08.1.26.23:56: GETTING OFF THE STREET

I hope this is my last post about Slavic words for 'street'.

The word that triggered the last few posts was Belarusian vulica 'street'. When I checked some other major Slavic languages, I couldn't find any cognates with v-. They all look alike except for the Czech form which ends in -e:

Rus ulica

Pol ulica

Cz ulice

Srb-Cr ulica

So why is Cz ulice the odd man out? My guess is that the Proto-Slavic form was something like *ulicja, and final *-ja became -e in Czech but was simplified to *-a in the other languages. Slavic c [ts] is not a retention from Proto-Indo-European but originated from palatalized *k (details here). I presume that *k became *c before *-ja.

Tonight, I found some more -a forms in Max Vasmer's Etymological Dictionary of the Russian Language:

Uk vulycja

another East Slavic language with v-

Slovak ulica

Serbian Church Slavonic (?) ulica

Bul ulica

Slovenian ulica

Upper Sorbian wulica [vulitsa]

with prothesis like Uk and Bel further to the east!

The Ukrainian form seems to retain the *-ja that I guessed would be in Proto-Slavic. Unfortunately, I can't check the actual Proto-Slavic form because indo-european.nl is down!

1.27.5:24: The noun 'will' is a clear example of Czech -e corresponding to -ja elsewhere:

Czech vůle < ?*woolja

but

Rus, Uk, Bel, Bul, Slovenian, Srb-Cr volja

Slovak voľa (ľ is palatal)

Polish wola < *wolja (*wola would have become Polish woła)

Upper and Lower Sorbian wola (also from *wolja?)

Kortlandt (p. 17) reconstructed Proto-Slavic *w^uol^ja.

Vasmer compared the Slavic forms to German Wahl 'choice', wollen 'to want', and Skt vara- 'choosing'. Other cognates are listed here.

08.1.25.23:59: A W-EIGHT-Y PROBLEM

Given that Belarusian has developed a v- < *w- before stressed initial back vowels -

вокны 'windows'

cf. Russian окна, Polish and Czech okna 'id.'

- one might expect the Russian cognate of Be восем to be осем with an initial o- like

Polish osiem

Czech osm

Serbian осам

But the Russian form is actually восемь.

Moreover, the Ukrainian form is вісiм with a front vowel in addition to a v- in the first syllable.

Did Proto-East Slavic strengthen *o- to *wo- in 'eight'? (Why couldn't the reverse be true?*)

And why did Ukrainian shift this *o- to vi- in вісiм - and even in вікно 'window'?

The Wikipedia article on Ukrainian phonology may have the solution to the i-mystery:

In a newly closed syllable, that is, a syllable that ends in a consonant, Common Slavic *o*e mutated into *i if the next vowel was one of the yers (*ǐ/ь or *ǔ/ъ).

If the Proto-East Slavic word for 'eight' was something like Old Church Slavonic osmǐ, then *o in the closed PES syllable *wos became Ukrainian i. This looks like long-distance assimilation:

*CoCCǐ (one front vowel) > *CiCCǐ (two front vowels)

However, if I am reading the above rule correctly, it also means that

*(Co)CCǔ (two back vowels) > *(C)iCCǔ (one front vowel and one back vowel)

which looks like long-distance dissimilation. I would have expected

*(C)oCCǔ (two back vowels) > *(C)uCCǔ (two back vowels of different heights)

The prothesis of *w- must have occurred before the *o to *i shift, since a change of *i- to *wo- would be bizarre.

Moreover, Russian, Belarusian, and Ukrainian вохра 'оchre' (< Greek ωχρα) must have been borrowed before *w-prothesis, whereas Rus охра 'ochre' must have been borrowed after *w- was no longer being added.

1.26.1:26: Early Slavic had *j-prothesis as well as *w-prothesis:

The hiatus between a word-final and a word-initial vowel was filled with a glide, which was j if at least one of the vowels was front and w if the preceding vowel was back and the following vowel was rounded.

(But what glide would have filled the gap between a word ending -a and a word beginning with a-?)

This would have occurred long before the *w-prothesis in East Slavic 'eight'.

Could this explain why some Slavic languages have j- in 'I' and others don't?

j-forms (with prothesis due to a front vowel at the end of a preceding word?): Ru/Be/Uk я, Po ja, Cz já

non-j-forms (without prothesis?): OCS azǔ, Bu аз

cf. Lith àš, Skt aham, Lat ego which all lack j-

and Gatha Azestan azə

but Serbian, another South Slavic language, has я!

(And what are the forms я and яз listed in parentheses for Bulgarian in the Wikipedia Swadesh list for Slavic languages? They're not mentioned in the article on Bulgarian pronouns.)

1.26.1:39: Until now, I had assumed that Slavic ja arose from the breaking of an earlier *e (cf. Lat ego), but I couldn't explain the glideless Bulgarian form. Beekes (1995: 207) seems to imply that breaking occurred, followed by (presumably irregular) j-loss in OCS:

PIE *eeg(H)om > *jazǔ > OCS azǔ

But I can't find the change of *ee to *ja or *j-loss in Frederik Kortlandt's "From Proto-Indo-European to Slavic".

1.26.00:52: I'm puzzled by the consonants of 'eight':

Was Proto-IE *kt > *s a regular change in Proto-Slavic? According to Kortlandt, PIE *kt became an affricate before high front vowels. Did it become a fricative in other environments?

Why do the Slavic forms for 'eight' have an m absent from the other IE langauges? Is this nasal a remnant of a PIE *-mo- (cf. Skt -ma- '-th' in aṣṭama- 'eight')? But PIE *-o- would have become a nonfront yer, not the front yer of OCS osmǐ.

*Proto-East Slavic couldn't have retained an original initial glide, because other forms for 'eight' throughout Indo-European have no initial glide: e.g.,

Lith aštuoni, Skt aṣṭa, Latin octo, Eng eight, Old Irish ocht, etc.

It's unlikely that all these branches of IE would have lost *w- independently in that one word but not before *o- in other words.)

1.26.00:24: I assume that Tsakonian, which preserves early Greek ϝ- as v-, does not have a v-initial word for 'eight'.

08.1.24.23:59: PLEOPHONIC* TORT-URE

David Boxenhorn suggested that Belarusian вораг is a native word inherited from Proto-East Slavic whereas Russian враг is a borrowing from Old Church Slavonic (i.e., a South Slavic language). I think he's right, because it fits this pattern (Comrie 1987: 330-331):

One of the main differences between East Slavonic and South Slavonic is the treatment of Common Slavonic sequences of o/e folowed by a liquid between consonants, symbolised *tort [is this an actual Proto-Slavic root or word?]. In East Slavonic, this sequence yields torot, while in South Slavonic it yields trat. In modern Russian, alongside East Slavonic golová 'head', there is also South Slavonic glavá 'chief; chapter'. (Note that in English head is of Anglo-Saxon origin [i.e., is native], whereas chief and chapter are of Romance origin [i.e., are borrowed].)

Compare:

This table makes Slavic look like Semitic with triconsonantal roots and vowel patterns! A little data can be very deceptive.

*1.25.1:22: The Wikipedia article on Russian phonology contained the solution to the Vladimir mystery:

Pleophony or "full-voicing" ('полногласие'), that is, the addition of vowels on either side of /l/ and /r/ between two consonants. Church Slavonic influence has made it less common in Russian than in modern Ukrainian and Belarusian ...

Uk: Володимір

Russian: Владимир ... (although the nickname form in Russian is still Володя).

Bel Уладзімір has a rounded vowel in the first syllable like the vo-forms of Ukrainian and Russian but no initial v-.

**1.25.00:50: Unstressed o is written as a in Belarusian, whereas it is still written as o in Russian. I assume that Rus мороз and Bel мароз are homophonous despite their spellings. (I've never heard Belarusian spoken.)

The other pairs in the CoCoC(V) column with initial г- would be homophonous were it not for the fact that Bel r- has lenited to [ɦ]. (This lenition is shared with Ukrainian; I assume it occurred independently in Czech and Slovak.)

After living in Belarus, Chris Marchant (author of "Fundamentals of Modern Belarusian") found it difficult to adjust to the Russian pronunciation of г as a velar stop:

My accent had changed, it was hard to say "mnoga" [i.e., Rus много] instead of "mnoha" [i.e., Bel многа].

This reminds me of my experience with Dutch, which lenited *g to [ɣ] ~ [x]. After living in Holland, it was hard to pronounce German g as [g]. (I had studied German before I moved to Holland, but I had never actually used German in daily life, whereas Dutch was an immediate necessity.)

08.1.23.23:59: STILL ON THE STREET

When I came up with the title "Gliding onto the Street", I was going to propose that Belarussian *ó- and *ú- went through an intermediate stage of *wó- and *wú- before becoming vó- and vú-. However, I rejected those transitional forms because Belarusian has a w (ў in Cyrillic), and I would have expected ўо- [wo] and ўу- [wu]. But an email from David Boxenhorn and some new findings (e.g., the near total- absence of words beginning with ў- [w]*) have made me reconsider my original idea.

As David pointed out, *ó- to *wó- makes more phonetic sense than going directly from *ó- to vó-. Here's what I think might have happened:

1. The Wikipedia article on Proto-Slavic is seemingly inconsistent. The PS word for 'star' is given as *gwozda with *w, but the inventory of consonants lists a *v but not a *w. Frederik Kortlandt reconstructed a *w instead of a *v for PS. (This *w was retained unchanged from Proto-Balto-Slavic.) For my purposes, I will assume that this phoneme was *[w] in early Polish and early Belarusian. Although those two languages belong to different branches of Slavic, a *[w] could have been an areal trait (either a shared retention of PS *w or a shared innovation from a PS *v).

2. Latin alphabets for Slavic languages typically have the symbol v for [v]. However, in Polish, [v] is written as w. I wonder if this w was once *[w].

(I suspect that German w [v] was also once *[w], and that German v [f] was once a real [v], which it still is in standard Dutch.)

3. In early Belarusian writings in the Latin alphabet from the 17th century, the sound corresponding to modern v was written as w, as in Polish (source). Could this w have represented *[w]?

4. The Belarusian Latin letter ŭ for [w] did not appear until the 19th century (source). Although the early Belarusian Latin alphabet was similar to that of Polish, Polish ł, currently pronounced [w] but with an archaic [ɫ] variant (Stone 1987: 356), was not used to write Belarusian [w]. This indicates that Polish ł was still a liquid and therefore inappropriate for [w].

The fact that w was not used for [w] tells us that w had already become [v] in Polish and Belarusian. It seems that similar chain shifts took place in the two languages:

Belarusian: *ɫ/w > *w > v

Polish: *[ɫ/w] > *[w] > [v]

Both may have shifted *w to v at about the same time, but Belarusian may have developed a new w before Polish.

If the above scenario is correct, the shift of *ó- to *wó- and *ú- to *wú- in Belarusian must have taken place at some point before stage 4 (the development of a new w to replace the old one that became v).

I wouldn't be surprised if much of the above turned out to be wrong. This is just mental exercise for me. It's not an attempt to prove that I'm a Slavicist. My answers are merely educated guesses, and I still have questions: e.g.,

If glides grew before stressed vowels, why does the Belarusian cognate of Russian учить 'teach' begin with v-: вучыць?

1.24.00:50: This v- is also present in prefixed cognates also lacking a stressed u: вывучаць and вывучыць, the imperfective and perfective verbs for 'learn'

Why does the Belarussian equivalent of Vladimir begin with a vowel u- instead of v-: Уладзімір?

1.24.00:24: Similarly, why does the Belarussian equivalent of Russian все 'all' begin with a vowel u- instead of v-: усе? Did *wC- always become uC-?

I expected the Belarusian equivalent of Russian враг 'enemy' to be ураг, but it's вораг with a stressed o!

1.24.00:40: The only example of a vC-cluster in this Belarusian-Russian dictionary is in a single word, встаць 'rag' (whose resemblance to the Russian verb встать 'get up' must be a coincidence).

This more recent monolingual spelling dictionary lists only acronyms and foreign names (Vladas and Vltava) with written vC- sequences. So it looks like *wC- (or *vC-) did shift to uC-.

*1.24.00:44: The only ў-initial word I can find is ў 'in', a variant of у corresponding to Russian в(о).

This Belarusian-Russian dictionary has no entries beginning with ў- other than ў itself (under the letter у!).

Foreign names beginning with wV- are all Belarusianized as uV-: e.g.,

Уінстан Чэрчыль (not Ўінстан) 'Winston Churchill' (Belarusian Wikipedia article)

08.1.22.23:59: GLIDING ONTO THE STREET

I don't expect to learn the right solution for the Hassaniya problem any day soon. I wish there were books outlining the histories and present states of language groups (e.g., the varieties of Arabic) which were accessible to nonspecialists.

J. Marvin Brown's From Ancient Thai to Modern Dialects and Other Writings on Historical Thai Linguistics (1985) is such a book. I first read it in 1993, before I could speak any Thai. It made sense to me then and it still seems to hold up now. I've been rereading passages from it tonight.

Books like Brown's with its many charts are useful for phonological tourists like me who compile mental catalogues of sound changes that can be recycled when we work on languages that we specialize in. We assume that if one group of speakers could change X to Y, then another group of speakers could make that same change. Enlarging our inventories of known sound changes enables us to see more possible explanations for phenomena that we observe. Moreover, if someone challenges us - 'that couldn't possibly happen!' - we can answer, 'yes, it did' and point toward cases documented in other languages by disinterested third parties.

Without third party guidance, I'm left to draw my own conclusions about languages I don't know. I enjoy the challenge, even though I might not be right. I can live with that. I might even be wrong about languages I am more qualified to talk about. I still am not sure that Tangut Grade II had pharyngealized vowels, for instance. But I keep hypothesizing, hoping that there will be some hits amidst the inevitable misses.

Is the following hypothesis about Belarusian vú- a hit or a miss?

A couple of days before I messed up with Hassaniya, I was intrigued by the second photo accompanying "Saying Nyet to Russian". It contained two Belarusian street signs:

ВУЛİЦА ЛЕНİНА

ВУЛİЦА УЛЬЯНАЎСКАЯ

(Is a Turkish-style capital İ acceptable in Belarusian?)

The Belarusian forms of 'Lenin' and 'Ulyanov' were less interesting to me than the word for 'street' which had an initial v- absent from Russian улица.

Seeing that unexpected v- reminded me of the fact that Belarusian seems to have developed vó- from stressed *ó-. Examples from Chris Marchant's "Fundamentals of Modern Belarusian" (PDF) with underlined stress:

stressed vó- : unstressed a-

возера 'lake' : азёры 'lakes'

cf. Russian озеро 'lake' and озёра 'lakes'

вокны 'windows' : акно 'window'

cf. Russian окна 'windows' and окно 'window'

(The exceptions that Marchant lists are all loanwords except for одум 'deep thought' which I assume is native, since it looks like a prefix о- plus the root дум [cf. Russian дума 'thought' as well as 'council'].)

(1.23.00:51: The plural endings of Belarusian and Russian don't match. I'll discuss them in a future post.)

Did stressed *ú- similarly become vú-? slounik.org confirms that вуліца has initial stress.

Two more examples I found at slounik.org are

'coal': вугал, corresponding to Russian уголь

'ear': вуха, corresponding to Russian ухо

But simply mechanically changing Russian stressed ú- to Belarusian vú- doesn't work, since Belarusian isn't simply Russian with different pronunciation: e.g., the Belarusian equivalent of Russian утка 'duck' is качка, and the Belarusian words for 'morning' (раніца, ранак) are not cognate to Russian утро.

''spark': іскра, corresponding to Russian искра

'truth': ісьціна, corresponding to Russian истина

(1.23.22:49: I had orignally listed

'name': імя, corresponding to Russian имя

but 'name' has final stress in Belarusian [why?], so I replaced it with 'truth'.)

Thus it seems that the change of *ú- to vú- has no front vowel counterpart.

08.1.21.23:59: HIGH-LY SPECULATIVE

In part 2 of "Wʔy I Wasn't Right", I mentioned a pair of related Chinese words:

GSR699c 仰 'look up': Md yang < Old Chinese *ŋaŋʔ

with semantic element 亻 'person'

GSR699b 昂 'lift high': Md ang < Old Chinese *ŋaŋ

with semantic element 日 'sun' (i.e., 'toward the sky'?)

They share a phonetic which also happens to represent their probable root:

GSR699a 卬 'high': Md ang < Old Chinese *ŋaŋ

It's obvious that 699c has a *-ʔ suffix absent from the other two words. But why does it have a nonemphatic initial?

I have yet to write a full account of my hypotheses about emphasis in Old Chinese. What follows is a sample:

Emphasis was probably not phonemic in pre-OC or in its parent Proto-Sino-Tibetan*. Those languages may have had uvular initials, and vowels after uvulars were automatically emphatic (uvularized?): e.g.,

*/qa/ *[qa] (underlining will indicate 'emphasis')

Emphasis may also have been automatic in syllables with nonhigh vowels: /a e o/. Thus the word 'high' had automatic emphasis:

卬 */ŋaŋ/ *[ŋaŋ]

There were at least two types of presyllables:

*Cɯ- (high vowel; automatically nonemphatic)

*Cʌ- (nonhigh vowel; automatically emphatic)

cf. gDong-brgyad rGyalrong, which has what appear to be three types of presyllables: Cɯ-, Cɤ-, Ca- (though no emphasis is involved).

When a presyllablc prefix was attached to a root, the root harmonized in terms of emphasis:

*Cɯ-CVC > *Cɯ-CVC (both prefix and root nonemphatic; no change)

*Cɯ-CVC > *Cɯ-CVC (emphatic root became nonemphatic)

*Cʌ-CVC > *Cʌ-CVC (nonemphatic root became emphatic)

*Cʌ-CVC > *Cʌ-CVC (both prefix and root nonemphatic; no change)

The assimilation of roots to affixes has a parallel in Germanic umlaut.

'Look up' was derived from 'high' via a prefix and a suffix:

卬 */ŋaŋ/ *[ŋaŋ] > 仰 */Cɯ-ŋaŋ-ʔ/ *[Cɯ-ŋaŋ-ʔ]

The bending of vowels may have started before the loss of prefixes:

仰 *Cɯ-ŋaŋ-ʔ > 仰 *Cɯ-ŋɨaŋ-ʔ

The first half of the nonemphatic root vowel *a has bent upward to match the height of the prefix vowel *ɯ. Other nonhigh nonemphatic vowels also partly bent upward:

*e > *ie

*ə > *ɨə

*o > *uo

The other nonemphatic vowels *i and *u were already high and couldn't get any higher.

Compare 仰 *Cɯ-ŋɨaŋ-ʔ with the unprefixed and emphatic

卬昂 *ŋaŋ

whose vowel did not bend upward. Emphatic vowels tended to bend downward:

*i > *[ɪ^ʕj]

*e > *[ɛ^ʕj]

*ə > *[ʌ^ʕɰ]

*u > *[ʊ^ʕw]

*o > *[ɔ^ʕw]

*a backed to [ɑ^ʕ] but it couldn't bend downward since it was already low.

After the loss of prefixes, emphasis disappeared, but the effects of vowel bending remained:

仰 *ŋɨaŋʔ vs. 卬昂 *ŋɑŋ

I wonder if 'brightening' in Qiangic (the raising of *a) also originated in prefix-triggered vowel bending:

*Cɯ-Ca > ?*Cɯ-Cɨa > ?*Cɨa > ?*Cɨ > Tangut Cji

1.22.???: Cf. how 'deemphasized' Old Chinese *a became Hakka -i:

'dwell': 居 Old Chinese*Cɯ-ka > *Cɯ-kɨa > *kɨa > *kɨə > *kɨ > Hakka ki

The phonetic of 'dwell' was emphatic:

'old': 古 Old Chinese *kaʔ > *kɑʔ > *kɔʔ > *ko > Hakka ku

*Although I have proposed that the Grade II vowels of Tangut were pharyngealized, I suspect that they are secondary and are not retentions from PST. If they were preserved from PST, they should strongly correlate with the vowels of OC emphatic syllables. However, OC emphatic syllables seem to correspond to Tangut Grade I (nonpharyngealized) syllables: e.g.,

'I': OC *ŋa : Tangut ŋa 2.14 (instead of ŋa^ʕ 2.15)

even if Tibeto-Burman did borrow 'I' from Old Chinese (Sagart 1999), Tangut should have a Grade II word if Proto-TB had emphasis

'fire': OC *hməjʔ : Tangut məə 1.31 (instead of mə^ʕ 1.28 [there was no long -əə^ʕ])

'three': OC *s-hləm : Tangut sọ 1.70 (instead of sọ^ʕ 1.71)

even if Tibeto-Burman did borrow 'three' from Old Chinese (Sagart 1999), Tangut should have a Grade II word if Proto-TB had emphasis

One case in which emphasis does match in the two languages is

OC *ɢoʔ 'ruler' : Tangut ɣu^ʕ 1.4 'head'

The correspondences and origins of Tangut Grade II require more investigation.

08.1.20.2:10: WʔY I WASN'T RIGHT (PART 2)

My incorrect solution for the Hassaniya problem may have reflected my orignal background in Japanese historical phonology.

Japanese borrowings from Chinese often have zero corresponding to glides in Chinese and in Sinoxenic 'dialects':

王 'king': Sino-Jpn ou

but Sino-Kor wang, Sino-Viet vương (< *w-), Cnt wong, Md wang

衛 'defend': Sino-Jpn ei

but Sino-Kor wi, Sino-Viet vệ (< *w-), Cnt wai, Md wei

圓 'yen', 'circle': Sino-Jpn en

but Sino-Kor wOn, Sino-Viet viên (< *w-), Cnt yun, Md yuan

In these cases, Japanese lost a glide that was preserved elsewhere. This can be confirmed by looking at the premodern kana spellings of those sinographic readings:

王 'king': ワウ wau

衛 'defend': ヱイ wei

圓 'yen', 'circle':　ヱン wen

Japanese lost w- after all vowels other than a. wau did not become au because *w-loss occurred after the monophthongization of *au to ɔɔ, which later merged with oo (spelled ou). Modern Japanese wi, we, wo are in recent loanwords and are not from earlier Japanese *wi, *we, *wo.

If wen became en, why does the English word yen have initial y-? Earlier wen may have been *[wyen] or *[w^yen] which became *yen after *w-loss. The y- of yen indicates that Japanese still had *ye when the word was borrowed into English. Earlier *ye became modern Japanese e, and modern Japanese ye is only in recent loanwords.

However, one cannot always assume that initial glides in one language corresponding to zero (or a glottal stop) in another language are retentions. One might think that Mandarin preserved a *y- lost in Korean on the basis of sets like

1. 雅 'elegant': Md ya but Kor a

2. 樂 'music': Md yue but Kor ak

3. 仰 'look up': Md yang but Kor ang

But there are also cases where Md y- corresponds to Kor y-: e.g.,

4. 約 'agreement': Md yue and Kor yak

5. 羊 'sheep': Md and Kor yang

Did Korean lose y- at random? Or did Mandarin 'grow' y- at random? Looking at other languages' readings leads to a third possibility:

1-3 have g- (< *ŋg-) in Japanese and nh- [ɲ] in Vietnamese

4 has y- in Japanese and an initial, unwritten glottal stop in Vietnamese

5 has y- in Japanese and d- [z] < *d^y- < *y- in Vietnamese

One might think that Mandarin shifted an earlier nasal to y- whereas Korean lost it entirely. This is close to the truth but still not quite correct. There are cases in which Mandarin has zero instead of an expected y-:

6. 昂 'hold up one's head': Md ang but Kor ang, Jpn gou (< *ŋg-), Viet ngang

obviously cognate to 3. 仰 'look up'

Here's my solution:

Mandarin lost *ŋ- without a trace. Its y- can reflect very late MC *-y-, *-ɨ-, *ʔy- as well as *y-.

The very late MC *ya of 1-2 is from earlier MC *æ. Cf. Japanese ya for Eng æ after velars:

Eng cat > Jpn kyatto

Eng gal > Jpn gyaru