Pre-PIE
(labializations) (delabializations) PIE 1 *-mu-átu *-m w ét w *-més w *-més (w) ~ *-mós 2 *-tu-átu *-t w ét w *-t w é *-té 1 *-mu-íku *-m w éx w *-m w éx *-wáh 2 2 *-tu-íku *-t w éx w *-t w éx *-táh 2
Note the irregular delabializations 2 which affected these forms.
When -i was added in the other forms to mark the present tense, *-mesi was no problem (we have it in Indo-Iranian), but the 2 pl. and 1/2 du. forms were apparently not capable of taking it 3 .
So when the non-Anatolian languages dropped the variants *-wén(i) 4 and *-tén(i) 5 , the resulting system may have looked temporarily like this:
There may have been pressure to distinguish primary and secondary endings in the 2 pl. and in the dual, and to distinguish the dual from the plural in the 3 rd person, but no obvious solution was available. However, when the perfect/aorist ending *-me was transferred to the 1 pl. imperfect (and, conversely, active *-té was transferred to the perfect/aorist), a pattern could be abstracted from the new opposition between past *-mé and present *-més 6 . In Italic, Celtic, Balto-Slavic and Armenian (possibly Germanic as well, but we cannot tell), a new 2pl. *-tés was created, and the new dual presents *-wh 2 ás, *-th 2 ás,*-tés 7 even spread to Indo-Iranian.
This gives a late PIE dialectal system (which does not include Anatolian, Tocharian, Greek or Indo-Iranian fully):
The former existence in Balto-Slavic of a contrast between secondary -w , -and primary -was, -tas is proven by the Lithuanian 1-2 pl. endings -and -, created analogically on the model of the dual (-w , -:: -was, -tas; -, -:: -mes, -tes), before the present endings were lost.
In Greek, and in part in Indo-Iranian and Tocharian 8 , the gaps in the system were either allowed to stand (Greek 1-2 pl. without primary/secondary distinction), or were filled instead with endings from the middle (marked in bold):
The stative endings
A completely different set of endings was employed in the "stative", i.e. the Hittite i-conjugation, the PIE perfect and the middle (medio-passive). Again there are two sets of endings, although here the "transitive" marker (*-s in the i-conjugation/perfect/aorist, *-t in the middle) appears in the 3 rd person only.
perfect -aorist 1 *-h 2 e *-h 2 e 2 *-th 2 e *-th 2 e 3 *-e *-s 1 *-me *-me 2 *-e *-e 3 *-ér *-ér-s
The similarities between the three paradigms are obvious. The only difficulties lie, as was the case in the active, in the 1-2 pl. endings.
In the 1 pl., Vedic -ma is (rarely) written -, and more often a long vowel is required by the metre. A reconstruction *-meH has sometimes been proposed. However, -also occurs in the active, and a metric long vowel is also found in the 1 sg. -a and 2 pl. -a. I would derive the long vowel ultimately from active 1 du. -, where the variation is to be expected: -vs. -vas and 1 pl. -ma vs. -mas leads to -va vs. -vas, or, as we also saw in Lith., to -vs. -mas. I see no reason, therefore, to reconstruct anything but *-me for the 1 pl. stative. As we saw above, this ending *-me was later transferred to the 1 pl. active as well (though not in Greek).
In the 2 pl., Vedic -a is weird enough that it must reflect the original ending. The *-te found elsewhere must have been secondarily imported from the active. This ending -a is really strange for a 2 pl., especially considering that the *-e in the stative endings is surely a secondary addition. It does not affect the stress or the Ablaut of the root, and seems to be simply an 'augment' to render the stative endings syllabic, which is of course the reason why it was not added to the already syllabic 3 pl. ending. This means that the 2 pl. stative ending was in fact *-ø, which makes no sense at all. The middle endings, so similar to the stative endings elsewhere, are strangely different in the 1-2 pl. The 1 pl. middle ending is sometimes reconstructed as *-medhh 2 (*-medhə 2 ), based on Gk. -μεθα and Ved. -mahi, which is of course phonetically possible, but does not really look like a plausible verbal ending. I would rather reconstruct *-medh-, with the vowel of the corresponding 1 sg. ending added in Greek (at a stage when the 1 sg. was still *-h 2 a c.q. *-a) and Indo-Iranian.
Miguel Carrasquer Vidal
