o-vocatives: Analogical account, Part I

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.

• • Mandatory
    • Penult-stressed
      • 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”
    • Final-stressed
      • M5: Multiple-syllabic final-stressed diminutive given names: ɣjanaˈkos, ðimiˈtros, manoˈljos)
  • Optional
    • Penult-stressed
      • 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
  • Antepenult-stressed
    • 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.