For the correspondence pattern TA tsp- ~ TB ts-, Adams (XXXX) notes the correspondence patterns TA tp- tsp- ~ TB t- ts-, and reconstructs PIE *wi-T-:

A tspäṅk-, B tsäṅk- ‘flay’ < *wi-tengʰ- (cf. OCS *tęgnǫti ‘pull’, Avestan θang- ‘pull’)
A tpuk-, B tuk- ‘be hidden, hide oneself’ < *wi-dʰeuk- (with metathesis of *keudʰ- also seen in PGmc. *daugana- ‘conceal oneself’)

A tspok ‘taste’, a verbal noun associated with the verb AB tsuk- ‘suck (out)’ (in which the *p was lost before -u-) < *wätsäuk- < *wi-deuk- (cf. Alb. nduk ‘pluck, pull out the hair, [dialectally] ‘suck’, Lat. dūcō ‘lead, pull’)

This is similar to the process Manaster Ramer (XXXX) calls “Hamp’s Law” in Albanian, namely PIE *wIḰ- > Alb. z-:

Or, following Manaster Ramer’s suggestion, *wIK-, where K is a palatovelar or a velar:

zëmër ‘afternoon meal’ < *wik Ha(ː)mVr “when the time is ‘the-hot-part-of-the-day’” (presumably rather *wik-Héh2mr̥; cf. Gk. ἦμαρ)

zemër ‘heart’ < *Hwi-kemH-r “(something coming) out of a (belt-like!) tight space”)

These sound laws can only be similar, not a shared inheritance: Hamp’s Law in Albanian operates on sequences of (word-initial) *wi- followed by a palatovelar, whereas the analogous law in Tocharian operates on *wi- followed by a coronal. But, as Orel (XXXX) points out, the operation of the law in Albanian can be explained with intermediate forms *wdz- > *dzw- > *dž, since inherited *ḱw- *ǵ(ʰ)w- also become alveolar sibilants in Albanian:

However it’s stated, Hamp’s Law requires voicing of the reflex of PIE *ḱ—but so does its Tocharian analogue, since the reflexes of *wi-tengʰ- in both TA and TB begin with ts- (< PIE *d-) rather than c- (< Pre-PToch. *tʲ-), t- (< PToch. *t-), or A ś-, B ts- (< Pre-PToch. *dʲ). Without irregular developments or interdialectal loans, the required form is PIE *wi-tn̥gʰ- with a zero-grade.^[1] The similarities suggest the possibility of diffusion, presumably in the form *wiT(S)- > *Dβ-,^[2] which implies close contact between Albanian and Tocharian at some stage; and because PToch. *wʲ > TA y but TB w, TB ytārye ‘road’ < *i̯ätār- < *h1i-tōr suggests that Hamp’s Law followed the fronting of palatovelars in Albanian, but preceded the emergence of palatalization in Tocharian.^[3]

At this point, there are three problems. First, Manaster Ramer’s treatment of Hamp’s Law includes also *wiK- (with a PIE velar) > Alb. z-, which has no apparent Tocharian correspondence, and which would suggest a pre-satemization date. Second, while the Tocharian forms seem secure, the Albanian examples of Hamp’s Law are questionable: all except -zet have alternate etymologies. Third, TODO

The second problem can be used to address the first, and at least zëmër ‘afternoon meal’ has inescapable issues: Manaster Ramer compares his *wik-Héh2mr̥ to Gk. ἕσπερος, but this has *we- rather than *wi- or *wey-. Orel’s suggestion of a borrowing from Gk. διήμερον ‘period of 24 hours’ is entirely phonologically regular (cf. gaz < gaudium, etc.), and the semantic development of ‘period of 24 hours’ > ‘period when the sun is up’ > ‘afternoon’ > ‘afternoon meal’ seems unobjectionable: cf. Russian ужин ‘dinner’ (a derivation of *jugŭ ‘south wind’) for the last part, and English day for the rest. Lacking support from zëmër, Manaster Ramer’s extension of Hamp’s Law to zemër ‘heart’ must also fall; this word is difficult, but Huld (XXXX) suggests, following Jokl, a derivation in *-mer- from PIE *gʷʰen- ‘swell’.

With Manaster Ramer’s extensions of Hamp’s Law rejected, we now turn to the standard examples, -zet, zot, and zonjë. (Hamp also suggested zog < *h2wi-ḱe-gwo-, in a paper titled “A far-out etymology” in the Festschrift of a Nostraticist, but Manaster Ramer correctly identifies this as a joke. At any rate, there are other options, such as Orel’s proposal of a Wanderwort connected to Armenian jag ‘chick’.)

Zonjë forms a gendered pair with zot, and can be interpreted as a straightforward derivative of it, with a -një element (presumably PAlb. *njā < PIE *-nih2) also seen in drenjë ‘doe’ < dre ‘deer’, enjë ‘dairy goat’ < *h2egʷ-nih2 (cf. Gk. *ἀμνός ‘lamb’, Lat. agnus ‘id.’ < *h2egʷ-nós), bilonjë ‘twig, branch, pretty girl or young woman’ < PAlb. *būlā-njā ~ Gk. φύλλον ‘leaf’, etc.^[4] Since *-njā may have remained productive into the Proto-Albanian period, any etymology of zot can also serve as an etymology of zonjë.

As for zot, Manaster Ramer correctly points out that Orel’s simple equation with Skt. viś-pati- (implying PIE *wiḱ-poti-) is invalid: not only is the *p troublesome, PIE *o should give Alb. a. Hamp’s proposal was instead *wiḱ-eh2-[s]-pti-, with the difficult replacement of the original genitive singular *wiḱ-ós with an ā-stem form *wiḱ-eh2-s. Manaster Ramer’s *wiḱéh1-p(o)ti- with an instrumental seems preferable, but better still would be a connection to *dyēw- > Lat. diēs, Skt. dyáuṣ-pitṛ́. (The long vowel is necessary; *dyew-t- would give **zet.)

Unfortunately, Huld rejects this equation as “without merit” with no further explanation. If the objection is that the semantic development from ‘lord’ to ‘god’ should be one-way, consider e.g. Mandarin huángdì and Japanese tennō ‘emperor’, as well as the phrase “God-Emperor Trump”—although these developments are not entirely unconnected, since the term ‘god-emperor’ passed into pop culture (Warhammer 40K < Dune) from its use as a WW2-era translation of tennō. If Huld’s objection is instead that a derivation from *dyēw- can’t explain the final -t of -zot, one possibility is the semantically repulsive but mostly regular^[5] *dyēw-atta ‘sky daddy’, or Demiraj XXXX’s similar suggestion of a later compound of an unattested *zo with atë ‘father’—and, notably, zot and atë decline identically except for umlaut in the latter.

This leaves only -zet. It would, at this point, be possible to reject Hamp’s Law for Albanian on the grounds that a sound law needs more than one witness; but since there are no other proposals, the attraction would remain, and -zet could, after all, be the only inherited Albanian word that reflects PIE *wiḱ-. So one last attempt at re-etymologization is necessary.

Albanian shows *st > sht, but *sd > dh^[6] (e.g. pidh ~ Rus. пизда́), presumably with the intermediate step of metathesis to PAlb. *dz.^[7] Since *ḱ > PAlb. *ts, the development of a hypothetical word-initial cluster *ḱd- > *tsd- to *sd- > *zd- would only require deaffrication, which raises the possibility of a form *h1ḱ-dwoh1-t- with an unclear *-t- element (possibly related to the -të of dhjetë ‘ten’^[8]) and *h1ḱ- < *dḱ, the zero-grade of the bare root *deḱ- (> *déḱm̥t ‘ten’) also seen in PGmc. *teguz > Eng. -ty: the regular development of TODO

Sequences of two plosives normally lose the first element (natë ‘night’ < PAlb. *nakti-, tetë ‘eight’ < PAlb. *aktōti, etc.), but it’s possible that the initial *ḱ- was protected by a morpheme boundary or an epenthetic vowel that was later lost, perhaps in conjunction with deaffrication. However this form *zdwō-t- is produced, its further development is regular: *zdwōt- > *dzwōt- > *džōt- > *džøt- > zet.

[1] The pathway *wi-t- > *wʲä-t- > *tyβ- works for tspäṅk-, but not tpuk-; PIE *dh- before a palatalizing vowel should give **c-.

[2] *-β- > PAlb. *-w-, TA -p- (possibly [ɸ], as Adams noted), TB 0. Surely not *-w-, given TB twere ‘door’ < *dʰworo-. This marginal [β] would then have survived long enough to have distinct reflexes in both TA and TB, but the requirement of preservation of the marginal is not a fatal objection: in the common ancestor of the major dialects of English, [ɑ] only existed in ‘father’, in words ending in -alm, and before a syllable-final /r/, but it survived long enough to be reinforced by ɒ > ɑ in some descendants and the development of nonrhoticity in others.

[3] Alternatives exist: a development in TB of specifically PToch. *wʲ- > 0 / #_C̥ (or some such environment), or irregular reduction of the (presumably unstressed) grammatical element *wi- to non-palatalizing *wä- (cf. English auxiliary ‘will’ > [wɫ̩] ~ [wʊˤ]), loss of PToch. *w- analogously to to the proposal for *wʲ- above, and alternate explanations for TB words in which w- appears word-initially before a plosive—analogical restoration in transparent derivatives such as wpelme ‘(spider’s) web’ < wāp- ‘weave, braid’? The wC̥- words in Adams XXXX seem easier to deal with than ytārye, at least.

[4] Demiraj XXXX notes the dialectal form zótnië with preserved -t-, although in principle this could be secondary.

[5] Maybe not entirely—there could be an objection based on stress rules, and, although Orel says that *VwV > ua, ue, ye, or y as a rule, I know of no examples of specifically *ēwa. *ewa with a short vowel and hiatus resolution (instead of uncompensated loss of *e) could be tested against Orel’s statement that hiatus resolution should produce a long vowel in Gheg. The assumption of a later compound (i.e. Late PAlb. *djē-atë > *zētë) avoids all such difficulties.

[6] Or th when preceded by voiced consonants, as in gjeth ‘foliage, green leaves’ ~ ON kvistr ‘branch’.

[7] Orel XXXX mentions that anlaut *sk commonly metathesizes to PAlb. *ks > Alb. h, and that *ps and anlaut *sp both give f.

[8] The final schwa is difficult, but the Albanian numerals from six to ten end in -të, and all but ‘two’, ‘three’, and ‘four’ end in -ë—so interference of some sort is possible, and could simply not have reached higher and non-adjacent numerals. It’s also possible that -zet lost its schwa. Any irregularity would have to be universal among the dialects Huld lists.

Posts