12.6.23.23:52: ALBANIAN GJ-ENESIS (PART 2)

In my last entry, I proposed that Albanian gj- [ɟ] was from *[ʒ] even though Albanian also has zh- [ʒ]. I realized that I could avoid alveopalatal [ʒ] in my derivation by replacing it with palatal [ʑ]. The relative chronology is not precise: e.g., the palatalization of *s- could have preceded or followed the fortition of *j-.

PIE *s- before an unstressed vowel became *ɕ- which then fronted to sh- [ʃ] instead of voicing to *ʑ- in stage 6. I still don't understand why stress conditioned voicing. And how could a monosyllable like shi 'rain' have sh- < *s-? I doubt that a content word like 'rain' was an unstressed monosyllable. Was it contracted from a disyllabic *sVCV́?

I also don't understand why PIE *gʷ(ʰ)- has two reflexes, gj- and z-. According to Beekes (1995: 261), gj- is secondary whereas z- is primary, so I had to derive gj- from *z- in stage 4 instead of directly from the cluster *gj- in stage 2. Neither Beekes nor Wikipedia give any examples of gj- from *gʷ(ʰ)-.

I assume the fortition of *z- was parallel to that of *s-. Both became palatal (stage 5) before hardening to a stop after stage 6.

I don't know where Albanian zh- [ʒ] comes from. I don't recognize any loanwords in the lists of zh-words in Hysa or in Wiktionary.

jay_albania_fan's gj-enesis

While writing part 1 of this series, I found this post and avoided looking at it until I was done with my own account of gj-enesis. jay_albania_fan got from *s- to gj- in 'six' in five stages (mostly rewritten in IPA):

*s- > *ʃ- > *tʃ- > *dʒ- > gj- [ɟ]

If this derivation is correct, modern Albanian

sh- [ʃ], ç- [tʃ], xh- [dʒ]

Çekosllovakia 'Czechslovakia' (ll = velarized [ɫ], not a palatal l as in Spanish!)

shah 'chess' < Ottoman Turkish şâh (< Persian shāh 'king'; the modern Turkish word for 'chess' is satranç < Arabic shatranj < Skt caturaṅga- 'four-limb')

xhaketë 'jacket'

xham 'glass' < Turkish cam 'id.'

xheloz 'jealous', xhelozi 'jealousy'

By avoiding sh-, ç-, and xh- in my derivation, I don't have to explain why they didn't also become gj-. But that doesn't necessarily mean my derivation is correct.

APPENDIX 1: A Similar Solar Situation?

The *s- > gj-/sh- shift may be somewhat parallel to

*sw- hardening to d- before stressed vowels: e.g.,

*swel- > diell 'sun' (cognate to Eng solar and Skt svar 'sun')

*swidro-tiā > dirsë 'sweat' (cognate to Eng sweat, Sanskrit svéda 'sweat', and Greek hydrṓs 'water')

but the Greek accent is not initial! - is Sanskrit more conservative?

*sw- > v- before unstressed vowels

*swe (a unstressed monosyllable?) > vetë 'self' (cognate to Eng self and Skt sva 'self')

The examples of *sw- are from Beekes (1995: 264). Is dirsë obsolete? I see it in works on historical linguistics, but other works have djersë for 'sweat'.

... is online but the pages are in the wrong order and some pages are missing. A shame.

APPENDIX 3: The Earlier Spellings of Gj

1814 gi

1855 g̅

1863 ǵ

1866 gki

1878 gy (cf. Hungarian gy for [ɟ])

Greek (no dates): γκι, γι, dotted γj, γj (gamma with a dot on top?)

Cyrillic (no dates): гї, гj, ђ (the same letter as in Serbian)

Arabic (no date): ك <k>

Thanks to David Boxenhorn who suggested that gj was once [gj]. Could gj have been [gj] or [gʲ] when some of these earlier spellings were devised? Could the *sw- > d- change be parallel to the *s- > gj- change?

*sw- > *dw- > d- before stressed vowels

but *s- > *hw- > *w- > v- before unstressed vowels

(There are no *dw-, *dv-, *hw, *hv- in the Albanian Wiktionary.)

*s- > *sj- > *dj- > gj- before stressed vowels

but *s- > *hj- > sh- before unstressed vowels

(Problem: dj- is possible in Albanian: e.g., djersë 'sweat' above. My scenario cannot account for this unless this dj- postdates the *dj- that became gj-:

Earlier Albanian	Modern Albanian
Something other than *dj-	dj-
*dj-	gj-

Table above added 6.24.11:19.)

12.6.22.8:02: ALBANIAN GJ-ENESIS (PART 1)

The Proto-Indo-European word for 'six' was *s(w)eḱs ~ *weḱs. Most Indo-European words for 'six' have s-like initials*, but Albanian gjashtë has the voiced palatal stop gj- [ɟ]. According to Beekes (1995: 261):

PIE *s- before unstressed vowel > Albanian sh- [ʃ]

PIE *s- before stressed vowel > Albanian gj- [ɟ]

Why would *s- undergo both voicing and fortition before a stressed vowel? In German and Dutch, *s- voiced to [z] (still spelled s- in German: G sechs [zeks], D zes 'six') but it didn't harden to [ɟ].

Albanian gj- has several other sources. All but *s- are voiced. Here's how they might have merged into gj-:

If that solution were correct, Albanian should have no initial [ʒ] since it hardened to [ɟ]. Yet Albanian does have initial [ʒ] (spelled zh-). Where does Albanian zh come from? I don't see it in Beekes' "From Proto-Indo-European to Albanian" after the initial table of consonant phonemes.

Next: Another Account of Albanian Gj-enesis

*Two major exceptions are Greek hex and Armenian vec'. PIE *s- became h- in Greek. I don't know whether Armenian v- in 'six' is from *sw- or *w-. Beekes (1995: 213) mentioned Old Prussian uschts 'sixth' but not Armenian vec' as a reflex of PIE *weḱs 'six'.

He proposed that the *s- of 'seven' may have spread to 'six'. Cf. how the initial of 'five' spread to 'four' in Germanic: four instead of *whour < PIE *kw-.

12.6.21.4:30: A CAESAR-I MYSTERY

In my last entry, I wrote that

Slavic c is from *k plus front vowels (but *ŭ is a back vowel!)

Slavic 'tsar' (a Latin loanword via some Germanic language) is an example of c < *k before the front vowel *æː:

(Modern Russian cezar' is presumably a later loan. There was no shift of -s- to -z- in Russian.)

Like 'father', Proto-Slavic *kaisarĭ ends in an unexpected -ĭ. Why wasn't it *kaisarŭ? Although Latin Caesar does have front vowels in its oblique endings

gen. sg. Caesaris

dat. sg. Caesari

acc. sg. Caesarem

abl. sg. Caesare

the nominative ends in -r. And only two out of eight forms of Gothic kaisar have front-vowel endings. Did Slavs borrow a Germanic form like Gothic genitive singular kaisaris as a nominative singular?

The Oxford English Dictionary lists other -i final versions of Caesar in its entry for tsar:

Old Norse keysari (with a front rounded vowel -y-!?; Wiktionary has keisari)

Finnish keisari (presumably through Old Norse)

(The OED also mentions Estonian keisri which is the gen. sg. of keisar.)

Oddly, Czech has car 'tsar' rather than *cař with an *-ř that I would expect from earlier *-rĭ. Polish similarly has car rather than *carz with an *-rz that I would expect from earlier *-rĭ.

Czech bohatýr 'hero' also has an -r that I assume is from *-rĭ (cf. Russian bogatyr' with a final palatalized -r' and older Polish bohaterz with -rz < *-rĭ as well as modern Polish bohater).

Czech does have words with final -ř: e.g., the loanword oltař 'altar' (cf. Rus altar' and Pol ołtarz) and bednář 'cooper' (cf. Rus bondar' and Pol bednarz). So I see three sets of correspondences:

The Polish words for 'hero' look like loans because they have -h- instead of -g-. *-g- to -h- occurred in Czech and Slovak but not in Polish. So maybe bohater is a borrowing from a language like Czech which had -r in 'hero'.

(Could Pol -h- reflect direct borrowing from Mongolian baɣatur? The first and last vowels in Pol bohater indicate otherwise. I would expect a direct borrowing to be *bahatur. The word is best known as the second half of Ulan Bator. The Mongolian word doesn't end in -i, so why does the Russian form end in -r' < *-rĭ? Analogy with other *-rĭ words?)

I thought Cz bednář 'cooper', etc. were native but Vasmer derived it from a Germanic *budin- (so where did the -ř come from?). Did *-rĭ in native words and early borrowings become -r in Czech? Is Czech -ř from *-rĭ in borrowings postdating the shift of *-rĭ to -r? One obvious counterexample is čtyř, the genitive of čtyři 'four', but that might be by analogy with all the other -ř-forms in its paradigm.

The correspondence of Cz -r (instead of -ř) to Pol -rz and Rus -r' reminds me of this correspondence in the infinitive endings of the three languages:

Cz -t (not -ť as in Slovak!) : Pol -ć : Rus -t', all < *-tĭ

Also note how Czech has -n (not palatal -ň!) < *-nĭ in den 'day' corresponding to Slovak deň, Pol dzień, and Rus den'. (How did Proto-Slavic *dĭnĭ 'day' develop *-ĭ? Its Sanskrit cognate dina- is an *-o stem, not an *-i stem. Lithuanian has diena, so *-i presumably cannot be reconstructed at the Proto-Balto-Slavic level.)

Did all early final coronals lose *-ĭ in Czech?

On the basis of the Slovak infinitive ending with -ť < *-ti, one might guess that Slovak has a special reflex for *-rĭ, but in fact Slovak has -r in

cár 'tsar' (with a long vowel unlike Czech)

bohatier 'hero' (with ie : Czech and Russian y < Mongolian u!?; ie < *ý?)

since Slovak lacks ř (or any other special reflex of *rĭ or *rʲ).

(Does/did Slovak have a cognate of Cz bednář 'cooper'? Slovak debnár 'cooper' happens to look like bednář with metathesis, but is from debna 'bin'.)

12.6.20.8:25: PATERNAL -C-ODA -C-ONUNDRUM

According to Vasmer, otec is from Proto-Slavic *оtĭcĭ which in turn is from *оtĭkŭ. Normally

- Slavic c is from *k plus front vowels (but *ŭ is a back vowel!)

- *-kŭ becomes Slavic -k

And in fact Vasmer does list the Russian dialectal forms

ótik 'male animal' (Olonets; is "Kulik." the author of a dialect dictionary?)

otjók 'father' (Ryazan, Russkij filologicheskij vestnik)

with the expected -k. Are there similar archaisms elsewhere in Slavic?

12.6.19.6:10: A PROTO-GERMANIC VOWEL CYC-*O?

Sunday was Father's Day in the United States, which got me thinking about atta-type words for 'father' in Indo-European: e.g., Albanian atë, Gothic atta, and Russian otec. Slavic o- is the o-dd man out. Why?

According to Schenker (1993: 66), Slavic and Germanic had "[s]imilar changes", yet one has *o where the other has *a and vice versa. 

Does the *ă- of late Proto-Indo-European 'father' go back to *ʕe-? Can *(ʔ)a- be reconstructed in PIE nursery words?

2. The PBS vowel subsystem above reminds me of Sanskrit: a length distinction for a but only long *ō. (Sanskrit ō is not from a PIE monophthong and is absent from the above table.) Could the PBS vowel subsystem have been like Bengali with quality instead of length distinctions?

(The above table is only meant to show structural parallelism; it should not be intrepreted as a table of sound correspondences.)

I could then reconstruct a pull chain in Lithuanian. *o broke to *uo, leaving a gap to be filled by *ɔ:

*ɔ > *o > uo

I don't know anything about Lithuanian, so this could be way off. What are the reflexes of those PBS vowels in Latvian?

And perhaps there was a two-stage shift in Slavic. First,

*ɔ > *o

No, wait, that new *o and the old *o would then merge with each other (stage 2 below) and then merge with *a (stage 3):

That can't be what happened. Stage 3 is wrong. Maybe there was nonphonemic length (stage 1 below) that briefly became phonemic (stage 2) and lowering only affected long *ō rather than short *o (stage 3):

On the other hand, perhaps Proto-Germanic had phonemic length that prevented short *a from raising and rounding like long *ā (stage 3 below):

But Wikipedia implies a different scenario for Proto-Germanic long nonfront vowels. (Stages and details are mine.)

Examples (with slight changes for consistency):

LPIE *dōmos > *dāmaz (stage 4) > *dōmaz (stage 6) 'judgment'

LPIE *swādus > *swātuz (stage 4) > *swōtuz (stage 6) 'sweet'

Did *ō really go full cyc-*ō in Proto-Germanic?

12.6.18.2:00: THE CRY OF THE MOVING SKY

On Saturday, I discovered wooroemae.com (< Korean uroe 'thunder' + mae 'hawk') and rediscovered thunderagents.com, so I've been thinking about the Korean words for 'thunder' and 'lightning' ever since.

1. chhŏndung 'thunder'

Korean has two basic words for 'thunder'. I always thought the first looked like a compound of Sino-Korean 天 chhŏn 'heaven' with a -dung that wasn't a Sino-Korean reading. I should have been suspicious because many or even most Korean roots with -ng are of Chinese origin. It turns out that the earliest attestation of the word I can find  is thyŏntong (in 新增類合 [1576]; > modern chhŏndong) which is clearly from SK 天 thyŏn 'heaven' plus SK 動 tong 'move'. chhŏndung looks like a harmonized form:

ŏ 'yin' + o 'yang' > ŏ 'yin' + u 'yin'

ŏ and u both belong to the 'yin' class in Korean vowel harmony:

(i is a neutral vowel. The breves are arbitrary diacritics distinguishing similar vowels; they do not represent shortness.)

2. ur(o)e 'thunder'

There are two major modern Korean versions of this word, uroe and ure. In Middle Korean, the word was urɣŏy with a HH (high-high) pitch pattern. I thought this might go back to an earlier *urŭ-kŏy with a LHH (low-high-high) pitch pattern, but

1. I don't know of any other compounds with *urŭ- LH 'cry' (as reconstructed by Martin 1996: 73, 76, 137).

2. I don't know of any other cases of LH being reduced to H rather than R.

3. What is *-kŏy H? It can't be kŏy L 'crab'. Also see below.

MK urɣ- L < *urŭk LL 'roar' (as reconstructed by Martin 1996: 76, 137) is a better semantic match than *urŭ- LH 'cry', but why would L become H in a compound? Sporadic assimilation: *LH > HH? And what was the suffix *-ŏy H added to *urŭk-? A nominalizer?

uroe < *uroy looks like an assimilated form of urɣŏy:

ŏ [-labial] > o [+labial] after u [+labial]

but such assimilation would violate harmony, since o is a 'yang' vowel and u is a 'yin' vowel. (Cf. the assimilation in chhŏndung above which is in accordance with vowel harmony.) Could the o in uroe be an archaism predating the development of vowel harmony in Korean?

譯語類解 (1690) has urăy which could be derived from an earlier *urɣoy. (ă is a reduced 'yang' vowel.) The hypothetical modern descendant of this variant would be urae. I wonder if any modern Korean dialects have such a word.

Another possibility is that uroe is due to a folk etymology deriving ure from Sino-Korean 雨 u 'rain' plus 雷 roe 'thunder', but that would not account for urăy. (Did Korean or Chinese ever have a word 雨雷 'thunder'?)

3. pŏn'gae 'lightning'

The earliest attestation of this word has vowel harmony: MK pŏnkŏy HH (rather than *pŏnkay which is the expected source of modern pŏn'gae).

This word has its own set of problems:

1. If its structure is parallel to ur(o)e 'thunder', it should be from

verbal root ('flash'?) + suffix

but there is no native root pŏn-.

2. According to my understanding of Vovin's lenition theory, *-k- lenited to Middle Korean -ɣ- between vowels whereas *-nk- simplified to MK -k- in the same environment. Yet -ɣ- is not common in MK, and MK pŏnkŏy has an unreduced -nk-. Could MK -nk- be from *-nVk-? But wouldn't *pŏnŭkŏy become *pŏnŭɣŏy? Was *-nk- retained to preserve morphological transparency (i.e., maintain an obvious border between the root *pŏn- and suffix *-kŏy)?

3. Is modern pŏn'gae with nonharmonic vowels

(a) the product of deharmonization

MK 'yin' -ŏy >  'yang' *-ay > modern 'yang' ae after 'yin' ŏ

(b) an archaism reflecting an original 'yang' vowel predating the development of vowel harmony in Korean?

'Thunder' and 'lightning' are not the only words which may have been subject to deharmonization. Perhaps the most important cases of deharmonization are the dative and genitive particles which lost their 'yang' allomorphs. (e is a 'yin' vowel.)

However, such deharmonization involves the spread of one allomorph to all environments which is distinct from deharmonization in words like 'thunder' and 'lightning' without such variation.