[tied] Morphology (7/20)

From: Miguel Carrasquer
Message: 14489
Date: 2002-08-24

7. The stative and the perfect

The original meaning of the stative was to denote a state. As such, the stative
was added to verbal and adjectival roots which of themselves denote a state.
Hittite still maintains this lexical distinction in that some verbs are
conjugated according the active conjugation, while others take the
"hi-conjugation", even though the semantic reasons behind the choice for one or
the other conjugation are not always clear any more. With its separate
categories of active and stative verbs, Hittite represents the older situation,
whereas in the rest of IE, the stative developed into a resultative aspect (the
perfect), open to all verbs, and then to a preterite/narrative tense (cf. the
same thing in e.g. Semitic where we have a historical development stative ->
perfective -> preterite). Traces of the old stative in its first phase,
however, may have been preserved here and there under the guise of o-grade
presents (transferred to the active conjugation), such as Arm. go- < *wos- "to
be there" or Slav. bojati "to fear". Other verbs (e.g. *woid- "to know", or the
preterito-presents of Germanic) retain the stative (perfect) endings almost
everywhere, but have present meaning. Where, the perfect developed into a
preterite/narrative tense, it was often amalgamated with other (active)
preterite formations such as the sigmatic and root aorists (e.g. in Celtic,
Latin or Germanic).

The PIE stative (perfect) endings are fundamentally different from the active
ones. At the deepest level of reconstruction, I would posit:

**-ka **-mu
**-tka **-tku
**-a **-u

This regularly became:

**-h2 **-mW
**-th2 **-dhWw > *-tW
**-0 **-0

[Note that *-tw < *-dhw does *not* become *-sw.]

Secondary addition of the thematic vowel *-e (originally unaccented, later
secondarily accented in the 1 & 2 pl., not always the 3 pl.) gives the attested
PIE perfect (missing 3pl. ending borrowed from the active preterite):

*-h2-a *-mW-é
*-th2-a *-t(W)-é
*-e(s) *-ér(s) > *-é:r, *-r.s

Note that in the stative, the thematic vowel *follows* the personal suffixes
rather than preceding them. That the perfect was in origin not a past tense is
illustrated by the fact that the 3rd person did not originally add *-s, except
in Anatolian (and there only in the newly created category of the stative past)
and in Tocharian (where the perfect had indeed become a preterite, and the
sigmatic aorist had not developed).

Besides the endings, the stative/perfect is characterized by (optional)
reduplication and a peculiar root Ablaut. The details are quite complex,
however. In the strong forms (the active singular), we find either no
reduplication or reduplication with *e. The root vowel is either *o or *e:. In
the weak forms, we find either lack of reduplication, reduplication with *e, or
reduplication with *e:. The root vowel can appear in *e-grade, *e:-grade or
zero-grade. It is hard to come up with a model that can explain all these
different variants. The closest I can come is the following:
cross-linguistically, reduplication of verbs is mostly used to denote either
intensive/iterative Aktionsart, or plurality of the subject and/or object.
Since the PIE stative/perfect is nothing like an intensive or iterative, it
stands to reason that the reduplication was initially used to denote plurality.
A common pattern in ergative languages (attested at least in Sumerian and
Basque) is for reduplication to agree with the absolutive. Supposing this were
also the case in a pre-stage of PIE, we would expect the distribution to be as
follows: intransitive verbs would have no reduplication in the singular, and the
root would be in lengthened grade (roots with **a/**u vocalism -> **a: > *o;
roots with **i vocalism -> **i: > *e:). In the intransitive plural, we would
have reduplication, with lengthened reduplication vowel *i -> *i: > *e:, and
zero grade of the root. Transitive verbs would have had no reduplication and
lengthened root vowel (*o or *e:) in case of a singular object, e:-reduplication
and zero-grade root in case of a plural object. Schematically for roots of the
shape **CaRC/**CuRC or **CiRC:

intransitive transitive transitive
sg. object pl. object
root vowel **a/**u **i **a/**u **i (any)
sg. subj. CóRC- Cé:RC- CóRC- Cé:RC- Cé:-CRC-
pl. subj. Cé:-CR.C- CóRC- Cé:RC- Cé:-CR.C-

When stressed endings were secondarily added to the 1/2 plural (but the 3 pl.
was liable to retain the stress on the reduplication vowel), this resulted in:

1pl. Ce-CR.C-mé CeRC-mé Ce-CR.C-mé
2pl. Ce-CR.C-té CerC-té Ce-CR.C-té
3pl. Ce-CR.C-ér CerC-ér Ce-CR.C-ér
OR: Cé:-CR.C-r. CóRC-r. Cé:RC-r. Cé:-CR.C-r.

The Hittite forms, with Ablaut o ~ e, can now be seen as the regular outcome of
the unreduplicated transitive singular paradigm. For <sakhi> "I know":

present past
sak-hi sek-weni sak-hun sekk-(u)wen
sak-ti sek-teni sak-t(a), -sta sek-ten
sakk-i sekk-anzi sakk-is, -sta sekk-er, -ir

The Hittite distinction between present and preterite of stative verbs,
analogous to what we have in the active, is an innovation, as the stative
originally denoted a tenseless state. It's clear that the new present tense was
formed by adding the presentive marker *-i, but the problem is to which form it
was added. The Hittite present forms -hi (-he), -ti (-te) and -i (-e) can be
derived equally well from *-h2-i, *-th2-i and *-0-i as from the extended
("classical" perfect) forms *-h2a-i, *th2a-i and *-e-i. Hittite orthography is
not very helpful here in barely ditinguishing between syllabic signs containing
/i/ or /e/. Perhaps both forms were possible initially, by the same scheme as
in the active, which also has both thematic and athematic forms. The Hittite
past tense forms are likewise open to various interpretations, although 1sg.
-hun is certainly from athematic *-h2 (+ -m. from the active). On the other
hand, Luwian has 1sg. -ha (there also transferred to the active preterite),
where -a does not seem to be merely ortographical (cf. Lycian alphabetic -Xa).

Outside of Anatolian, the perfect forms tended to regularize along the lines of
an analogical paradigm with (e-grade) reduplication in both singular and plural
forms, o-grade of the strong root, and zero grade of the weak root, i.e.
regularization along the lines of:

3sg. **CóRC-e ~ **Cé:-CRC-e -> *Ce-CóRC-e
1pl. **Ce-CR.C-mé ~ **CeRC-mé -> *Ce-CR.C-mé

However, forms without reduplication (Tocharian, Germanic, Latin, etc.), with
e:-grade in the reduplication (Indo-Iranian) or with e:-grade of the (weak
and/or strong) root (Germanic, Latin, Albanian) are not uncommon, as can be seen
in the following overview:

Tocharian B (preterite):
-wa: (< *-mw-h2a) -(ä)m(o) (< *-me)
-(ä)sta: (< *-s-th2a) -(ä)s(o) (< *-s-te(s))
-0/-(ä)s -är, -re (< *-or, *-ro)

The 3sg. zero ending appears to be from active preterite *-t rather than from
perfect *-e. Tocharian A too shows intrusion of active preterite endings in the
perfect (zero ending in th 1sg., from aorist *-m., besides normal 1sg. -wa:.
For the development *-mwh2á > *-wá, cf. the 1st person dual.

At first sight, the Tocharian forms bear a striking similarity to the Latin
perfect endings. First person sg. -wa: reminds one of the Latin perfects in
-u-/-v-, while the -s- in 2sg. -sta: and 2pl. -s(o) (< *-s-te(s)) can be
compared to Latin -isti: / -istis. I believe, however, these resemblances are
only superficial. In Tocharian, -w- is limited to the 1sg., and is better
compared to the PIE dual endings in -wá:/-wá/-wás < *-mW-eh2, *-mW-h2á,
*-mW-hás, and to the Tocharian 1sg. middle preterit endings (Toch. A -we ~ -e,
Toch B. -mai; middle present Toch A -a:r ~ -ma:r, Toch B. -ma:r), from *-mw-h2ái
and *-mw-h2ár, respectively. No connection with the perfect active participle,
which is what underlies the Latin (and Albanian) -v- in the perfect. Likewise,
the -s- in the second person is better compared to Hittite mixed forms such as
-sta (act. *-s + stat. *-th2a) and to the processes that led to the sigmatic
aorist outside Anatolian and Tocharian (3sg. *-s transferred in Anatolian and
Tocharian to the stative-as-preterite).

The Tocharian perfect sares with Hittite a complete lack of reduplication. It
shares with e.g. Latin and Celtic the merger that occurred with the other
preterite categories (imperfect, aorist). Not all perfects/statives have ended
up in Tocharian as preterites (some are conjunctives and imperatives), and not
all forms that have perfect endings are etymologically perfects. Those that
*are* show, as far as it is possible to tell and no Ausgleich has taken place,
the expected Ablaut of o-grade in the singular vs. zero grade in the plural.
But Tocharian also has a number of preterites with e:-grade of the root.

Germanic (preterite)

The endings of the Gothic strong preterite are:

-0 -u -um
-t -uts -uT
-0 -un

The Germanic strong verbs are traditionally classified into the following
classes (of the three Ablaut forms given, the first is that of the present sg.,
the second that of the preterite sg,. the third that of the preterite pl.):

e- group:
A1 (ei ~ oi ~ i -> i: ~ ai ~ e/i)
Goth. greipan [/gri:pan/] ~ graip ~ gripum
A2 (eu ~ ou ~ u -> iu/eo ~ au ~ o/u)
Gothic biudan ~ bauT ~ budum
A3 (eRC ~ oRC ~ R.C -> eRC/iRC ~ aRC ~ uRC/oRC)
Gothic bindan ~ band ~ bundum
A4 (eR ~ oR ~ R.C // eR/iR ~ aR ~ e:R [!])
Gothic niman ~ nam ~ ne:mum
A5 (eC ~ oC ~ (e)C // eC/iC ~ aC ~ e:C [!])
Gothic mitan ~ mat ~ me:tum

R1 (oi -> Gothic ai + Reduplication, NW Gmc. ai ~ e:2)
Gothic haitan ~ haihait [/hehait/]
OE pret. hét
R2 (ou -> Gothic au + Reduplication, NW Gmc. au ~ eo)
Gothic hauan ~ haihau [/hehau/]
OE pret. héow
R3 (oRC -> Gothic a + Reduplication, NW Gmc. a ~ e/e:2)
Gothic staldan ~ staistald [/stestald/]
OE pret. feng
A6 (oR // aR ~ o:R ~ o:R ; oC // aC ~ o:C ~ o:C)
Gothic slahan ~ slo:h ~ slo:hum

long vowel group:
R4 (e: -> Gothic e: + Reduplication, NW Gmc. a: ~ e:2)
Gothic sle:pan ~ saisle:p [/sesle:p/]
OE pret. slép
R4/A7 (e: -> Gothic e: ~ o: + Reduplication, NW Gmc. a: ~ e:2)
Gothic le:tan ~ lailo:t [/lelo:t/]
OE pret. lét
R5 (o: -> Gothic o: + reduplication, NW Gmc. o: ~ eo)
Gothic wo:pan ~ waiwo:p [/wewo:p/]
OE pret. wéop

The most interesting cases are Ablauting classes 4 and 5, where in the preterit
plural (and also, in West Germanic, in the 2sg.) we find *e: (WGmc. -a:-) for
expected zero-grade. Ablauting class 6, with Ablaut *o ~ *o:, is similar, but
here o: is found also in the strong forms.

This e:-grade in the preterite root has been compared with Latin (and Albanian)
verbs which show e:-grade in the perfect (preterite). These mostly conform to
the same root structure as Germanic Class IV and V verbs (roots ending in a
single obstruent or resonant): ce:pi:, e:mi:, ve:ni:, fre:gi:, etc. But in
Latin, these forms appear throughout the paradigm, not only in the plural,
weakening the parallel with Germanic. Remarkably, this is exactly where we have
an almost exact match between Germanic and Lithuanian, where a large number of
verbs ending in -VR and some that end in -VC (C=obstruent, R=resonant) take the
e:-preterit with lengthening of the root vowel (lekiù ~ le:ke:, geriù ~ ge:re:),
instead of having zero grade in the root as expected. We would not expect a
Lithuanian preterite to be an old perfect etymologically, but the lengthening of
the root vowel certainly seems to suggest a common origin with the Germanic
forms. If the Baltic e:-preterite is indeed, as discussed below, derived from
the stative conjunctive (a stative paradigm with zero-grade of the root), this
should not be surprising. The presence of e: in the weak form of the root is
most likely explained as the haplological elimination of a reduplication pattern
*Cé:-C(e)C- ~ *Cé:-C(e)R- (with long reduplicative vowel) to unreduplicated
*Cé:C ~ *Cé:R-.

Returning to the Germanic perfect, it should be noted that Germanic has also
retained an important category of "preterit-presents", old stative verbs with
present/resultative meaning, such as Gothic wait "I know (it)", kann "I know
(him)", skal "I should" ,mag "I can", etc.


I will discuss the Armenian active aorist endings here, even though it is far
from clear whether the endings of the Armenian aorist have something to do with
the perfect:

-i -ak`
-er -êk`, -ik`
-0 -in

The 1sg. -i could be from *-h2ai (cf. the 1sg. pf. ending of Latin (-i:) and
Slavic (-e^)). Alternatively, it could be an optative form, either athematic
*-ye:(m), or thematic *-oih1m. A third alternative is that is derives
analogically from such aorist forms as eki, edi, where the /i/ is part of the
root (e-gwem, e-dhe:-).

The 2sg. -er could be from *-(e-)th2a, if *-th2- (*th) had been weakened to *dh
> Arm. /r/ (cf. the optative 3sg. -yr < *-yeth ?). Alternatively, it could just
be a thematic aorist in *-és > -ér, although we would not expect either the
vowel or the consonant to have been preserved in Armenian (the genitive in *-os
or *-és gives -0), at least not in polysyllables. Perhaps again the answer lies
in roots like ekir, edir, etur ((e-)gwem-es, (e-)dhe:s, (e-)do:s), where -r <
*-s may have been retained in a monosyllabic root (not counting the augment).

The 3rd sg. -0 could be from just about anything: perfect *-e, aorist *-(e)t,
e:-preterite *-é:t.

The 1pl. form, like all past tense 1pl. forms in Armenian comes from the middle
(passive) aorist ending -ak`.

The 2pl. has -e^k` (*-etes) or -ik` (e:-preterit *-e:tes or optative *-i:tes).
The 3pl., finally, can be either from *-ent, *-e:nt or *-i:nt.

Against the identification of 1 and 2 sg. -i, -er as perfect endings are the
facts that the Armenian aorist shows neither reduplication nor o-grade (except
perhaps in a few forms like c^`ogay "I went" < *kyow-). A form like eber "he
carried" is formally an imperfect (present-stem preterite) *h1e-bher-et, while
other forms are comparable to thematic (lk`- < *likw-e/o-) or athematic aorists
(1sg. eki, 3sg. ekn < *h1e-gwem-, 1sg. etu, 3sg. et < *h1e-doh3-, 1sg. edi, 3sg.
ed < *h1e-dheh1-). The weak aorist in Armenian is built on -sk^e/o-. A trace
of the e:- and a:-preterites may lie in the passive aorists in -ea- < -ia-, if
from -e:-a:-.

We have one certain case of a preterit-present (conjugated as an active present)
in Armenian, the defective verb goy "there is", gon "there are" (*wos-e-ti,
*wos-enti), etymologically connected to the Germanic past tense of "to be"

Old Irish
-0 -(a)mmar
-0 -id
-^ -(a)tar

The Old Irish perfect does comply with the prerequisites: reduplication,
o-grade. As to the endings, 1sg. and 3sg. are as expected. In the 2sg. we
would expect -t (< *-th2a) after vowels, instead of -0. This may have been lost
in a process similar to the loss of -t in Dutch phrases with pronoun inversion
such as <loop je> vs. <je loopt>. It's true that in Old Irish, tú "thou"
cannot directly follow the verb: the particle -s(i)u, -so is used instead. But
that particle itself probably originates in the amalgamation of the verbal
ending -s + -tu. In the same way, a preterit like *reragt tú could have become
*rerag (t)tú, thus explaining the loss of *-t.
The 1pl. -(a)mmar is analogical after 3pl. -(a)tar, itself a mixed form composed
of *-nt + *-ró.
Forms without reduplication have root vocalism /a/ and /i/ < *o: and *e:.


The Latin perfect is the result of the falling together of at least three
different formations: the PIE perfect, the s-aorist and a "weak" periphrastic
preterit based on the ptc. pf. act. in *-wo:ts (obl. *-us-). The sigmatic
aorist forms simply adopted the preterit endings after the aoristic /s/ (e.g.
scrip-s-i:, scrip-s-isti:, scrip-s-it, scrip-s-imus, scrip-s-istis,
scrips-e:runt). The weak and strong forms must originally have had the
following paradigms:

weak (ptc. + "to be") strong

ama:vu sm. e-em-h2a-i
ama:v(u) es e-em-(e)-th2a-i
ama:v(u) est e-em-e(t)
ama:vu sm.os e-em-(o)-mos
ama:v(u) estes e-em-(e)-tes
ama:vu sn.t e-em-e:r-i (or: e-em-e:re)

The resulting preterit endings are:

ama:v-i:, e:mi: < *-h2ai (strong)
ama:v-isti:, e:misti: < *-es + *-th2ai (combined)
ama:v-it, e:mit < *-et (thematic/strong)
ama:v-imus, e:m-imus < *-omos (thematic/strong)
ama:v-istis, e:m-istis < *-estes (weak)
ama:v-erunt < *-u sn.t (weak),
e:mere < *-e:ri (strong)
ama:ve:runt, e:me:runt < -e:runt (combined)

The rest of the Latin perfect system is based on the periphrastic forms (i.e.
the endings are the endings of the verb "to be");

pf. conj. = pf.stem + sim (fuerim; ama:verim, ama:veri:s, etc.)
pqpf. = pf.stem + eram (fueram; ama:veram, ama:vera:s, etc.)
pqpf.conj. = pf.stem + essem (fuissem; ama:vissem, amavisse:s, etc.)
fut.ex. = pf.stem + ero: (fuero:; amavero:, amaveris, etc.)

The futurum exactum was later confused with the perfect conjunctive, hence such
forms as pf. conj. ama:veris or 3pl. fut.ex. ama:verint.

Reduplication and o-grade are features of the strong verbs (momordi:, tetigi:,
etc.), but, as mentioned earlier, e:-grade perfects also occur (ce:pi:, e:mi:,
le:gi:). The sigmatic forms sometimes have maintained lengthened grade (rego ~


The Albanian aorist has the following endings:

-a -më/-ëm
-e -të/-ët
-0, -i -në/-ën

The root shows Ablaut /o/ < *e: in a number of verbs. Most of the non-Ablauting
verbs use suffixes to form the aorist stem, including -v-, -n- and -t-, which
point to periphrastic formations based on the pf. act. ptc. (-v-) or on the pf.
pass. participles in *-nó- and *-tó-.


The perfect is not attested in Baltic, but see above and below (under "stative
conjunctive") for the e:-grade e:-preterites.


Slavic conserves only one perfect form, the irregular 1sg. present ve^de^ <
*woid-h2ai, "I know" (cf. Latin -i:). No other forms attested. See below under
"stative conjunctive" for the e:- and a:-preterites.


Greek has a weak perfect with 1sg. in -ka (after vowels, liquids and dentals) or
-ha (after labials and velars). Some verbs ending in a consonant have a
semi-strong perfect in -a. The strong perfect has survived relatively
unchanged only in a handful of verbs (e.g. the preterito-present oida "I know").

weak strong
-ka/-(h)a -kamen/-(h)amen oid-a is-men
-kas/-(h)as -kate/-(h)ate -katon/-(h)aton ois-tha is-te is-ton
-ke/-(h)e -ka:si/-(h)a:si -katon/-(h)aton oid-e is-a:si is-ton

The Greek perfect generally reduplicates, has o-grade in the singular and zero
grade in the plural. The k/h-perfect clearly originates in the 1sg. in *-h2a,
where, especially in combination with another laryngeal, *-H-h2a gave *-ka.
Note also the transfer (or independent development out of *-H-m.?) of this
ending to the aorists of laryngeal roots (aor. ethe:ka, he:ka, edo:ka). In the
2sg. imperfect of the verb "to be" (e:stha) and, sporadically, "to go",
(e:ieis-tha, e:ieis), we are certainly dealing with transfers from the prefect.


-a -má -vá
-tha -á -áthur
-a -úr -átur

The singular forms are straightforward: *-h2a, *-th2a and *-e. Note that *-h2a
prevented the working of Brugmann's Law (1sg. cakara, 3sg. caka:ra < *ke-kor-Ha,
*ke-ko:r-e). Reduplication and o/0 Ablaut are the norm, but verbs ending in a
single consonant may have /e:/ (< *ei) in the weak forms (se:d-má < *sazd-má).
Roots in -H have -a:u in the 1 and 3 sg., presumably generalized from roots in
*-h3 (dada:u "I/he gave").

The 1st person plural *-mé and 3rd plural *-r.s are as expected, as is the 1du.
*-w(h2)á. The 2/3 dual forms look like transferrals from the plural: a mixed
form *-nt-é(:)rs (cf. Old Irish -tar < *-nt-ro) > -atur, and an analogical from
-áthur, ultimately reshaped from *-te(:)rs (cf. Tocharian B -cer).

The 2nd person plural form in -á is mysterious. One would expect *-tá (< *-té).
Perhaps this continues a form where *-dh(w) > *-t(w) was dropped in the Auslaut
before the addition of *-e?

Remarkable in Old Indic are the perfects with long reduplicative vowel (usually
in the weak plural stem): kan- -> ca:-kan-, gr.- -> ja:-gr.-, kl.p- ->
ca:-kl.p-, dhi:- -> di:-dhi:-, tu- -> tu:-tu-, s'u:- -> s'u:-s'u-. Except in
the case of a small number of verbs with initial laryngeal HV-HC- > HV:C-, no
explanation is known for this phenomenon, but as mentioned above, it's possible
that this is a retention of the long grade of the reduplicative vowel where it
retained the stress (as often in the 3 person plural), analogically extended to
the rest of the weak forms.

Miguel Carrasquer Vidal