modification that the self-referential vowel æ
, and not the distal vowel a
, is the first-person pronominal root,
Caldwell's deep insight indeed turns out to be correct and is brought to a positive logical conclusion.
Only a couple of unfortunate mis-steps such as an incomplete analysis of Tulu pronominal forms and a misconception about the pronunciation of the
palatal nasal consonant ñ
seem to have prevented Caldwell from reaching this positive conclusion himself.
But this positive result was only possible by building on the continuous tradition of deep, broad and relentless research over the centuries
by all the great scholars including but not limited to Masti Venkatesha Iyengar, Caldwell and kēśirāja.
I cannot but feel grateful for such a great tradition and heritage.
As Caldwell remarks, the related set of processes acting on the root vowels, a
, i
, u
and æ
to form the first, second and third person pronominal forms
as well as the corresponding grammatical person/gender suffixes of verb finite forms of such simple, logical and beautiful regularity
in the Dravidian languages are second to none and probably unmatched.
Because traces of this system is found, but with less regularity and consistency, in the other language families (for example, the Indo-European) as well,
it can be speculated that the system is at least as old as a possible common ancestor of these families if they indeed have a shared ancestry.
It is fortunate that the modern Dravidian languages have largely retained this system with sufficient regularity and consistency,
which sheds light on some aspects of the deep ancestry of pronouns across language families.
Further investigation of the pronominal forms in the other language families in the light of this result might be productive.
The self-referential vowel æ
can found quite a bit in the first-person pronominal forms of Indo-European languages.
But there are variations with other vowels, including the vowel a
, as can be seen in the sample below.
Please note that this list not complete by any means and is presented here only for the purpose of comparison.
Language | First-person pronouns, with vowels e or ai |
First-person pronouns, with the vowel a |
First-person pronouns, with other vowels |
---|---|---|---|
Sanskrit | अहम् (ahaṁ), आवाम् (āvāṁ), वयम् (vayaṁ) | ||
Hindi | मैं (mēm̐) | हम (haṁ) | |
Gujarati | હું (haṁ), અમે (amē) | ||
Punjabi | ਮੈਂ (mēm̐) | ਅਸੀ (asīṁ) | |
Bhojpuri | हम (haṁ) | ||
Bengali | আমি (āmi), আমি (āmarā) | ||
English | I | We | |
German | Ich, Wir | ||
Greek | ἐγώ (egó, ēgo), εμείς (emeis, emīs) | ||
Latin | ego (ēgo) | nos (nōs) | |
Proto-Indo-European | heǵ(oH/Hom), eǵoH, uei, weh |
In the above list, the first-person pronominal form of the reconstructed Proto-Indo-European h₁eǵ(oH/Hom)
, eǵoH
is closer to the Greek and Latin forms egó
, ēgo
, ego
etc. rather than the Sanskrit ahaṁ
.
These forms clearly have a vowel that is close to the self-referential root vowel æ
.
The same self-referential root vowel æ
is also found in the reconstructed first-person plural forms uei
, weh
.
I do not know the reason behind choosing these forms while reconstructing the Proto-Indo-European language
and I will be grateful if knowledgeable readers point me to relevant works.
But choosing these forms seems to be consistent with the result that the vowel æ
is the first-person pronominal root,
because in the pronunciation of the Greek, Latin and Proto-Indo-European ἐγώ
(egó
, ēgo
), ego
(ēgo
)heǵ(oH/Hom)
, eǵoH
,
the first vowel sounds almost the same as æ
.
There is no similarity in the interrogative pronominal forms in the Dravidian and the other language families like the Indo-European, but among the Dravidian languages there is a lot of commonality. An investigation of the interrogative pronominal roots turns out to be complementary to the investigation of the first, second and the third person pronominal roots. There is not much conflict amongst the views of scholars regarding the interrogative pronominal roots and where conflicts exist, they are easily resolved in light of the discussion on the first-person pronominal root. So, let us briefly see the main points about the interrogative pronouns mentioned in the "Introduction and Context" and proceed to resolve any conflicts that might be there.
-
As Caldwell mentions, the processes producing the interrogative pronominal forms seem identical to those producing the third-person pronominal forms. I.e., interrogative pronominal forms are produced when the third-person pronominal root is substituted by the interrogative pronominal root in the corresponding third-person pronominal forms. For example,
avanu
/yāvanu
,avaḷu
/yāvaḷu
etc. -
Old Kannada has interrogative pronominal forms starting with the vowel
ā
. For example,āvudu
etc. As seen in the "Introduction and Context", kēśirāja only mentions these forms in the verses of śabdamaṇidarpaṇaṁ. -
Interrogative pronominal forms starting with the vowel
e
orē
are found in all the Dravidian languages. For example,eṁta
,ēnu
,ēke
in Kannada;enna
etc. in Tamil;ērŭ
,eṁca
etc. in Tulu;ēmi
,ē
=which
etc. in Telugu. -
Interrogative pronominal forms starting with
yā
are found in Kannada and Tamil. For example,yāvan
/yāvanu
,yāvaḷ
/yāvaḷu
,yāvar
/yār
/yāru
etc. -
Interrogative pronominal forms starting with
ō
orvā
are found in Tulu. For example,ōvu
=yāvudu
,vā
=yāva
. Caldwell doesn't seem to have noticed these forms.
The variations ā
, e
/ē
, yā
, ō
and vā
of the starting component of the interrogative pronominal forms provide a clue about what might be the interrogative pronominal root.
It was already seen in the section on first-person pronominal roots that,
of these variations, ā
, e
/ē
, yā
can be produced from æ
/ǣ
.
I.e.,
æ
=> a
-æ
turns intoa
- This form is not found in the interrogative pronominal forms.=> e
-æ
turns intoe
.=> ya
-æ
turns intoya
- This form is not found in the interrogative pronominal forms.
ǣ
=> ā
-ǣ
turns intoā
.=> ē
-ǣ
turns intoē
.=> yā
-ǣ
turns intoyā
.
As discussed in the sections on the first, second and the third person pronominal roots,
nasalisation either is part of pronominal roots
or mediates the appending of suffixes to the pronominal roots, which leads to different pronominal forms.
The process of producing ō
from ǣ
was also seen in the section on
first-person pronouns,
which is the same process naturally found in Tamil and Tulu.
I.e.,
In ISO 15919,
ǣm̐
=> ām̐
-ǣ
turns intoā
.=> ōm̐
- The ending nasalised vowelā
turns into the nasalised vowelō
.
In IPA,
æ̃ː
=> ɑ̃ː
-æː
turns intoɑː
.=> õː
- The ending nasalised vowelɑː
turns into the nasalised voweloː
.
The ō
, thus produced, can further produce the vā
of Tulu,
because, the vowels u
and o
turning into v
is a common process in not just Tulu but also in most other languages.
The reason for this is the fact, as mentioned in the research article on nasalisation,
that the pronunciation of u
, o
and v
are closely related and involve the use of the lips.
I.e.,
o => va
-o
turns intova
- This form is not found in the interrogative pronominal forms.ō => vā
-ō
turns intovā
.
Taking all these processes together, the processes producing all the variations of the starting component of the interrogative pronominal forms could be as follows.
In ISO 15919,
ǣm̐
=> ēm̐
-ǣ
turns intoē
in most of the Dravidian languages.=> ēn
- The nasalisation turns inton
in Kannada and Tamil.ēn + u
=> ēnu
- Kannada appends an euphonic vowelu
to the consonant ending.
=> ēm
- The nasalisation turns intom
in Telugu.ēm + i
=> ēmi
- Telugu appends an euphonic voweli
to the consonant ending.
=> yām̐
-ǣ
turns intoyā
in Kannada and Tamil.yām̐ + am̐
=> yāvam̐
- The initial nasalisation turns intov
.=> yāvan
- The final nasalisation turns inton
.yāvan + u
=> yāvanu
- Kannada appends an euphonic vowelu
to the consonant ending.
=> ām̐
-ǣ
turns intoā
in Old Kannada.ām̐ + udu
=> āvudu
- The nasalisation either has turned intov
or has disappeared and the euphonicv
has infixed between the vowelsā
andu
.
=> ōm̐
- The ending nasalised vowelā
turns into the nasalised vowelō
in Tulu.ōm̐ + u
=> ōvu
- The nasalisation turns intov
.
ōm̐ + a
=> ō + a
- The nasalisation disappears.=> vā
-ō
turns intov
and elongates the vowel animating it in Tulu.
In IPA,
æ̃ː
=> ẽː
-æː
turns intoeː
in most of the Dravidian languages.=> eːn̪
- The nasalisation turns inton̪
in Kannada and Tamil.eːn̪ + u
=> eːn̪u
- Kannada appends an euphonic vowelu
to the consonant ending.
=> eːm
- The nasalisation turns intom
in Telugu.eːm + ɪ
=> eːmɪ
- Telugu appends an euphonic vowelɪ
to the consonant ending.
=> jɑ̃ː
-æː
turns intojɑː
in Kannada and Tamil.jɑ̃ː + ə̃
=> jɑːʋə̃
- The initial nasalisation turns intoʋ
.=> jɑːʋən̪
- The final nasalisation turns inton̪
.jɑːʋən̪ + u
=> jɑːʋən̪u
- Kannada appends an euphonic vowelu
to the consonant ending.
=> ɑ̃ː
-æː
turns intoɑː
in Old Kannada.ɑ̃ː + ud̪u
=> ɑːʋud̪u
- The nasalisation either has turned intoʋ
or has disappeared and the euphonicʋ
has infixed between the vowelsɑː
andu
.
=> õː
- The ending nasalised vowelɑː
turns into the nasalised voweloː
in Tulu.õː + u
=> õːʋu
- The nasalisation turns intoʋ
.
õː + ə
=> oː + ə
- The nasalisation disappears.=> ʋoː
-oː
turns intoʋ
and elongates the vowel animating it in Tulu.
With this, it seems appropriate to say that the interrogative vowel æ
/ǣ
is the interrogative pronominal root,
but the same vowel also being the first-personal pronominal root seems confusing; this will be picked next for discussion.
The same vowel æ
/ǣ
being both the first-person and interrogative pronominal roots might seem problematic.
The difference of mostly the short vowel æ
being used in the first-person and the long vowel ǣ
being used in the interrogative pronominal forms
may not be the full solution to this problem, because the long vowel ǣ
does appear in the first-person nominative case.
However, the solution for this problem is easily found,
which is that in the first-personal pronominal forms the vowel is pronounced with a normal tone,
whereas in the interrogative pronominal forms the vowel is pronounced with an interrogative uptick in intonation.
The fact that such an interrogative tone is heard not only in interrogative pronominal forms of Dravidian languages,
but also in other languages when a nasalised æ
/ǣ
(i.e. am̐
/æ̃
and ǣm̐
/æ̃ː
) might be used naturally in an interrogative sense,
seems to show a deep and general linguistic feature.
This strengthens not only the argument that the nasalised vowel æ
pronounced in an interrogative tone is the interrogative pronominal root,
but also the general insight that the nasalised vowels a
, i
, u
and æ
are variously the third, second, first and interrogative pronominal roots
in the Dravidian languages.
The fact that these vowels are used systematically in the Dravidian languages (and less systematically in the other language families)
and the close resemblance these root vowels have to the initial utterances of babies seem to bridge the gap between the ancient roots of language
development and the roots of language learning in developmental biology,
which makes it both ancient and ever-green.
So, Caldwell using the adjective Japhetic to describe the antiquity of these root vowels seems beautifully appropriate;
not literally, of course, but metaphorically.
The processes behind the interrogative/speculative/skeptical suffixes ē, ā, ō and the emphasising suffix ē in Kannada
In Kannada and other Dravidian languages, the vowels ā
, ō
are used as interrogative, speculative/skeptical suffixes
and ē
as interrogative, speculative/skeptical and emphasising suffix.
The interrogative, speculative/skeptical suffixes might be used to express surprise as well.
For example,
Interrogative, speculative/skeptical ā
,
baṁdanu + ā? => baṁdanā?
Interrogative, speculative/skeptical ō
,
baṁdanu + ō? => baṁdanō?
Interrogative, speculative/skeptical ē
,
baṁdanu + ē? => baṁdanē?
Emphasising ē
,
(dēvaru + ē) gati => dēvarē gati
As already seen above, if the root vowel æ
, pronounced in an interrogative tone, is the interrogative pronominal root,
then it seems reasonable to suppose that it might be the root of the interrogative, speculative/skeptical suffixes ā
, ō
and ē
.
I.e.,
In ISO 15919,
ǣm̐
=> ēm̐
-ǣ
turns intoē
.=> ē
- The nasalisation disappears in the interrogative, speculative/skeptical suffixē
.
=> ām̐
-ǣ
turns intoā
.=> ā
- The nasalisation disappears in the interrogative, speculative/skeptical suffixā
.=> ōm̐
- The nasalised ending vowelā
turns into the nasalised vowelō
.=> ō
- The nasalisation disappears in the interrogative, speculative/skeptical suffixō
.
In IPA,
æ̃ː
=> ẽː
-æː
turns intoeː
.=> eː
- The nasalisation disappears in the interrogative, speculative/skeptical suffixeː
.
=> ɑ̃ː
-æː
turns intoɑː
.=> ɑː
- The nasalisation disappears in the interrogative, speculative/skeptical suffixɑː
.=> õː
- The nasalised ending vowelɑː
turns into the nasalised voweloː
.=> oː
- The nasalisation disappears in the interrogative, speculative/skeptical suffixoː
.
These set of processes producing the interrogative, speculative/skeptical suffixes ā
, ō
and ē
strengthen the argument that the root vowel æ
, pronounced in an interrogative tone, is the interrogative pronominal root.
Though it can be said that the emphasising suffix ē
is also derived from the interrogative root vowel æ
,
from the perspective of linguistics and semantics, it might be more appropriate to say that it is derived from the self-referential root vowel æ
.
I have used IPA in the image because I could not get the ISO 15919 character ಁ
to render properly in the image.
The processes producing the interrogative pronominal forms in the Indo-European languages do not seem to be as systematic or consistent as
those in the Dravidian languages.
But in the interrogative pronominal forms of Sanskrit, the languages derived from the various Prakrits and Latin,
the consonant k
(or something close to it) abounds.
For example, कः (kaḥ), का (kā), किम् (kiṁ), कुत्र (kutra), etc. in Sankrit, कौन (kōn), क्या (kyā), कहाँ (kahām̐) etc. in Hindi,
ਕੀ (kī), ਕਿੱਥੇ (kithē) etc. in Punjabi, का (kā), कहाँ (kahām̐) etc. in Bhojpuri, কি (ki), কোথায় (kōthāya) in Bengali
and quis
, quid
in Latin.
But the consonant ś
is found in Gujarati. For example, શું (śum̐).
The consonant v
(or something close to it) abounds in English and German.
For example, who
, what
etc. in English, was
, wo
etc. in German.
The consonant t
(or something close to it) is used in Greek. For example, τι
.
So, the Indo-European languages seem to have taken their own different paths for the interrogative pronominal forms.
The root vowels being long in the nominative case but short in the inflexional prefixes of the non-nominative cases and in the grammatical person/gender suffixes of verb finite forms is already mentioned in the "Introduction and Context" as well as in this document. What could be the reason for this peculiarity? I.e.,
āṁ
/ān
/ānu
/nānu
/nā
- The vowel is elongated in the nominative case (ā
).- In Tulu,
ēnŭ
/yānŭ
- The vowel is elongated in the nominative case (ē
/ā
). nīnu
/nī
- The vowel is elongated in the nominative case (ī
); in Tulu, the equivalent nominative form isī
.nannaṁ
/nanna
/nannanu
/nannannu
,nanniṁ
/nanniṁda
/nanniṁde
,nanage
/naṁge
,nanna
,nannoḷ
/nannoḷu
/nannal
/nannali
/nannalli
- The vowel is short in the non-moninative cases (a
).ennaṁ
/enna
/ennanu
/ennannu
,enniṁ
/enniṁda
/enniṁde
,enage
,enna
,ennoḷ
/ennoḷu
/ennal
/ennali
/ennalli
- The vowel is short in the non-nominative cases (e
).
But it would be incorrect to say that the root vowel is elongated only in all the nominative case,
because only the short vowel is seen in the third-person demonstrative pronominal nominative cases (avanu
, avaḷu
, adu
, avu
etc.)
and the root vowel is elongated in the demonstrative determiner
(ā
, ī
, ū
, in Tulu vā
etc.).
As already seen above, the aggregating suffix in Modern Kannada is also
the elongated presentative pronominal root vowel ū
.
All the examples of the root vowels being elongated, if taken together, would be as follows.
āṁ
/ān
/ānu
/nānu
/nā
- The root vowel is elongated in the first-person nominative (ā
).- In Tulu,
ēnŭ
/yānŭ
- The root vowel is elongated in the first-person nominative (ē
/ā
). nīnu
/nī
- The root vowel is elongated in the second-person nominative (ī
); in Tulu, the equivalent nominative isī
.ā
,ī
,ū
, in Tuluvā
etc. - The root vowel is elongated in the third-person demonstrative determiners.ū
- The root vowel is elongated in the aggregating suffix in Modern Kannada.
It is significant that all these forms are made up of only one or two syllables and never more than that.
If we consider only the monosyllabic forms from this list, most of them are frequently seen at the beginning of sentences,
except for ū
, the aggregating suffix in Modern Kannada.
I.e.,
āṁ/ān/nā baṁdeṁ/baṁden
nī baṁde
ā/ī/ū biṁdige
If the root vowel was not elongated in these sentences, the daDUM
rhythm (i.e., a light syllable followed by a heavy syllable)
would have formed at the beginning of the sentence, which, as already seen above, is undesirable in Kannada
and also in the other Dravidian languages though to a somewhat lesser extent.
I.e.,
aṁ/an/na baṁdeṁ/baṁden
- daDUMDUMni baṁde
- daDUMdaa/i/u biṁdige
- daDUMdada
In a previous research article
(please note that the document is in Kannada and is yet to be translated to English),
it can be seen that elongating the vowel in the first syllable is one of the ways of avoiding such an undesirable daDUM
rhythm
in Kannada and other Dravidian languages.
By elongating the vowel in the light first syllable, it is turned into a heavy syllable, thus avoiding the daDUM
rhythm.
I.e.,
āṁ/ān/nā baṁdeṁ/baṁden
- DUMDUMDUMnī baṁde
- DUMDUMdaā/ī/ū biṁdige
- DUMDUMdada
In the bisyllabic forms from the above list, if the first syllable remains light (i.e., the root vowel is not elongated),
then dada
rhythmic unit would have formed.
As can be seen in the same research article, this form too is undesirable in Kannada and other Dravidian languages.
I.e.,
anu
/nanu
- dada- In Tulu,
enŭ
/yanŭ
- dada ninu
- dada
Therefore, the vowel is elongated in the bisyllabic forms also to avoid this.
Since the dvitvasaṁdhi and the other related processes operate in the processes involved in the production of
the first, second person non-nominative pronominal forms and the third-person demonstrative pronominal forms (except the demonstrative determiner forms),
the problem of daDUM
rhythm gets addressed already, obviating the need to elongate the root vowel in the first syllable.
The reason for the elongated vowel in the aggregating suffix ū
of Modern Kannada, which never appears at the beginning of the sentence,
must be that if the vowel is not elongated, it becomes almost indistinguishable from the euphonic vowel u
appended to consonant endings in Modern Kannada.
Regarding the problem, why is the root vowel not elongated in the aggregating suffix uṁ
of Old Kannada,
I can offer no better solution than to merely observe that, as listed in the same research article,
forms starting with daDUM
rhythm occur more in Old Kannada in comparison with Modern Kannada
and Modern Kannada has intensified the removal of such forms.
In the light of the insight that the nasalised vowels im̐
/ɪ̃
, am̐
/ə̃
are, respectively, the proximal and distal demonstrative pronominal roots,
the derivation of the demonstrative determiner forms ikō
/igō
and akō
/agō
in Kannada also becomes clear.
Here, the proximal and the distal demonstrative pronominal roots im̐
/ɪ̃
, am̐
/ə̃
are suffixed with verb kō
in its imperative form.
That is,
In ISO 15919,
-
im̐ + kō => ikō
- The nasalisation disappears.=> igō
- The consonantk
turning intog
(i.e. voicing of stops) is a common euphonic change in Dravidian languages.
-
am̐ + kō => akō
- The nasalisation disappears.=> agō
- The consonantk
turning intog
(i.e. voicing of stops) is a common euphonic change in Dravidian languages.
In IPA,
-
ɪ̃ + koː => ɪkoː
- The nasalisation disappears.=> ɪgoː
- The consonantk
turning intog
(i.e. voicing of stops) is a common euphonic change in Dravidian languages.
-
ə̃ + koː => əkoː
- The nasalisation disappears.=> əgoː
- The consonantk
turning intog
(i.e. voicing of stops) is a common euphonic change in Dravidian languages.
The plural and honorific forms of these being ikoḷḷi
/akoḷḷi
makes the presence of the verb kō
in the sense of koḷḷu
crystal clear.
So, unlike the ŭṁda
of Tulu, ikō
, akō
of Kannada,
are complete sentences in themselves (like here/there you go
in English);
losing the pronoun part of the sentence (in this case, the second-person) is natural to Kannada in such cases, especially in the spoken dialects,
because the verbs are always inflected/conjugated for, among other things, grammatical person.
Kannada and other Dravidian languages systematically use the nasalised vowels am̐
/ə̃
, im̐
/ĩ
, um̐
/ũ
, æm̐
/æ̃
(and their longer forms ām̐
/ɑ̃ː
, īm̐
/ĩː
, ūm̐
/ũː
, ǣm̐
/æ̃ː
) as distal, proximal, medial demonstrative third-person, second-person, first-person and interrogative pronominal roots.
Other language families (for example, Indo-European), also use these but less systematically.
In other words,
- The distal demonstrative
am̐
/ə̃
is the distal demonstrative pronominal root. In Kannada, it is also the accusative suffix. - The proximal demonstrative
im̐
/ĩ
, medial/presentativeum̐
/ũ
are the proximal and medial/presentative demonstrative pronominal roots correspondingly. They both (im̐
/ĩ
andum̐
/ũ
) are also the second-person pronominal roots. Among them, in Kannada,im̐
/ĩ
is also the root of theiṁ
suffix which is widely known as the instrumental suffix but really is also the ablative suffix. - The presentative
um̐
/ũ
is also used as the aggregating suffix (uṁ
in Old Kannada andū
in modern Kannada) - Self-referential
æm̐
/æ̃
is the first-person pronominal root. It is also the root of the emphasising suffixē
. æm̐
/æ̃
with an interrogative tone is the interrogative pronominal root. It is also the root of the interrogative/speculative/skeptical suffixesā
,ē
,ō
.
This hypothesis, on one hand, takes the provenance of of the pronouns back to a pre-historical (perhaps even a proto-linguistic) time and on the other, to baby-talk. That these nasalised vowels are naturally suited to signify the above mentioned semantics can be pictorially represented as follows.
- Producing all the various interrogative forms is added.
- Producing all the various second-personal forms is added.
- Producing all the various first-personal forms is added.
- A small section on the derivation of the akō/ikō in Kannada is added.
- A small note about indications of mediating nasalisation after word fragments ending with vowels other than
a
in Tulu is added. - Interrogative Pronominal Roots and Some Loose-ends sections are translated.
- First-person Pronominal Roots section is translated.
- Second-person Pronominal Roots section is translated.
- Changed the terminology collectivising/conjugating suffix to aggregating suffix. I am grateful to Pooja P for the suggestion.
- Third-person Pronominal Roots section is translated.
- Background and Summary sections are translated.
A more detailed revision history can be seen here.