15.5.30.13:51: WHY IS THE KHITAN SMALL SCRIPT SO LARGE ON MY NEW LAPTOPS?

For nine and a half years, I have been using 96 x 124-pixel images to represent the Tangut, Khitan, and Jurchen (TJK) scripts. These images are almost always screen captures of TJK characters in 72-point fonts in BabelPad. (In a few cases, they are screen captures of Chinese characters that I have modified by hand to represent Khitan large script characters that I don't have in a font.) This system has worked since January 2006 on five different laptops: two with Windows XP, one with Windows Vista, and two with Windows 7. However, it no longer works on my new Windows 8 machines. 72-point characters no longer fit in a 96 x 124-pixel space: e.g., the Khitan small script character for 'twenty'

was 85 x 82 in 72 point in Andrew West's font on previous machines but is now 107 x 104 on one Windows 8 machine and 129 x 125 on another. The screen resolution is 1920 x 1080 on both machines. Is there anything I can do to make characters appear at the same old size again? I am reluctant to post about TJK if future character images can't be consistent with previous ones.

Ah, I see now. Windows 8 has nothing to do with this problem. Resolution is the key. My previous machines had screens set to 1024 x 768, 1280 x 800, and 1366 x 768.  If I reset my current machines' resolution to one of those smaller formats, 'twenty' is 85 x 82 in 72 point. Why is it 26% and 52% larger on two different 1920 x 1080 screens? Intel should know the answer; both machines have an Intel HD Graphics Control Panel.

15.5.29.22:45: IN-S-ERTED IN TIME

Having mentioned Middle Korean (MK) ᄣᅢ pstay 'time' in my last post, this might be a good, um, time for a short note on John Whitman's (2012: 32) etymology which I just discovered on Wednesday:

Proto-Koreo-Japonic (PKJ) *pə(n)tə >

Proto-Korean *pət-ay > MK pstay > modern Korean ttae 'time'

I thought Whitman's split of 'time' into two parts was ingenious.*-ay is a locative suffix reanalyzed as part of the stem, so modern Korean 때에 ttae-e 'time-LOC' has the same suffix twice: i.e., it is etymologically *'(time-LOC)-LOC'. Although the sequence *pət-ay is not harmonic according to Middle Korean rules, earlier Korean may not have had any harmony.

Proto-Japonic *pə(n)tə 'interval' > Japanese hodo

First, if the PKJ form had *-n-, that consonant corresponds to nothing in MK unless *pnt- became pst- (via *pzt-?). I know of no parallel for such an unusual change.

Second, if the J form did not have -n-, there is no source for MK -s-, unless *pt- became pst-, which is not only strange but also raises the issue of how MK could also have pt- (e.g., in ᄠᅢ ptay 'dirt'). I would rather not posit a chain shift with syncope in 'time' and 'dirt' occurring in different periods to explain why both pst- and pt- exist in MK:

I did not specify a syncopated vowel in 'dirt' because I have no evidence for its quality. One could hypothesize that *pət- became pst- whereas *pVt- (in which *V was a vowel other than *ə) became pt-, but I don't know why *ə would be more s-friendly (sigmaphilic?) than, say, *ʌ.

15.5.28.21:32: SICILIAN GEMINATION

What is the origin of initial geminates in Sicilian which may or may not be written? Has gemination from syntactic doubling been carried over into isolated forms: e.g.,

è bonu [ebˈboːnu] 'is good' (?) > bonu [bboːnu]?

(And is [ebb] in turn from *es b ...?)

In any case, I assume the phenomenon is a Sicilian innovation. Although nomu [nnomu] 'name' originally might have had a consonant cluster in Proto-Indo-European (*ʕʷnomn*), that cluster was gone in Latin, so no parallel with Korean 'tense' consonants from earlier clusters (e.g., pstay > ttae 'time') can be drawn.

Is gemination in words like mmàggini 'image' from earlier *VC-sequences (*imàggini > mmàggini), or does it postdate apheresis (*imàggini > *màggini > mmàggini)?

*I assume Greek o- in onoma 'name' is from PIE *ʕʷ- rather than a prothetic vowel as proposed by Cowgill and Beekes (1969). What would be the motivation for prothesis? Greek does not have a constraint against initial n-. (Does any language have such a constraint before o? In Korean, n- was lost before i and y: e.g., 李 Ri > Ni > I 'the surname Lee'.)

Another possible initial PIE cluster is *ʔn- which would normally become Greek en-. Greek o- could be due to *e- assimilating to a following *o.

15.5.27.22:39: SOLVING FOR X IN MALTESE

-a remnant of a convention that once existed on the Italian peninsula (and perhaps still exists there, albeit in a language other than Italian or Sicilian)?

- influenced by the orthography of some European language which Maltese is not in direct contact with: e.g., Portuguese and Catalan?

- a Maltese-internal innovation possibly motivated by a one-sound-per-symbol principle (though the digraph għ from my last post is not compatible with that principle).

15.5.26.23:57: MYSTIFIED BY MALTESE VOWEL BENDING

I was surprised to see this in the Wikipedia article on Maltese:

/ɐɪ ɛɪ/ represented by għi, and /ɐʊ ɛʊ ɪʊ ɔɪ ɔʊ/ written għu.

I only had a vague memory of għi standing for /ɛɪ/ and għu standing for /ɔʊ/. I was particularly surprised by the equation of għu with /ɔɪ/. So I went to the source (Borg and Azzopardi-Alexander 1997: 299) and found that the passage should be rewritten as follows:

/ɐɪ ɛɪ/ represented by għi, and /ɐʊ ɔʊ/ written għu.

/ɛʊ ɪʊ ɔɪ/ are written as ew, iw, and oj.

I have long assumed that the spellings with għ are historical and point to a time when there was an 'emphatic' consonant (a voiced uvular fricative corresponding to Arabic gh?) that conditioned the bending of the following vowel before disappearing. Cf. how 'emphasis' (pharyngealization) conditioned the bending of *i and *u in Old Chinese before disappearing:

*Cˁi > *Cˁei > *Cei

*Cˁu > *Cˁou > *Cau

I wish I could confirm my guess by consulting a work on Maltese historical phonology.

I also wish I knew why għi and għu each have two readings. Those readings can't be allophones because they are in brackets: i.e., they are phonemic. Do the spellings reflect a period before a phonemic split conditioned by a factor that has now been lost?

15.5.25.23:59: WHAT IS THE ORIGIN OF UVULARIZATION IN QIANG? (PART 2)

In part 1, I asked,

Did MLQ [Mawo and Luhua Qiang] merge its equivalents of 'Grade I' and 'Grade II': i.e., is MLQ QV^ʁ from *QV with a plain vowel and *QV^ʁ with a uvularized vowel?)

I suspect that uvularization is secondary in at least some MLQ words with uvular initials: e.g., Mawo and Luhua Qiang qaʶ 'I' whose external cognates lack any trace of a medial *-r-. Other possible examples are 'afraid/fear', 'fish', 'Chinese', and 'chisel' below.

Luhua Qiang χuʶ 'tiger' looks like a loan from modern Mandarin 虎 hu 'tiger'. Could secondary uvularization be very recent in MLQ?

4046 1khi2 'bitter'

Normally I reconstruct medial *-r- as the source of Grade II in Tangut, so it is initially tempting to regard uvularization in the Luhua form as having beem conditioned by a lost *-r-. However, once again there is no external evidence for a medial *-r- (e.g., the cognate Tibetan root is kha, not *khra), so I think some Tangut velar-initial Grade II syllables originally had uvular initials with secondary uvularized vowels:

*qʰa > *qʰVʶ > 1khi2 [kʰiʶ]

(I do not specify the vowel in the intermediate stage since I don't know if raising in Tangut preceded or followed uvularization.)

Possible Tangut cognates of other uvularized MLQ words in Evans et al. (2015) are not Grade II: i.e., they lacked medial *-r-. (Syllable-final numbers indicate Tangut grades: e.g., 1khi2 above is Grade II.)

Some of those words may be unrelated: e.g., the Tangut word for 'chisel' is probably a loan from Middle Chinese 鑿 *dzak. The sound correspondences between MLQ and Tangut are not yet known, so I am not able to easily distinguish between true cognates and mere lookalikes.

Could a uvular affix absent in Tangut have conditioned MLQ uvularization?

5.26.2:34: Mawo Qiang nuʶ and Luhua nuʶ-tə 'ram' might be from MLQ nu 'sheep' plus a uvular affix: e.g., *ʁ-nu. However, Evans (2001: 298) listed the Mawo word for 'sheep' as ȵu with a palatal initial instead of n-. I wish I had Qiangyu jianzhi on hand to check the word.

5.26.2:39: Luhua Qiang suʶ 'ten' has no Tangut cognate, but it does resemble Pyu <sū> (Krech 2012's <sav>) ~ <sau> 'ten'. Pyu had initial <sr> in native words (e.g., <srūḥ> 'relative'), so the simple <s> of 'ten' cannot be from *sr- unless there was a chain shift: *Xr- > *sr- > s-.

(Pyu has <h> in <hoḥ> 'three' corresponding to s- elsewhere: e.g., Tangut 1soq1 'three'. That may imply Pyu <s> was once something else that filled the gap left by original *s- when it lenited to *h-.)

15.5.24.23:59: WHAT IS THE ORIGIN OF UVULARIZATION IN QIANG? (PART 1)

I forgot to make one point in my last entry. It seems that a lot of Chinese historical phonological studies are conducted in a vacuum without much reference to other Sino-Tibetan languages, let alone general phonological typology. Even Sinoxenic (Sino-Vietnamese, Sino-Korean, Sino-Japanese: i.e., systematic borrowings of Chinese) and transcriptive data are not getting as much as attention as I think they deserve. A better (I dare not say 'true' or 'correct') reconstruction of the history of Chinese should take into account the bigger picture.

One of the reasons I like Norman's pharyngeal theory for Old Chinese (OC) is that it makes sense both areally and typologically; it makes OC like its 'Altaic' neighbors (see Norman's 1994 article for details) and it allows nongenetic parallels to be drawn between OC and Semitic. It was a chance look at Maltese that convinced me Norman was right; pharyngeals conditioned vowel lowering in both languages: e.g.,

Imġarri Maltese [anté͜ik] < *antˁiːk 'ancient' (loan from Italian; Camilleri & Vanhove 1994: 104)

MC *tek < *tejk < OC 弔 *tˁikʷ 'arrive' (but *tˁekʷ is also possible; is there any rhyming evidence pointing to one or the other vowel? Baxter and Sagart 2014 regard the vowel as ambiguous. My guess is that the word was originally *tekʷ with pharyngealization developing before the lower series vowel *e. If the word was *tˁikʷ, its pharygealization would reflect a lost presyllable with a lower vowel: *Cʌ-tikʷ > Cˁʌ-tˁikʷ > *tˁikʷ.)

The vowel changes in OC are also similar to those in Khmer, though the conditioning factor in Khmer was voicing rather than pharyngealization: e.g.,

Khmer [əj] < *iː after voiceless consonants

Late OC *ej < *i after pharyngealized consonants

Last night I proposed that uvularization was conditioned by pharyngealized (= 'emphatic') initials followed by uvular allophones of */r/ in OC and Tangut. Does uvularization in Qiang have a similar origin?

The distribution of uvularized vowels in Mawo and Luhua Qiang (MLQ) as described in Evans et al. (2015) suggests that exact parallels cannot be drawn between Qiang on the one hand and OC and Tangut on the other:

1. Chinese and Tangut contrasted plain and uvularized vowels ('Grade I' and 'Grade II') after reflexes of *uvulars, whereas only uvularized vowels can occur after uvulars in MLQ (Evans et al. 2015: 24). (5.25.1:50: Did MLQ merge its equivalents of 'Grade I' and 'Grade II': i.e., is MLQ QV^ʁ from *QV with a plain vowel and *QV^ʁ with a uvularized vowel?)

2. If I understand Evans et al. (2015) correctly, MLQ permits both plain and uvularized vowels to occur in uvular QC-clusters. In theory, both QRV and QRV^ʁ might exist in MLQ, though I cannot find any examples in Evans et al. (2015). On the other hand, Chinese and Tangut only had uvularized vowels after reflexes of *QR-clusters.

3. Some uvularization in MLQ is due to right-to-left spreading: e.g.,

Luhua kʰɹa 'eight' + suʶ 'ten' = kʰɹaʶ-suʶ 'eighty' (Evans et al. 2015: 29)

(5.25.1:51: I think uvularization in vowels after velars is exclusively secondary in MLQ: i.e., there are no isolated monosyllabic roots combining velar initials with uvularized vowels.)

This phenomenon has no known parallel in Chinese or Tangut. (I reconstruct left-to-right emphatic spreading in those languages.)

Although Tangut is more closely related to Qiang than to Chinese, Tangut areally aligns with Chinese at least as far as uvularization is concerned if my interpretation of Grade II is correct for both languages.