Subscribe to Blog via Email
October 2023 M T W T F S S « Jul 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
My original post on the clumsy coinages of terms for paraphilias was a bit of careless venting on Quora, and did not bother researching the creation of the words too deeply. It was, as commenters at Nikos Sarantakos’ blog correctly identified, some xavales, “goofing off”. (Sarantakos did say when I sent him the post, “I think you are being a little harsh on John Money.”) And it was not the kind of thing I should be saying with my linguist hat on: with my linguist hat on, autassinophilia was not intrinsically worse than automobile as a hybrid form (though, by invoking the Asāsiyyūn, it was a more bizarre form.) Eproctophilia is, in fact, intrinsically worse; but it’s not the first dumb coinage that has become a real word in language, and it won’t be the last.
When I updated the post for Sarantakos (which he titled “Fifty shades of paraphilia”), I thought I should put a somewhat less goof-off addendum, about what the clumsy coinages were demonstrating:
For centuries, the scientific terminology of Western science had its footing on classical languages, because classical languages were the languages of learning. In the 20th century, that footing dropped away: classical languages are not particularly known outside of classical philology (and often enough are barely known even there.) Consequently, newer Western scientific terminology does not pay particular attention to Latin and Greek. The terminology of IT in particular is playful, simple, and Anglo-Saxon; and that results in clumsy solutions when we make those terms learnèd again in Greek. Bubblesort is much more spontaneous than ταξινόμηση φυσαλίδας “classification by gas globule”.
Not knowing Latin and Greek is a sign of the times. It’s the poor command of Latin and Greek that shows how marginal classical languages have become in scientific terminology. There are plenty of experts who want to come up with impressive Latin and Greek terms, because impressive Latin and Greek terms are still an institution in the sciences; but they don’t do it right, they come up with broken forms, and noone bothers to correct them. (At least, noone within the English-speaking world.)
So speciesism was coined in 1973 by a psychologist who was not particularly interested in Latin morphology, and so treated species as an uninflected form. As I analysed in ουγκανιά and αντισπισισμός, at least Romance languages corrected the word when they borrowed it (espécisme, specismo, especismo ). (Greek also changed it to spisismos, when it doesn’t nativise it as iðismos; but I believe that was just haplology.)
The case of coulrophobia, the phobia of clowns, is much sillier. Noone knows where the first person to launch it online in 1995 came up with it from, and after it was launched, noone took the time to check it. It might be a distortion of Modern Greek klooun < clown. It might have something to do with kōlobathron “stilts”, as is often claimed, because someone dug up that word, and thought that the closest Ancient Greek ever got to clowns was stilt-walkers. (That corrected etymology has made it back to Greece: Τι είναι η κωλοβαθριστοφοβία; Μήπως πάσχεις και εσύ και δεν το ξέρεις; “What is kolovathristofovia? Might you be suffering from it without even knowing abut it?”) But the opinion of etymonline rings truest: “Coulrophobia looks suspiciously like the sort of thing idle pseudo-intellectuals invent on the internet and which every smarty-pants takes up thereafter.”
Of course, as that article also points out, the word is not an official term in psychology, though the phenomenon is acknowledged there; and if it were an official term, the requisite plausibility check might have happened…
… Maybe, but Money’s acrobatics, and the unbelievable sloppiness of Aggrawal’s catalogue of paraphilias (in an otherwise serious monograph from the reputable publisher Taylor & Francis) shows that it probably wouldn’t have. Sloppiness, misinterpretation and solecisms in classicising terminology now go unnoticed, because classicising terminology is no longer a living process with many participants actively checking it. It is a dry convention, with little attention paid to it. And when we need to borrow its results back into Greek, we end up patching them up as best we can.
As several commenters chimed in (Diver of Sinks first),
Bad news for those of us who boast about the words we have given the Dumb Franks: we’ve already offered anything there was to offer (ό,τι δώσαμε δώσαμε).
There was a bit of discussion on that between me and commenter Neo Kid: yes, coinage based on Graeco-Latin roots is still going on. But it no longer particularly cares how true to Greek and Latin the results are; much less so than in the past. The Graeco-Latinity of the coinages is a much more empty gesture now. And commenter Nikiplos made the justified (if painful) observation that the Asian and African future pioneers of science are going to have a lot less patience for Graeco-Latin coinages than their European colleagues.
I was also justly asked by Neo Kid whether my expression of revulsion was a scholarly opinion or merely a matter of personal taste. It could only, of course, be a matter of personal taste, because as a linguist, all I get to say is, “shit happens”. That’s why I’d linked to Professor Higgins in the original post: Professor Higgins might well have said that Liza Doolittle should have been “taken out and hung for the cold-blooded murder of the English tongue”; but no linguist in the past century would issue those kinds of aesthetic (and classist) judgements ex cathedra.
Though that’s not the kind of lack of privilege in evidence with these coinages. And I’m not going to forgive eproctophilia soon. If you’re going to go to the bother of making up a Graeco-Latin word for “fart fetish”, at least take the time to look up an actual Greek or Latin word for “fart”, instead of coming up with “out-of.(Latin)-anus.(Greek)-philia”.
The crowd at Sarantakos’, I have to say, is tougher than on this obscure blog, or even on Quora. (And I’ve been blocking commenters much more on my return to it!) Commenter Ioanna was critical of the unserious part of my post, and the criticism is not unwarranted. I reproduce it here.
Dr Nick Nicholas, who is acknowledged to be an excellent linguist, should not have rushed to conclusions about the competence in Greek of a scholar, whose brilliant career he is completely unfamiliar with. The late John Money was one of the most preemiment American sexologists, professor for years at John Hopkins and Baltimore Hospital. Professor Money had Greek learning, had boundless admiration for the Greek language, and was in no way only familiar with it via dictionaries, as Dr Nicholas maliciously implies.
I would have expected the esteemed Dr Nicholas to have at least leafed through some of Prof Money’s books before writing this article. It is embarrassing for a linguist with Dr Nicholas’ reputation to have used Wikipedia to look up the meanings of kinds of paraphilias, and ignored John Money’s masterpiece Gay, Straight, and In-Between: The Sexology of Erotic Orientation (1988), where all the words are explained in detail. If he can’t afford to buy the book, he can say so and I’ll mail it to him.
For instance, this is what Money writes about erotophonophilia, which so offended the linguistic sensitivities of Dr Nicholas, clearly acting as a snob:
erotophonophilia: a paraphilia of the sacrificial/expiatory type in which sexuoerotic arousal and facilitation or attainment of orgasm are responsive to and contingent on stage-managing and carrying out the murder of an unsuspecting sexual partner (from Greek, eros, love + phonein, to murder + -philia). The erotophonophile’s orgasm coincides with the expiration of the partner. The reciprocal paraphilic condition is autassassinophilia. Syn. lust murder.
Is that anything to do with Wikipedia’s definition?
I’ll take my lumps. Erotophonophilia is clearly defined by Money as a murder fetish and not a death fetish, the way Wikipedia said it was. And using Wikipedia was sloppy; yes, it was deliberately sloppy, because I was writing a flippant post, but it was impugning a scholar in doing so. (Although I still don’t think autassassinophilia, let alone olfactophilia, are evidence of boundless admiration.) I also can’t confirm that eproctophilia was Money’s doing (although olfactophilia is on his 1988 list). But if he didn’t want to be judged by classicist snobs, he shouldn’t have used olfactophilia instead of smell fetish to begin with. The snobbishness is inherent in Graeco-Latin coinages.
Money’s reputation won’t suffer that much by me grousing about autassassinophilia, even if it was a cheap shot. There are far more serious controversies to his research anyway.
I have had an updated version of my old Quora post What is it called when you get aroused by watching people die? published in Greek on Nikos Sarantakos’ blog, as Οι πενήντα αποχρώσεις της παραφιλίας, “Fifty shades of paraphilia”.
I’m sure someone somewhere has already written about this. In fact, I’m sure multiple people have. But I enjoy reinventing the wheel.
(And getting inspiration from Quora, as well as Nikos Sarantakos’ blog, for articles here. My thanks to Evangelos Lolos for prodding me on this.)
Modern Greek used to have a lot of ethnonyms ending in -ezos, -eza, a suffix taken from Italian -ese, -esa. Those suffixes originally came as part of ethnonyms that were borrowed wholesale from Italian: englezos, ɣenovezos, frantsezos, olanðezos, kinezos “English, Genoese, French, Dutch, Chinese”.
With the advent of Puristic, many of those ethnonyms were replaced with Classical terms (anɡlos, ɣalos “Angle, Gaul”), and others just had the -ezos suffix stripped (olanðos).
The current status of the -ezos is therefore a messy mix:
- There are some ethnonyms for which the -ezos suffix is unavailable: frantsezos is long forgotten, and no form like ɣalezos has ever developed.
- There are some ethnonyms for which the -ezos suffix is the only available form: maltezos, kinezos, filipinezos “Maltese, Chinese, Filipino”. (Puristic did come up with malteos, sinas, but those variants did not stick for whatever reason)
- For a large number of Western ethnonyms, the -ezos suffix was retained, as a colloquial variant. So the normal forms are anglos suiðos olanðos, norviɣos, polonos, kanaðos “English, Swedish, Dutch, Norwegian, Polish, Canadian”, but the variants enɡlezos, suiðezos, olanðezos, norviɣezos, polonezos, kanaðezos are possible. They are however quite marked now: they sound folksy, potentially derogatory, and they are seldom used. In fact, the suffix is so marked now, that people even pause on instances like maltezos, where it is the only form in widespread use, or ɣiaponezos “Japanese”, where the alternative form i.aponas should be unacceptably archaic.
The oddity is with the female counterparts to those -ezos forms. While the masculines in -ezos are now marginal, the feminines in -eza are still in widespread use. So engleza, suiðeza, olanðeza, norviɣeza, poloneza, kanaðeza are nowhere near as restricted in usage as their masculine counterparts, and can turn up productively in unexpected quarters. (kanaðeza “Canadian chick”, for example, is the term for a small truck in military slang, because the first such vehicles in the Greek army were WWII surplus Canadian Military Pattern trucks. The term persists though they haven’t been used for decades.)
In fact, the -eza is being applied innovatively to ethnonyms where it has no historical warrant. “Albanian” for example was traditionally arvanitis, arvanitisa (now restricted to the Christian ethnic Albanians living in Greece), replaced by the learnèd alvanos, alvani. There has never been a form alvanezos, but the feminine alvaneza has emerged recently.
So a suffix that now sounds vaguely pejorative in the masculine flourishes in the feminine. Is the explanation “because Greeks are sexist”? Greeks are sexists, sure. But there’s a little bit more going on linguistically.
First, the new learnèd forms were learned by people by exposure, as much as by paradigm. The learnèd masculine forms featured in the press, because the press talked about newsmaking foreign men. The learnèd feminine forms did not feature anywhere near as often, because the press didn’t talk anywhere near as often about newsmaking foreign women. That’s sexism, but it’s the sexism of the entire world blocking women from making news, as well as blocking the papers from reporting it if they did. So there was less opportunity for the learnèd forms to displace the vernacular -eza forms, than there was for their male counterparts.
Second, the new learnèd masculines used the bog-standard inflectional ending -os, with zero-derivation; they posed no particular morphological challenge in Modern Greek. But some new learnèd feminines used the Ancient Greek ending -is, which is alien to Demotic; Demotic nativised it as -iða, and it does turn up in some vernacular forms (nɛːrɛːis “Nereid” > neraiða “fairy”), but there was some discomfiture with it, and it isnʼt always taken up. Generations of Greeks now have had no problem nativising platɔːn “Plato” to platonas and tʰɛːseus “Theseus” to θiseas, but feminine proper names like artemis, alkistis “Artemis, Alcestis” have resisted nativisation to the regular artemiða, alkistiða—to the extent that speakers would rather keep them indeclinable.
When there was no alternative, because the original vernacular form was banned, people shrugged and used the -iða form: eliniða “Greek woman” was obviously going to be hard to resist, and ɣaliða, ɣermaniða, rosiða, ispaniða “French woman, German woman, Russian woman, Spanish woman” were the only forms available after the elimination of the vernacular, Italian frantseza, alemana, rusa, spaniola. But if the old -ezos form survived at all, people seem to have jumped on it, as something more comfortable than the -iða form: thus poloneza for poloniða, enɡleza for anɡliða, suiðeza for suiðiða.
The zero-derivational feminines that also turned up as learnèd forms, like irlanði as the feminine counterpart of irlanðos “Irish”, clearly sounded odd to people, that they weren’t comfortable with; presumably, as marked feminine forms, they were too close for comfort. But I’m not sure it was just the zero-derivation that was the issue; there are plenty of zero-derivational feminines in the vernacular: eliniða replaced romja “Roman woman”, “Bulgarian woman” is vulɣara.
Nikos Sarantakos’ article on the issue (which I stumbled on after almost finishing this post) points out a more plausible reason why people hesitate over formal feminines in general.
Feminine ethnonyms are often difficult, and most of the time we find them in two or three forms, precisely because they have not been settled by usage [NN: my first point above on exposure], whereas one form dominates in the masculine. For example, we have alvanos “male Albanian”, but for a female Albanian we have alvani, alvaniða, alvaneza; the same goes for a female Ukranian [ukrani, ukraniða, ukraneza]. The -eza suffix is considered sometimes overly colloquial, and sometimes dismissive; but how else can you call an Irish woman or Dutch woman? irlanði of course in formal language, but also irlanðeza, just as you can’t do away with olanðeza, despite the existence of olanði; the same goes with suiðeza.
Feminines in general in Greek present an embarrassment of riches, and when novel feminines emerge, like “female doctor” or “female member of parliament”, there is considerable uncertainty as to what to use: the vernacular feminine endings often sound dismissive, because of past sexism, and there’s lots of them; the formal endings often sound awkward, because they’re learnèd (or because they use seemingly masculine endings, which can make sense for -os in Ancient Greek, but not in the vernacular).
In the instance of feminine ethnonyms, the formal language offers you a choice of zero-derivational -i, which may be a bit too adjectivey (because morphological markedness, too, is sexist); or -iða, which sits poorly in the vernacular.
That confusion can lead to mere avoidance. The article in fact is premised on the peculiar fact that Greeks aren’t sure what the feminine form of aravas “Arab” should be, at all; and the feminine of “Arab” is plagued by even more uncertainty:
- The original feminine is arapina, feminine of arapis. But that vernacular form no longer means “Arab”; because Greeks conflated the Sudanese Ottomans slaves they encountered with Arabs, its meaning has changed to “black person”: unmarked two centuries ago (“moor”), pejorative now. And arapina only survives in an orientalising song by Tsitsanis from 1946, that would have given Edward Said a heart attack. (Αραπίνες λάγνες ερωτιάρες, “Lustful erotic Arab women”. It really does sound so much worse when you translate it, and realise what he’s actually saying, as opposed to just singing along to the nice tune.) So aravas has been revived as the Demoticisation of the classical form araps, gen. arabos. But that revival did not bring a suitable feminine along with it.
- aravina is the analogy from arapina; but the -ina suffix is too colloquial for a learnèd stem
- aravisa is the most plausible vernacular form, but there isn’t enough precedent for -isa in learnèd ethnonyms either (whereas it is extremely common in demonyms within Greece)
- araviða is the expected learnèd form. But because noone actually got around to formally proposing it, it gets greeted by peals of laughter. As Sarantakos’ commenter Pepe found in high school.
- The situation is so bad that the first edition of Babiniotis’ dictionary, normally a den of conservatism, actually capitulates and gives ɣineka aravas “Arab woman”. Sounds fine in English. Is utterly humiliating in Greek, since aravas is masculine. (The second edition allows aravisa.)
A similar confusion on where to accent the genitive plural for feminine nouns has led Greeks to just avoid them completely, as my friend Dion Mertyris found in his thesis. For that matter, the controversy around how to interpret Esperanto tense in compound verbs led Esperantists to simply abandon the passive, which can only be expressed as a compound tense.
Faced with that conundrum of learnèd forms, Greeks are likely to run away too. At least they have somewhere to run away to: -eza has the advantage of familiarity and productivity, even if it is visibly informal. (And yes, sexism helps people get over the inconsistency between masculines and feminines; but that isn’t the only factor involved.) So people are happy to lean on it. That familiarity has in fact helped it expand to instances like alvaneza, ukraneza, where the corresponding masculines alvanezos, ukranezos have never existed.
Though noone to date has come up with araveza. That -eza suffix, being Italian in origin, is associated with the West too strongly, even if it also gets applied to the Far East (Chinese, Japanese); it can be applied to the Near West (alvaneza), but there seems to be a block to applying it to the Near East.
Updated post on “If you were allowed to add a symbol to unicode, what symbol would it be, and what would it mean?”
I have had an updated version of my old Quora post If you were allowed to add a symbol to unicode, what symbol would it be, and what would it mean? published in Greek on Nikos Sarantakos’ blog, as Πώς χαντάκωσα τα Παμφυλιακά, “How I ruined Pamphylian”—referring to how I’m responsible for the psi-like Pamphylian sampi not getting a distinct codeset from the regular alphabetic sampi.
Many posters were non-plussed by the highly technical discussion of Ancient Greek orthography and Unicode; a few made it through to the end. My especial thanks to commenter Efi-Efi, for chiming that she too lamented (θrino) the fate of the pitchfork (in Cretan dialect: θrinaki).
I’ve finally taken the time to read Günther Henrich’s 1976 thesis on the spread of the -o vocative and -o genitive in Greek.
My blog series has been something like 15 pp written off the cuff, with minimal research. Henrich’s is 270 pp of meticulous historical and dialectal research. He has orders of magnitude more detail than I provided, and his tracing of developments is much more fine-grained than I could have done. (Of course. After all, I spent a couple of evenings on this, and he spent several years.)
It’s an impressive piece of work, all the more so because it was done in 1976; and I’m always impressed with the kind of work people used to do with index cards instead of computer corpora. (My own PhD was at the tail end of that.) I haven’t been well disposed the recent Bible Code-like attempts by Henrich to read cryptogram signatures into the works of the Cretan Renaissance; but the quality of research in this work has more than made up for that.
And I’ll allow myself the gratification that he hasn’t contradicted me. The developments I posited, like the neuter for Manolio(s) “Manuel”, were confirmed. The claims of analogical factors that I’ve made are all there in his thesis: formality, penult stress, proper names, all given with quantitative justification. (Of those original third declension nouns like ɡerɔːn that have switched to the second declension, and which are the core of the o-vocative, 3/4 are penult-stressed, and 2/3 are proper names.)
He also a quirky-looking derivation for the o-vocative (and its related o-genitive), which seemed implausible to me at first, but which has grown on me:
|Early Mediaeval (ca. 1000): Indeclinable||ɣeron||ɣeron||ɣeron||ɣeron|
|Later Mediaeval; Cretan||ɣeros||ɣero(n)||ɣero(n)||ɣero(n)|
In other words, the old third declension in -ɔːn did not switch wholesale to the second declension (-os, -u, -o, -e); it moved one case at a time, and the vocative (subject to less analogical pressure than the genitive, as a more rare case) is the contemporary battleground. That means that Cretan, with its o-genitive corresponding to the o-vocative, is archaic rather than innovative, which is consistent with the o-genitive being more common in Mediaeval Greek. (That also makes more sense because it’s harder to think of the vocative or accusative influencing the genitive, which is what would be required if the o-genitive were innovative.)
Because Henrich had a lot more old data to work from than I did, and because he’s taken more of an old school historical linguistics approach, he spends a lot more time on how the phenomenon got started than how it spread. Not only does he analyse how the original nom. ɡerɔːn, gen. ɡerontos, voc. ɡeron gave rise to the o-vocative; he also introduces a lot of pathways for the o-vocative to have been continually imported into the language, mainly through borrowings (proper names from every Balkan people there ever was, and both proper names and common nouns from Venetian/Italian).
It is a subtle game to establish the relative importance of the factors that bring about language change: the initial reanalysis or borrowing, that brings the change into being, versus the subsequent analogy that generalises it through the language. My own opinion is that Henrich trusts etymology too much: I’m not convinced all those Albanian and Bulgarian names were imported complete with an o-vocative from the very beginning, and without the powerful analogy of proper names already in place, I don’t think the loanwords would have lined up so promptly.
And not all of Henrich’s etymological accounts have the same explanatory power. For example, he thinks that the o-vocative is particularly popular after velars, because it avoids the switch between palatal and velar, which makes the morphology less transparent (that’s how he explains [kakurɣos ~ kakurʝe] “criminal”, [likos ~ lice] “wolf”); but those allophonic switches happen in verbs and nouns constantly, and native speakers are not aware of them at all.
He also posits -ios > -os as a class of nouns that takes the o-vocative: that accounts for eɣiptios > ɣiftos “gypsy”, ɣeorɣios > ɣiorɣos “George”, and several diminutives like -akos > -akios. But that phonological simplification does not particularly explain why the vocative -ie would switch into -o. In fact the laundry lists of reanalysed and borrowed names that take the o-vocative (names formed from verbs! names formed from adverbs! Albanian diminutives! Slavic diminutives!) have a simpler underlying explanation—which he does mention in passing: novel name forms in general will tend to take the analogically spreading form, since the conservative force of the inherited e-vocative is not present there to push it back. All because the o-vocative is now the unmarked strategy for penult proper names (as he himself concludes, already in 1976: “all penult-stressed proper nouns ending in –os can potentially take an o-vocative.”)
And where the o-vocative pushes into previously unavailable terrain (like common nouns), I think semantic similarly counts for a lot more than phonology in how the analogical spread works. The vocative kakurɣo after all shows up next to the vocative ðolofono “murderer”, which has no velar—but a lot semantically in common with the main contemporary class of common nouns taking the o-vocative (murɣos “uncouth”, bufos “fool”, rufianos “informant”).
On the other hand, I probably underestimated the contribution of loanwords: there certainly appear to be a disproportionate number of Italian words among the common nouns that do use the o-vocative, and the dialects where the o-vocative has prospered most (including extending to antepenults) have been the dialects with the longest Italian contact, Cretan and Heptanesian. The fact that Nicolò was indeclinable in Cretan Greek notary documents up until 1500 (appearing as ŋikolo even in the nominative) shows that the o-vocative was certainly going to be reinforced by the influx of Italian names. I had guessed as much, but I hadn’t realised how much that extended to common nouns as well (like say filiotsos < figliozzo “godchild”). And many of the contemporary common nouns that take an o-vocative can be traced to Italian (kamarotos, kouniaðos, bufos, rufianos, tsarlatanos).
The other overall thing I noticed about our different accounts was that too much detail gets in the way of the big picture. That’s not necessarily a fault in an historical account, of course; but analogy does not work by appeal to etymological detail; it works by squinting the eyes, and blurring all the detail into large, arbitrary generalisations. Sort of like I did with my account. Which, after all, was seeking to explain not what had happened in the 11th or 16th century, but what generalisations explain what has been taking place over the last century.
It’s interesting to see both what hasn’t changed and what has. Renaissance Cretan had anticipated a lot of the developments taking place now in Standard Modern Greek (since the o-vocative was more advanced there); the modern military slang neo “newbie”, for example, turns up in a tragedy around 1600, and Phoebus (which Cretan dramatists had only heard of via Italian) gets an o-vocative, like any Italian name: Φοίμπο fibo. (That’s not the same thing as o-vocatives now spreading to learnèd names; but that’s how it gets started. In fact, a much more modern play, Petros Markaris’ translation of Brecht’s Life of Galileo, makes a point of having scholars in the play use the vocative ɣalilee, while commoners use the vocative ɣalileo.)
On the other hand, Henrich still found very little evidence of o-vocatives spreading to antepenult names, let alone common nouns; that’s something that has changed in the last four decades.
It’s also frustrating to realise the examples you don’t come up with if you’re thinking off the top of your head, instead of doing research. I wouldn’t have made the connection between my name, Nikos, and the Mediaeval name Nikōn; it’s too hard coded in my head that Nikos is short for Nicholas, but that doesn’t mean it can’t have initially been the regular development of Nikon. But I should certainly have realised that the Grim Reaper, Charon < Charos, was a canonical example of the shift of -ɔːn to the second declension (and its o-vocative is attested as early as the late 15th century, in Emmanuel Limenites’ Plague of Rhodes. Which, I notice, was recently published in a popular edition by one Günther Henrich.)
Oh, and Philip Newton? Good catch on the e-vocative in γέρε being associated with adjectival use. Henrich argues that the o-vocative does not stick with adjectives, because there is much more analogical pressure to align adjectives with the well-established first and second declensions, since adjectives decline in all genders and numbers. In fact, it’s why third declension adjectives like askʰɛːmɔːn “misshapen” did not keep an o-vocative when they switched to the second declension: nom.masc.sg asximos, voc.masc.sg asxime “ugly”.
In the last few posts, I’ve worked through the analogies that have extended the o-vocative into proper names: M1–M5, O2–O6. There was to-ing and fro-ing, there was nebulous definition and redefinition of rules, there was a whole ballet of criteria. But the ballet orchestration can be formulated: the rules for the analogy are sweeping, even when they seem to be contradictory, and could be distilled in a handful of features: NAME, VERNACULAR, FAMILIAR, PENULT-STRESSED. (Those features are at work even when the rules were redefined: the diminutive -akos suffix for surnames in 1941 was an analogy based on the FAMILIAR feature; the etymological transparency criterion used now is an analogy based on the NAME feature—surnames that look like common nouns are less NAME-like than opaque surnames, so they tend to take the o-vocative less.)
That leaves us with the common nouns of O1. The story for those is going to be less satisfactory, because the analogy does not involve entire well-defined classes of name: they are more typical of analogical spread, in that they involve individual words, and the commonalities at play are harder to discern. This is an area where I may need to be enlightened by Henrich’s thesis on the o-vocative, once I do read it.
The one exception documented in Triantafyllidis’ 1941 grammar, for example, is O1a: kapeˈtanios “captain”, kamaˈrotos “porter, cabin body”. Why the navy would be fertile ground for o-vocatives early on, to the extent of extending them to common nouns, is not obvious. They are both Venetian words, and their foreign origin might have encouraged some sort of grammatical streamlining (like I surmised for final-accented names like nikoˈljo < Niccolò); but for those two naval terms, there are dozens of Venetian nouns like kuˈnjaðos “brother-in-law” or maraˈnɡos “carpenter”, which form their vocatives regularly.
The temptation is always there, whenever maritime vocabulary is involved in the Mediterranean, to invoke the Mediterranean Lingua Franca: maybe the o-vocative with maritime vocabulary reflects some sort of pidginisation of Greek, that would have been all too happy to dispense with its now anomalous e-vocative. That’s a little too speculative for my liking, though.
It’s safer to posit that the source of all the analogical extensions of the o-vocative, ˈɣeros “old man”, extended to random other honorifics, one of them being “captain”, and that “cabin boy” patterned after “captain” because it was another naval term.
That’s safer. It’s not more satisfactory, I’ll grant you.
Commenter Pepe had an intriguing hypothesis: “captain” and “old man” are both routinely prefixed before given names (ˈɣero-ˈnikos “old man Nick”, kapeˈtan-ˈnikos “captain Nick”). The prefixed form of “captain” is truncated, but he thinks the o-vocative appears with those common nouns in isolation, because they are so strongly associated with given names. We have seen that historically ˈɣero probably came first (at least, Henrich’s third declension theory explains the parallel ðjako, ðrako “deacon, ogre”); but the association Pepe posits is real enough. I hesitate over it because it requires analogy to apply even when the name is absent, and to generalise from truncated kapeˈtan to the complete vocative kapeˈtanjo. But it’s not impossible.
If there were a foreign-language bias towards the o-vocative, that would also explain it showing up in kumˈbaros “god-sibling”, which is also Venetian. (147 hits for re kumˈbaro on Google, 14.4k for re kumˈbare.) But the explanation is likelier that ˈɣeros “old man” brought it about, as an extension of honorifics to familial terms.
In fact, the most controversial extension on the thread in Sarantakos’ blog was ˈθios “uncle”. There was much discussion about whether it was a regionalism, whether it was slang, and what its precise semantics was with an o-vocative. As Evangelos Lolos insightfully pointed out, there is nothing rural about ˈθios: with its V-V hiatus, it is clearly a learnèd form, and the traditional vernacular forms are either θjos (which forms its vocative regularly as θje), or the Italian loan barbas.
But the learnèd form has been pressed into service in a vernacular function: “uncle” is also used traditionally as an honorific for any older stranger. (That survives jocularly in barbas, but of course the etymology of the word, “beard”, shows that it did not start out as a joke.) My own experience is that if anything, ˈθios is likelier than θjos to be used as a title of respect: it is after all a more polite-sounding word, being phonologically learnèd. And as a title of respect towards elders, ˈθios was more susceptible to influence from ˈɣeros “old man”.
If ˈθios is slang (μάγκικο), and less likely to be used with actual relatives, as it is for Sarantakos, that is all the more reason for the analogy with ˈɣeros to take hold. But this is messy enough that commenter Maria claims the opposite pattern, with the o-vocative used for the relative, not for the honorific directly analogous to ˈɣeros:
If I addressed my uncles as ˈθie, or older men in the neighbourhood as kirie “sir” instead of ˈθio, they would have thought I considered them strangers. I only addressed one uncle as ˈθie, after being duly tutored. But he was a judge.
The miscellaneous terms in O1c are harder to explain because they are miscellaneous; but I suspect they are all tied together by the feature FAMILIAR, which we have already seen at play in given names, and which in common nouns is at its most pronounced when it is used as an overt signal of disrespect.
The miscellany are:
- ˈɣiftos “Gypsy, blacksmith”
- kaˈkurɣos “criminal”
- ˈbufos “horned owl; fool”
- ˈmurɣos “shepherd’s dog; uncouth man”
- slang ˈfilo “buddy!”
- military slang ˈneo “newbie!”
In the 1973 song Come, gypsy/blacksmith, gather hammer and anvil, which I have already posted on here, there is nothing necessarily racist going on: blacksmiths were traditionally Roma, and the song only refers to Roma in that capacity. But of course, there has been no shortage of contempt from ethnic Greeks towards the Roma, and the o-vocative is consistent with that. A word used as a slur (even if it is the unmarked traditional term for Roma) is going to choose the familiar-coloured vocative over the default vocative. That also applies to the other instances of invective on the list, ˈbufos, ˈmurɣos, and the surprising entry in the list, kaˈkurɣos, which ostensibly looks too learnèd to tolerate an o-vocative.
Contemptuous familiarity, for that matter, is overtly behind the military slang ˈneo, as Pepe remarked explicitly; and it’s not that far away either from ˈfilo (Pepe’s example in the same comment is τι μας λες ρε φίλο; “what are you trying to tell us, buddy?”) In fact, we have already seen contempt in the diminutive anθroˈpako “puny little man” (although its parallel filaˈrako “buddy” is amicable, not contemptuous, in its familiarity.)
So it is possible, just, to discern some unifying trends in the few common noun instances to date of the o-vocative—other than the inherited constraint that they are all penult-accented:
- There’s the initial category of third declension nouns (M1);
- Two maritime terms (possibly as a one-off extension of ˈɣeros);
- Two familial terms (possibly as another one-off extension of ˈɣeros, with the secondary sense of ˈθios “uncle” in play as an honorific towards elders);
- Terms likely to be used as disparaging terms, either because of their literal content (“criminal, fool, uncouth”), because of racism (“Gypsy”), or because of slang register (ˈfilo, ˈneo).
The military slang of ˈneo, by the way, can be interpreted in a different contemptuous way: as a neuter, which also ends in -o. The neuter with reference to humans in Greek has a connotation of infantilisation (since children are pre-sexual). It can be used as a sign of affection; and indeed, as Pepe had pointed out and I had failed to acknowledge, the neuter could in fact be behind the putative masculine final-stressed given names in M5, like manoˈljos, nikoˈljos: manoˈljo, nikoˈljo could equally be the vocatives of the neuter hypocoristics manoˈljo, nikoˈljo. But the infantilising neuter can also be used dismissively, particularly in a hypermasculine domain like the army.
There is a different variant vocative which also comes from the military, and which is also used negatively: the use of the nominative instead of a vocative. This extends to first as well as second declension, and is independent of accentuation. In the military, it is de rigeur with surnames, and is how orders and punishments are barked at soldiers: Παπαδόπουλος! Δέκα μέρες κράτηση! “Papadopoulos! Ten days confined to quarters!” The point of the nominative is that the army grunt should feel talked at, not talked to, as part of breaking down their sense of individuality. (What Wikipedia politely refers to as resocialisation.) The military nominative-as-vocative turns up now and again in more general slang, most notably in the not particularly respectful ˈkirios “hey Mister!” As elucidated at slang.gr:
Form of address towards men, usually strangers, not particularly polite nor outright rude. It has long been used to express restrained irritation, combined wiht distance. It is not so familiar as to imply a provocation to fight, but it won’t say no to a few cross words. The counterpart for women is maˈdam.
Nowadays ˈkirios is even used by students towards their high school teacher, without necessarily being provocative. By analogy, the vocative ˈfilos, which can be completely decorous and friendly. [The corresponding entry in slang.gr notes that it can be “either negative or positive”.]
Note: the use of the nominative instead of the vocative, which can only happen grammatically with masculines, expresses distancing. It is used routinely in the military.
- M1: Bisyllabic common nouns that used to be third declension: ˈɣeros “old man”, ˈðjakos “deacon”. (Ancient ɡérɔːn, diákɔːn).
- M2: Bisyllabic truncated, informal given names: ˈɣjorɣos, ˈnikos, ˈðimos (corresponding to the formal forms ɣeorɣios, nikolaos, ðimitrios) “George, Nick, Dimitri”
- M3: The trisyllabic (truncated) name aˈlekos “Alec”
- O2: Bisyllabic formal given names (which are used in both formal and informal language): ˈpavlos, ˈpetros, ˈstavros, ˈmarkos “Paul, Peter, Stavros, Mark”
We accounted for M3 in passing: the bisyllabic restriction on o-vocatives does not appear to have been particularly important; so the class of bisyllabic truncated informal given names in M2 was straightforwardly extended to trisyllabic truncated informal given names. The grammarians have puzzled over aˈlekos as an exception; but the economic way of stating the rule for given names was not about them being bisyllabic (although it’s understandable that grammarians would prefer that criterion, as more linguistically clear cut); it is about them being vernacular and familiar.
So the features that define the o-vocative so far are: GIVEN name, PENULT-ACCENTED, TRUNCATED and VERNACULAR/FAMILIAR. We have just seen that the BISYLLABIC feature was quickly ignored.
The names in O3, which have escaped grammarians’ notice, have dropped the TRUNCATED feature, as indeed O2 have: avɣuˈstinos, maˈrinos, puˈlikos, leoˈnikos “Augustin, Marinos, Poulikos, Leonico”. These names have both e-vocatives and o-vocatives, just like O2. O2 keeps the e-vocative because they are FORMAL as well as VERNACULAR; O3 have taken up the o-vocative because they too are PENULT-ACCENTED, GIVEN, and particularly VERNACULAR: they have no Hellenic heritage, and you’d be hard put to find an Orthodox saint’s name behind them. They keep the e-vocative as well, because they are not as typical of the category of o-vocatives: they are not BISYLLABIC, and unlike M3 they are not TRUNCATED, nor especially FAMILIAR.
What is promoting the o-vocative for those words appears to be that they are disconnected from formal names as used by the Orthodox church, just as the truncated forms like nikos are. Poulikos appears to be a (Pontic?) diminutive of Paul, which means it’s in the same bucket as Alekos and Nikos. There are a couple of St Marinus commemorated by the Greek Orthodox Church (Marinus the Elder, born in Tarsus, martyred in 304; Marinus born in Rome, and a Roman senator, late 3rd century), and Augustine of Hippo is officially an Orthodox saint, although not popular among Orthodox theologians; but it is fair to say they are obscure saints in Orthodoxy (unlike St Marina), so there would likely have been less occasion to hear e-vocatives from them. Leonico is Ladino, and Sephardic Jews would not have an equivalent to liturgical Greek to promote the e-vocative. (The minute commenter Leonicos hellenised his name from leoˈnikos to leˈonikos, he noted, he also hellenised its vocative.)
Friend to this blog commenter Diver Of Sinks raised roˈðolfos “Rudolf” as a counterexample: a trisyllabic given name that does not have an o-vocative. “Rudolf” is clearly not a Hellenic name, nor a name commemorated in the Orthodox church; so the e-vocative isn’t pushing back on the o-vocative because of Koine Greek. But it doesn’t need to. Greek speakers recognise that roˈðolfos is a learnèd loan into the language: the -lf- cluster is enough to indicate that (cf. puristic aðelfos vernacular aðerfos “brother”), and the fact it’s German, inflected, and with a [ð]. Linguistically, roˈðolfos is not German: it’s learnèd. It might as well be Ancient Greek linguistically. And speakers know that learnèd names do not have o-vocatives.
The names in M5, ɣjanaˈkos, ðimiˈtros, ˈnikoljos, manoˈljos “John, Dimitris, Manuel” are given names, vernacular, and familiar, all like O2. But they violate the pervasive restriction of the o-vocative to penult-stressed forms: they are stressed on the final syllable. The analogy with familiar forms has proven stronger than the phonological constraint on stress; but given how pervasive the phonological constraint is elsewhere, one would think that it has had some help to push through. My own speculation is that these forms are based on Italian diminutives: Niccolò, Manolò—and hearing Niccolò, Manolò as vocatives would have reinforced the analogy to use o-vocatives with them when they were taken up in Greek, even though they were the only such instance to use the o-vocative on the final syllable.
The words in M4 and O4, ending in diminutive suffixes like -akos, reflect a different and more far-reaching breakdown, that of the restriction of o-vocatives to given names. In the 1941 Triantafyllidis grammar, those diminutive suffixes are the only instances where the o-vocative shows up in surnames (O4) and common nouns (M4b); so it is reasonable to assume that this class of words started off as given names (M4a), and then extended to surnames and common nouns with the same suffixes.
At the beginning, presumably, came diminutive given names like ɣjorˈɣakos, nikoˈlakos “George, Nick”; they call in the same class as O3, and we just say puˈlikos as a similar name with a slightly different diminutive suffix, and both of them share the feature FAMILIAR with the core class of O2. The o-vocative then would have spread to given names that ended in -akos without any diminutive meaning (kiriˈakos “Cyriac” < “of the Lord”—though you could argue that the suffixes are ultimately related).
And once the o-vocative was associated with diminutive suffixes in given names, like Poulikos and Nikolakos and (seemingly) Kiriakos, it is straightforward analogy to extend the o-vocative to other words ending in the same suffixes: surnames like Dimitrakos and Sarantakos, and common nouns like anθropakos. In the case of the vernacular diminutives, there was not a strong enough pushback by an e-vocative, even if they are common nouns: the vocatives are universally anθropako, filarako “puny little man; buddy”.
Surnames however as they are used in the Modern Greek State are formal Greek, which the State has had its hand in, and speakers know they are formal Greek: they retain Puristic phonology, and Puristic inflections. Because they are formal, and because they are a more recent category, speakers will tend to use e-vocatives for them more than for given names; which is why the o-vocative is still optional with surnames like Sarantakos and Dimitrakos—and is not optional when those same forms are used as given names.
The more recent trend in surnames, reported in Katsoudas’ grammar, is a new rule (O5): etymologically transparent surnames take e-vocatives, etymologically opaque surnames take o-vocatives. There’s an obvious analogy behind that rule too: o-vocatives are characteristic of proper names, therefore e-vocatives are characteristic of common nouns; and if a surname is identical to a common noun, then the analogy with the common noun is pressure to retain the e-vocative. In the absence of that countervailing pressure, the default pattern takes over; and it allows us to see that there has been a markedness shift in names since 1941. In 1941, only diminutive-looking surnames took the o-vocative. Now, the o-vocative has become the default for surnames.
Just as the o-vocative has in fact become the default for given names: the given names that retain the e-vocative have in reality become just those names where the e-vocative is reinforced from Koine or Ancient or Puristic Greek. Which would explain the hitherto sporadic instances reported for given names with o-vocatives accented in the antepenult, like ˈmario (O6): a Romance, unchurchlike given name like in O3, and the stress is no longer a barrier to the o-vocative. For that matter, just like highly Hellenic names like liˈkurɣos “Lycurgus”: penult-stressed like Nikos and Marinos, but Ancient Greek, which should have blocked the o-vocative from ever appearing. Yet Panos in Lowercase reports that he has heard the vocative liˈkurɣo, which suggests that that constraint, too, is starting to break down. And there will be more: commenter Manousos reports an o-vocative for anˈðronikos “Andronicus”, which is not only learnèd (complete with unvernacular -nðr- cluster), but accented on the antepenult—breaking both the restrictions to date. (“Speakers know that learnèd names do not have o-vocatives”, I said above with relation to roˈðolfos “Rudolf”. Not so much any more.)
These analogies are spreading at different rates for different people; it causes me some glee to note that Nikos Sarantakos himself is clearly on the conservative side when it comes to most of these, as well as the extensions of the o-vocative to common nouns. That doesn’t make him a fuddy-duddy blocking language progress. (Or at least, that in itself doesn’t. 🙂 Nor for that matter is guaranteed that all names will end up dropping the o-vocative: the last two decades has seen a resurgence of archaising purism in Greek, and the o-vocatives could yet be stigmatised once the purists actually notice them. But it does mean that the system of Greek vocatives is a lot more in flux than anyone seeking to codify Greek grammar can be comfortable with.
More in flux than anyone trying to decline words consistently can be comfortable with, in fact, and they clearly aren’t comfortable. The use of o-vocatives is unsteady, and strongly influenced by context, because people’s language instinct is uncertain about how to form the vocative for any of these. Thence the spontaneous debates about the proper vocative of “Augustin”. Thence forms like ðimo verikio, where the o-vocative on the etymologically opaque surname confounded Giannis Haris; but if the rule around O5 is new and unsteady to begin with, the mere presence of an O2 o-vocative in ðimo will be enough to undo it.
So we have the messy data on the distribution of the o-vocative in Greek. And we have the tools to try and make sense of that distribution, in terms of features that classes of nouns with the o-vocative have in common.
We also, as it turns out, have an entire PhD thesis on the o-vocative: Günther Henrich‘s 1975 thesis from Aristotle University, Κλητικές και γενικές σε -ο από αρσενικά σε -ος στα μεσαιωνικά και νέα ελληνικά. There is an online summary available, and I have a copy of the dissertation in my mailbag.
AND I WILL NOT READ IT YET. Where’s the fun in that? I’m going to try and work out what happened on my own, given the contemporary data reported by Triantafyllidis, Katsouda, and the commenters from Sarantakos’ blog. And then I’ll read what Henrich found, and see the extent to which my analysis (and his) holds up.
I do know one thing from Henrich’s thesis: where he thinks the phenomenon started. I’ll take it as my starting point too, because it makes sense, although I have a vague misgiving about it, and I will advance a secondary mechanism for it.
I’m going to recap the noun classes identified for the o-vocative, and label them. There is a distinction to make between categories where the o-vocative is optional, and where it is mandatory. All other things being equal, it is reasonable to assume that the classes where it is mandatory have had the o-vocative longer, and they are more salient classes from which the o-vocative spreads further to new classes.
- M1: Bisyllabic common nouns that used to be third declension: ˈɣeros “old man”, ˈðjakos “deacon”. (Ancient ɡérɔːn, diákɔːn).
- M2: Bisyllabic truncated, informal given names: ˈɣjorɣos, ˈnikos, ˈðimos (corresponding to the formal forms ɣeorɣios, nikolaos, ðimitrios) “George, Nick, Dimitri”
- M3: The trisyllabic (truncated) name aˈlekos “Alec”
- M4: Given names (M4a) and common noun diminutives (M4b) ending in the diminutive suffix -ˈakos: kirˈjakos, anθroˈpakos “Cyriac, contemptible little man”
- M5: Multiple-syllabic final-stressed diminutive given names: ɣjanaˈkos, ðimiˈtros, manoˈljos)
- O1a: Common nouns, maritime: kapeˈtanios “captain”, kamaˈrotos “porter”
- O1b: Familial terms: ˈθio “uncle”, kumˈbaros “god-sibling”
- O1c: Others: ˈɣiftos “gypsy, blacksmith”, kaˈkurɣo “criminal”, slang ˈfilo “buddy!” and ˈneo “young man!”
- O2: Bisyllabic formal given names (which are used in both formal and informal language): ˈpavlos, ˈpetros, ˈstavros, ˈmarkos “Paul, Peter, Stavros, Mark”
- O3: Plurisyllabic given names (mostly Romance in origin): avɣuˈstinos, maˈrinos, puˈlikos, leoˈnikos “Augustin, Marinos, Poulikos, Leonico”
- O4: Surnames ending in diminutive suffixes: -akos, -ukos, -itsos
- O5: Surnames that are not etymologically transparent: e.g. Venizelos
- O6: Plurisyllabic given names (mostly Romance in origin): ˈmarios “Mario”
These are almost all names, so NAME is the feature that bind almost all of them together. But the starting point Henrich claims for the o-vocative is not a name: it is M1, third declension nouns like ɡérɔːn, diákɔːn, drákɔːn.
Third declension nouns regularly went to the first declension, preserving their full stem from the genitive. In fact, this happened to these nouns: ɡérɔːn, gen. ɡérɔːntos “old man” has the modern form ˈɣerondas “old man, elder”; drȧkɔːn, gen. drȧkɔːntos “dragon” has the modern from ˈðrakondas “dragon”, and diákɔːn, gen. diákɔːnos “helper, deacon” has the modern form ˈðiakonos “deacon” (which has shifted to second-declension). But these forms also have second-declension variants, formed from their shorter nominative stems: ˈɣeros, ˈðjakos, ˈðrakos “old man; deacon; ogre”.
The third declension original nouns had vocatives that ended in -on: ɡéron, diákon, drákon. When they shifted to the second declension based on the shortened stem, the argument would be, they took those vocatives with them: ˈɣeros ˈɣeron, ˈðjakos ˈðjakon, ˈðrakos ˈðrakon. Those vocatives then looked identical to the regular second declension accusative; and when the accusative dropped its final -n, so did the vocatives: ˈɣero, ˈðjako, ˈðrako.
That’s certainly possible, although I’m a bit nervous about the notion of such an archaic vocative surviving like a time bomb in just that very narrow class of nouns. But it is a fact that for ˈɣeros in particular, the o-vocative is almost universal; that implies that it is a core instance of the o-vocative.
… Almost universal; but one will find counterexamples. Not a lot of counterexamples: Google gives 112 hits for ɣere, almost all of them from one song lyric, a New Year’s Day carol (Γέρε χρόνε φύγε τώρα “old year, go away”). For the colloquial vocative particle re, Google gives 7600 hits for re ɣero, and just four for re ɣere:
Τι LIFE COACH ρε γέρε άνθρωπε μίλα ξεκάθαρα!!! Τι έχασα???
What do you mean, “life coach”, old man?! Speak clearly! What have I missed?
— Marrie_Qrie ♏️ (@marrie_qrie) January 26, 2018
That’s an analogical change in the reverse direction, and that’s plausible. Recall that almost all the o-vocatives are names. ɣeros might be historically Ground Zero for the o-vocatives, but analogy doesn’t have a historical memory, speakers form analogies case by case as they notice word similarities. And analogy doesn’t just happen at the local level, of individual words: it also happens with big sweeping classes of words.
There is a spread out of o-vocatives that is based around proper names. That means that an overall rule has emerged, that o-vocatives involve proper names, and therefore e-vocatives involve common nouns. And that leads to an analogy running in the opposite direction: since ɣeros is a common noun, it should have the e-vocative ɣere, like other common nouns. It is not a strong analogical pressure: ɣere is rare. But the pressure is there. And this kind of analogical backwash, of analogies going in the reverse direction to the overall trend, when the overall trend is strong enough, happens a lot. It’s why there was a resurgence in strong verbs in American English (sneaked > snuck), after centuries of strong verbs retreating in English.
ɣero “old man!” as a vocative is common in Greek; more common than its counterpart in English, and certainly more common than “ogre” or “deacon”. As a common vocative, it was clearly a candidate for extension by analogy to given names like nikos “Nick”.
There was an additional pathway for ɣero “old man!” to influence the vocatives of given names. ɣero is used as a prefix before names: ɣero-ðiˈmitris “Old Dimitris”, ɣero-ˈɣjanis “Old John”. The vocative of those compounds would not inflect the prefix: ɣero-ðimitri, ɣero-ɣiani. But those vocatives can be reanalysed as two separate words: ɣero ðimitri, ɣero ɣjani. And if the prefixed given name was itself second declension (ɣero-ˈnikos, ɣero-ˈɣjorɣos “Old Nick, Old George”), there will be additional pressure on the prefixed names to rhyme with the prefix: ɣero-nike, ɣero-ɣjorɣe could easily be remodelled to ɣero niko, ɣero ɣjorɣo. (The rhyme is itself a local instance of analogy, operating on its immediate context.)
In any case, once the o-vocative hits M2, it hits jackpot: even more than the vocative and prefix “old man”, given names like “George” and “Nick” are a common, well-defined group of nouns, that let the vocative be entrenched as an easily learned exception to the global pattern of e-vocatives, and that is distinctive enough to form the basis of further analogies.
If the o-vocative is going to spread further from M2, we need to work out what the features in play are, that will be the vehicles for it to spread further.
Words like ˈnikos,ˈɣjorɣos, ˈðimos are:
- NAMEs of people. More specifically, they are
- GIVEN names. They are
- PENULT-ACCENTED. That feature carries through most of the subsequent instances of the o-vocative, but not all. They are also
- BISYLLABIC. That’s likely how the analogy from ɣeros to M2 carried across, whether remotely (this word sounds like that, and they both refer to a person), or in the context of compounds (this word sounds like that, and they are used together to refer to a person). The names are also
- VERNACULAR: ˈnikos,ˈɣjorɣos, ˈðimos correspond to the official forms niˈkolaos, ɣeˈorɣios, ðiˈmitrios, which traditionally Greek-speakers would have heard in ecclesiastical contexts. And related to that, they are
- FAMILIAR. They are not all familiar to the same extent, and individuals’ connotations will differ. ˈɣjorɣos “George” for example is the unmarked vernacular ways of saying “George”. But ˈðimos is not the only vernacular way of saying “Demetrius”: the unmarked form is ðiˈmitris. And because short Greek given names are rare in the formal version of the language, there will be a correlation between short vernacular forms of names and familiar, diminutive forms of names.
These features overlap, of course, for a particular class of words like M2; and different overlapping features can trigger different analogies with different words. Not all features are as useful in explanation either. For example, grammatical descriptions make a lot of BISYLLABIC. There’s a reason they would: BISYLLABIC is a purely linguistic criterion, that can be diagnosed directly from the linguistic data, without having to appeal to social context or diglossia. It makes the grammarian’s job easier. But BISYLLABIC seems to have run out of steam as an analogical criterion after M2: we have seen those grammarians struggle to make sense of M3 aˈlekos “Alec”, and the bisyllabic restriction is ignored almost all other classes of the o-vocative (although I suspect it plays a part in O1c: ˈɣiftos “gypsy, blacksmith”, ˈfilo “buddy!”).
The actual criterion that seems to be in play, VERNACULAR/FAMILIAR, is more nebulous to diagnose, and less-language internal: you can’t tell just by looking at the word in isolation. The vernacular criterion presupposes a self-consciousness about diglossia. Yet, as Greek grammarians know, and routinely appeal to in their grammars, Greek speakers are highly conscious of which words in their language are vernacular and which words are learnèd, and it’s usually easy to tell from their phonology and their morphology.
Similarly, the familiarity of names is a semantic criterion, and accounts of morphology would rather operate on just other morphological and phonological criteria: it makes for a much simpler model. But it’s analogy, and nothing about analogy is simple: any factor in language and language context can be brought to bear.
For what it’s worth, I don’t think speakers became aware of the VERNACULAR and FAMILIAR feature immediately in M2: I don’t think that, the minute they started forming the vocative of “George” as ˈɣjorɣo, they thought, “this name is familiar”. The way those features became apparent can be seen by contrasting them with O2: given names that are bisyllabic, but have preserved their e-vocative.
The names that have kept their e-vocatives are names that are the same in formal Greek and vernacular Greek, like pavlos “Paul”. Names like nikos were vernacular, and once the o-vocative was entrenched, there was no reason for an e-vocative to reemerge, as we have seen with ɣere by analogy with common nouns. But for a name like, say, petros “Peter”, which are the same in Koine Greek and Modern Greek, there is an obvious place that an e-vocative can come back into the language from.
ὁ δὲ εἶπεν· Λέγω σοι, Πέτρε, οὐ φωνήσει σήμερον ἀλέκτωρ ἕως τρίς με ἀπαρνήσῃ εἰδέναι.
And he said, I tell thee, Peter, the cock shall not crow this day, before that thou shalt thrice deny that thou knowest me. Luke 22:34
e-vocatives are associated by speakers with common nouns. e-vocatives are also associated with Ancient Greek and the learnèd language; speakers are aware that the o-vocative is colloquial, and cannot be used in formal language. So if a given name is the same in colloquial Greek and in the language of the church, they will have occasion to hear that name with an e-vocative at least occasionally. Formal forms like pavle, petre, marke will be heard, whereas purely colloquial names will not present forms like nike, ɣjorɣe, ðime. Because names like pavlos, petros, markos are also used in vernacular Greek, the vernacular o-vocatives will also show up: pavlo, petro, marko. And after that, the relative preponderance of e-vocatives and o-vocatives is a matter of contingency and happenstance.
And when speakers realise that o-vocatives are vernacular and not formal, they will correlate them with familiar forms—which are never formal.
We have seen the data on the spread of o-vocatives in Modern Greek. I will post how I make sense of the data. But first, some preliminaries about analogy.
How analogy works
Analogy in language change takes a linguistic rule that applies to one word or paradigm or category, and starts applying it to another word or paradigm or category. To take an example that’s rather removed from o-vocatives: analogy in non-rhotic dialects of English takes the rule “keep the word-final [ɹ] before a vowel, as in more or less [mɔː-ɹ ɔː lɛs]”, and starts applying it in novel contexts, like law and order [lɔ-ɹ ənd ɔːdə].
There was no historical reason for there to be an r after law; that’s creating a new rule where there is no historical justification for it (there is no r to keep). That’s what makes it a linguistic innovation, after all. But there is a linguistic justification for it: it is that more looks like law. When they’re not followed by a vowel, the two words end in the same vowel, and they have the same syllable structure: [mɔː, lɔː].
And that’s how analogy works in general. If a linguistic rule is generalised from A to B, it doesn’t happen at random. It’s because A and B have a feature in common. And linguistics is not powerless to make sense of analogical change: linguistics can identify what features A and B have in common, and use that to explain the change.
This is a diagnostic, not a predictive explanation: it’s the kind of explanation at home in historical linguistics, not synchronic linguistics, because we’re not dealing with a rule here, but a tendency, that could have happened, and could have not happened. But analogy is a bunch of contingent tendencies, and not rules; that’s why language undergoing analogical change is not particularly rule-bound. Or, to put it more informally, a mess.
As analogies spread, if an analogy spreads from A to B and from B to C, that spread can be explained because A, B, and C all have a feature in common. But again, analogy is not that rule-bound or predictable, and the analogy between A and B can be quite independent from the analogy between B and C. The feature B and C have in common may not be a feature A and B have in common; it may well not be present in A at all.
To give a sneak preview of how the o-vocative spread: I will argue that it spread from “old man” to “Nick” to “Venizelos”. The words “old man” and “Nick” [ˈɣeros, ˈnikos] have phonological features in common: they are bisyllabic, and penult-accented. The words “Nick” and “Venizelos” have one out of the two phonological features in common: they are penult-accented, but there is no longer a requirement for the word to be bisyllabic. Yet that’s not enough to explain the analogy: there are lots of penult-accented words that don’t have an o-vocative. The feature that “Nick” and “Venizelos” have in common is that they are names. And that’s a feature that “old man” does not have: as the o-vocative spread from class to class of words, the NAME feature is something it picked up when it started applying to given names like “Nick”. It was not a feature present at the start.
So there are features involved in the spread of a phenomenon by analogy. Those features are how the distribution of the analogy can be made sense of, while it is in progress. And when the analogical spread settles down, they will be the basis of the new rules that govern the phenomenon.
And if an analogical change involves more than one feature, not all features are necessarily equal. The more common or “basic” a particular feature is in a language, the more likely it is that new instances of analogy will be based on it. If an analogy is spreading among, say, names of ethnic groups, there are limits to how far the change is going to spread. If the analogy hops across from ethnonyms to, say, proper names in general, the analogical change has the potential to take over much more of the language.
Similarly (though this is not as obvious), a very common or psychologically salient word, like “old” or “mother”, is likelier to form the model for further analogies than a less common or salient word, like “vague” or “ogre”. I haven’t seen this particular notion articulated explicitly anywhere, though I suspect something to it will have been stated by Kuryłowicz or Dressler, who have theorised extensively on analogy.
What analogy does
Analogy is the mechanism that spreads language change, and makes it make sense. Analogy takes a local change in one instance of language, and spreads it further afield. Analogy spreads such changes in a way that, eventually, can be made sense of by a simple rule, that ends up replacing whatever other simple rule was there before—as the features that enabled the analogy are generalised.
Eventually. Because until the language change goes all the way through, analogy in action is a mess. It’s two rules, the old and the new, clashing, and the context in which one rule or the other applies is completely idiosyncratic: it varies from social group to social group, from semantic category to semantic category, from individual to individual. You’ll have seen a little of that in the disagreements among commenters I related, as to whether “uncle” or “Marinos” had an o-vocative, and what the connotations were if it did.
And that’s not an idiosyncrasy of Greek vocatives: that happens whenever there’s a language change underway. Moreover, while analogy usually runs to completion, and you have a clean new status quo, sometimes it remains messy. Hence the unpredictability of the pronunciation of <ea> in English, for example.
Analogy is the culprit for language learning being exasperating for adults, from a book: it’s where the attempts to formulate simple, learnable rules for language founder. That’s where the infuriating laundry lists of exceptions, and long lists of ifs and unlesses in textbooks come from. Language in preliterate societies, and child acquisition of language, deal with these exceptions and conditions. In fact, they are valuable in tidying them up for the next generation, through logical abduction: confronted with a whole lot of messy data that doesn’t follow nice rules, language learners make up their own rules, which tend to be simpler, and more learnable, and wrong: they overgeneralise from the data they get. Which ends up being a good thing for the language.
Which is fine if you’re two years old, or if you’ve married into a tribe, and have a couple of years on your hands. If you’re in a bit more of a hurry, grammars and textbooks give you shortcuts: they’ve worked out the rules for you in advance. Except where the rules are a mess, because that aspect of the language is in flux, and being pummelled back into non-mess over generations through drift of analogy.
So to any language learner who has groaned about why they have to learn laundry lists of genders, and why there’s so many exceptions, and why there are so many irregularities, I present to a cartoon villain for your dartboard. Commenter Kostas in the discussion of o-vocatives at Sarantakos’ blog:
I rejoice that the unruly people finds a way to escape the norms and shackles set by learned linguists, and in their own unique way they transform, develop, and enliven their language.
And I don’t think that’s by accident. As they learn their mother tongue, they create their own internal mechanisms and rules and they then apply them in its development. I call that language instinct, and I follow it in my life, in my spoken language. In my written language I conform more with the rules posed by linguists and other specialists.
Kostas is of course correct in the second paragraph, in how “language instinct” works: that’s logical abduction in the face of messy linguistic input. Linguists will be shocked to see themselves caricatured in the first paragraph: all descriptive linguists are trying to do is make sense of the rules that people do carry with them in their “internal mechanisms”, which is what the logical abduction is producing: they aren’t setting any shackles for anyone. But of course prescription does have a role to play in the social aspect of language, and linguistic descriptions are recruited for prescriptions. That’s not an evil; that’s just what happens.
Language learners can keep Kostas on their dartboard though. His convenience and unruliness is your inconvenience and rote-learning. And those language learners do the same in their own native language. (“English is a mongrel language, English has no Academy”, etc. etc.)
And that’s how speakers like Kostas, and everyone else who speaks Greek, comes up with rules for second declension vocatives that look like this:
- Most of the time use -e
- The word for “old man” uses -o
- “Nick” uses -o
- Words like “old man” and “Nick” use -o
Words like “old man” and “Nick”?! What kind of rule is that?
A very porous, ambiguous rule. That’s why speakers disagree so much. And that’s what happens when one aspect of a language is in flux.
This is a story of analogy, based on an article at Nikos Sarantakos’ blog (because the traffic between our two blogs has ever been two-way). Sarantakos’ article in turn cites an older blog post by Giannis Haris, which cites two grammars of Modern Greek.
The story is the retreat of the vocative in Modern Greek. The vocative is a case that strikes many a learner of Classical languages, because most modern European languages don’t have them. The vocative has survived in Modern Greek…
… but it has been restricted. Even in Ancient Greek, the vocative did not show up in all numbers or declensions: a distinct vocative appeared only in the singular, and it did not appear in feminines of the first declension, in neuters of the second declension, or in most subclasses of the third declensions:
- First declension: masc. nom. polítɛːs “citizen”: voc. polîta; fem. nom., voc. kórɛː “maiden”
- Second declension: masc. nom. ántʰrɔːpos “human”, voc. ántʰrɔːpe; neut. nom., neut. paidíon “child”
- Third declension: nom. ɡérɔːn “old man”, voc, ɡéron; nom. sɔːkrátɛːs “Socrates”, voc. sɔ́ːkrates; nom. voc. pʰýlaks “guard”; nom. voc. poimɛ́ːn “shepherd”
The vocative has retreated substantially in Modern Greek.
- In the first declension, masculines use the accusative/genitive as the vocative: nom. politis acc. gen. voc. politi. Puristic Greek did revive the ancient vocative for honorific titles: nom. kaθiɣitis “professor”, voc. kirie kaθiɣita “Herr Professor”. But you won’t find anyone under 70 now using vocatives like that.
- In the vernacular, the survival of the third declension is vestigial: most forms had switched to the first declension. Even where Puristic revived the third declension, it did not bring back with it the distinctive vocatives with the short final vowels.
In fact, the first declension (which in the vernacular expanded to include all five final vowels, in order to deal with loanwords) only has two distinct cases in the singular: nominative vs the rest in the masculine, genitive vs the rest in the feminine.
The second declension masculines, however, hold out. The vocative of anθropos “human” remains anθrope, with the same inflection as in Ancient Greek.
(That’s the masculines: the feminines in the vernacular had switched to acting like first declension nouns: Corinth was nom. acc. voc. korθo. gen. korθos. Puristic has undone this development, and restored some feminines in the second declension, although speakers occasionally stumble on them.)
But the vocative is starting to retreat in Modern Greek: there are nouns where the vocative in -e is replaced by the accusative ending -o, analogous to what happens in the first declension. It is retreating in specific classes of nouns, and the retreat is spreading from subclass to subclass of those nouns. The data, which I’m going to present in this blog post, looks at first somewhat random. But if you identify the classes of nouns, and the categories people think they belong to, the retreat makes a lot more sense. As does the confusion and vacillation of speakers about which vocative to use; because this is an ongoing change in Greek, and there are plenty of grey areas among those subclasses.
(The categories are established in the grammars that have been cited; but the analysis is not particularly parsimonious, and I’m going to try and make it more general.)
So where is the vocative retreating now in Modern Greek?
The -o vocative is mandatory in:
- Penult-stressed nouns
- A small number of bisyllabic nouns that used to be third declension: ˈɣeros “old man”, ˈðjakos “deacon”. (Ancient ɡérɔːn, diákɔːn). (As commenter Panos in Lowercase points out, when those nouns are compounded they go back to the old vocative: paˈljoɣere “stupid old man”.)
- (Etymologically truncated) bisyllabic given names: ˈɣjorɣos, ˈnikos, ˈðimos (corresponding to the formal forms ɣeorɣios, nikolaos, ðimitrios) “George, Nick, Dimitri” (Astonishingly, noone picked up on the truncation being the explanatory factor until commenter and fellow Esperantist Angelos, 119 comments down.)
- The trisyllabic (truncated) name aˈlekos “Alec” (corresponding to the formal form aleksanðros)
- Given names and common noun diminutives ending in the diminutive suffix -ˈakos: kirˈjakos, anθroˈpakos “Cyriac, contemptible little man”
- Final-stressed nouns
- Multiple-syllabic final-stressed diminutive given names: ɣjanaˈkos, ðimiˈtros, manoˈljos (corresponding to neutral ɣjanis, ðimitris, manolis: “John, Dimitri, Manuel”)
The -o vocative is optional in:
- A few familiar penult-stressed nouns: the grammars give kapeˈtanios “captain”, kamaˈrotos “shipmate”
- A few more nouns not mentioned in the grammars, although these seem to be much more contentious. Commenter Panos in Lowercase gives ˈɣiftos “gypsy, blacksmith”. Friend of this blog Pepe adds kumˈbaros “god-sibling” as having a rare o-vocative (which Sarantakos rejected). Pepe also adduces the vocatives ˈfilo “buddy!” in contemporary slang, and ˈneo “young man!” in older military slang. A few commenters mentioned kaˈkurɣo “criminal!” from literature or old movies.
- The case of ˈθio “uncle!” turned out particularly controversial: Pepe adduced it, Sarantakos rejected it as slang; Pepe insists that ˈθio is widespread, and commenter Alexis believes it is a Northern Greek form, avoided in the Peloponnese.
- Penult-stressed given names that are already bisyllabic in their formal forms: ˈpavlos, ˈpetros “Peter, Paul” (ˈpetre is much rarer than ˈpavle). Stavros is also not truncated; commenter Alexis considered the vocative ˈstavre to be jocular, but commenter SP reports it as extant, though more used by men than women.
- Although the grammars do not say so, longer penult-stressed given names may take an o-vocative too: commenter avɣuˈstinos “Augustin” reported a friend hesitating on what vocative to use for him, Panos in Lowercase reported Marios, and friend of this blog Diver of Sinks reported Marinos and Poulikos (but not e.g. Rodolfos.) Commenter Leonicos accents his name both after the original Ladino as leoˈnikos, and as the more hellenised leˈonikos; following this trend, his vocative of leoˈnikos is leoˈniko, and his vocative of leˈonikos is leˈonike.
The greatest confusion is around penult-stressed surnames. Sarantakos acknowledges this in the title of his post: Κύριε Σαραντάκο ή κύριε Σαραντάκε; “Mr Sarantako, or Mr Sarantake?” Triantafyllidis 1941 grammar claims that the surname Dimitrakos has a mandatory -o vocative, like the other instances of the -akos suffix, and he also extends that rule to two other diminutive endings on surnames, –ukos and –itsos.
Georgia Katsouda (who I’ve had the pleasure of meeting recently at the Historical Dictionary of Modern Greek in the Academy of Athens) wrote a grammar in 2007, which makes a more recent distinction: surnames that are etymologically transparent—that is, surnames that look like common nouns—use the -e vocative: Mr kaˈmene “burnt”, Mr ðefteˈree “second”. Surnames that are etymologically opaque use the -o vocative: Mr aleˈvizo, Mr veniˈzelo. Sarantakos confirms that the distinction is recent: the politician Eleftherios Venizelos a century ago was addressed in the press as veniˈzele, while the vocative veniˈzelo is used for the contemporary politician Evangelos Venizelos (no relation, though Sarantakos could not resist the temptation to stick his partisan boot in). He also cites a mid-19th century work by Panagiotis Soutsos, using e-vocatives in surnames that would now be unacceptable (Karatase, Soutse, Diake, Giatrake, Kanele).
But Giannis Haris, who Sarantakos is citing, attests widespread confusion with surnames: as both the difference in the two grammars and in the treatment of Venizelos shows, surnames appear to be where the vocative is currently volatile. There are lots of surnames that Haris heard take -o on the TV to his surprise:
- Giakoumatos; but that could be because of the patronymic suffix -atos (by analogy with -akos)
- “Dimos Verykios”, ˈðimo veˈrikio; I agree with Sarantakos that veˈrikio is likely influenced by the preceding vocative ˈðimo, and where there is confusion between grammatical alternatives, context is going to have a much bigger sway than normal.
- Alimonos: the vocative aˈlimono sounds exactly like the interjection aˈlimono “alas!” (which is of course the origin of the surname: there is no common noun aˈlimonos for the surname to sound like, so the -e vocative would not make sense here)
- Karamanos: maybe because it was explained as Kara-Manos, where Manos is a truncated given name (short for Manuel), of the type that takes an -o suffix.
- Marinos: an -o vocative when it is used as a surname; but always an -e vocative when used as a given name. (We’ve just seen that Diver of Sinks disagrees!)
Haris’ concluding comment is a despairing “Arbitrariness and chaos?” Sarantakos’ conclusion (after dispensing with Evangelos Venizelos) is “this indicates a more general, but very gradual tendency towards o-vocatives.”
There is no unitary set of rules that will explain the distribution of vocatives in Greek, because speakers themselves are clearly confused. But as I indicated, the confusion is explicable, and I will attempt at least some of that explication in the next post.