--- In
cybalist@yahoogroups.com, Piotr Gasiorowski <gpiotr@...> wrote:
>
> On 2009-05-28 23:10, dgkilday57 wrote:
>
> > Old Norse <taka>, <to:k>, <tekinn> 'take' and Gothic <te:kan>,
> > <tai'to:k>, <te:kans> 'touch' are problematic since we don't expect PIE
> > roots of the form *deg- or *deHg-. Simple ideas occurred to me and I'd
> > like to know if they can be easily dismissed, so I don't waste time
> > pursuing faulty hypotheses.
>
> For my view, see:
>
> http://tech.groups.yahoo.com/group/cybalist/message/45736
Having looked at Kortlandt's take on 'take' in more detail, I find more difficulties. As already noted, the assimilation of reduplication in Latin <bibit> against Vedic <píbati>, Old Irish <ibid> 'drinks' from *píbeti (earlier *pi-pH3eti, root *peH3-) is not a good parallel to the last stage of the development postulated by K. for pre-Grimm's Law Germanic, namely *tetH2g- > *tetg- > *tedg- > *dedg-. Latin, unlike Greek and Sanskrit, rejected reduplicated verbal forms in which soundlaws had regularly produced different single consonants. Thus Latin has <fe:cit> 'made' instead of the *fedicit expected on the basis of the Oscan 3sg. perf. subj. <fefacid>, and <finxit> 'fashioned' instead of the *fidigit expected from the Faliscan 3pl. perf. ind. <fifiqod>. Faliscan, whose alphabet uses <p> to denote both /p/ and /b/, has two forms of the 1sg. fut. 'I will drink', both used by the Foied painter, <pafo> and <pipafo>. Of these, <pafo> agrees in form with Lat. <dabo:> 'I will give', while <pipafo> has a reduplicated stem like that of Vestinian <didet> 'gives' which is obsolete in Latin as the simplex, but continued in <reddere> from *re-didere. Thus it appears that Fal. <pafo> is the earlier future form, with <pipafo> acquiring reduplication from the inherited present stem; we may compare the Epic Greek fut. <didó:so> with the usual <dó:so:>. We cannot know whether <pipafo> was pronounced /pibafo:/ or /bibafo:/, but we can be reasonably sure that <pafo> was pronounced /bafo:/, since it reflected *p&3-bho: the way Lat. <dabo:> does *d&3-bho:. Latin replaced *didere, *didit, etc. with zero-grade <dare>, <dat>, etc. agreeing with the future and imperfect, but instead of replacing *pibere with *bare, it assimilated it to *bibere. Probably this happened at a time when *didere, etc. were still current, so the relation between *didere, *didit and <dabo:>, <dabis> provided the model for assimilating *pibere, *pibit to <bibere>, <bibit> beside *babo:, *babis. Then the verbs went their separate ways, the inherited present *didere, *didit yielding to the shorter <dare>, <dat>, and the inherited future *babo:, *babis yielding to the new forms <bibam>, <bibe:s> which rendered <bibere> a regular 3rd-conj. verb.
In Kortlandt's scenario, the reduplicated aorist *tetg- and nasalized present *tagna:- are part of the same paradigm arising from *teH2g- 'to touch', with *tagna:- regularly becoming Gmc. *þakko:- by Grimm's and Kluge's Laws, and reflected in Old English <þaccian> 'to pat'. Now, K. argues that if *tetg- became *tedg-, further assimilation to *dedg- has a "perfect analogue" in Latin <bibit> from *pibit. In fact there is no analogue at all, since the only plausible basis for this unique Latin assimilation is the model of *didit with <dabo:> against *pibit with *babo:, and K. has not provided any model paradigm which would encourage *tedg- to become *dedg-. Instead *tedg- has an analogue in the situation which must have occurred at some stage of Proto-Latin when the perfects *dhe-dhaked and *dhi-dhiged developed into *fedaked and *fidiged, upsetting the usual pattern of reduplicated perfects. Rather than being retained in their old paradigms, or forming suppletive paradigms (like K.'s scenario for later *te:k-), these unusual perfects were simply discarded and replaced by alternative non-reduplicating forms, *fe:ked and *fingsed, leading to <fe:cit> and <finxit>. I believe *tedg-, if ever there was such a form, met a similar end.
In order to explain the long vowel of Gothic <te:kan>, Kortlandt brings out a theory on which he has written extensively, in which PIE voiced unaspirated stops were preglottalized, and this feature was retained even after Grimm's Law devoiced the stops, indeed retained by some modern English and Danish dialects. Whether some modern Gmc. dialects actually preserve preglottalization inherited from PIE is not something I wish to debate. I note here that his scenario has *dedg- ['de'd'g-] losing its second dental articulation to become ['de''g-], which he supposes equivalent to *deH1g-, thus regularly becoming *de:g-, after Grimm's Law finally *te:k-. The geminated preglottalization is ingenious but tough to swallow, as is the notion that while ['] was phonetically equivalent to [H1] everywhere, it was phonologically equivalent to [H1] only when preceding another [']. We are all in trouble if we need new soundlaws to explain every difficult word.
Anyhow, in Kortlandt's scenario Gothic created a new 7th-class present <te:kan> from the isolated aorist stem *te:k-, but in NWGmc *te:k- underwent "considerable pressure" to change into *to:k- in order to conform to the 6th-class pattern, even though he states that this class was no longer productive in late Proto-Gmc. Somehow the 6th class reached out of the grave to form *takan, *to:k, *takinaz reflected in Old Norse <taka>, <to:k>, <tekinn> which managed to displace *þakko:- here. In Middle Low German *þakko:- somehow adopted the t- from this new verb yielding *takko:-, realized as <tacken>. This cacophony of forms and replacements does not inspire confidence.
Finally, Kortlandt and other scholars seem oblivious to the serious semantic problems resulting from referring all these words to the same root *teH2g- 'to touch'. Verbs describing physical action tend to lose denotative intensity with time, not gain it. As their intensity wanes, verbs find themselves being collocated with adverbs or other complements, affixated, iterated, otherwise intensified, or replaced by vivid, novel verbs. The presumed semantic transition 'touch' > 'seize, grab, take' flies in the face of this experience. The OED suggests 'touch' > 'hold' > 'lay hold of', but the latter is intensified as a collocated phrase, not at all equivalent to 'hold', which is itself more intense than 'touch'. ON <taka> 'seize, grab, take' is a simple verb, and its intense sense could hardly have arisen spontaneously from 'touch'.
Given all these difficulties, I prefer to return to my earlier suggestion that ON <taka> has nothing to do with PIE *teHg-, but instead reflects *deH3- with a /w/-extension, the cluster */Hw/ being fortited in NWGmc as proposed by Lehmann. For our purposes the best illustrations are provided by the lexemes for 'alive' and 'riverboat'. PIE *gweiH- often has a /w/-extension which appears with the zero-grade in Latin <vi:vus>, Oscan nom. pl. <bivus>, etc., but OE <cwicu>, ON <kvikr> beside Goth. <qius> seem to require a PGmc *kwikwaz beside *kwiwaz (which moreover ought to be *kwi:waz). This set is explained if we assume that intervocalic */Hw/ became */kw/ in NWGmc, but */w/ in EGmc. Likewise OE <naca>, OS <nako>, OHG <nahho> presuppose NWGmc *nakwan, *nakwo:n which Pokorny refers to a doubtful root *nogWo-/*nagWo- 'Baum?' whose only support outside Gmc. is Skt. <nagah.> 'Baum, Berg'. If we again assume */kw/ from */Hw/ we can derive these 'boat' words from PIE *neH2w-o(:)n-, the root colored to *naH2w- before the fortition to *nakw-. However, the same PIE root *neH2w-, best known for root-nouns (Skt. <náuh.>, Grk. <naûs>, Lat. <na:vis> with secondary /i/-stem declension), also underlies some NWGmc words without this fortition. ON <no:r> m. 'ship' apparently reflects *no:waz, either thematized from the Gmc. root-noun *no:wz or continuing *na:was thematized from the root-noun *na:ws before the characteristic Gmc. shifts, in any event showing that */s/ inhibited the clustering of preceding */Hw/ into a phonetic unit, allowing the */H/ to lengthen the preceding vowel instead. ON <No:atu:n> 'Njord's dwelling' i.e. 'Shiptown' contains as first element the gen. pl. of either *no:wz or *no:waz, and OE <no:wend> 'sailor' must be an indirect derivative of the root-noun. ON <naust> 'boathouse' appears to contain the reduced grade *nau < *n&2w- as first element, showing that a sonant laryngeal does not unite with following */w/.
Now, the root *deH3- means 'give' in most branches of IE but 'take' in Hittite, where the 1sg. <da-ah^-h^i> /dah^h^i/ and 3sg. <da-a-i> /da:i/ are attested; it seems otherwise absent from Germanic. The semantic shift 'take (for someone)' > 'give (to someone)' is not difficult and is found with other words for 'give'. Hence there is no serious objection to regarding 'take' as the original sense of *deH3-, preserved in Gmc. and Anatolian. As with *gweiH-w, forms derived from extended *deH3-w- are known in several branches of IE. If this was the only form of the root available, it would effectively be the root in pre-shifted "Protero-Germanic", and if the verb formed a zero-grade root-present, it would fall into the 6th strong class. The pre-shifted principal parts would be *d&3won, *doH3w (cf. Skt. 1/3sg. perf. <da-dáu> 'I/he have/has given'), *doH3wn., *d&3wonos, leading to EGmc *tawan, *tow, *towun, *tawanaz and NWGmc *tawan, *tokw, *tokwun, *tawanaz. This unusual paradigm would be unlikely to survive for long. Probably the preterit vowel would quickly be lengthened to correspond to other 6th-class verbs. Then the 1/3sg. pret. (the most commonly used form, 'I/he took') would be reduced in NWGmc from *to:kw to *to:k, the result being generalized throughout the preterit. Finally the remaining principal parts *tawan and *tawanaz would be replaced by *takan and *takinaz after the preterit, leading to the regular 6th-class paradigm reflected in ON <taka>, <to:k>, <tekinn>. Competition with other verbs based on Gmc. *taw- 'make, fashion' likely killed off EGmc *tawan 'take' entirely and helped suppress NWGmc *tawan, *tawanaz. OE <tacan>, <to:c> do not occur until very late and are clearly borrowed from ON. Also the ON deverbal noun <tak> 'taking, take' was borrowed as OE <getæc>, and I suspect it made its way southward as well, forming the denominatives *takjan (Old Frisian <tetsia> 'to appropriate') and *tak-sko:- (OHG <zasco:n> 'to rob').
For Gothic <te:kan>, I feel compelled to abandon my earlier suggestion of derivation from a Schallwort as far-fetched and inherently untestable. An Indo-European explanation is preferable and I now agree with Kortlandt and others that <te:kan> is indeed related to Lat. <tangere>, but as detailed above I disagree about the mechanism by which /t/ and /e:/ got into Gothic. Regarding the /t/, everyone seems to have overlooked the fact that the simplex <te:kan> is not even found in the Gothic Bible, only the compound <atte:kan> which regularly renders Greek <háptesthai>. I propose that this compound was not put together in Gothic, but inherited from pre-shifted Protero-Gmc. as a compound, independently of the simplex *tagna:- > Gmc. *þakko:-. The prefix *ad- < *H2ad- < *H2ed- would have undergone assimilation here as it does in Italic and Phrygian, thus *H2ad-te:gon > *H2atte:gon, and the -tt- would have escaped shifting, hence Gmc. *atte:kan. With no simplex in use and *þakko:- not recognizably related, there would be no basis for analogical restoration of anything.
Regarding the /e:/, the only thing supporting the reconstruction of 'touch' as *teH2g- is the connection, accepted by most scholars, of <tangere> with the Epic Greek 2nd aor. part. <tetagó:n>, usually rendered as 'having seized'. We have already seen the implausibility of the semantic shift 'touch' > 'seize', so this participle must be examined. It occurs in the phrase <rhîpse podòs tetagó:n>, referring to a clear physical act. If someone is to be thrown by the foot (more accurately the ankle; this use of <poús> is parallel to that of <kheír> in the sense 'wrist'), the crucial thing is a strong grip, so this phrase is well rendered as 'he threw (him) having grasped (him) by the ankle'. The zero-grade 2nd aor. stem <tag-> could represent *tn.g- as well as *t&2g-, since we have (for example) 2nd aor. <épathon> 'I suffered', 2nd perf. <pépontha> 'I have suffered', with full-grade *penth- in <pénthos> 'grief, sorrow, misfortune'. I propose that in fact <tetagó:n> belongs to the root *teng- 'to grasp', by a widespread metaphor also 'to comprehend, understand', whence in /o/-grade Lat. <tonge:re> 'to know, master, prevail over', Goth. <þagkjan>, OE <þencan> 'to think', in zero-grade Osc. abl. sg. <tanginúd> 'by opinion', Goth. <þugkjan>, OE <þyncan> 'to seem'. Thus the nasalized zero-grade of Lat. <tangere> need not be *t&2ng-, but could equally well be *t&1ng-, in which case the full grade *teH1g- and /o/-grade *toH1g- would regularly give the vocalism of Goth. <atte:kan> and <attaíto:k>, naturally falling into the 7th class like <le:tan>, <laílo:t> from *leH1d-. Recognition of the correct root as *teH1g- might also simplify the situation with Proto-Tocharian *te:k- 'to touch', but since I know little about Tocharian, I will not challenge what Kortlandt has written about its derivation.
MLG <tacken> 'to touch' remains to be addressed. The OED notes a current South Netherlandic <tokken> 'to touch' apparently derived from something like Old North French <toquer>, <touker>, from Common Romance *toccare, usually referred to a Schallwort *tocc 'knock!'. In the absence of a better idea, my guess is that the loanword ancestral to <tokken> crossed in MLG with *þakken, of identical meaning but derived regularly from Gmc. *þakko:-, thus producing <tacken>.
The other material is summarized as follows:
PIE *deH3-w- 'take' > NGmc *takan > ON <taka> > some WGmc words for 'seize', 'take', 'rob', etc.
PIE *teH1g- 'touch' > PToch *te:k-, Goth. <atte:kan>
PIE *t&1-n-g- > Lat. <tangere>
PIE *t&1g-n-eH2- > *tagna:- > PGmc *þakko:- > OE <þaccian>, etc.
PIE *teng- 'grasp, understand' > Lat. <tonge:re>, OE <þencan>, etc., Grk. *téggo:, <tetagó:n>
DGK