[tied] Morphology (4/20)

From: Miguel Carrasquer
Message: 14375
Date: 2002-08-18

4. The sigmatic aorist

In the preceding section, I pretended that the preterite endings were completely
parallel to the present endings, except for the absence of the final element
*-i. There is one form, however, where the parallel is not quite exact, and
that is the 3rd person plural. The absence of a form *-ort in the thematic
forms can be explained away by invoking perhaps a more recent origin of the
thematic formations, or by pointing out that in the conjunctive, which is likely
to have had quite an impact on the thematic indicative conjugation, the *-i is
optional, thus favouring the alternation *-onti ~ *-ont rather than *-onti ~
**-ort. More difficult to explain, however, is the complete absence of expected
3pl. **-ért in the athematic conjugation. All attested forms (including those
in the perfect) point to *-er or *-e:r, the latter by Szemerényi's Law from
*-ér-s, or, with accent retraction in the reduplicated forms, *-r.s (> Avestan
-rs^, Vedic -ur).

The 3rd. plural then points to a preterital 3rd person ending *-s rather than
*-t. Cf. also the Osco-Umbrian opposition between 3pl. present -nt, preterite
-ns. The same ending *-s is of course characteristic of the 3rd person singular
in the Hittite preterite of the hi-conjugation, and of the 3rd person sg. of the
Tocharian Class III preterites. Now the Hittite hi-conjugation as well as the
Tocharian Class III preterites are linked to the PIE perfect, both because of
the personal endings and because of the root Ablaut (o-grade in the singular),
so at first sight this Anatolian-Tocharian 3rd sg. *-s would appear to be a
thing peculiar to the perfect/stative. However, outside of Anatolian and
Tocharian, the perfect 3sg. ending is unanimously *-e, not *-(e)s. The presence
of *-s in the 3rd plural *-ér-s, where the stative/perfect ending was surely
borrowed from the active (the stative lacked or had lost a 3rd. plural ending of
its own), decides the issue in favour of an origin of the *-s in the active
preterite. So where the present added *-t to the 3rd. person forms, the
preterite added *-s. Whether this was in origin the lost 3rd person pronoun
**su (surviving only in the oblique reflexive form *swe) or an old
nominative/ergative of the pronoun *to (**tu-a: > *swo > *so vs.
accusative/oblique *ta-a: > *to), the ending in both cases would have resulted
in labialized *-sW. The resulting preterite paradigm would have been:

Set I:

**-mW **-mWén **-mWáh2
**-s(W) **-t(W)ér **-t(W)áh2
**-s(W) **-ér-s(W) [**-éh2-s(W)]

Set II:
**-mWé **-mWh2á
**-t(W)é **-t(W)h2á

Such a paradigm, with 2sg. and 3 sg. both ending in *-s(w) would have been an
open invitation to create an analogical paradigm using *-s(W)- as the preterite
marker. Which is precisely what we have in the shape of the sigmatic aorist.

Hittite and Tocharian, by transferring or restricting the *-s ending to the
stative preterite, where it did not clash with 2sg. *-th2(a), did not take this
step (but the mixing of mi- and hi-conjugation forms eventually led to Hittite
2/3sg. preterite by-forms in -sta, for -s and -t(a), and likewise the Tocharian
Class III preterites, in a move similar to the creation of the sigmatic aorist
elsewhere, transferred the *-s- (extended by general preterite *-a:-) of the
3sg. to the whole of the middle as well). Besides Anatolian and Tocharian, only
Germanic and Armenian show no trace of the sigmatic aorist, although the
characteristic Armenian aorist in -c'- (< *-sk^e-) is sometimes interpreted as a
(reinforced) s-aorist.

The element *-s of the 2+3 sg. had a further effect: by Szemerényi's Law. 2/3
sg. *bhér-s became *bhé:r-s. From there, the long vowel was transferred to the
other persons of the indicative active, which after all are built analogically
on the model of the 2+3 sg. We have *bhé:rsm; *bhé:rsme, *bhé:rste, bhé:rsr.s
etc., with columnar (static) accent on the lengthened root. Although the long
grade of the root vowel was largely lost in Celtic and Greek, it was retained in
Latin, Slavic and Indo-Iranian.

Skipping over Anatolian, Tocharian, Germanic and Armenian, the forms occurring

Old Irish (conjunct):

-us -sam
-is -sid
-0 -sat

As in Greek, -s- should have been lost in Old Irish when following a vowel, but
the sibilant was probably restored after the forms where a consonant preceded
the *-s. The endings in Old Irish are the thematic present endings (*-so:,
*-si: etc.), which is surely not original. Only the 3rd p.sg. derives from
athematic *-s(t).


The s-aorist in Latin survives in those forms of the perfect which are made with
*-s- (aux-i:, scri:ps-i:, etc.). The endings are those of the perfect. The
root vowel is normally lengthened (e.g. veho, ve:xi:), as it is in Slavic and


Albanian has an *-s- (> -sh-) in the imperfect, but that is more likely to be
from an agglutinated form of the verb "to be". The 1sg. aorist of suppletive
verbs has -sh- (rashë "I fell", pashë "I saw"), but this -sh- was absent in the
oldest attested forms and seems to be an innovation. The s-aorist is retained
in the Albanian optative in -sh- or -fsh-, which is usually interpreted as
derived from an s-aorist conjunctive.

Whether the Baltic future in -si- derives from an s-aorist conjunctive or, what
seems more likely, from a desiderative in *-si-, the inclusion of Baltic in
Balto-Slavic implies that the sigmatic aorist must once have existed in Baltic
as well.


-xU, -sU -xomU, -somU -xove^, -sove^
-0 -ste -sta
-0 -s^e~, -se~ -ste

The 1st person forms have been thematized (*-sW-o-m, -*sW-o-mos, *-sW-o-we:),
the others are athematic (*-s[-s], *-s-té, *-s-té:; *s[-t], *-s(W)-ént, -s-té).
Slavic also preserves the characteristic lengthening of the root vowel.
Ironically, it is precisely the pivotal 2+3 sg. of the s-aorist which in Slavic
are replaced by short-vowel thematic preterits, as in the paradigm of tes^ti "to
flow" (root tek-):

te^-xU te^-xomU te^-xove^
tec^-e te^-ste te^-sta
tec^-e te^-s^e~ te^-ste


-sa -samen
-sas -sete -saton
-se -san -sate:n

The 1st sg. is regular from *-s-m., while 3pl. might be from *-s-n.t. The
resulting -a- vocalism was carried over to the 2sg. -sas (for expected *-s(s)).
Other forms (-se, -sete) have been thematized. As pointed out by Szemerényi,
Greek retains some traces of long grade in the ind. act. of the s-aorist (e.g.
érre:xa "I broke"), although usually e-grade has been restored. As in Old
Irish, the /s/ of the sigmatic aorist should have disappeared in Greek when it
followed a vowel, but it was everywhere restored analogically after forms where
/s/ followed a consonant.



á-bha:r-s.-am á-bha:r-s.-ma á-bha:r-s.-va
á-bha:r-s.-0 á-bha:r-s.-ta á-bha:r-s.-t.am
á-bha:r-s.-0 á-bha:r-s.-ur á-bha:r-s.-t.a:m


á-vr.k-s.-a-m á-vr.k-s.-a:-ma *á-vr.k-s.-a:-va
á-vr.k-s.-a-s á-vr.k-s.-a-ta *á-vr.k-s.-a-tam
á-vr.k-s.-a-t á-vr.k-s.-a-n *á-vr.k-s.-a-ta:m

Note the retention of long grade throughout the paradigm in the athematic forms.
The thematic forms are rare (only ten roots take it in Vedic). If the Latin
-re:- (*-se:-) past conjunctive is derived from the same form, it is perhaps
originally a conjunctive (no conjunctives of the sá-aorist occur in Vedic),
albeit a peculiar one with accented thematic vowel.

Miguel Carrasquer Vidal