Re: Crist's /z/-deletion (was: Gmc. w-/g-, j-/g-)

From: dgkilday57
Message: 68217
Date: 2011-11-18

--- In cybalist@yahoogroups.com, "dgkilday57" <dgkilday57@...> wrote:
>
> [...]
>
> The paper can be found by searching for "crist_z_loss".

Yesterday's post requires some additions and corrections, which I have inserted today.

> Sean Crist should have consulted the OED and Skeat's etymological dictionary before submitting this paper. Both these sources and Feist, "Die sogenannten reduplicierenden Verba im Germanischen" (PBB 32:447-516, esp. 498, 502-3 [1907]), have additional material relevant to the /z/-deletion problem.
>
> Crist argues that the OHG words for 'meed' were borrowed from Ingvaeonic. I agree, and it must have happened at least twice, before and after the High German consonant-shift. In addition to the forms which Crist cites from Torp, Feist has OHG <meida>, with the same vocalism as OFris <meide>. This brings up another matter. Crist's /z/-deletion and compensative lengthening must follow Ingvaeonic lowering of */i/ to */e/ before */z/, as he notes in §11 (wrongly marked as a second §10), in order to get OE <me:d>, OFris <me:de>, and OS <me:da> out of Gmc. *mizdo:. But OS <li:no:n> 'to learn', from Gmc. *lizno:n, flies in the face of this chronology, and moreover OE <leornian> did not delete /z/. I propose the following modifications:
>
> 1. Crist's /z/-deletion between a high vowel and a consonant, with compensative lengthening of the vowel, is regular and went to completion in Old Saxon, and spread into some of the Old Frisian dialects, but did not reach Old English. The original quality of the high vowel was maintained. Thus, OS <mieda> and OFris <mi:de> are regular outcomes of Gmc. *mizdo:, but so is OE <meord> (earlier *miord). Likewise, OS <li:no:n> and OE <leornian> are regular outcomes of Gmc. *lizno:(ja)n, while OFris <lirna>, <lerna> (wrongly replaced by the OHG forms in §2 and §9, but correctly cited in the Appendix) are from dialects closer to OE which did not undergo Crist's deletion.
>
> 2. OE <me:d>, OFris <me:de>, and OS <me:da> do not come from Gmc. *mizdo:, but from the Nordwestblock cognate, in which */z/ had already been deleted after */i/ had been lowered. Old (East?) Frisian <meide>, which was borrowed into OHG after the second consonant-shift as <meida>, reflects a late diphthongization in an OFris dialect. The alternative is to invoke dialectal differentiation in NWB, which would amount to "obscurum per obscurius".
>
> My proposal (1) is practically the same as Crist's first possibility for explaining <leornian> (§9, top of p. 4). His second possibility, analogical restoration of */z/ from the related causative 'to teach' (OE <læ:ran>, OFris <le:ra>, which at the time still had */z/) loses plausibility when one considers OE <weornian> 'to pine away', from Gmc. *wizno:n, briefly discussed by Skeat s.v. <wizen>. No OE causative *wæ:ran 'to cause to pine away' is attested; the closest relative to <weornian> is <wisnian> 'to shrivel, dry up, wizen'. It defies all sense to imagine early OE-speakers (or Anglo-Frisians or whatever) restoring */z/ to *wi:no:(ja)n (and decompensating the long */i:/ too!) on the basis of the /s/ in <wisnian>, even if they all knew about Verner's Law a millennium and a half before Verner.
>
> Linen, canvas, and rope are trade items, so regarding OE <twi:n> as a borrowing from Old Saxon (or southern Old Frisian) is as good as Crist's dialectal explanation (top of p. 4). Whitehall (WNWD s.v.) takes the English verb <twine> as a probable borrowing from Dutch <twijnen>, not an inherited word like Torp, and <twijnen> can easily be regarded as the outcome of OS *twi:no:n, regularly from Gmc. *twizno:n by Crist's deletion. Köbler (on-line Ae. Wb. s.v.) and some other scholars explain OE <twi:n> as the neuter of a Gmc. adjective *twihnaz beside the secure *twiznaz, but this ad-hoc form is difficult to justify. Latin <bi:ni:> 'two each' corresponds exactly to *twizna- from IE *dwisnó-, but there is no Lat. *bigni: corresponding to *twihna- from *dwík(^)no-, and I am not aware of any parallel numerical adjectives elsewhere.
>
> The OED gives OE <weard>, <werd> as glosses for 'sandix', which is also the usual way of glossing 'woad': "Hic sandyx, þis wad" (Ælfric, c. 1000). "Sandix, wod" (early ME voc., c. 1200, cited from Wright-Wülcker). The connection between <weard>, <werd> and <wa:d> is considered doubtful in the OED, but it is too similar to <meord> against <me:d> to dismiss. Medieval Latin forms like <ouisdelem> acc. sg. 'woad', said to be from Gothic, go back as far as the 6th-cent. Oribasius Medicus, from which Souter (Gl. of Later Latin) cites <guisdil> or <auisdil>. Like modern Rx-men, the compiler of the Oribasius was not noted for orthographic accuracy or neat handwriting, and <ouisdil> may have been intended. At any rate, since we cannot know the exact source of the ML words, or how far they were deformed before scribal usage stabilized them, Crist is correct in rejecting the supposed Gothic zero-grade *wizdila as a basis for usable conclusions.

Torp, Feist, Crist, and I neglected to consider the possibility of /e/-grade in this purported Gothic word. Gothic regularly raised inherited */e/ to /i/ unless /r/, /h/, /w/, or /hW/ immediately followed. (The reduplicating prefix of the class VII strong preterit never shows raising, presumably because the /e/ was restored by analogy every time the raising, which acted in stages, moved on to a new preconsonantal environment.) Moreover, Common Gmc. raised /e/ to /i/ when a high front vocoid (/i/, /i:/, /j/) occurred in the next syllable. Thus, associated with the Gmc. verb *bregdan 'to brandish, pull, weave' (OE <bregdan>, MnE <braid>) we have the noun *brigdilaz 'device for pulling a horse's head, bridle' (OE <bri:del>, early OE <brigdils> (Erf. Gl.)). Possibly the ML forms of 'woad' then represent Goth. *wizdils, Gmc. *wizdilaz, implying a verb *wezdan (Goth. *wizdan).

> On the other hand, Anglo-French <waisde> (13th cent.) and Old French <guesde> (later <waide>, <gaide>, <voide>, etc., now <guède>) were recorded as living forms, and suggest that Frankish also had a form with *-zd- from which they were derived. This form was apparently *wazd- (cf. OFr <guespe>, Fr. <guêpe> 'wasp' from Frk. *wasp-, but with Fr. -ê- from *vêpe, Lat. <vespa>, replacing expected -è-). Incidentally, the words which Crist could not find in dictionaries, Da. <vaid>, <veid>, Sw. <veide>, are stated by Skeat to be borrowed from German, which nullifies Crist's objection of expecting NGmc *-dd- from *-zd-. Presumably these words became obsolete two or more centuries ago. (Torsten?)

Thanks for the information, Torsten. I did not want to think that Skeat would cite phantom words.

The form *wazd- would have to be very early Frankish for the /z/ to remain unrhotacized. Perhaps *wazd- also came from Gothic, showing /o/-grade of the Indo-European root.

> If OE <weard> is West Saxon, it also reflects Gmc. *wazd-, and <werd> could be from a non-WS dialect or the result of sloppy transcription. Thus (under reasonable assumptions) we have to do with Gmc. protoforms *wazd- and *waid-; there is no basis for *waizd-, and *wizd- is unusable. My view is that *wazd- is the regular Gmc. form of 'woad', while *waid- is the NWB cognate, with /z/-deletion and compensative diphthongization preceding borrowing into Ingvaeonic. I do not share Crist's objection to 'woad', like 'meed', being further borrowed into High German. Dyestuffs and the plants used in making them can certainly be trade items.
>
> If *wazd- is formed like *mizd-, Grk. <misthós> etc., presumably it has zero-grade of the root, which would then be *weh{x}s- with undetermined laryngeal. I have no further etymological ideas.

But if *wizdil- represents Gmc. /e/-grade, *wazd- Gmc. /o/-grade, and *waid- the NWB form of /o/-grade, we could be dealing with IE *wezdH- 'to put into clothing, to dye', from *wes- 'to be clothed' plus the extension *-dH- 'to put, place' (cf. *steh2- 'to be standing', *steh2dH- 'to place standing'). In this view Gmc. *wizdilaz originally meant 'material useful for dying' and referred to the plant itself, while Gmc. *wazdam (and NWB *waida-) meant 'substance applied in dying, dye' and referred to the finished product. But such distinctions need not have been maintained.

> I agree with Crist in reconstructing *hezdo:n, not *hizdo:n, as the Gmc. protoform of OE <heorde> (used in pl. <heordan>, glossed 'stuppa', i.e. 'hards of flax, fibrous residue'). Skeat cites Middle (actually Early Modern) Dutch <heerde>, <herde> (Kiliaen), later <he:de> (Hexham), also East Frisian <he:de>, which is identical with Crist's OFris form. MnHG <Hede> beside the usual <Werg> is yet another Ingvaeonic loan. Like OE <heordan>, <he(e)rde> must be from a dialect which the /z/-deletion did not reach. The IE root is evidently *kes- 'to comb, card', and the noun may have taken the form *kézdHoh1n, *kézdHeh1n- (cf. the Sabine river <Anio:>, acc. <Anie:nem> and war-goddess <Ne:rio:>, acc. <Ne:rie:nem>; elsewhere the gradation has been contracted or levelled out).
>
> Since Crist recognizes that non-final /z/-deletion never occurs after a back vowel in WGmc, he is not guilty of connecting OE <ha:d> n. 'hair' with ON <haddr> m. 'a woman's long hair' (Gmc. *hazdaz, IE *koz-dHo-), but Campbell (cited in §7) and Köbler are. I believe we should refer OE <ha:d> to Gmc. *haidam, IE *kh{a}i-tóm or *kh{a}i-dHóm, related to Lat. <caesar> 'full head of hair' (with its intervocalic -s-, obviously a Sabinism), <caesarie:s> 'bushy-hairedness; bushy head', Skt. <késara-> m./n. 'hair, mane' (with -s- for expected -s.-, probably borrowed from Middle Indic; cf. Uhlenbeck, Kgf. et. Wb. der ai. Spr. s.v.), all reflecting an IE root *keh2i- or *keh4i- (i.e. *keh{a}i-, traditionally *ka:i-) but with different formations. Morphology of the Sabine and Indic words is difficult, and my best guess is that the second element is based on a root-noun from IE *sreu- 'to flow' from which a thematic noun was extracted in Indic, yielding *kóh{a}i-sara- 'hair-flow' > 'flowing hair, mane'. Italic operated with the zero-grade of the first element, getting something like *kh{a}i-saru-, Sab. nom. sg. *kaisars. This attempt to explain the non-Gmc. words is admittedly sketchy and unsatisfying, but has no bearing on the fact that OE <ha:d> can be assigned to another root than *kes-.

The Indic protoform with /o/-grade cannot be correct. Lubotsky has shown that laryngeals triggered Brugmann's lengthening in Indo-Iranian before they disappeared between vowels, so *-oh{x}i- would eventually yield the long diphthong -ai- in Sanskrit. Since <késara-> is very likely Middle Indic, its recessive accent need not be original, and its first syllable probably reflects zero-grade *kh{a}i- like the Sabine word.

> In the Appendix (p. 8), Crist questions how Torp can assign the OE compound <wunden-heord> 'having curled hair'(?) to Gmc. *hazdaz without a high front vocoid to produce umlaut before breaking. Köbler cites (apparently) the same word as <bunden-heord> 'having bound hair', so it appears that scholars now read <b> instead of <w> in the MS. Regarding the umlaut, all I can suggest is that the Gmc. compound adjective was an /i/-stem, *bundana-hazdi-, even though the simplex *hazdaz was an /a/-stem. This sort of thing is common enough in Latin, where we have (for example) <arma> 'arms, weapons', an /o/-stem noun, yielding /i/-stem adjectives like <inermis> 'unarmed' and <se:m(i)ermis> 'half-armed'.
>
> Further down on p. 8, Crist thinks Torp may have accidentally written <Quelle> intending <Welle>. But Greek <kré:ne:> (Doric <krá:na:>), which Torp considered cognate to Gmc. *hrazno:, is rendered very well by 'Quelle, source, spring'. Mallory and Adams have it both ways, hedging their bets by rendering IE *kr.sneh{a} as 'spring, wave'.
>
> Finally, Crist is completely right about OE <ce:n> 'pine, fir, spruce' and its WGmc cognates. Since this word has no secure relatives outside Gmc., there is no basis to suppose it underwent any /z/-deletion.

Unless it came into Gmc. from the Nordwestblock. By Kuhn's original hypothesis, the NWB language did not undergo Grimm's consonant-shift, but I have argued elsewhere that it probably had */a/ for IE */o/, and voiced stops for IE mediae aspiratae (e.g. */d/ for IE */dH/), like the Messapic language, and belonged to the centum group. And NWB loanwords into Ingvaeonic postdated Grimm's shift. Therefore, Russian <sosná> 'pine' (IE *k^osnó-, *k^osnéh2-) could be cognate with Proto-NWB *kazna-, which would yield NWB *kaina- if it agreed with the sound-changes I consider necessary for 'woad', IE *wozdHó- > PNWB *wazda- > NWB *waida-. If *kaina- then entered Ingvaeonic, it would become Torp's MLG <ke:n> (in Crist's §10), which agrees with OHG <ke:n> (presumably an Ingv. loan). The other HG forms, OHG <kien>, <chien> and MHG <kien>, have the vocalism of <Kiefer>, a much more common word for 'pine' with which the Ingv. loan presumably crossed. OE <ce:n>, in place of expected *ca:n, is probably borrowed from OS or OFris rather than continuing the direct NWB loan.

DGK