H1 and t??

[Jens Elmegaard Rasmussen]

On Tue, 13 Apr 1999, Glen Gordon wrote:

> ME (GLEN) IN RESPONSE TO JENS' SOUND RULE FOR *yu:s "you":
>   Still don't get it. How does *t- become *y-??
[...]

Dear friends,

I had not anticipated that the list would be discussing what I might
possibly mean by a passing comment. Maybe I can save you some trouble by
specifying what I had in mind.

Twenty years ago I was giving a course in IE morphology and, just before
we were to deal with the personal pronouns, there was a one-week vacation.
I prepared a series of handouts containing all attested forms from the
diverse branches together with their probable history. For each branch I
tried to specify a intermediate proto-stage (like, Indo-Iranian or Proto-
Germanic), and for the whole thing, of course, an IE stage. It then became
obvious that the result lent itself to further analysis, and it was now my
idea to get as far into that as I could.

I was working on the assumption that there ought to be some connection
between the pronominal forms and the personal markers of the verb. I also
considered it probably, not to say evident, that the personal markers
recur in neighbouring linguistic families like Uralic and Eskimo-Aleut (I
knew too little of the other Nostratic branches).

The IE system has been basically correctly reconstructed by Cowgill in
Evid.f.Laryng. as

*eg' *tu    *we:   *yu:   *wey  *yu:s (nom.)
*me  *t(w)e *nH3we *uH3we *nsme *usme (acc.)

I have some difficulty only with the 2du acc. for which the Skt. stem
yuva- (with y- from the nom.) rather points to *uH3e, probably with
dissimilatory loss of the *-w-.

The other cases are formed from the acc. by the addition of postpostions,
cf. Vedic dat. asma-bhyam 'to us', abl. ma-t, tva-t, asma-t, yuSma-t etc.

The disyllabic acc.s have enclitic variats consisting in the first
(underlying) syllable, *noH3, *woH3, *nos, *wos, for med a time when there
was a vowel /o/ in that syllable.

Possessive adjectives are formed from the acc. by vrddhi: *tew-o-s,
*no:H3-o-s, *wo:H3-o-s, *no:s-o-s, *wo:s-o-s. The 1sg *me had to prefix
the vrddhi vowel because there was only one consonant, *emo-s 'my'. The
reason is that there is no variant *mwe parallel with *t(w)e and *s(w)e,
but that is obviously due to simple sound change, *mwe > *me completed
before the poss.adj. was derived.

>From the poss.adj.s were formed substantivized uninflected forms used as
genitives of the personal pronoun: *tewe 'of thee', *no:se 'of us' etc. In
the 1sg, *eme was adjusted to *meme by adoption of the same beginning as
the other cases, then dissimilated to PIE *mene (Av. mana, OCS mene).

Now it appears that there are so many unexpected u- or w-elements that
something just has to be included into the underlying forms to account for
this. Note that, while *tu:, *yu:s are 2nd person, there is /w/ in the
1st person du./pl. *we:, *wey. That must be accounted for. I can suggest
only the presence of a STEM consonant /w/ originally present in ALL forms
and thus meaning something like "person" or "speech-act participant",
because that's what all the forms have in common.

It is plain that *nH3we : *nsme contain the number markers, /H3/ for dual,
/s/ for plural, while the -me is the casemarker of the acc.; the -we
occurring after the dual markers appears to represent a development of -me
in that position (cf. the VERBAL 1du in -we as opposed to 1pl -me which
were probably also originally accompanied by du. vs. pl. markers).

As number markers we find *H3 (something like a voiced labiovelar
spirant), not at all out of style with what is found in Eskimo-Aleut (a
velar spirant) and Uralic (a velar spirant in Vogul). For the pl. we have
in IE *-y and *-s, perhaps in complementary distribution, and not quite
unlike Esk.-Al. /D/ (dental spirant), cf. also Finn. -j-/-t. A therefore
calculate with /G/ (velar spirant) for the dual and /D/ dental spirant)
for the plural.

The vocalization appears to be simple: The nominatives have a vowel before
the last consonant, while the acc. adds *-m to the full nom. form, this
triggering the addition of *-e after the resulting final cluster. There is
final accent throughout.

We cannot explain *eg' 'I' as part of the original system, but all the
rest can be relatively artlessly derived from a completely regular
underlying inflection:
     1. person         2. person
    nom.  acc.       nom.   acc.
Sg. (m-w) m-w-m      t-w   t-w-m
Du. m-w-G m-w-G-m    t-w-G t-w-G-m
Pl. m-w-D m-w-D-m    t-w-D t-w-D-t

With vocalization - and suggested stages of the phonetic development:

mew-me tew tew-me mweG mweG-me tweG tweG-me mweD mweD-me tweD tweD-me
       tuw             nweGme  twuG         weD  nweDme  twuD
mewe       tewe   weG  neGme   DwuG DweGme       neDme   DwuD DweDme
                               DuG                       DuD
       tu              neGwe        DweGwe
                                    weGwe                     weDme
                               juG          wej  nezme   juz  wezme
mowe       towe   we'  noG-we  ju'  woG-we       noz-me       woz-me
mwe    tu  twe    we   nH3we   yu   wH3we   wey  nsme    yus  wsme

> PIE

me     tu  twe    we   nH3we   yu   uH3(w)e wey  nsme    yus  usme

with enclit.
me         te           noH3         woH3         nos          wos

The changes are in part pure invention, but not unnatural and not without
intrinsic coherence:

1. The vocalization is normally with /e/, only assimilated to /u/ in the
nom.s of the 2nd person *tuw *twuG *twuD. No such assimilation took place
in the forms beginning with *m- (1st person), which must be a dissmilatory
brake on the rounding influence ("ma non troppo").

2. *mw-Gm- and *mw-Dm- changed the initial to *n- (dissimilation).

3. wm > w, mw > w, nw > n, tw > Dw

4.5.6. wu > u, Gm > Gw, Dw > w

7. D > j- initially, -z- medially (before consonant?), but word-finally -z
after high vowel (-uz in juz, cf. instr.pl. in -bhis), -j after non-high
vowel (-ej in wej, cf. toy 'they'); -z (> -s) finally after consonant does
not apply here.

8. On its way to zero, unaccented /e/ becomes /o/ as already known. Final
G is lost (presumably through change into a feature on the preceding
vowel, perhaps glottlization). After that time, the first-syllable
stretches  of the du. and pl. accusatives are singled out as enclitics,
this giving *noH3 and *woH3 with retained spirants, pl. *nos, *wos.

9. Unstressed short vowels are completely lost, and the cluster mw thereby
arising is reduced to m.

10 (not in the chart). Monosyllabic forms develop long-vowel (emphatic?)
variants: *tu:, *t(w)e(:), etc.

Viewed in this fashion, 'us' IS the acc.pl. of 'me', 'ye' the nom.pl of
'thou, thee', etc. The system may be fully regular, based on normal
inflection. The use of dual and plural number is elliptic: the dual of
'me' means 'me and someone else', the plural 'me and some others', i.e.
what is normally called 'us'; note that this is normal IE syntax.

The sound-change postulates for the oldest periods cannot be verified
because we can analyze nothing else that far back. But the second half or
so of the chart consists of changes that are already known; therefore, teh
suggestions at least integrate the forms of the personal pronouns into a
general picture of IE language history.

Until a few day ago, this analysis had only been published in the
Copenhagen institute papers APILKU, vol. 6 from 1987, but now my old and
hidden papers have been collected in two volumes of "Selected Papers in
Indo-European Linguistics", published by Museum Tusculanum in Copenhagen
(700 pp., $70, www.mtp.dk), and this one is among them. It may not be
true, but no one shuld call it impossible.

Jens