Re: Dating *e > *i in Germanic

[...]

> On point 3, I arrived at a similar conclusion supporting "late" /i/-umlaut of *e (and
> *en,C > *in,C), and was forced to regard Fi. _teljo_ and _rengas_ as borrowed from Early
> NGmc, not Late PGmc, on the basis of studying 'wolf'.

The shift *o > *a in Germanic is not a lower bound for the loaning of _rengas_ and _kuningas_, though. Proto-Uralic did not allow 2nd syllable *o - this arose in Finnic by *aw > *o (difficult to date, could have included an intermediate stage *ou) and *aj > *oi (from which *oi > *o only during the early 2nd millennium in individual dialects), and seems to have even after these remained a disfavored stem vowel, until the influx of loans from Germanic with PG *-o:. Finnic would for a long period have adopted any word ending in *-os or *-oz with the ending *-as. There are indeed no **-os-stems in the Finnic languages (as distinguished from *-o-kse-stems such as _tule-_ "to come" : _tulos, tulokse-_ "result"; these only occur in derivatives anyway, never in loans.) Cf. _porsas_ "pig" (not **porsos), whose 1st-syllable *o is generally considered an archaism. Although I suppose loaning not directly from IE but from a precedessor of Mordvinic *purc´@..., which can go back to not only *porc´Vs but also *parc´@..., cannot be ruled out...

 

DGK:  I found three examples of archaic /o/ in Baltic loanword roots by skimming through some literature.  First, Finnish _lohi_ (Lappish _luossa_) 'salmon' against Old Lithuanian _lâs^is_ (Latvian _lasis_) points to Proto-Baltic *lós^is.  Since the salmon is a spotted fish, its name has been plausibly connected with Lith. _lâs^as_ 'drop', and Pokorny reconstructs a root *lak^- 'sprenkeln, tupfen' (IEW 653) under which 'salmon'-words of various formations are included.  On the basis of Fi. _lohi_ I would correct this to *lok^- 'spot(ted)', and I would further identify P.'s tentative root *lok- 'tadeln, schmähen' (IEW 673, only Celtic and Germanic) as the same *lok^- under a straightforward semantic shift 'mark, spot' > 'blemish, fault' which has several independent parallels.  The /o/-grade is presumably deverbal but I can find no direct evidence for any *lek^- or *h{x}lek^- 'to mark, spot, dot' vel sim.

 

Second, Fi. _morsian_ 'bride' against Old Prussian _ma:rtin_ acc. sg. 'bride', Lith. _mar~c^iaN_ acc. sg. 'bride; daughter-in-law' requires PBlt *mórtjan.  This is a de:vi:-noun (Lith. nom. sg. _martì_) and was presumably borrowed into Proto-Finnic from the accusative because brides were property and usually mentioned in connection with transactions (i.e. 'taking a bride').

 

Third, Finnish/Estonian _oinas_ 'ram' against Lith. _âvinas_ requires PBlt *ówinos.  Like _porsas_, this maintains /o/ in the stem but shifts *-os to -as.  Clearly Fi. _porsas_ (Est. _pôrzas_) cannot be a Proto-Baltic loan, since PBlt *s^ and *z^, which are generally maintained as such in Lithuanian, become Finnish /h/ (and Estonian zero in this position).  From PBlt *pórs^os (Lith. _par~s^as_) one would expect Fi. *porhas (or better *porha, as below, with Est. *pôra).  I would be surprised if Proto-Mordvinic *purc^@s yielded Fi. _porsas_ rather than *porha(s), since Proto-Finnic would have to borrow *c^ like *s, and it did not borrow *s^ like *s even after PBlt had shifted *o to *a, as shown by Fi. _hammas_ 'tooth' from PBlt *z^ámbas (earlier *z^ómbos, cf. Greek _gómphos_ 'wedge-shaped bolt, dowel'; Lith. _z^am~bas_ 'sharp edge of a beam').  Pokorny (IEW 841) regards Fi. _porsas_ as borrowed from an Iranian dialect.  Kurdish _purs_ has both the required secondary vowel-decentralization and secondary *-rs^- > *-rs-.

 

The other sibilant-borrowing issue is Fi. _hanhi_ 'goose' not *hansi (evidently *-is > -i regularly in these borrowings) from PBlt *z^ansís (OLith _z^ansis_, Lith. _z^aNsìs_, OPruss _sansy_).  It appears that Proto-Finnic assimilated borrowed *s to the reflex of a preceding *s^ or *z^ unless a nasal cluster intervened as in *z^ámbas.  Thus PBlt *z^ansís yielded Fi. _hanhi_ (Est. _hani_), *dárz^as 'enclosure, garden' (Lith. _dar~z^as_) yielded *tarhah, then Fi. _tarha_ (Est. _tara_) not *tarhas, and *káus^as 'dipper, ladle' (Lith. _káus^as_) yielded Fi. _kauha_ not *kauhas.  Of course, this small sample is not definitive.  Other factors may be at work.  Nevertheless the agreement between Finnish and Estonian with the 'garden'-auslaut is striking, given the /s/-retention in Fi. _taivas_, Est. _taewas_ 'heaven' (OLith _deivas_ 'god') and the like.

 

Proto-Baltic *-á: (from PIE *-áh2) yields Lith. -à but Fi. -o or -a.  Thus Lith. _s^eimà_ 'family' = Fi. _heimo_, Lith. _salà_ 'island' = Fi. _salo_, but Lith. _s^alnà_ 'frost' = Fi. _halla_.  My guess is that the words with -o were borrowed earlier, when the final vowel was still long, and those with -a reflect a shortened PBlt final vowel.