The hiatus of διαζύγιο "divorce"

Eighty year old recordings of popular music should tell you, for a normal language, how that language has changed in the interim. And so it is for Greek, as I'm finding by listening to the collected recordings of Markos Vamvakaris, 1933–1937. The catch is, diglossia has meant Greek is not a normal language; and the artificial influence of Puristic has meant that 80-year-old Greek is in some words more archaic, and in some words less archaic, than the modern Standard.

This is not the first time someone has thought of mining rebetika recordings for linguistic evidence, btw; Amalia Arvaniti & Brian Joseph have published a study of what old recordings tells us about the shift in Athens fron [mb] to [b]—another shift which the hand of Puristic has weighed heavy on. Whether this has been looked at as well—well, I'm behind in my reading.

The phenomenon noticable in Vamvakaris, as an intrusion of Puristic, is hiatus (χασμωδία). Hiatus is the phenomenon of a vowel next to another vowel counting as two syllables; for example, συκέα /si.ˈke.a/ "fig tree" or Τουρκία /tur.ˈki.a/ "Turkey" being three syllables. Hiatus is something the Modern vernacular does not do.

In fact, hiatus was something that Attic Greek wasn't comfortable with either— which is why a lot of vowel pairs and triples in Homeric Greek ended up reduced in Attic, in various forms of contraction and elision. The more vernacular the Attic, the more reduction in vowels; that's why only in Aristophanes, our most vernacular of Attic authors, do we get contractions like πρωὐδᾶν [prɔːudân] from προαυδᾶν [proaudân] "to decalre beforehand".

The modern vernacular went further in getting rid of hiatus than the ancient vernacular. In particular, the modern vernacular reduces vowels next to other vowels, making them non-syllabic. So while the medieval form for "fig tree" is συκέα /si.ˈke.a/, the modern form is συκιά /si.ˈkja/, and just two syllables long. Likewise, διαβάζω "I read" is pronounced /ðja.ˈva.zo/, not /ði.a.ˈva.zo/. In fact, it's a universal rule of the vernacular (outside of some archaic dialects, notably Zante and Mani), that /e/ or /i/ before another vowel, stressed or not, turns into [j]. In phonetic notation, e, i > j /_V.

(The Attic form of "fig tree", incidentally, was not συκέα /sykéaː/, but συκῆ /sykɛ́ː/. Attic used vowel contraction as its own way to address hiatus.)

The universal rule e, i > j /_V means that the vernacular form of "Turkey" should be not Τουρκία /tur.ˈki.a/, but Τουρκιά /tur.ˈkja/. Tell that to a Greek now, and they'll guffaw: calling Turkey Τουρκιά has the prerequisite of the speaker wearing a kilt and brandishing a musket. (That is to say, people now only know of the form Τουρκιά from folksong.) Tell that to a Greek two hundred years ago, when people were wearing kilts and brandishing muskets—and they'd agree that Turks indeed came from Τουρκιά. That was the word for Turkey back then, and it followed the normal development of Greek vowels.

So why has the modern standard gone backward? What has changed in the past two hundred years to make the language evolve backwards? Puristic, of course; that's the kind of thing Puristic does.

Puristic pronounced words as they were spelled in Ancient Greek (though it conveniently ignored the fact that the pronunciation of the letters had changed, which made several spelling pronunciations awkward). Iota was a vowel, so in written Greek it should not be pronounced as anything but a vowel: it couldn't be pronounced as a [j]. When in fact people did start writing down the vernacular in the 19th century, under the shade of Puristic, the vernacular's [j] were such a nuisance for how Puristic treated spelling, that they got a diacritic added to them, as I have written up elsewhere: vernacular δι̯αβάζω /ðja.ˈva.zo/ vs. learnèd διαβάζω /ði.a.ˈva.zo/. As Yannis Haralambopoulos has astutely pointed out, the vernacular wasn't so much written down in the 19th century, as transcribed.

So if a word was vernacular, it had /j/'s; if a word was learnèd, it had /i/'s. Even if was exactly the same word, the pronunciation difference means Standard Modern Greek now has minimal pairs to distinguish the two.

• άδεια as a vernacular word, "empty", has two syllables, [ˈa.ðja]; the selfsame άδεια as a learnèd word, meaning "leave of absence" (an "empty" day), has three syllables [ˈa.ði.a].
  
• βιάζω in its vernacular sense of "hasten" is two syllables, /ˈvja.zo/; the verb originally meant "to force", and its learnèd meaning of "rape", it has three syllables, /vi.ˈa.zo/. (The vernacular word for "rape" is bound to the society that used it before Puristic: ατιμάζω "dishonour", so it no longer has the same connotations as βιάζω.)
  
• Ancient παιδία /pai.dí.a/ "children" has survived in the vernacular as παιδιά /pe.ˈðja/. Ancient πεδία /pe.dí.a/ "fields" has been revived in Puristic (denoting fields of expertise, or force fields); being learnèd, it is pronounced /pe.ˈði.a/.

And because the modern linguistic standard is a messy mixture of Puristic and vernacular, the two phonological systems coexist: some registers, and some words, follow the rule e, i > j /_V, some do not.

Having two phonological systems coexist is workable; in fact, that's how Standard Modern Greek works. To a contemporary speaker of Greek, it is inconceivable that "fig tree" should be anything but /si.ˈkja/, and it's just as inconceivable that "Turkey" should be anything but /tur.ˈki.a/. To a neogrammarian like Psichari, adamant on the Exceptionlessness of sound change, this jumble was unacceptable: the language of Greece should be the vernacular, and it should follow vernacular rules. It's what makes linguistic sense.

In a Greece not under the shadow of Puristic, linguistic sense could have prevailed. Vincenzo Cornaro, three centuries before Psichari, managed to slot into his Erotokritos a few learnèd words in the vernacular garb Psichari called for: μοιότη /ˈmjo.ti/ "similarity" (V 126), as an updating of ὁμοιότης /o.mi.ˈo.tis/, is pure Psichari, and βέβαιο "certain" in IV 1599 is clearly meant to be two syllables, /ˈve.vjo/ (ό,τι γενή σ' αυτούς τσι δυο βέβαιο να το κρατούσι).

But the 1880s wasn't the 1590s; Psichari's call for consistency was dismissed as extremist, not least when the Demotic was finally enshrined as the official language of Greece, as "a Demotic without dialect features or extremism" (άνευ ιδιωματισμών και ακροτήτων). The words now are βέβαιο /ˈve.ve.o/, and ομοιότητα /ˈo.mi.ˈo.ti.ta/: the inflection is vernacular. but the phonology isn't. What has been enshrined, not just by the state but also by the people, was the compromised language Psichari abhorred.

Not just Psichari abhorred it, either. Compromises don't just offend against neogrammarian linguistics, they also offend against literary aesthetics. Cornaro is renowned for avoiding hiatus studiously. But not completely: learnèd words have always been a perturbance in the vernacular. Even in the seeming Eden of Erotokritos, some of the learnèd loans were still too much under the shadow of antiquity to escape hiatus: the 1590s wasn't really the 1590s either.

• Cornaro's εβεβαίωσεν "he assured" IV 1512 is the same stem as βέβαιο, but here is clearly /e.ve.ˈve.o.sen/.
• Cornaro uses περικεφαλαία "helmet" three times, each with hiatus as /pe.ri.ke.fa.ˈle.a/.
• As Nikos Sarantakos has written, Georgios Souris mocked the Psicharist Alexandros Pallis' περικεφαλιά /pe.ri.ke.fa.ˈlja/ in Pallis' translation of the Iliad, which modernised the word Cornaro didn't dare modernise. But even Pallis vacillated between hiatus and non-hiatus for this stubborn learnèd word. In the modern standard, it has remained with a hiatus.

So when Demotic started being manumitted as a literary language, authors started hankering after an uncontaminated, pure Demotic. As Peter Mackridge likes to argue, literary Demotic is itself an artificial, puristic code. And Greeks have been looking since for a holy grail of untutored, internally consistent vernacular, clean and plain-spoken. Like Cornaro's Erotokritos (if you ignore his hiatus—and he really does not have much of it.) Like Makriyannis' Memoirs (so long as you concentrate on his vibrantly oral syntax, and don't notice the telltale final /n/s that Makriyannis took in from the Puristic he heard every day.)

Or just like the song lyrics of Markos Vamvakaris, the abbatoir's apprentice and hashish addict who begat a new style of music. The world of street thugs and addicts he depicted was as common as dirt; surely it should be as vernacular as common dirt, too.

But of course, Markos was not brought up in cotton wool in Arcadia; he went to school in Puristic, and interacted with officialdom in Puristic, not least in his brushes with the law. His baseline is vernacular, of course; but Puristic words do show up in the lyrics, and their phonology stands out all the more because the baseline is more consistently vernacular than the contemporary language is.

After all that preamble, I'll give just one example here; I have a more extensive example next post. First, Markos dedicates a song to his divorce in 1936, and the word for divorce, διαζύγιο, turns up twice in the song (0:25, 2:57).


The word διαζύγιο is learnèd, as you would expect of the Modern Greek legal system; the vernacular has a verb for divorcing, χωρίζω "to separate", but the noun χωρισμός "separation" is generic enough to mean "parting". Being a learnèd word, διαζύγιο has learnèd phonology; so it has five, Puristically-correct syllables: /ði.a.ˈzi.ɣi.o/. That is how Vamvakaris sings it, both times in the song. The hiatus is completely out of place against Vamvakaris' vernacular.

The hiatus is also out of place, it turns out, against the standard of 2010, because the word for "divorce" is now familiar enough to yield a syllable. In a concession to vernacular phonology, διαζύγιο now has four syllables: /ðja.ˈzi.ɣi.o/. Its first syllable follows the vernacular rule, like διαβάζω. Its last syllable does not.

Admittedly, the Triantaphyllidis dictionary keeps the old pronunciation, and gives it as [δiazíjio] (i.e. [ði.a.ˈzi.ʝi.o]). But a longitudinal study of YouTube says differently:

Πέτρος Κυριάκος: Η γραμματική του μάγκα (1932) (1:30: 5 syllables)

Μητσάκης-Τατασόπουλος: Το Διαζύγιο (Να πας) (1953) (2:26: 5 syllables)

Κώστας Βίρβος, performed by Βαγγέλης Περπινιάδης: Διαζύγιο Θα Πάρω, (1950s?) (0:15: 5 syllables)

Πάνος Γαβαλάς: Διαζύγιο (1950s?) (0:46: 5 syllables)

Γρηγόρης Μπιθικώτσης: Τι τραβάω (1963) (2:31: 5 syllables)

Γιώργος Ζαμπέτας: Ο Πιτσιρικάς. (1966) (1:04: 5 syllables)

Θέμης Ανδρεάδης - Νατάσσα Γερασιμίδου: Το διαζύγιο (1977) (2:36: 5 syllables)

Παπαδοπουλος-Δημητριου: Το διαζύγιο (1991). (0:26: 4 syllables)

ΣΥΓΑΠΑ ΦΑΚΕΛΟΣ "ΔΙΑΖΥΓΙΟ ΚΑΙ ΠΑΙΔΙ" Νο 1 (2009) (0:12: 4 syllables)

Conn-x And the star is you- Θέλω διαζύγιο. (2010) (0:12: 4 syllables)

Το διαζύγιο της Μιμής Ντενίση. (2010) (1:03: 4 syllables)

Getting dates for some of the tracks was not possible; and because songs get rerecorded in Greece a lot more than in the Anglosphere, there was a high margin of confusion. Before I found Mitaskis' 1953 song, for example, I found Glykeria's 1993 cover—which of course used the same number of syllables for διαζύγιο (2:26), by now anachronistically:


Still, there seems to be a story here. Before the abolition of Puristic, διαζύγιο is sung like in Puristic; by the 1990s, it loses its first syllable. What I'm missing is songs from the period when the transition must have happened, looks like between 1977 and 1991. In other words, the '80s, when Demotic was enshrined as the state language, and the fervour of the socialist government to do away with Puristic swept away several formerly accepted shibboleths—such as the archaic first declension vocatives like καθηγητά.

This doesn't mean Psichari has had his revenge. In fact, what has happened rather messier linguistically, though unsurprising sociolinguistically. Instead of two phonological systems, one Puristic and one vernacular, we now have a spectrum between the two systems, and words of learnèd origin are adrift between the two rules. So διαζύγιο now starts vernacular, and ends learnèd: /ðja.ˈzi.ɣi.o/, not /ði.a.ˈzi.ɣi.o/ or /ðja.ˈzi.ɣjo/. διάβρωση "erosion" can lose its hiatus (1:22, 2:25, /ˈðja.vro.si/):

or keep it (0:04, 0:13, /ði.ˈa.vro.si/):

The speaker at 2:50 in the first video does not use the word διάβρωση, but he is overloading on hiatus anyway, in a rather pompous kind of public speaking. So hiatus is now clearly a register feature—and unpredictable: it's a sociolinguistic variable, rather than a purely linguistic variable.