--- In
cybalist@yahoogroups.com, "stlatos" <stlatos@...> wrote:
>
> [...]
>
> This uvular R usually > r in historical Gmc. Also, at:
>
> http://www.utexas.edu/cola/centers/lrc/books/piep07.html
>
> Winfred P. Lehmann, when discussing OHG r-preterites says: "The sequence assumed here for seventh class verbs is PIE /eXw/ [eXu], for first class verbs /yX/ [iX]. I suggest that in these sequences the laryngeals were preserved, and that their reflexes fell into the OHG r-phoneme."
>
> (note: Lehmann uses X to symbolize any "laryngeal"; I use it for the uvular fricative)
Lehmann has done a good job establishing that this OHG /r/ originated from a laryngeal in the sequence *-eXu- or *-iXu-. However, I think he was sidetracked by the apparent restriction of the /r/-preterit to OHG. I believe we have examples of the same phenomenon in Old Norse and Northumbrian 3sg. pret. forms:
1. ON <bnere> 'rubbed'. Gothic <bnauan>, only attested in the present, has the same formation as <bauan> 'to dwell', OHG <bu:an>, to which <biruun> 3pl. pret. belongs. Koebler reconstructs Gmc. *b(i)no:wwan as the pres. inf.
2. ON <gnere> 'rubbed'. If this is not modelled after the foregoing, it likely belongs to the root cited by Pokorny (IEW 436-7) as *gHne(:)u- 'nagen, kratzen, reiben'.
3. ON <grera> 'greened up, grew'. Koebler reconstructs Gmc. *gro:an, but the corresponding OE forms have /w/ throughout the paradigm.
4. ON <rera> 'rowed'. Usually considered reduplicated, but again the OE forms have /w/ throughout.
5. ON <sera> 'sowed'. Usually considered reduplicated, but Go. <saiso> shows that the pret. sg. did not voice the root-anlaut. Again OE forms have /w/ throughout.
6. ON <snera> 'turned, wound'. Pok. (977) considers that *sne:r-, *sno:r- in nouns may be an extension of *sne:-. The ON pret. may thus involve *sneh1-w- where the */w/ formed extended root-aorists like Lat. <(g)no:vi:> (cf. OE <cna:wan> with the /w/ carried through the paradigm).
7. NU (and Old Kentish) <on-dreord> 'came to fear', questionably, since it may have been remodelled after <reord>.
8. NU <leort> 'allowed'. Usually considered reduplicated like <leolc> 'jumped', but the dissimilation required is ad hoc.
9. NU <reord> 'advised'. Usually considered reduplicated.
In all the examples we have good reason to place laryngeals (mostly *h1) in the roots. The following */u/ or */w/ either belongs to the root proper, or can be understood as the root-aorist extension seen in Lat. <gno:vi:>, <se:vi:>, etc.
Along with Lehmann's <scorra>, I would cite OHG <erran> 'to plow' as an example of *-rX- > -rr-. Unlike other geminates, WGmc *-rr- cannot arise from *-rj-, so <erran> cannot be compared sound-for-sound with Go. <arjan>, a /j/-present. And we know from Greek that the IE root was *h2erh3-.
DGK