13.2.2.23:56: A HOPEFUL FAMILY OF FOUR

The tangraph (Tangut character) 2ʂwii 'to need, want, require' consists of two halves whose functions and sources are unknown:

=+

2ʂwii 'to need, want, require' = ? + ?

It combines with other components to form three other tangraphs:

+=

'top' or 'female' + 2ʂwii 'to need, want, require' = 1võ 'to wish' (< Chinese 望)

+=

? + 2ʂwii 'to need, want, require' = 2gii '(to) hope'

+=

? + 2ʂwii 'to need, want, require' = 2kiʳw 'wide, roomy*, shoe last'

If I did not know any Tangraphic Sea analyses, I might think that the structure of these characters could be explained by assuming that they reflect the structure of 'Tangut B', a hypothetical non-Sino-Tibetan language implied by the script but otherwise leaving no direct traces (!):

Tangraph	Tangut (A) reading	Tangut B reading (A-E are algebraic symbols)
	2ʂwii	AB 'want'
	1võ	CAB 'wish' (verb AB 'want' with prefix C-)
	2gii	DAB 'hope' (verb AB 'want' with prefix D-)
	2kiʳw	EAB (adjective E 'wide' plus suffixes -A, -B or a suffix resembling Tangut A 2ʂwii)

However, I don't think Tangut B is necessary to explain this set.

The first tangraph vaguely resembles the Chinese characters 欲 'desire' (cursive forms) and 須 'need' (cursive forms). Could it be a distortion of a whole Chinese character resembling two components found in other tangraphs?

The second tangraph is analyzed as a semantic compound in the Tangraphic Sea:

=+

1võ 'to wish' = top of 1kiụ 'to pray' + all of 2ʂwii 'to need, want, require'

The analyses of the other two tangraphs are unknown, but I suspect they might be

=+

2gii '(to) hope' + top of 1dzɨu 'love, affection' + all of 2ʂwii 'to need, want, require'

=+

2kiʳw 'wide, roomy, shoe last' = right of 1zie 'wide' + all of 2ʂwii 'to need, want, require'

1zie 2kiʳw

*These first two glosses are from Kychanov and Arakawa (2006: 677). I don't know of any textual evidence for them.

13.2.1.8:30: (X)EGG

Watkins (2011: 7) regarded Proto-Indo-European *ʕōw-yo- 'egg' (> Latin ovum, English egg; more descendants at Wiktionary) as a possible derivative of PIE *ʕewi- 'bird' (> Latin avis). PIE *ʕ normally becomes zero in Iranian: e.g., in Iron Ossetic 'egg' is айк ayk. So why is 'egg' خایه xāye with x- in Persian?

Also, why does Spanish huevo have an h- absent from Latin ovum and its other descendants (e.g., Portuguese ovo, French oeuf, and Italian uovo)? Hypercorrection: i.e., adding a nonetymological silent h- out of fear of omitting an etymological silent h-?

13.2.1.2:11: DARKENING (ALPHACISM?) IN NORTH FRISIAN AND SOUTHEAST ASIA

The Wikipedia article on North Frisian describes the lowering of ɪ to a as "vowel reduction", but that term makes me think of the reduction of vowels in unstressed syllables, not stressed monosyllabic content words like 'fish': e.g., Mooring North Frisian fasch and Fering-Öömrang North Frisian fask corresponding to English fish, German Fisch etc. I prefer the term darkening, an antonym of brightening which Matisoff used in his 2004 paper on raising and fronting of *a in Tangut: e.g.,

1mi 'not' < *ma (cf. Old Chinese 無 *ma 'not exist')

I also made up the term alphacism by analogy with terms like zetacism.

Whatever it's called also occurred in the ancestor of Cantonese: e.g.,

立 'stand'

Middle Chinese *lip > Mandarin li, Sino-Korean rip

but Cantonese lap

The largest stratum of Sino-Vietnamese was borrowed from a southern Late Middle Chinese dialect that was in transition: e.g., 'stand' is SV lập [ləp].

Similarly, there are transitional North Frisian dialects with partial darkening:  e.g., Söl'ring fesk 'fish'.

Conversely, Tangut has partly as well fully brightened forms: e.g., 'not exist' is

1mie < *ma (cf. Old Chinese 無 *ma 'not exist')

with a partly mid diphthong ie instead of the high monophthong i.

13.2.1.1:12: GERMAN -RR- : SATERLAND FRISIAN -DD-

I have been reading about Frisian lately. This photograph in the Wikipedia article on Saterland Frisian caught my eye because the German place name Scharrel corresponds to Saterland Frisian Schäddel. Is this a regular or even an occasional correspondence? If I knew nothing about German, I might guess that Saterland Frisian preserves a [d] that lenited to [r] between vowels in German. However, there is no such lenition in German. Was there a fortition in Saterland Frisian, or is this an irregular correspondence defying explanation? Could the two be separate attempts to represent some (substratal?) third name?

13.1.31.1:21: 'HTER

I initially thought Slovene hči 'daughter' couldn't be a descendant of Proto-Indo-European *dhugʕtḗr 'daughter', but I noticed that its oblique stem hčer- ended in -er, just like the oblique stem mater- of mati 'mother' which is obviously from PIE *méʕtēr. Then I found that Wiktionary derives hči from Proto-Slavic *dŭtʲi which is of course from PIE *dhugʕtḗr. č is obviously from *tʲ, but how did *d become h?

One might expect other cases of Slovene h- corresponding to d(h)- in other Indo-European languages*, but at a glance it seems that the normal Slovene reflex of *d(h)- is d-: e.g.,

An unusual consonant in 'daughter' is not unique to Slovene in South Slavic:

Serbo-Croatian kći

Macedonian ḱerka [cɛrka]

Bulgarian is the 'odd man out' because its dǎšterja ironically looks more 'normal' from a general IE perspective. (Do Bulgarian dialects have Macedonian-like forms?)

One might think that Slovene h-, SC k-, and Macedonian ḱ are traces of PIE *-gʕ-, but that cluster was gone in Proto-Slavic *dŭtʲi. (5:36: Or was it? The Wiktionary page for Proto-Indo-European *dhugʕtḗr 'daughter lists the Proto-Slavic word for 'daughter' as *dŭkti with *-k-, though clicking on the link for that word goes to the entry for *dŭtʲi without *-k-.)

I think the Macedonian form is from *tʲer-ka < *dtʲer-ka: cf. in West Slavic how Polish córka 'daughter' corresponds to Czech dcera and Slovak dcéra.

However, I can't explain Slovene h- or SC k-. I guess Slovene h- might be from *k-, but where did that *k- come from? Is it the result of dissimilation and assimilation?

*dtʲ- > *gtʲ- > *ktʲ-

Assimilation not preceded by dissimilation would have led to a geminate that probably wouldn't have been dissimilated:

*dtʲ- > *ttʲ- (> *ktʲ- doubtful)

Bogadek's (1944) Croatian dictionary lists kćeti as a variant of htjeti 'to want'. In this case k must be from h rather than the other way around because it is from Proto-Slavic *xŭtěti (cf. Russian xotet', etc.).

ADDENDUM: The Interslavic word for 'daughter' is dočera. This word is transparent to Russian speakers who will recognize it because it resembles the oblique stem dočer- of doč', but I wonder how recognizable it would be to West and South Slavic speakers. Would they guess from context that moj syn i moja dočera means 'my son and my daughter'? (Syn is pan-Slavic with minor variations: e.g., Slovene and Serbo-Croatian sin.)

*Such a correspondence is rare but not impossible. Toisanese h- is partly from *dh-: e.g., 唐 hɔŋ 'Tang Dynasty' from Late Middle Chinese *dhaŋ.

13.1.30.23:15: THE AFRICAN ARGUMENT FOR PROTO-INDO-EUROPEAN EJECTIVES

I have long been bothered by the glottatic theory for Proto-Indo-European because I didn't know of a language whose voiceless ejectives had become voiced plain stops. But on Monday I learned of two languages with voiced allophones of ejectives.

Judging from these examples, Blin in Eritrea has voiced allophones of ejectives intervocalically and before voiced consonants. However, /kʷʼ/ deviates from this pattern: it is voiced initially in the same position as /kʼ/ [kʼ] and is debuccalized intervocalically. I presume /kʷʼ/ surfaces as [kʷʼ]  in other positions: e.g., perhaps before voiceless consonants.

In Kwa'dza in Tanzania, "/kʼ/ and /kʼʷ/ are voiced [ɡ, ɡʷ] if a preceding consonant is voiced" according to Christopher Ehret (1980). Why aren't other ejectives like /tsʼ/ voiced in that position? Because they're affricates? I am reminded of how the Korean fricative /s/ doesn't voice intervocalically unlike the stops /k t p/ and the affricate /c/.

I also found this passage in the Wikipedia article on ejectives (emphasis mine):

In the languages where they are more obvious, ejectives are often described as sounding like "spat" consonants [which is what they sound like to me]; but ejectives are often quite weak and, in some contexts, and in some languages, are easy to mistake for tenuis or even voiced stops.

So it would not be surprising for Proto-Indo-European ejectives to become plain voiceless stops in Hittite, Tocharian, and Germanic or voiced stops in the other branches. I almost included Armenian, but on Monday I also learned that the Armenian reflexes of Proto-Indo-European stops are a lot more complex than I thought. More on them later.

This morning I had a dream about Armenian which made me check Beekes' (1995: 130) table of Proto-Indo-European, Proto-Germanic, and Armenian consonants. Along the way I rediscovered this passage on page 133 (emphasis mine):

... the glottal feature probably preceded the consonant: it was pre-glottalized, 'p, etc. Understood in this way, Lachmann's law for Latin is explained. This law states that a PIE [Proto-Indo-European] voiced (non-aspirated) stop b d g gʷ before a consonant lengthened a preceding vowel: for example: ag-ō, āc-tus, but veh-ō : vĕc-tus (with *ģʰ). The solution is that the glottal stop (ʔ) of the g = kʼ lengthens the preceding vowel: aʔg-tos > āctus. The glottal stop works, then, in the same way as a laryngeal [which leaves a trace as a vowel length]: cf. eh₂C > āC. One of the laryngeals was probably a glottal stop.

On page 126, Beekes identified the three laryngeals of PIE (*h₁, *h₂, *h₃) as *ʔ, *ʕ, and *ʕʷ, so PIE *eh₂C could be rewritten as *eʕC.

(Does any language have ʕʷ without ʔʷ? UPSID lists only one language with ʔʷ - Kabardian - and no languages with ʕʷ. I think Old Chinese might have had both. Modern Vietnamese has [ʔw], but that's a cluster /ʔ/ plus /w/, not a unit phoneme.)

If PIE 'voiced' stops (in the traditional reconstruction) are reinterpreted as preglottalized stops, then PIE *ʔeǵHom from "Me-tathes-I-s" could be reinterpreted as *ʔeʔǵHom and the second glottal stop was preserved in Kortlandt's (2012: 2) Proto-Balto-Slavic (PBS) *ʔeʔźun 'I'.

The *H after *ʔǵ apparently vanished without a trace in Proto-Balto-Slavic. This *H is needed to account for Sanskrit -h- in aham 'I'. Beekes (1990: 307) mentioned that Rix (1976: 177) reconstructed that uncertain laryngeal as *h₂ and guessed that Rix thought only *h₂ could condition aspiration in Indo-Iranian. Lunt (2001: 232) also reconstructed *h_2, citing Greek ego, Latin ego, and Sanskrit aham, but not explaining his reasoning.

Maddieson (2011) wrote that sounds similar to implosives "are often referred to as “pre-glottalized voiced stops” by linguists working on Asian and Pacific languages." Could traditional PIE *b *d *g *gʷ be reinterprted as implosives ɓ ɗ ɠ ɠʷ? I doubt it, because languages with implosives tend to have ɓ but not ɠ (e.g., Vietnamese), whereas PIE has the opposite pattern: traditional b is rare, but traditional g is common.

I also happened to open Beekes' (1988) Gatha Avestan grammar today and find a section on preglottalization on page 71:

In Indo-Iranian the preglottalization was still present at the time of Lubotsky's Law (see 53.2) and is preserved in modern Sindhi.

Modern Sindhi does have implosives in addition to the usual four types of Indic stops: e.g., ɓ as well as p ph b bh. Khubchabdani (2003: 627) mentioned that Turner (1924) derived Sindhi implosives "from geminated voiced plosives": e.g., əɠʊ 'before' < Sanskrit agra- 'top, front' (with an intermediate stage like Pali agga). However, this does not account for initial implosives: e.g.,

ɗ̣ohʊ 'fault' : Sanskrit doṣa- 'id.'

ɗ̣əhə 'ten' : Sanskrit daśa 'id.'

ɗ̣is- 'see', ɗ̣ekh- 'show' : Sanskrit √dṛś 'see'

ɗ̣y- 'give' < *dH; cf. Sanskrit √dā < *deʕʷ 'id.'

ɗ̣əndʊ 'tooth' : Sanskrit danta- 'id.'

Note how the Sindhi reflexes of Sanskrit d are retroflex (indicated with the non-IPA subscript dot) as well as implosive. Does Beekes view Sindhi implosives as conservative? Ah, I see now. Kortlandt (2012: 1) wrote (emphasis mine):

I have argued that the Sindhi preglottalized voiced stops are an archaism (2010: 121-124). In this language, the unconditioned reflexes of the d and dh series are glottalic and aspirated, respectively, while dissimilation of the dh series before aspirates of recent origin has given rise to a plain voiced series, e.g. ’gāh^u ‘bait’ < grāsa versus gāh^u ‘fodder’ < ghāsa-. The glottalic articulation cannot be attributed to external influence because the surrounding languages do not present anything comparable.

One could derive ’gāh^u from *ggāsa with an initial geminate from *gr, but that cannot account for the implosives corresponding to Sanskrit nongeminates in 'fault', etc. above.

I would add that loans from Sanskrit are also a source of voiced stops without preglottalization: e.g., Sanskrit duḥkha- 'pain' corresponds to native Sindhi ɗ̣ʊkhʊ and was borrowed as dʊkhʊ (forms from Khubchabdani 2003: 637).

I found that article when looking for Kortlandt's 2004 article about preglottalization in English and Scandinavian. Could this phenomenon be a remnant from Proto-Indo-European?

When a phoneme is accompanied (either sequentially or simultaneously) by a [ʔ], then one speaks of pre-glottalization or glottal reinforcement. This is common in most varieties of English, RP included; /t/ and /tʃ/ are the most affected but /p/ /k/ also regularly show pre-glottalization. In the English dialects exhibiting pre-glottalization, the consonants in question are usually glottalized in the coda position. E.g. "what" [ˈwɒʔt], "fiction" [ˈfɪʔkʃən], "milkman" [ˈmɪlʔkmən], "opera" [ˈɒʔpɹə]. To a certain extent, there is free variation in English between glottal replacement and glottal reinforcement.

This makes me wonder if those preglottalized voiceless consonants were ever voiced in pre-Germanic. Did Proto-Germanic preserve Proto-Indo-European *ʔt, etc.?

The preglottalization of English codas reminds me of the use of บ <ʔb> and ด <ʔd> for final [p] and [t] in Thai. Were those codas preglottalized when the Thai script was created? (There is no <ʔg> in Thai, so final [k] is written as ก <k> in native words. Final [k] in loanwords is written etymologically: e.g., Thai [roːk] 'disease' from Sanskrit roga- 'id.' is written as โรค <rōg>.)