I mentioned the Latin and Slavic imperfects, and upon closer
inspection, the similarities between the two are quite strong,
something that I have never seen mentioned before, and that I hadn't
previously noticed myself.
Both in Latin and in Slavic, the algorithm seems to be: if the verb
already has an extension *-a: or *-e:, do nothing, else add *-e:. Then
add a past tense of the verb "to be" (and make sure to subsequently
replace the independent form of this past tense of the verb "to be"
with a form based on the stem used in the *other* language).
Hence:
*bh(u)a:m ~ eram *e:xom ~ be^xU
*bh(u)a:s ~ era:s *e:xes ~ be^
*bh(u)a:t ~ erat *e:xet ~ be^
*bh(u)a:mos ~ era:mus *e:xomos ~ be^xomU
*bh(u)a:tes ~ era:tis *e:x[e]te ~ be^ste
*bh(u)a:nt ~ erant *e:xont ~ be^s^e~
ama:-bam de^la-axU
de:le:-bam vele^-axU
audi-e:-bam xvalj-e^-axU [> xvaljaaxU]
em-e:-bam nes-e^-axU
Note further that it seems quite impossible to associate this ending
*-e: (optionally *-a:) with any of the known PIE participles or verbal
adjectives. Rather, it seems to correspond to the past tense stem of
finite preterites (like the Baltic e: [e.] and a: [o] preterites).
I.e. not a periphrastic contruction like "I was doing" but something
like a "serial verb": "did, I was".
=======================
Miguel Carrasquer Vidal
mcv@...