From: dgkilday57
Message: 71736
Date: 2014-05-08
> > Do you mean to suggest that this would have been the same change as *eN > *iN inWell, yes, I am under the impression that no raising *eN > **iN occurred in open stressed syllables, and there seems to be no reason to expect *oN > **uN in this position either. I am only aware of examples such as _dung_, _hound_ etc. But this does not mean that the change could not have happened even in open syllables, if unstressed.
> > stressed syllables, which would then have been the earliest change of its sort?
>
> DGK: No. If that were so, /a/-umlaut of *i resulting from prenasal *e would be necessary
> to explain OE _cwene_, OHG/OS _quena_ 'woman' (remodelled in Gmc. as wk. fem.
> retaining /e/-grade, cf. OIr _ben_, OPr _genna_, etc.), OHG _neman_, OIce _nema_ 'to
> take' (cf. Grk. _némein_ 'to allocate'), OHG _breman_ 'to roar, growl, murmur' (cf. Lat.
> _fremere_ 'id.'). But Tacitus already has Got(h)ones against Pliny's Gutones, so /a/-umlaut
> of *u had occurred by T.'s time, and in all probability /a/-umlaut of *i also. This makes me
> doubt that stressed *eN > *iN describes an actual event (apart from the occluded-nasal
> situation *eNC > *iNC which occurs in borrowings from classical Latin as well as native
> words). Stressed /e/ in the words mentioned above could just as easily be a retention.
>
> Likewise, I doubt that /a/-umlaut of *u was blocked by a simple nasal (as opposed to a
> nasal cluster), cited at the start of point 1 in your earlier post. A counterexample is
> provided by ON _kona_, MHG _kone_ wk. f. 'woman' which must be built on the
> zero-grade *gWn.h2- > PGmc *k(W)un(h2)o:n-. With this word there is no "later raising *o
> > *u" as you suggested for other examples (none actually quoted).
(English _honey_ looks like an exception but something must've happened here secondarily, given German _Honig_ etc?)
·
DGK: 'Honey' is sticky. Attic-Ionic _kne:kós_, Doric _kna:kós_ 'yellow, safflower-colored' and OPruss _cu[n]can_ 'brown' point to a protoform *kn.h2kó-. What the laryngeal in this position would yield in PGmc is not entirely clear to me. PIE *h2enh2t- 'duck' (Lat. _anas_ etc.) has given PGmc *anuD- (OE _æned_, OHG _anut_, ON _o,nd_). PGmc *aniD- appears to be a ghost form based on OHG _enit_, which could simply be back-formed from the pl. _enti_. Some other few examples suggest that non-absorbed laryngeals in non-initial syllables yielded PGmc *u also. This would lead to Gmc. *xunuGa-, with /a/-umlaut *xunoGa-. But if the laryngeal was effectively part of the first syllable (yet not absorbed), one might expect *xunaGa-, with umlaut *xonaGa-. This is good for OHG _honag_ and OS _honig_, but OHG _honang_ seems to have been contaminated with something. I cannot explain the lack of /a/-umlaut in OE _hunig_, OIce _hunang_, and OSwed _hunagh_. This is a real stumper.
> DGK: 1. If these 'neck' words are not assigned to *kWel(h1)-, they require a new rootI see, you're explaining these words thru a similar "turning point" semantic development as in Baltic *kaklas?
> *k(^)el- 'to turn' vel sim. just for them. This is not particularly parsimonious and leads to
> bad public policy. Anyone who disagrees with a soundlaw (or a borrowing) can simply
> invent a new root. The result is etymological anarchy.
DGK: Yes. That is a good parallel.
Is this the only word to show *Ka- from *KWo-, though?
DGK: For the "voiced aspirates", 'tongue' may be compared with 'sing': Gmc. wk. fem. *tun,G(W)o:n- > Go. _tuggo_, ON _tunga_; *sen,GWanaN > Go. _siggwan_, ON _syngva_, OFris _siunga_ with /w/-umlaut (elsewhere in WGmc the labial component regularly leaves no trace, OE/OS/OHG _singan_).
> 2. In Gmc. lgs. not reflecting *ko:N, 'cow' rhymes with 'sow': OE _cu:_, _su:_ (moreOK, but this still provides no explicit evidence that *kWo: > _ko:_ occurred in Proto-Germanic and not only during the separate evolution of Old Saxon / OHG.
> commonly _sugu_ of different formation); OIce _kýr_, _sýr_; OSwed _ko:_, _so:_ (OIce
> raising of *u: by /R/-umlaut; East Norse *o: from NGmc *u: retained in West Norse as in
> _gno:a_ vs. _gnúa_, _bo:a_ vs. _búa_, etc.). This can hardly be a coincidence.
> Moreover there is no trace of the expected Late PGmc nom. sg. *kauz from *gWó:us (as inIt also gives no evidence to expect such a reduction, if I'm following correctly. If anything, this even seems to provide a motivation for partial analogical leveling of *kWo: to *ko:.
> ON _naust_ 'boathouse', Osthoff's shortening preceded *o > *a in this combination, *o:us > *ous > *aus). Evidently this nom. sg. was replaced by *ku:z after *su:z 'sow', since *u: was
> felt to be closer than *au to *o: of the acc. sg. Some of the individual Gmc. lgs.
> generalized the vowel from the acc. sg., others from the new nom. sg. This irregular
> development gives no reason to doubt the earlier reduction of *kWo(:) to *ko(:) in PGmc.
DGK: I do not follow that. At any rate, my earlier explanation was inadequate. I now believe the correct solution to the replaced nom. sg. of 'cow' involves analogy with the pattern of *u in the nom. sg., *o in the acc. sg. of masc. /a/-stems resulting from /a/-umlaut not occurring in the nom. sg. because the stem-vowel had already been syncopated (as outlined in my draft on 'wolf'). That is, before any root-vowel levelling, it was nom. sg. *wulfs (reg. from *wulxWs) vs. acc. sg. *wolfaN (from *wolx(W)aN with -f- from nom. sg.) 'wolf', *fuGl.s vs. *foGlaN 'fowl', *bukks vs. *bokkaN 'buck', masc. sg. *fulls vs. *follaN 'full'. Individual dialects generalized either /u/ or /o/ for each word, and some languages attest both. 'Wolf' was covered in the draft. 'Fowl' is OE _fugol_, OFris _fugel_, OS/OHG _fugal_, _fogal_, OIce _fugl_, _fogl_. 'Buck' is OE _buc_ (with wk. m. _bucca_), MD _bok_, OHG _boc_ (gen. _bockes_), OIce _bukkr_, _bokkr_ (wk. m. _bokki_). 'Full' is OE _full_, OFris/OS _ful_, _fol_, OHG _fol_, OIce _fullr_. This variation has nothing to do with labials and laterals, since neuters like 'folk' lack it (nom./acc. sg. *folk(k)aN). OIce _gull_ nt. 'gold' beside expected _goll_ can be due to an obsolete adj. *gullr (like _fullr_) 'shining, gleaming' whose nt. sg. was substantivized as 'gleaming metal'.
Osthoff's shortening is seen in Go. _naus_, OIce _na:r_, OE _ne(o)_ 'corpse' from Gmc. *naviz, earlier *ná:vis (IEW 756). The nom. sg. 'cow' should have given Early PGmc *kWó:us, later *kóus by Osthoff and delabialization, finally *kauz against the acc. sg. *ko:N. This pattern was aberrant, but on the model of the therionyms above with nom. sg. *u against acc. sg. *o in the root, a new nom. sg. *ku:z was created to fit *ko:N. As with the other therionyms, different dialects generalized one vowel or the other.
Since 'buck' had to be oxytone in order to get the geminate from Kluge's Law, I will need to make a revision to the timing of the accent-shift in my draft.
> 3. Analogical levelling is as common as dirt. Interrogative-relative words like Go. _hwan_I'd expect high-frequency function words like these to be particularly resistant to analogy.
> 'when' (PIE *kWóm, Old Latin _quom_, Lat. _cum_) reflect Gmc. *xW- restored from forms
> in which it preceded *-i- or *-e- and remained as such, e.g. Go. _hwis_ 'whose' < Gmc.
> *xWes(s) < PIE *kWésjo.
DGK: Whatever the abstruse metalinguistic status of "high-frequency function words" might be, the plain truth is that interrogative-relatives are not exempt from analogical processes. Thus we have Skt. _kím_ 'what?', _kimartham_ 'why?', _kim.cid_ 'anything', _ná-kih._ 'nobody' with k- not c-, OPruss _quai_, _quoi_ 'which' fem. sg. with qu- (from _quei_ 'where') not k-, and MLG _wu:_ 'how?' against OE/OFris _hu:_. Attic-Ionic extended *tín (inherited from *kWím) 'whom?' acc. sg. to _tína_ after the acc. sg. of C-stem nouns, and Attic generalized a new stem, dat. sg. _tíni_, gen. sg. _tínos_, etc. (against Ion. gen. sg. _téo_, _teû_ = Go. _hwis_ from *kWésjo).
> That */w/ in the PIE sequence *k^wo- was not deleted is shown by 'wheat', OE _hwæ:te_Approximately no language in the world has a contrast between /Cw/ and /CW/. To plausibly claim that *k^wo- > *xwa- but *kWo- > *xa-, you'd be better off dating the latter change before the centum merger of *k and *k^.
> etc., referred to Gmc. *xwait(t)ja-, derived from *xwaitta-, this by Kluge's Law from PIE
> *k^woit-nó-, from *k^weit- 'to bleach, blanch, whiten' vel sim. Thus, although Early Proto
> Germanic as a centum language reflected both PIE *k and *k^ as *x (later *G under
> Verner's conditions), it kept *xWo- distinct from *xwo-. The subsequent delabialization in
> the former sequence did not entail /w/-deletion in the latter.
DGK: That would not work. The apparent typological difficulty is easily sidestepped anyway by assuming uvular articulation for PIE deep plosives. In this view the PIE "labiovelars" were actually labio-uvulars, the "plain velars" were uvulars, and the "palatals" were velars. This has the considerable advantage that a highly improbable backing of palatals to merge with velars in the centum group is not required. Instead, the satem group lost the labial component of the labio-uvulars (though its lip-print remained in some environments), fronted the velars to palatals, and fronted the uvulars (including old labio-uvulars) to velars. The centum group fronted the plain uvulars, merging them with the velars, but maintained the labio-uvulars as such. Thus, although we write *k, *kW, etc. (i.e. a K-series and a KW-series), the actual articulation in early centum languages was [k] against [qW], etc. (a K-series and a QW-series).
It is highly relevant that Ulfilas used the single character _q_ (not _qu_ or _qw_) to represent Gmc. *kW. To him, this was a single phoneme, not a sequence of two phonemes. The same goes for the ligature formed from _h_ and _w_, a single character for Gmc. *xW. That U. saw no need for a special character to denote *GW is explained by its limited distribution. It occurred in Gothic only after a nasal, and U. wrote _ggw_ for the whole cluster.
The OE spellings with _hw_, _cw_, _gw_, OHG _(h)w_, _kw_, _gw_, ON _hv_, _kv_, _gv_ suggest that by the time the Roman alphabet was adopted, the North and West Gmc. lgs. no longer had true labio-uvulars like Ulfilan Gothic, but sequences of two phonemes.