19.6.15.23:50: WHY DON'T FINAL FRICATIVES DEVOICE IN TURKISH?

In my last post, I didn't comment on the final consonants of Arabic Muḥammad and Turkish Mehmet. Turkish final stops and affricates devoice in final position: e.g., Arabic kitāb > Turkish kitap 'book' (but acc. sg. kitab-!). Note, however, that the etymological -d of Mehmet does not survive in the spelling of the accusative singular: Mehmet'i [mehmedi]. (The apostrophe separates a proper name from a suffix. The rule is to keep the spellings of proper names intact regardless of actual pronunciation.)

Note also that I spoke of final stops and affricates devoicing but not fricatives: /ʒ z v/ remain voiced in final position unlike their Russian counterparts.

Tonight I realized that /z v/ are phonetically fricatives but behave like sonorants. Final /z/ in native words comes from Proto-Turkic *-r. It is a former sonorant that still behaves like one. /v/ acts as if it were /w/. I think /ʒ/ is only in borrowings like bej 'beige' and garaj 'garage'; it may retain its final voicing by analogy with /z/. And/or such borrowings postdate devoicing. (When did devoicing occur?)

(I am reminded of how traditional Tangut phonology groups z- and zh-sounds with liquids in consonant class IX rather than with s- in consonant class VI and sh- in consonant class VII.)

One problem with the above analysis is that /r/ devoices in word-final position. So if /z/ is really like /r/, why doesn't it devoice like /r/? And if I understand Kornfilt (2009: 524) correctly, speakers who devoice /r/ also devoice palatal /lʲ/ and may even devoice velar /ɫ/. (Göksel and Kerslake 2005: 8-9 do not mention the devoicing of laterals.)

19.6.14.23:50: HOW DID MUḤAMMAD BECOME MEHMET?

Originally this post was titled "Why Doesn't Muḥammed Have Ü?". But the answer to that question is simple: Arabic /u/ was borrowed as Ottoman u both before and after Arabic pharyngeals. I mistakenly thought vowels in Ottoman borrowings from Arabic were determined only by preceding Arabic consonants.

Originally the intermediary vowels in the Arabic Muhammad were completed with an e in adoption to Turkish phonotactics, which spelled Mehemed, and the name lost the central e over time. Final devoicing of d to t is a regular process in Turkish. The prophet himself is referred to in Turkish using the archaic version, Muhammed.

I thought Mehmet was a Turkish version of Arabic Maḥmūd, but they are only related because they share the same M-Ḥ-D root. The two names are distinct in Arabic spelling: Meḥemmed (now Mehmet) and Muḥammed are both محمد <mḥmd> like Arabic Muḥammad (Buğday 2009: 220), whereas I suppose Turkish Mahmut (Ottoman Maḥmūd?; the name is not in Buğday 2009) is محمود <mḥmwd> like Arabic Maḥmūd.

Turkish Mahmut < Arabic Maḥmūd has a for the same reason that Ottoman Muḥammed has u: a neighboring Arabic pharyngeal. (Contrast with Ottoman mühimmāt < Arabic muhimmāt 'important matters' in which /u/ has no pharyngeal neighbor.)

On the other hand, Turkish Mehmet < Ottoman Meḥemmed < Arabic Muḥammad has a first e where I would expect an u before an Arabic pharyngeal. And the second e of Ottoman Meḥemmed occurs where I would expect an a after an Arabic pharyngeal.

The key word is "Arabic". Turkish doesn't have pharyngeals. Here's what I think might have happened. Turks heard Arabic [muħammæd] and borrowed it in harmonized form ("in adoption to Turkish phonotactics" as Wikipedia put it) as *Mühemmed. (I assume the borrowing of Arabic /a/ as a in the presence of pharyngeals was a learned practice only possible to those who were literate: i.e., aware of a graphic if not a phonetic distinction between Arabic glottal ه <h> and pharyngeal ح <ḥ>, both borrowed as [h] in Turkish.) The first vowel was then irregularly assimilated to the other two e: Mehemmed (written etymologically in Ottoman as <mḥmd>, transcribed here with vowels and unwritten gemination as Meḥemmed, a compromise between the pronunciation and the spelling).

The relationship between Mehmet and Muhammed is slightly like that between the Korean and Japanese words for 'Buddha' on the one hand and the Sino-Korean and Sino-Japanese morphemes for 'Buddha':

The two columns represent two kinds of borrowing. All of the above forms are based on Chinese 佛 'Buddha' (itself a borrowing from Indic Buddha). But the forms in the first column cannot be mechanicaly derived from Chinese like those in the second column. The former were idiosyncratically borrowed as single items and not as part of an entire lexicon complete with systematic conventions of pronunciation. (Chinese is to Korean and Japanese what Arabic was to Ottoman.)

Adding to the idiosyncracy are suffixes absent from Chinese. Early Korean *-ke and early Japanese *-ka- seem to be a Koreanic morpheme 'ruler' which may have continental origins: cf. Khitan qa 'khan'. Japanese *-i is a noun suffix.

I cannot explain the *o in Japanese. Perhaps there was a lowering of *u in Paekche, the likely donor language. But there is no other evidence of such lowering. The general tendency in early Japanese was toward raising, not lowering: pre-Old Japanese *o became Old Japanese u, not the other way around.

19.6.13.21:13: WHY DOES MÜHACIR HAVE Ü?

After the ethnic cleansing of Phocaea, muhacirs settled in what is now Foça.

The Turkish word muhacir [muhadʒir] 'migrant' is from Arabic muhājir. I was surprised that the Azerbaijani counterpart is mühacir with ü. I would understand fronting a foreign u to make a word conform to vowel harmony, but mühacir is even less harmonic than muhacir (which would be ˟mühecir or ˟muhacır if it were fully harmonic).

On the basis of these two words (dangerous!), I expected that Arabic u after nonemphatic consonants such as m was be borrowed into Turkish as back u and Azerbaijani as front ü.

And I was wrong. Buğday's The Routledge Introduction to Literary Ottoman (2009: 11) explains:

The pronunciation of short vowels in Persian and Arabic words is generally governed by which consonants appear before and after the vowels. Arabic vowel graphs are as a rule interpreted as front vowels in Ottoman (üstün = e, kesre = i, ötre = ö, ü). There is nonetheless a group of consonants that cause front vowels in their environment to shift their point of articulation and become back vowels (a, ı, o, u).

Those consonants that shift vowels from front to back are: ح ḥ, خ ḫ, ص ṣ, ض ż, ط ṭ, ظ ẓ, ع ,` غ ġ, ق ḳ. The remaining consonants retain the front articulation of the vowels:

ب b, پ p, ت t, ث s, ج c, چ ç, د d, ذ z, ر r, ز z, j, س s

ش ş, ف f, ك k, ل l, م m, ن n, و v, ه h,ی y

I have long known about Turkish e for Arabic a, and that has never surprised me since [æ] is an allophone of Arabic /a/ and is the phonetic value of Persian short a.

Arabic [æ] > Persian [æ] > Turkish e

But neither Arabic nor Persian have front rounded vowels, so I didn't expect this shift:

Arabic [u] > early New Persian [u] > Turkish ü (less commonly ö and rarely o)

(Modern Persian has lowered [u] to [o].)

ö is particularly odd in `Ömer after `ayn which normally should favor a back vowel: e.g., in sā`at [saːʔat] 'clock'. (Turks could not pronounce `ayn [ʕ], but they did replicate the backness of /a/ after /ʕ/ in Arabic.) Did the first vowel front to match the frontness of the second vowel?

o in `osmān 'Uthman' is understandable since a mid [o] approximates the lowered allophone [ʊ] of /u/ after `ayn.

So although I initially thought that Turkish mücahit 'jihadi' (cf. Azerbaijani mücahid) < Arabic mujāhid was irregular, in fact it is regular, and the real question is: why isn't Turkish muhacir 'migrant' ˟mühacir with a front vowel?

Another question is: Why does the word 'jihadi' have u in Uzbek mujohid (cf. Tajik mujohid with the Tajik-internal shift o < ā) and modern Uyghur mujahit? Is there an east-west split in the way Arabic u is borrowed in Turkic? Do Uzbek and Uyghur reflect Chagatai borrowing practices? Did Chagatai and early Turkish speakers perceive Arabic /u/ in nonemphatic environments differently?

Turkish fronting of nonemphatic vowels interests me because it reminds me of the Mandarin palatal reflexes of Middle Old Chinese nonemphatic vowels in Mandarin: e.g.,

(Not all *k-nonemphatic vowel sequences have palatal reflexes in Mandarin. *k- that palatalized early became *tɕ- which in turn became [tʂ]: e.g., 支 *ke > *kie > *tɕie > *tɕi > [tʂɻ̩] 'branch'.)

Norman (1994) was the first to make the connection between Arabic emphasis/nonemphasis and what Pulleyblank called the type A/B contrast in Old Chinese (which Norman interpreted in terms of pharyngealization).

19.6.12.22:42: HOW DID PHOCAEA BECOME FOÇA?

Today is the centennial of the massacre at Φώκαια <Phṓkaia> /fokea/ [focea] 'Phocaea', now Turkish Foça /fotʃa/ [fotʃa]. (What would its Ottoman spelling have been? فوچا <fwčʔ>?).

I was surprised by the correspondence betwen Greek /kea/ [cea] and Turkish /tʃa/ [tʃa]. In theory Greek /fokea/ [focea] could have become Turkish ˟Fokea /fokea/ [focea]. But maybe the local Greek and Turkish versions of the name are closer: e.g., if the local Greek dialect had shifted *ea to [ja] and if the local Turkish dialect had merged [c] and [tʃ], etc. Or maybe I'm just seeing regular borrowing conventions at work reflecting an earlier time: e.g., if Greek /k/ had palatalized to [c] before /e/ before Turkish /k/ did, then the closest Turkish equivalent of Greek [c] at that time would be [tʃ].

Having spent so many years studying Sinoxenic - systematic Chinese borrowings in Vietnamese, Korean, and Japanese - I'm accustomed to regularity in borrowings. And unusual features are usually not random noise. They generally reflect lost features: e.g., dentals in Sino-Vietnamese reflecting old southern palatalized labials, the -l of Sino-Korean reflecting an old northern final liquid absent from any living Chinese language, etc.

Middle Chinese 必 *pit 'necessarily'

> Sino-Vietnamese tất [tət] < *sət < *psət < *pʲət in Annamese Middle Chinese

Ferlus (1992) reconstructed earlier Sino-Vietnamese *pz-, but I have never seen that cluster in any Mon-Khmer language

the schwa is an interesting deviation from the Chinese norm I'll explore later

> Sino-Korean phil < *pir in northern Middle Chinese

the Korean aspiration is irregular and may be due to hypercorrection

So I'd like to think there's some significance in the correspondence betwen Greek /kea/ [cea] and Turkish /tʃa/ [tʃa]. But maybe there isn't any. The elite of Vietnam, Korea, and Japan looked up to Chinese and wanted to closely emulate Chinese pronunciation, whereas Turks had no motivation to closely emulate the pronunciation of their Greek subjects. Greek εἰς τὴν Πόλιν [is tim bolin] 'to the city' became İstanbul, not ˟İstimbolin.