Re: Rozwadowski's Change

--- In cybalist@yahoogroups.com, "Torsten" <tgpedersen@...> wrote:

>
> It turns out that this mysterious /a/ for /e/ in the Northern Substrate (hm, vel sim.) that Douglas and I discussed and which Udolph mentioned for Slavic river names, actually has that above name.
>
> Henning Andersen
> Slavic and the Indo-European Migrations, in
> Language Contacts in Prehistory,
> ed. Henning Andersen,
> pp 62-64
>
> '2.4 Word-initial laryngeals
> The regular reflexes of initial laryngeal + *e in Slavic and Baltic are PS, PB e- for PIE *h1e- and PS, PB a- for *h2e- and *h3e-. However, in a number of lexemes Proto-Slavic and Proto-Baltic have irregular reflexes of such initial sequences. There are two cases to consider.
>
> 2.4.1 Rozwadowski's Change. In a number of lexemes, Proto-Slavic and/or Proto-Baltic have initial e- or doublets with initial e- || a- for PIE *h2e- and *h3e-, a peculiarity first described by Rozwadowski (1915). See Table 6.
>
> It must be mentioned that both language groups have had a change in recent prehistory (perhaps around the beginning of our era) of initial e- to a- with characteristic geographical distributions of the reflexes, disturbed, however, by the Slavic territorial expansion as well as by the westward displacement of the Lithuanians (cf. Section 2.2). The reflexes of this recent change are: in Slavic, mainly o- in Russian, otherwise commonly o- in central dialects, je- in peripheral Slavic dialects; in Baltic, mainly a- in Old Prussian, in eastern Lithuanian dialects a-, elsewhere in Lithuanian and in Latvian, e-). Although the recent changes obscure the reflexes of the proto-language initials somewhat, the distinction between PS and PB e- and a- is clear enough, and it is clear as well that the recent changes affected PS, PB e- from *h1e- and from *h2e- and *h3e- on equal terms (Andersen 1996:88-112).
>
> Remarkably, most of the examples of Rozwadowski's Change show morphological differences between the Slavic and Baltic languages. Consider the difference between PS al-k-u-ti- and PB el-k-u:-n-e:- "elbow", both apparently sharing one layer of derivation and then diverging. Or consider the difference between PS el-i-x-a:- || al-i-x-a:- and PB el-s-ni- || al-s-ni- "alder", where the morphological difference provided different environments for the Ruki Change.
>
> Or note the different ablaut grades in PS el-au-a- || al-au-a- and PB el-u-a- || al-u-a- "tin", or the sat&m and centum (pre-sat&m) reflexes of PIE *k^ in PS es-e-ti-"rack" and PB ek-e-ti-a:- "harrow". All these differences must have developed subsequent to Rozwadowski's Change. If one assumes the contrary, it is impossible to understand why a change in a word-initial vowel would have affected predominantly (actually seven out of eleven) synonymous lexemes with morphologically distinct by-forms in different (ante-)Slavic-Baltic dialects while leaving dozens of other lexemes with initial PS, PB a- untouched. The morphological differences clearly go back to before the Sat&m Change (cf. "harrow") and the Ruki Change (cf. "alder") and remind one of the morphological differences among the centum (pre-sat&m) accessions mentioned in Section 2.1.2. One can conclude, then, that Rozwadowski's Change is older than the dialectal differentiation reflected in these morphological differences.
>
> Even more remarkable, considering the early date of the change, is this: if the Rozwadowski lexemes are plotted on a virtual map - assuming the same geographical disposition of (pre-)Latvian, (pre-)Lithuanian and Common Slavic as at the beginning of our era - one can discern (or construe) spatial relations among the e- || a- reflexes which, if they are not a mirage, amount to a pale reflection of the change's extension in a central region of a presumable ante-Slavic-Baltic dialect continuum. In Slavic, the e- doublets in PS elaua- and erila-, which are limited to one language each, may have been northern before the migrations, that is, contiguous to Baltic (or quasi-Baltic) dialects with e-. Otherwise a- variants occur mostly in western Slavic dialects; in Baltic, there are more a- doublets in Latvian than in Lithuanian, and more in Old Prussian than in the East Baltic languages (Andersen 1996:99-101).
>
> This virtual geographical difference is similar to the differences in the distribution of lexical doublets with velar and sibilant reflexes of PIE *k^, *g^(h) (Section 2.1.2), where a central area represented by Lithuanian shows a greater concentration of discrepant dorsals than the peripheral areas to the north (Latvian) and south (Common Slavic).
>
> There is no way of accounting for Rozwadowski's Change as a purely phonological change in pre-Slavic and pre-Baltic (Andersen 1996:103). But of course the discrepant PS, PB e- forms may reflect a regular change in some ante-Slavic-Baltic dialect. If they are to be understood as intrusions, as their geographical distribution suggests, there are several possible interpretations. The most likely seems to be that (i) the e- forms reflect substratum dialects with a markedly different realization of open vowels than the prevailing dialects. If the substratum had, say, e [è] vs. a [a], but the prevailing dialect e [æ] vs. a [o], individual substratum forms with a [a] might have been interpreted by speakers of the prevailing dialect as having e [æ]. Or (ii) perhaps they reflect a substratum that after the loss of laryngeals had merged its low vowels in [æ]. In such a situation, substratum variants without the (initial) vowel distinctions might easily intrude into the tradition of the prevailing dialect. (See further Andersen 1996:111-112.) Be this as it may, the fact of the change and the pale reflection of its apparent geographical distribution are data that point to a distant ante-Slavic-Baltic substratum.
>
> ...

Clear evidence for a pre-Baltic IE substrate, in my opinion.

I have argued elsewhere that the language of Kuhn's Nordwestblock belonged to the Illyro-Lusitanian branch of Indo-European, which includes Messapic and Macedonian. Beside the rather common reflection of IE mediae aspiratae as plain mediae, these languages are characterized by gemination of postvocalic *Cw and *Cy to *CC. Thus IE *h1ek^wo- 'horse' is reflected in Lusitanian as *ikko- (the goddess <Iccona> = Gaulish <Epona>), in Illyrian and Messapic as *(h)ikka- (Epidaurian <íkkos> 'horse', <-ikkos> in Tarentine personal names), and in Macedonian as *(h)ippa- (<-ippos> in personal names, Attic <híppos> 'horse' as a borrowing). Here we also see the raising of initial *h1e- to *(h)i- characteristic of this branch. This is clearly not happening in the pre-Baltic substrate, so this "Nordostblock" cannot be closely related to the NWB. At a greater time-depth, we might assign river-names east of the Oder belonging to Krahe's Alteuropäisch system to the NOB. The absence of AEur names between the Elbe and Oder would then indicate that while the Illyro-Lusitanian people came in from the west, and pre-Baltic IE-speakers from the east, the Elbe-Oder interfluve remained a non-IE refugial area. One of the difficulties in pursuing this matter is distinguishing reliably between true Baltic and presumed NOB hydronyms. As we know, Schmid (who happened to be a Balticist) regarded Alteuropäisch, Proto-Baltic, and Proto-Indo-European as identical, leaving no room for a pre-Baltic substrate. We need a finer comb to sort through all this stuff.

In his paper on 'thousand', "Akzentstudien" §1 (IF 6:344-9, 1896), Hirt argued forcefully against the Vigfusson-Bugge explanation as a compound 'Schwellhundert(schaft)' vel sim., seeing the attested forms with /h/ (which are restricted to Frankish and part of North Germanic) as resulting from folk-etymological contamination with 'hundred'. He analyzed Gothic <þu:sundi> (originally 'Fülle, Menge') as derived from an adjectival *þu:s- like <hulundi> from *hula-; this *þu:s- in our notation reflects an extended zero-grade IE *tuh1-s- from *teuh1- 'to swell', otherwise agreeing with the compound explanation. He then found that the resemblance among Germanic, Baltic, and Slavic forms of 'thousand' admitted neither a satisfactory IE protoform nor a satisfactory borrowing from Gmc. into Balto-Slavic, and concluded thus:

"Ich kenne in der That keinen Punkt, der für eine nähere Verwandtschaft des Lit.-Slavischen und Germanischen spräche, ja mir scheint sogar eine recht bedeutende Kluft zwischen beiden zu bestehen, eine Kluft, die auf alte Trennung durch ein anderes Volk schliessen lässt."

Whether the difficulty with 'thousand' can be resolved by assuming NOB origin, I cannot say.

> PS al-k-u-ti- "elbow", LCS olkUtI, R lokot'.
> PB el-k-u:-n-e:- || al-k-u:-n-e:- "elbow", OPr. alkunis, La. è,lkuons, Li. alkú:ne., d. elkú:ne.. PIE *hxh3-el-.
> Cf. Skt. aratní-, Av. ar&thna-, Gr. o:léne:, o:llon, Lat. ulna (*olena:), OIr. uílen, Go. aleina, OHG elina.
>
> PS el-au-a- || al-au-a- "lead (Pb)", Bg. o. elav(o), elsewhere *o-: R olovo.
> PB e:l-u-a- || a:l-u-a, OPr. elwas 'tin', alwis "lead", La. al^vs "tin", Li. álvas "idem".

Illyro-Lusitanian should have had *-ll- from *-lw-.

> PS elix-a:- || al-i-x-a:- "alder", R ol'xa, SC jels^a, d. jelha.
> PB el-s-ni- || al-s-ni- || al-is-ni- "alder", OPr. alisknas («Abskande»), La. àlksna, Li. e~lksnis, alìksnis.
> Ante-IE *al(V)s-.
> Cf. Mac. álidza (Hesych.), Lat. ulnus (*alisnos), OHG elira, Gm. Erle, ON o,lr, jo,lstr "willow", Fr. alise "rowanberry" (< Gaul. *alisia).

The Latin form is <alnus>. Oscan *alsin(o)s is inferred from South Italian dialects. Note also the river-name <Alisontia>, probably *Alisa:-went-ya: 'Full of Alders'. For 'thousand', Li./La. (but not OPr.) also has "/k/-Einschub" before the /s/.

> PS epsa:- || apsa:- "aspen Populus tremula", R osa, osina. PB ep(u)s^e:- || apse-, OPr abse, La. apse, Li. ãpus^e, e~pus^e (contaminated with pus^ìs "pine"). Ante-IE (?) *asp-.
> Cf. OHG aspa.
>
> PS erila- || arila- "eagle"
> LS jerjol/, elsewhere *o-: R orël. PB erelia-, OPr. arelie, La. èrglis, Li. ere~lis. PIE *h3er-. Cf. Go. ara, Eng. erne, Gk. órnis "bird", Hitt. haras^, haranas^ "eagle".
>
> PS esera:- "prickly stuff, P d. jesiora "fish bone".
> PB es^eria- || as^eria- "perch Perca fluvialis", La. asers, Li. es^ery~s. PIE *h2ek^-er-o-.
> Cf. OHG ahira, Gm. Ähre, Eng. ear (of grain), PIE *h2ek^-er-a:-. Slavic and Baltic have a- in the underlying adjective and all other derivatives: OCS ostrU "pointed", Li. astrùs "idem". PS eseti-, P d. jesiec´ "grain sieve", R oset' "grain rack". PB eketia:- "harrow", OPr. aketes, La. ece:s^as, Li. ake.´c^ios, d. eke.´c^ios.
> PIE *h2ek^-.
> Cf. OHG egida, Lat. occa (< *oteka: < *oketa:), Gk. oksina (Hesych.).

Again, one would expect *-tt- from *-ty- if we were dealing with Ill.-Lus. forms.

> Table 6: PS, PB e- for PIE *h2e-, *h3e-'
>
> It is interesting that the PB and PS "lead(Pb), etc" word seems to have the same relationship to the South and West word for the same metals
> http://tech.groups.yahoo.com/group/cybalist/message/36844
> http://tech.groups.yahoo.com/group/cybalist/message/36854
> http://tech.groups.yahoo.com/group/cybalist/message/36874
> as the Northern "apple" words have to Latin ma:lum etc.
> http://tech.groups.yahoo.com/group/cybalist/message/64469
>
> Should a common "lead(PB), etc" be reconstructed *aNlau- instead?

Friedrich in _PIE Trees_ (1970) considered and rejected connecting the two 'apple' words (and <Abella> in Campania, famous for filberts, has nothing to do with the Northern 'apple' word anyway).

In my opinion the Greek 'lead' word variation results from Pre-Greek dialectal differentiation plus crossing with <boli's> 'glans missilis; fishing sinker' (both commonly made of lead), thus the /m/-forms are the older. I derive Lat. <plumbum> from Etruscan *plumpe, extracted from *plumpna 'lead-works', Old Etr. *p(u)lumpt(i)na, from NW Greek *bolu'mbdina neut. pl. tant. 'id.' Basque <berun> can also be derived from Greek, but not the Celtic or Germanic words. It seems futile to lump together 3 or 4 distinct words for 'lead'.

DGK