10.1.15.23:59: IGNIS IMMORTALIS

Thanks to Guillaume Jacques for clariying the history of 'fire' for me. Contrary to my last post, the Proto-Indo-European forms *h₁engʷ-ni- and *h₁ongʷ-ni- are unnecessary because the Sanskrit, Latin, and (Balto-)Slavic words for 'fire' can all be derived from *h₁ngʷ-ni- without a vowel in the first syllable.

In Sanskrit, syllabic n became a: PIE *h₁ngʷ-ni-s > Skt agnis. Another example is my name Amritas < PIE *n-mr-to-.

In Latin, syllabic n normally became en (Beekes 1995: 136, Watkins 2000: 147), but was reduced to i in 'fire': PIE *h₁ngʷ-ni-s > Lat ignis (via ?*engnis with *-n- loss and vowel raising - why?). immortalis < *en-m ... < *n-mr-to- also has i instead of e.

In Proto-Balto-Slavic, syllabic n became un: PIE *h₁ngʷ-ni-s > PBS *ungnis. *un was lowered to *on and reduced to *o (details here): PBS *ungnis > Proto-Slavic *ongni > *ogni.

Next: liauqu, a very different sort of fire.

10.1.14.23:59: PYRO-CYC-O

Thinking about Schenker's account of Proto-Slavic vowel history led me to realize that the vowel of 'fire' has lost, regained, and lost its rounding between Proto-Indo-European and Russian:

Proto-Indo-European *ogni- (Watkins 2000: 22) > Proto-Slavic ?*agni > ?*ogni > Russian огонь [ʌgonʲ]

In Russian, unstressed o was reduced to [ʌ].

Proto-Slavic ?*agni looks like Sanskrit agni- 'fire'. I don't know if Skt agni- is from PIE *ogni- or *egni- (also reconstructed by Watkins) since Skt a can be from PIE *o or *e. Both Proto-Indo-Iranian and Schenker's Proto-Slavic shifted PIE *o to *a, but PS retained *e.

Watkins presumably reconstructed PIE *egni- to account for Latin ignis 'fire'.

I know little about PIE, but if I knew nothing about it, I might reconstruct *øgni- as the single source of both PS *agni- and Latin ignis. Of course, PIE had no *ø, but I don't understand why 'fire' had both *e- and *o-variants. Or did it?

Derksen's online Slavic inherited lexicon reconstructs it as PIE *h₁ngʷ-ni- with two ns (the first is syllabic) and without any initial vowel. Beekes (1995: 126), also of the Leiden school, equated PIE *h₁ with *ʔ. (Does any real language have a syllable like ʔn̩gʷ with syllabic n̩?) I presume that in a Leiden school reconstruction, the root is *h₁ngʷ-, and *h₁engʷ-ni- and *h₁ongʷ-ni- are equivalent to Watkins' *egni- and *ogni-.

Kortlandt (1979: 60-61) reconstructed Proto-Balto-Slavic *ungnis with an -n- after *u to block Winter's law (i.e., to prevent the short vowel *u from lengthening before *g). Lithuanian ugnis and Latvian uguns still have u-, but why does Derksen's Proto-Slavic have *o- instead of *u-? According to Derksen,

The sequence *un was lowered to *on before a tautosyllabic stop, with subsequent loss of the nasal as a result of dissimilation [...] Apparently, the latter development occurred in Baltic as well.

And in Sanskrit and Latin. Did any language retain the first *n?

I am puzzled by *gʷ in PIE *h₁ngʷ-ni-. Which language indicates that a labiovelar should be reconstructed instead of a velar?

10.1.12.22:05: YER *O(U)RIGINS

Before I move on to Proto-Slavic *o and the yers, I want to reconsider the origin of PS *yy which I discussed in my last post.

In Greek, ου [ou] became [oo]. υ [u] had fronted to IPA [y] (not to be confused with PS *y [ɯ]), leaving a gap that was later filled by a new [u] from [oo].

Greek stage 1

Greek stage 2 (excluding original front vowels)

Modern Greek

Similarly, early Proto-Slavic *a(a)u could have become an *ɔɔ that rose to *oo. (au underwent the same process in Japanese.) Proto-Slavic *uu delabialized to *yy, leaving a gap to be filled by a new *uu from *oo.

Proto-Slavic stage 1 (excluding short vowels; cf. Greek stage 1)

Proto-Slavic stage 2 (cf. Greek stage 2)

Proto-Slavic stage 3 (cf. modern Greek)

Here's the rest of the stage 3 chart excluding nasal vowels:

The short high vowels *i and *u lowered slightly to become the yers [ɪ] and [ʊ].

The short nonhigh vowel *a became *o.

The long nonhigh front vowel *ee became yat [æ] which I discussed last week.

Proto-Slavic Stage 4 (? - this is the system implied by Schenker [1993: 79], assuming that vowel length was lost)

If there was no vowel length at stage 4, where did modern Slavic languages like Czech get their long vowels from? In a later stage, -j- between vowels was lost, creating new long vowels: e.g.,

*-aja: Czech nová [novaa] 'new' (nom. sg. f.; cf. Russian новая novaja which still has -j-)

Schenker's (1993: 82) final stage of Proto-Slavic has a nearly full set of these secondary long vowels. Only the yers have no long counterparts.

(All of the above Proto-Slavic tables only contain oral vowels. I have left out the nasalized vowels *ẽ(ẽ) and *õ(õ).)

Before I can discuss Schenker's (1993) origin of Proto-Slavic o, I should cover his account of the origin of y. I use y here to represent a back unrounded vowel [ɯ] that corresponds to Russian ы which is usually transcribed as y.

My last post ended at the early point where Proto-Slavic had only eight vowels:

Further long vowels developed from earlier vowel-glide sequences. Schenker indicates these new long vowels with subscript numbers. The phonetic values are my speculations.

*a(a)j > *ee₂ [ɛɛ]? (later merging with original *ee)

*e(e)j > *ii₂ [ɪɪ]? (later merging with original *ii)

*a(a)w > *uu₂ [ʊʊ]?

*e(e)w > *juu₂ [jʊʊ]?

According to Schenker (1993: 72),

The Early Proto-Slavonic [= Proto-Slavic] back vowels were redundantly and, hence, weakly labialized.

Did he meant to say that even the low back vowel a(a) was labialized: e.g., was rounded *[ɒ(ɒ)]? Does any language have only palatal and labial vowels without any neutral vowels like a or ə? The only such language I can think of offhand is reconstructed. Proto-Indo-European is often reconstructed with two basic vowels: palatal e and labial o (disregarding length).

Schenker described *uu₂ as "fully labialized". I am puzzled because he wrote that*uu₂ comes from *a(a)w. Even if *a(a) were labial *[ɒ(ɒ)], a sequence *ɒ(ɒ)w would still be less labial than original *uu. However, Schenker viewed original *uu as less labial than *uu₂, and proposed that the former delabialized to *yy while the latter remained labial.

I assume *uu was more labial than *uu₂ before the following chain shift:

*ʊʊ (= *uu₂) > *uu > *yy

In a pull chain scenario, *uu moved to *yy, leaving a gap to be filled by *ʊʊ.

In any case, the labial origin of Proto-Slavic *yy explains why Russian ты ty 'thou' has y corresponding to the labial vowels of English thou and French tu.

Schenker reconstructed a nine-vowel system for late early Proto-Slavic

Note the gaps. *yy has no short counterpart and there are no nonhigh back rounded vowels.

Next: The origin of *o. Also: Enter the yers.

I haven't posted lately because I fell asleep early on Friday and my computer was installing updates on Saturday night.

Before I went to bed on Friday, I came up with an answer for the question I asked in my last entry.

According to my understanding of Schenker's (1993: 66) reconstruction of Proto-Slavic, early Proto-Slavic syllables could theoretically begin with one of eight vowels:

There was no *o or *oo because those vowels had merged with *a and *aa.

Schenker (1983: 68) reconstructed prothetic glides before all syllable-initial front vowels and long back vowels (except in some unspecified dialects which preserved aa-) for a slightly later stage of Proto-Slavic:

I have no idea why short *u- and *a- were exempt from prothesis.

The glides generally match the vowels that follow them with one exception:

palatal glide *j- : palatal vowels *i(i), *e(e) but also the nonpalatal vowel *aa

labial glide *w- : labial vowel *uu

Why did *aa- become *jaa- with a mismatched glide? Perhaps aa- briefly became *ɰa- with a matching velar glide *ɰ- that then shifted to the much more common palatal glide *j- (j occurs in 84.92% of UPSID's languages, whereas ɰ only occurs in 2.66%.)

Middle Chinese had a medial *-ɰ-/-ɨ- (depending on notation) that developed into -j-/-i- in modern Chinese languages. MC loans into Vietnamese preserve the original nonpalatal medial: e.g.,

將 'general': Sino-Vietnamese tướng [tɨəŋ] but Mandarin jiang < Middle Chinese *tsɨaŋh

Next: Where did Slavic languages get their o from?