PROTO-NOSTRATIC: A GRAMMATICAL SKETCH

Phonology

Based on the lexical evidence gathered so far (despite a number of basic
disagreements about sound correspondences), the PN phonological system
must have been something like the following:

*p *b *p' *m (*f) *w *u *u:
*t *d *t' *n *s *r *l *a *a:
*k *g *k' *ng *x *y *i *i:
*q *G *q' (*N) *X
*h *3 *?

As far as the morpholoical evidence is concerned, we have to deal
only with a subset of this inventory (*b, *m, (*w), *u, *t, *n, *s,
(*l), *a, *k, *ng, (*y), *i, (*q'), (*?)).

For a better understanding of the following grammatical presentation,
we need to introduce one Proto-Nostratic soundlaw right away: the
Nasal Rule. I have borrowed this from Selqup Samoyed and Eskimo-Aleut,
but it fits Proto-Nostratic like a glove: postulating this soundlaw at the
PN stage solves a large number of otherwise intractable morphological
problems. The rule is simply that in the absolute Auslaut, PN *-p, *-t,
*-k become *-m, *-n and *-ng, respectively.


THE PN NOUN

I reconstruct an ergative system, with absolutive *-a, and ergative *-u.
In Semitic, the endings are *-a and *-u, but also *-0 in the construct
state. Kartvelian and PIE also require an ending *-0 for their "strong
cases" (Kartv. nom, voc, erg, dat; PIE nom, voc, acc). In most other
languages, due to phonetical wear, we cannot tell if the word ended in a
vowel or not. Presumably, most forms go back to absolutive *-a, while the
evidence for ergative *-u is thin outside of Afro-Asiatic.

Afro-Asiatic (Semitic) has a genitive in *-i, while an adjectival
suffix *-i or *-ya is found in other languages (e.g. Basque and PIE).

The Hausa/Egyptian relativizer *na can be linked to the Uralic genitive
in *-n, perhaps the Basque genitive in -en, and the PIE adjectival suffix
*-n-.

Another common case ending is *-si, found in Semitic as a dative
(p.pronouns) or a directive, in Sumerian as a directive (-s^e), in
Kartvelian as a dative/accusative (*-s) or as a genitive *-is, when added
to the absolutive stem. In Luwian *-a-si- is an adjectival formant, and in
Etruscan and PIE it is a genitive.

The ending *-ma appears as a focus marker in Basque and Japanese, a dative
in Hausa, an ergative in Kartvelian (*-ma-n), a genitive/ergative in EA,
and an accusative in Etruscan (-n), PIE and Uralic.

A locative ending *-[?]a (*-a-[?]a added to the absolutive) can be
identified in at least Hausa, Sumerian and PIE (also Basque allative -a).

Whether *-na was another locative ending is unclear: we have Uralic -na and
Basque -n, and perhaps also PIE *-r in fossilized pronouns such as "where".

A suffix *-ka is perhaps present in the Sumerian genitive, the Basque
ergative and the EA instrumental.

The suffixes *-ta, *-da (often difficult to keep apart) generally make
ablatives and instrumentals in AA, Kartvelian, Sumerian, PIE, Uralic and
EA.

In summary:

base form *-0, *-a
ergative *-u
genitive? *-i
genitive? *-na (*-nu, *-ni)
allative? *-si
focus marker? *-ma

The other forms discussed were probably added to the absolutive, and at the
PN stage, it seems better to speak of postpositions rather than case
endings:

genitive? *-a si
locative *-a a
instrumental? *-a ka
locative? *-a na
ablative? *-a ta
comitative? *-a da

Now while this is all very interesting, it (and more) has all been said
before, (cf. Illich-Svitych's list of: abs. *-0, oblique *-n, accusative
*-mA, lative *-K.V, locative/allative/ablative *-na, *-da, *-LA), and truth
be told, it doesn't carry too much convincing weight. These are all
single-consonant (*-C or *-CV) affixes, using universally unmarked
consonants: the chances of finding, say, something resembling a locative
containing the letter *-n, is rather high, especially if only a few or a
single example from a large language family is acceptable.

If Proto-Nostratic was once a reality, then what we should be looking for
are highly marked phenomena which just so happened to be part of the
proto-language, and which otherwise are unlikely to have happened
independently by chance.

With regards to the nominal declension in the singular, only a handful of
such telling examples can be found, and both of them, if accepted, would
only prove that a link exists between PIE and Afroasiatic:

1) the ergative (> nominative) case *-u can be reconstructed for AA
(Semitic and Berber, perhaps also some relicts in Egyptian). Some PIE
phenomena can be explained if a stage prior to PIE proper had an
ergative/active suffix *-u vs. an absolutive/passive suffix *-a. In the
first place the pronominal stem *so (masc.) ~ *seh2 (fem.) vs. *to-d
(neuter), *to- (oblique). We know that *-tu gave *-s(W) in the Auslaut
(e.g. 2sg. *-s, 2pl. *-te), so it would not be surprising if the same thing
happened in the Anlaut: ergative **tu-a: > *s(W)o, absolutive *ta-a: >
*to-. Another case where we see active *-u vs. passive/oblique *-a is in
the personal pronouns: 2sg. *tu-, *twe- vs. acc. enclitic *te, 3refl. *swe-
vs. enclitic *se.

2) the "genitive" suffix *-n, in Indo-European an adjectival formant, would
have had the three case endings *-nu (ergative), *-na (absolutive), *-ni
(genitive). In Indo-European we would expect those to have given *-m(W),
*-n- / *-r and *-i, respectively (the same thing happens with suffixes in
**-un-, **-an- and **-in-, which become u-stems [e.g. **gá:nun, *ga:núnâs >
*g^ónu(r), *g^énwos "knee"], r/n-stems [if neuter: **wá:dan, *wa:dánâs >
*wódr, *wédnos "water"] and i-stems [**pá:tinz, **pá:tinm, **pa:tínâs > *
*pótyo:n, *pótim, *pétyos "lord"], respectively). Now an alternation *m ~
*w ~ *n ~ *r ~ *y immediately brings to mind the "Caland system" of PIE
adjectival forms (e.g. Avestan xrú:ra-, xrú:ma-, in compunds xrvi-).
If we can turn up evidence for erg. *-u, abs. *-a and gen. *-i, as well as
relational *-na in pre-PIE, while at the same time solving the Caland
riddle, that should be a good thing.


THE PN PLURAL

We have seen various ways in which plurals can be formed in the various
languages discussed.

1. Reduplication (e.g. Sumerian, Hausa). This can be full reduplication,
as in Sumerian (ensi-ensi) or partail reduplication, as in the Hausa
examples. I won't go into other phenomena, like verbal reduplication
(e.g. Sumerian, Basque, PIE) at this point.

2. "Broken plural". This procedure follows on logically from the kind of
reduplicative plural we see in Hausa. If two-consonant words (awà: ->
awo:wi:, dabbà: -> dabbo:bi:, bak'i: -> bak'à:k'e:) make their plural by
partial reduplication of the final (second) consonant, then applying this
process to three-consonant words yields Hausa k'arfè: -> k'arà:fa: "metal",
and the Semitic "broken plural", which in its procedure is practically
identical with the PIE collective (*wódr ~ *udó:r "water", like Semitic
nahr, nhar (> ?anhar) "river").

3. The "long vowel" plural of Hausa and Semitic again looks like an
application of the three-consonant "broken plural" to two-consonant words
(*CáCaC-u -> CaCá:C-u, therefore *CáC-u -> CaCú:). In e.g. Akkadian, it
almost completely replaced the "broken plural" and three-consonant words
regularly make their plural on the model (nom.) CáCaC-u -> CáCaC-u:.

The "broken plural", as attested in Afro-Asiatic and the PIE collective,
can thus reasonably lay claim to PN status.

However, it was not the only plural formation, and may indeed, as in PIE,
have originally had an exclusively collective function (Arabic broken
plurals are grammatically singular).

4. An apparently very old pattern of plural formation is the Ablaut sg. *i
(*a) vs. pl. *u. This is found in the Hausa and Basque personal pronouns
(Hausa 1. ni ~ mu, 2. ka/ki ~ ku, 3. si ~ su, Basque 1. ni ~ gu, 2. hi ~
zu). More about this below in the section on personal pronouns.

Next, we come to suffixed plurals.

THE PLURAL SUFFIX *-(a)n, *-atu/*-ati

The "construct state" variant had *-an (by the Nasal Rule), the erg/nom
*-atu, oblique *-ati. This suffix occurs in practically every Nostratic
language.

That the form *-n arose in the absolute Auslaut is made plausible by its
occurence in places where it's unlikely that any additional material was
ever present, e.g. in the Hausa personal pronouns mun, kun, sun (in Semitic
with case markers added: s^un-u, s^in-a, like sg. s^u(w)a, s^i(y)a), or the
Semitic "definite" nunation in -n.

Of course, suffixes could secondarily be added to *-n, as we see in the
Hausa nominal pl. -una: (reflecting what is perhaps the oldest vowel of the
suffix, with "Ablauting" plural *-u), Semitic plural -a:nu, etc.

Other forms with -n: Basque 1pl. gen-, 2pl. zen-; Berber 3pl. -s&n, nominal
pl. -&n, -in; Egyptian 2pl. -kina, 3pl. -sina; Semitic 2pl. -tunu, -kunu,
-tina, -kina 3pl. -s^unu, -s^ina, ?1pl. *nina > nia-, nominal pl. -a:nu,
obl. -a:ni; "definite" nunation/mimation; Kartvelian 1pl *c^wen, 2pl
*tkwen, nominal pl. -n (but oblique -t), verbal 3pl. -en; Sumerian pl.
-(e)ne, PIE verbal -mén(i), -tér (-téni), -ér (-én-ti); Uralic (Samoyed)
3pl. -n.

As to plural *-atu/*-ati, the most fascinating thing is that the difference
between oblique and non-oblique was maintained not only in Semitic
-a:tu ~ -a:ti (also -u:tu ~ -u:ti with u-vocalism), but also in Georgian
oblique -t (vs. non-oblique -n), PIE pronominal -es (*-atu) ~ -ey/-oy-
(*-ati), and Uralic/EA nom/erg. -t (< *-d < *-atu), oblique -j (< *-d^ <
*-ati), as shown at length earlier.


THE PLURAL SUFFIX *-(a)m, *-abu/*-abi

With u-vocalism Hausa *-úba: > -úwa:. Not as widely attested as the
previous one, this plural suffix was mainly used for nominal forms (as
opposed to *-atu/*-ati/*-an for pronominal forms) in Hausa, Ancient
Egyptian (aw-plural < *-abhu ?), Kartvelian -eb(i) and PIE (pl. oblique
*-bhi, G.pl. *-om). In Uralic and EA, it was completely (and in PIE
largely) replaced by the pronominal *-atu/*-ati-plural.
There seems to be no occurrence in absolute Auslaut (*-m), but I haven't
been looking.


THE DUAL SUFFIX *-ing, *-iku/*-iki

With u-vocalism the Hausa plural marker -úka:. AFAIK, not found in AA as a
dual marker (Semitic uses -a:/-ay instead, but note the Egyptian dual
ending (pp's, verbal endings) -ij), but present in at least PIE, Uralic and
Eskimo-Aleut. The form affected by the Nasal Rule is *-ing, which is
present in the Uralic dual endings *-mVn', *-tVn' (*-ming, *-ting), and
perhaps in part also in IE pl. *-mén/*-wén etc.

The forms with *-k show the old alternation between erg. *-iku and
abs./oblique *-iki in PIE (o-stem *-oh3 < *-iku vs. C-stem *-ih1 < *-iki,
p.p. *wéh1 vs. *juh3, like *wey-es vs. *jus). In Uralic and Eskimo-Aleut,
*-g^ and *-g both are reflected as -G-/-k-. The Uralic nominal dual marker
-kan'/-G&n is from *-g, extended with the verbal dual marker *-ing (but
Vogul has simple -G).


THE SUFFIX *-ka/*-qa (?)

The Uralic plural markers are: nominal *-t, obl. *-j. In the verb, there's
a pattern of 1/2pl. *-t vs, 3pl. *-n in Samoyed, which is replicated
throughout Uralic, only with different combinations (-n -n -t; -t, -t, -n;
and most often -k, -k, -t) The origin of the markers *-n, *-t and *-j was
discussed above. Remains the marker *-k, which seems to belong originally
in the 1/2 stative plural (in Hungarian it has replaced *-t completely even
as a nominal plural). It is not a dual marker. We can perhaps compare it
with the PIE collective marker *-h2 < **-k (or **-q). Altermatively, we
can think of a post-Nostratic analogical stative ending (collective *-h2
can seemingly in Sanskrit also be added to a verbal stem to make the
so-called "aorist passive"). The original plural of the stative
was as will be given below, but the *-k of the personal endings 1. *-k, 2.
*-tk may have been transferred analogically to the plural (*-d-me-k,
*-d-te-k), perhaps to avoid confusion with the dual (*-men', *-ten' >
*-men, *-ten).


THE PN PERSONAL PRONOUN

The oldest forms of the personal pronouns appear to be:

pl.
1. *mi *mu(n)
2. *ki *ku(n)
3. *si *su(n)

In a part of Afro-Asiatic, this system was retained with little change.
*mi was assimilated to *ni everywhere in AA (and Basque), and a gender
distinction was created in the 2sg. (m. *ka, f. *ki or reduplicated
*ki-k > *king by the Nasal Rule, and further *kim) and the 3sg. (Hausa
3m. si:, 3f. i-ta).

Egyptian and Semitic innovated by adapting the personal pronouns in part to
the nominal case endings erg. (masc.) *-u, abs. (fem.) *-a.

Egyptian has:

2m. -k-u 3m. -s-u
2f. -k-i(-m) 3f. -s-i
2p. -k-i-n-a 3p. -s-i-n-a

Semitic has, in the suffixed forms:

2m. -k-a 3m. -s^-u
2f. -k-i 3f. -s^-a
2pm. -k-u-n-u 3pm. -s^-u-n-u
2pf. -k-i-n-a 3pf. -s^-i-n-a

And in the independent (oblique) forms:

1. *ni- 0-a > *n^a > ya-
2m. *ki-u-0-a > *k^wa > ka-
2f. *ki- 0-a > *kya > ka-
3m. *si-u-0-a > *s^ua > s^uwa-
3f. *si- 0-a > *sia > s^iya-

1p. *ni- n-a > *n^in^a > niya-
2pm. *ki-u-n-u > *k^unu > kunu-
2pf. *ki -n-a > kina-
3pm. *si-u-n-u > s^unu-
3pf. *si -n-a > sina > s^ina-

In the case of the 3rd. person pronoun, Semitic *s^ seems to indicate
that the case endings were not simply substituted, but added on to the
base *si (*si-u > *s^u, *si-a > *s^a), with *s^ spreading analogically
to e.g. the plural. If this was also the case in the 2nd. person
(as shown in the table above) is uncertain.

Outside of Afro-Asiatic, the same innovation was apparently (and
independently) carried out by (pre-)PIE, where the pronouns have
the shape:

**mú **má pl. **muátu **muáti > -- *mé *me *mes *ns- *wey-
**tú **tá pl. **tuátu **tuáti > *tu *twé *te *jus *us- *swey-
**sú **sá pl. **suátu **suáti > -- *swé *se -- -- --

Etruscan and Uralic have the old *mi, *ti, although the 3rd. person
suffixed form is *sa in Uralic.

As to the exact shape of the 2nd. sg. pronoun in PN, we can choose
between *ki and *ti. The apparent antiquity of the 2sg. feminine
pronoun *kim (*ki-k) and its verbal suffix *-m (Basque, Berber) < *-ng <
*-k decide the issue in favour of *k (and *ki > *ti is an expected
assimilation, comparable to *mi > *ni).


THE PN VERB

The only pseudo-verbal system reconstructable for PN is the stative.

The active conjugations show a basic split between Afro-Asiatic and the
rest. Afro-Asiatic has a prefix conjugation *?a-/*na-, *ta-, *ya-, except
for Egyptian (which has a suffixed pronominal conjugation in the style of,
say, PIE and Uralic), assuming the Hausa pronominal marker ya- is a
leftover from the prefix conjugation in Chadic.

The suffixed pronominal conjugations of PIE, Uralic and EA all show the
same person markers: 1. *m(V), 2. *t(V) and less regularly 3. *s(V).
The duals and plurals are made with the pronominal dual and plural markers
*-n/*-t (> *-d- in Uralic and EA) and *-ng'/*-k (> *-g- in Uralic and EA).
The difference between Uralic/EA 1sg. stative *-k and du. *-g-, and 2sg.
*-t and pl. *-d- is due to the fact that the dual/plural suffixes were
intially intervocalic (*-atu, *-iku), while the 2sg. was a separate
pronoun with *t- in the Anlaut, before it was agglutinated. Suffixal *-ka
was perhaps likewise not (always) intervocalic at the time of the lenition.


THE PN STATIVE:

The stative is attested in Afro-Asiatic, Kartvelian, PIE, Uralic and
Eskimo-Aleut.

The endings are:

PAA Kartv. (PP) PIE (middle) Uralic EA
1 *-ku *xw- *-h2(a) *-k *-k
2 *-t(k)a: *x- *-th2(a) *-r *-n *-tket
*-t(k)i:
3 *-a *-(e) *-0 *-0
*-at(a)

1 *-nij [?] *-h2- *-(g)meñ *-tkmeg
2 *-tkunij [?] *-h2- *-(g)teñ *-(g)teg
3 *-unij [?] *-h2- *-g(eñ) *-g

1 *-nu: *xw- *-m(w)é *-dmek *-tkmed
2 *-t(k)unu *x- tkven *-té *-dhw- *-dtek *-ci < *-d^td^
*-t(k)ina
3 *-un ~ *-u: *-er-s *-t *-t (*-j)
*-an ~ *-a:


Being stative forms, we expect these forms to be built along the
following model: (1) number/case of predicate, (2) person marker,
(3) number of person marker. With the components we have at hand
for Proto-Nostratic that would result in a theoretical reconstruction:

1sg *-a-k(a) *-(a)ng / *-(a)k(a)
2sg *-a-tk(a) *-(a)n / *-(a)tk(a)
2sg *-a-0-0 *-(a)

1du *-ik-mi-ik(i) *-(i)kming, *-(i)kmik(i)
[*-(i)kning, *-(i)knik(i)]
2du *-ik-ki-ik(i) *-(i)kking, *-(i)kkik(i)
[*-(i)kting, *-(i)ktik(i)]
3du *-ik-0-0 *-(i)ng, *-(i)k(i)

1pl *-at-mu(-at(i)) *-(a)tmu(n), *-(a)tmwat(i)
[*-(a)tnu(n), *-(a)tnwat(i)]
2pl *-at-ku(-at(i)) *-(a)tku(n), *-(a)tkwat(i)
[*-(a)ttu(n), *-(a)ttwat(i)]
3pl *-at(i)-0-0 *-(a)n, *-(a)t(i)

This model goes a long way to explain the actual attested forms.


To summarize:

The Proto-Nostratic noun had a base form (absolutive) in *-0 / *-a.
The ergative was *-u. The genitive perhaps *-i. In the plural, only an
ergative ending *-u and an oblique ending *-i were distinguished.

Collectives could be made by (partial) reduplication and by the
"broken plural" method (CáCaC- => CaCá:C-, CáC-u => CaC-ú:).

The (personal) pronouns retained an "Ablauting" pattern sg. *-i (or *-a),
plural *-u. The pronominal plural was otherwise erg. *-atu ~ abs. *-ati ~
*-an (older *-utV ~ *-un ?).

The nominal plural was erg. *-abu ~ abs. *-abi ~ *-am (older *-ubV ~
*-um?).

The dual was erg. *-iku ~ abs. *-iki ~ *-ing (older *-ukV ~ *-ung?).

The stative verb had special 1/2 sg. endings *-kV (*-ng) and *-tkV (*-n),
and was otherwise regular (see paradigm above).

No other verbal forms can be reconstructed.


=======================
Miguel Carrasquer Vidal
mcv@...