Re: *-men- stems in PIE and Germanic (2)

Proterokinetic *-men- neuters (with a PIE nom./acc.sg. in *-mn.) were
eliminated from Proto-Germanic, as was the hysterokinetic type with a
nominative in *-mé:n, which left only indirect traces of its existence
(though some other types of hysterokinetic nasal stems fared well in
Germanic). The only varieties of *-men-stem nouns directly attested in
Germanic were old amphikinetic neuter collectives reinterpreted as
singulars (a dying-out type, with just a few known representatives) and
masculines (a productive type, treated as singular from the outset and
partly taking over the fuctions of the old *-men- neuters). In both
types the nom.sg. ended in *-mo:: (with the *n lost already in PIE times
and the exposed final long vowel vowel gaining an extra mora, as in
Balto-Slavic); in masculines, the acc.sg. masc. ended in *-manuN <
*-mon-m., the nom.pl. masc. in *-maniz < *-mon-es. In the remaining case
forms, inflectional endings were added to a weak stem in *-mn-. It seems
that the ablaut of the root was given up already in PGmc. and that,
since *-mn- was prone to phonological mutilation, a prop vowel was
inserted to restore the transparency of the suffix: *-mn- -> *-men- >
*-min- (its ultimate source may have been the old locative in *-men-i,
which had a full vowel even in the amphikinetic type). The process was
underway but not yet complete at the time of the breakup of Germanic
linguistic unity.

*-mn- itself, recessive as it was, represented an analogical innovation
for "regular" PIE *-n-. It can also be found in derivatives of *-men-
stems, including formations that occur only in Germanic and have no
direct cognates in other branches (e.g. *-mn- + *-(i)jo/ah2). In fact,
Germanic probably provides more secure examples of *-mn- than any other
branch.

Example: PGmc. *leux-man- '(beam of) light, radiance'

nom.sg. *leux-mo:: (OE le:oma, OIc. ljómi)
acc.sg. *leux-man-uN
gen.sg. *liux-min-iz (with palatal assimilation)
nom.pl. *leux-man-iz

The analogical weak stem replaced *l(e)ux-mn-, visible in derivatives
like *lux-umn-(i)ja- (ME levene < *luxuBnija-) ~ *lux-mun-jo: (Goth.
lauhmuni). Opinions vary as to which pattern of syllabification is
original before *j. Gothic, at any rate, has only a few cases of -mun-
against the majority type in -ufn-/-ubn- < *-umn- < *-m.n-. The
variation od *-mn-/*-Bn- is common throughout Germanic, and the Gothic
devoicing in the allomorph <-ufn-> is conditioned by Thurneysen's
voicing dissimilation rule. The 'lightning' word is often reconstructed
with *au (the Gothic spelling is ambiguous), but since ME levene does
not require *au (it can reflect OE *ly:f(e)ne), a zero grade is easier
to justify. Another characteristic example showing a zero grade in the
root is Gothic witubni 'knowledge' < *wit-umn-ijaN (PIE *wid-m(e)n-).

At the time when the type of *leuxmo:: was really amphikinetic, the weak
cases must have had final accent. If the reduced form of the suffix
still existed at the time, we would expect gen.sg. *luk-n-és to have
yielded early PGmc. *luGniz, hence, by Kluge's Law, *lukkiz. The variant
*lukk- is not attested for this particular stem, but scattered traces of
similar kinds of alternations can be found in Germanic. For example, PIE
*bHudH-me:n became Germanic *BuDmo:: through the conflation of the
hysterokinetic and amphikinetic types, but the weak stem variant (and
derivational base) *bHudH-n-' yielded *Butt-, hence the contaminated
variants *Butma- beside *BuDma- (shifted to thematic nouns). Another
example is Goth. hiuhma < *xeux-man- 'pile' vs. MLG hocke 'haystack' <
*xukk- < *xuG(m)n-'.

Piotr