nasal pres / root aorist II

[Jens Elmegaard Rasmussen]

On Tue, 3 Aug 1999, petegray wrote:

[...]
> ****A: in PIE
[...]
> Now not all verbs show an aorist, but the claim is that verbs with a nasal
> infix present go with a root aorist.   (Does this mean all nasal infix
> verbs, or only those that have an aorist?)

J:
Not everything present in PIE need have been retained down to an attested
language millennia later, so the question makes little sense.

P:
> I checked a random sample of 16 verbs with nasal presents.   Nine of them
> had no aorist; the other 7 all had root aorists somewhere in PIE.    A
> miniscule sample, but suggestive.   I'm sure you will be pleased by it,
> Jens.   In any case, it is clear that there is a need for Strunk's claim to
> be based on a proper statistical analysis, not just on a theoretical
> assertion supported by his ridiculously small group of 8 selected examples.

J: Statistics may be totally misleading, if one type runs crazy and become
productive. It seems the s-aorist has done that (luckily then often
forgetting some of its old characteristics so that we can still spot it as
secondary).

P:
> ****B: in an individual languages
> Strunk and Jens do not make the claim that nasal presents are associated
> with root aorists within a particular langauge (which would be a much more
> significant claim), but it is worth testing nonetheless.

J:
The "same-language" ideal is indeed met here and there, see the other
posting, simultaneous with this one.

P:
> Strunk says (p128) "[This system] appears precisely in Indo-Iranian and
> Greek"   He says that this proves the archaic character of these two
> languages.

J:
I believe you have a word of criticism hanging in the air. To me (too?),
the inference is no quite logical. Others would say that it proved the
archaic character of the system, since it is retained in precisely Greek
and Indo-Ir. which are by widespread consent the most archaic branches of
IE. But even that is of course open to question.

> I [petegray] said:

>>> (a) In Greek it is largely true, but there are exceptions;   so a bald
>>> statement would need qualification.
>>> (b) In Latin it is largely untrue, since the aorists are either sigmatic or
>>> the rare reduplicative  aorist (tango tetigi, claimed by some as an aorist
>>> on the basis of Homer tetago:n, or the lengthened vowel: pango pe:gi (~
>>> perfect pepigi).   I can only find cumbo cubui which supports the claim in
>>> Latin.

> Jens said:

>> You might also have thought of cerno:/cre:vi:, fundo:/fu:di:,
>> linquo:/li:qui:, rumpo:/ru:pi:, sino:/si:vi:, sperno:/spre:vi:,
>> sterno:/stra:vi:, vinco:/vi:ci:.

> li:qui and  vi:ci probably have the same origin as vi:di (which you
> significantly did not include, because its origin is known), that is to say
> in an -o- grade perfect.   (*-oi- > *-ei- after v, or between /l/ and labial
> or labiovelar, then *-ei- > i: as normal)

Jens now says:
vi:di cannot be the perfect which with this verb means 'know': vi:di:
means 'I saw' which is the meaning of the aorist (I have, however, seen
vi:di: _quoted_ to mean 'I know', but I know of no textual basis for it;
in case I have overlooked something, it is in reference to the meaning 'I
saw, I have seen' I insist that it reflects an aorist). I grant you that
li:qui: and vi:ci: could be perfects (the "li:ber" rule might work in
li:qui:, good point; and vi:c- could even be *wi-wik-, aor. or pf.).
However, in that case it is a _very_ odd principle that the o-grade
perfect is never retained de- (or un-) reduplicated when the vowel would
be retained as /o/ into Latin times.

P:
> fu:di and ru:pi could equally be from o grade perfects (we know that
> iu:vi is).

J:
Nothing excludes that iu:vi: has IE *-ew- (Italic cannot distinguish eu
and ou, cf. also novus : Gk. ne'os), but the example is useless in this
context, sure.

P:
> In any case, If they are aorists, they would have to be full grade,
> and I excluded them because I thought that "root" meant "root", not full
> grade - though I see the Strunk uses "root" to mean full grade, too.

J:
A root may alternate between any of its ablaut grades. Root presents and
root aorist often have full grade in the daughter languages, originally
the alternated, as *{gw}em-t, 3pl *{gw}m-ent 'came' (still Vedic a'gan,
a'gman)

P:
>  cre:vi, spre:vi, si:vi, stra:vi are clearly -vi formation (Latin) perfects,
> on a laryngeal base.  The origin of these -v-forms is still unknown.
> In what sense do you make them either a root form, or an aorist?

J:
The Latin v-"perfect" is obviously an innovation, but that does not make
the verbs forming it innovations themselves - they must have had
_some_form in the protolanguage. And the locus of diffusion of the Latin
-v- is not ard to find: in /fu[v]i:/ there was an automatic glide which
could be regarded as phonemic or subphonemic, sincefui:/ and /fuvi:/ would
be pronounced the same. Thus, if fu- formed fu-vi:, what would other verbs
in vowels like ama:- form if not ama:-vi: and the like? A comparable extra
-v- is seen in Sanskrit in 3pl aor. abhu:van, pf. 1/3sg babhu:va, 3pl
babhu:vur., only in Sanskrit the additional -v- was not utilized as a
pattern for other verbs, while in Latin it was. I guess the Old English
verba pura in -wan have a comparable origin: since bu:wan 'dwell' is
simply bu:- + -an (possibly from *bu:-j-an with loss of intervocalic *-j-
by phonetic rule, itself a back-formation based on the aorist *bhuH-t,
Ved. a'-bhu:-t), the root form *kne:- (whence North/West-Gmc. *kna:-)
'know' formed cna:wan; and *se:- formed sa:wan 'to sow' etc.
   Therefore there ought to be no reservations against the right of an
expected root aorist with 3sg *kreH1-t to turn up as cre:-v-i:. The form
*kreH1-t itself is from *kreH1y-t which corresponds to a nasal present of
the shape *kri-ne'-H1-ti/*kri-n-H1-e'nti by the rules I have been able to
work out; therefore, cerno:, cre:vi: looks very good. The same goes for
se:vi: 'sowed' and, with bigger or smaller footnotes, for the others as
well.

P:
> I note that
> elsewhere in your posting you claim that the Greek sigmatic aorists can be
> counted as root forms.   This means you count zero grade, full grade, Greek
> sigmatic, and Latin -vi formations all as root aorists for the purposes of
> your claim.

J:
That's right, I've got no inhibitions at all. A Gk. aorist like
epe'ras(s)a from pe'rne:mi (Dor. -a:-) 'sell' cannot be an old s-aorist,
for they have lengthened grade. Instead it must represent a reformation of
*e-pera from *e-per at 2-t with commonplace addition of productive morphemes,
this giving 3sg e-pe'ra-s(s)-e just as lu'o: forms e'-lu:-s-e. By the same
token, a Vedic aor. like a-s'ami:t 'laboured, fatigued' cannot be an
s-aor., but only a set.-root's root aorist from *e-k^em at 2-t, and this is
not overthrown by the existence of a 1sg a'-s'ami-s.-am which is simply
due to backformation from a wrong analysis of the for in -i:-t as being
sigmatic. Thus, the old aor. to go with s'amna:ti was 1sg *as'amam, 2sg
a's'ami:s, 3sg a's'ami:t, just as gr.bhna:'ti formed aor 1sg a'grabham
(from *e-grebH2-m.), a'grabhi:s, -i:t. The true s-aorists from set.-root
do have lengthened grade, cf. a'-ka:ri-s.-am. -ka:ri:-t 'commemorated'.
This stratification has been worked out very clearly by Johanna Narten in
a solid monograph of lasting value and seems now disputed by nobody
informed. - Note that the Gk. prs. pe'rne:mi already reveals the
paradigmatic companionship of the nasal present with something having
non-lengthened full grade and so rather obviously points to an IE set of
n-prs. and root-aor.

P:
> (a) The wider the definition of "root aorist" the less meaningful your
> claim becomes.

J:
Not if the re-definitions apply to reformations that have kept their
diagnostic value as these have.

P:
> (b) You would need to show that there is indeed a proper root aorist
> underlying these forms.   But (i) this cannot be done in Latin - the best we
> can do is show a zero grade form (as in many Latin perfects);

J:
There is no way cre:vi:  or se:vi: could be based on IE perfects.

P:
> (ii) how would
> you distinguish a sigmatic aorist in Greek from a root aorist?

J:
If the forms has no /s/ (or its reflex), there is no problem. Thus e'kamon
can only be a root aorist vis-a-vis the n-prs. ka'mno: 'fatigue, wear out'
(the same paradigm as in Sanskrit). If there is -s-, significant (i.e. not
Osthoff-triggered) short vocalism points to root aor. A well-establised
s-aorists like *de:'ik^-s-m. (Lat. di:xi:, Avest. da:is^-) is thus opaque
in Gk. e'-deik-s-a because of Osthoff's shortening. I know of only
Barton's brilliant analysis of ege:'ra: 'I grew old' as a provable s-aor.
in Gk.: Proto-Greek *e-ge:ra-h-a from IE *g^e:'r at 2-s-m. - an analysis I
would have wished I had made myself.

P:
> For Sanskrit, I gave examples which contradicted the claim, and Jens gave
> some which supported it.  Neither of us had (or has ?) complete accurate
> figures.   In my sample I took random instances, so I restricted myself to
> just one Skt class.   Jens said:
> [J:]
>>why ignore those of class nine (too good?) - ?
> [P:]
> That was a petty jibe, Jens, and unworthy of you.  In fact the figures for
> class nine are not good.  Of those Whitney lists as "older", 7 show no
> aorist, 10 do not show a root aorist, and 13 do.   You get 13 / 30, about
> 43% - not bad, but not totally convincing, either.

P:
I looked through Macdonell's lists for class nine, finding:

11 work fine:
as'na:'ti	as'i:t
gr.bhn.a:'ti	a'grabhi:t
ja:na:'ti	jn~eya:'s
pr.n.a:'ti	a'pra:t
vr.n.i:te'	sbj. va'rat
s'r.n.a:'t	as'ari:t
s'rathni:te'	as'ranthi:t
sina:'ti	a'sa:t
skabhna:'ti	askambhi:t
stabhna:mi	a'stambhi:t
str.n.a:'mi	a'stari:s

3 have thematic prs. based on the sbj. of the root aor. thus recovered:
jina:'ti	ja'yate
juna:ti		ja'vate
puna:ti		pavate

2 have this in the other languages:
badhna:ti	Gmc. *binda-
ubhna:ti	Gmc. *weba-

2 won't behave:
mina:ti forms mes.t.a, while hrun.a:ti has red. and s-aor. forms

I have left nothing out on purpose, but I have not made _very_ thorough
inquiries into the accuracy of Macdonell's data. This ought to mean tat
corrections may be expected to go both ways and outweight each other. The
picture this gives is not bad: Out of 18, 11 still behave as they should,
3 have clear traces of having done so earlier, 2 do so elsewhere, only 2
don't fit. Again, it looks like the opposite of random distribution.

P:
> So for PIE we need someone to crunch numbers and get a proper statistical
> analysis, and for individual languages only Greek and Sanskrit have a hope
> of showing a genuine correlation, but for both it is, if anything, nothing
> more than a general tendency.   It certainly can't support the original
> claim that since Sanskrit shows the form tundate, it must once have had a
> root aorist to go with it.

J:
But it does do that, for tund- recurs in Lat. tundo:, and in neither
language is the infixing of nasals productive, ergo this stems from PIE,
and in the very same root at that.

Jens