Analogy in third declension -ης nominals

If you're blogging about language, and want a readership broader than two linguists to follow you, lexicon is easy to blog about: people get words. Grammar is harder to blog about: people get grammar only when they've been told they're doing something wrong. And the operation of analogy on the declension of Ancient Greek—well, that's an uphill sell at the best of times. Especially as convoluted as this is going to get. This... is going to lose people fast.

Still, I said I owe the Hattics more Ancient Greek, so here's an observation of a well-known oddity in Ancient Greek declension, an obvious and wrong explanation for it, and some reasons why it might not be so wrong after all. I'm sure someone's come up with it before—me, for example, in a presentation on analogy by Brian Joseph four years ago. But it's a nice grammatical story. If you're into that kind of thing.

Declensions: I'm going to have to start with sketching declensions, so bear with me for a while; the Classicists among you can start skimming. Ancient Greek has three declensions for its nominals, which make sense once you do internal reconstruction and follow it through. (The Roman and Byzantines started but didn't follow through, which is why they thought they had fifty declensions. Once they saw Latin had it down to five, they revised it to ten. Greek grammar took a while to get straight.)

The first declension has stems ending in -ā-. The second declension has stems ending in -o-. The third declension doesn't have a vowel ending; and when it does, it acts like it doesn't. (I'm going to keep saying first declension and third declension here, so I'm going to abbreviate them to Decl1 and Decl3.)

With masculine nouns, there is an ambiguity. Masculine Decl1 nouns in Attic normally have their singular nominatives end in -/ɛːs/ (ης). There are masculine Decl3 nouns that also end in -/ɛːs/. But because the declensions are different, the other cases of the nouns are quite different.

Erm, I'll use IPA for this, and Attic. θύτης "sacrificer" and αἱρησιτείχης "taker of cities". The Greek-speakers and Classicists will be non-plussed, but come on, you already know how these are declined:

	Decl1	Decl3
Nom Sg	tʰýtɛːs	hairɛːsiteíkʰɛːs
Gen Sg	tʰýtoː	hairɛːsiteíkʰoːs
Dat Sg	tʰýtɛːi	hairɛːsiteíkʰei
Acc Sg	tʰýtɛːn	hairɛːsiteíkʰɛː
Voc Sg	tʰýta	hairɛːsíteikʰes
Nom Du	tʰýtaː	hairɛːsiteíkʰeː
Gen Du	tʰýtain	hairɛːsiteikʰoîn
Nom Pl	tʰytai	hairɛːsiteíkʰeːs
Gen Pl	tʰytɔ̂ːn	hairɛːsiteíkʰɔ̂ːn
Dat Pl	tʰýtais	hairɛːsiteíkʰesi
Acc Pl	tʰýtaːs	hairɛːsiteíkʰeːs

If you learn your noun paradigms the old fashioned way, you dutifully learned that there are Decl3 nouns in -ης, and the example you used was probably τριήρης "trireme", and you got names like Ἀριστοφάνης and Εὐμένης mentioned for good measure.

You then learned your adjective paradigms, and you found that there are no Decl1 masculine adjectives, because Decl1 is used to mark feminine adjectives. But there are Decl3 adjectives, with their masculine and feminine in -ης (and their neuter in -ες). The adjectives in -ης decline just like αἱρησιτείχης does: ἠχώδης ἠχώδους ἠχώδει ɛːkʰɔ́ːdɛːs, ɛːkʰɔ́ːdoːs, ɛːkʰɔ́ːdei.... Well, that's no big deal, there's adjectives and nouns ending in -ος too: there was just one bunch of declensions to go around.

Decl3 Distribution: But if you look back critically at the examples you're given of nouns in -ης, you'll notice something odd about them. Let me see if I can show how. I'm going to go through the noun stems registered in the TLG lemmatiser that would end in -ης or -ος, and eliminate all post-classical stems. See what you notice:

	Proper names	Common nouns	Adjectives
Decl1 (-ης/-ας)	5614	5531	0
Decl3 (-ης)	665	40	5531
Decl2 (-ος)	12569	8826	52825

That should make you pause.

• In Decl2, adjectives to common nouns are 4:1, and proper to common nouns are 1.5:1.
  
• Decl1 nominals don't get to be adjectives in the masculine, and are equally split between proper and common nouns
  
• Decl3 nominals in -ης are 138:1 adjectives to common nouns, and 16:1 proper to common nouns.

So -ης in Decl3 is overwhelmingly about adjectives. When it's not about adjectives, it's about proper names. And those proper names are descriptive compounds. Ἀριστοφάνης: "Excellent-looking". Εὐμένης: "Good-minded". Σωκράτης: "Healthy-strengthed". When you look at the common nouns left, well, they're also descriptive compunds. αἱρησιτείχης: "city-taking". τριήρης: "three-fitted". In fact, LSJ defines τριήρης "trireme" in terms that should not by now be a surprise: τριήρης (sc. ναῦς), ἡ "trimeme (namely 'ship'), fem." Sounds like a three-fitted ship. And indeed: -ήρης: "an Adj. termin." Decl3 nouns in -ης? No such thing. Those are adjectives. Including the proper names. The handful of common nouns are just adjectives that get used as nouns.

This is no great news: Smyth's grammar almost tells you as much:

Masculine stems in ες with the nominative in -ης are proper names; the feminine τριήρης trireme is an adjective used substantively (properly, triply fitted; ἡ τριήρης (ναῦς) ‘ship with three banks of oars’).

And Kühner-Blass, a real grammar, does tell you. (I'll link to version 3 and not to version 4 of Perseus for the text, which is far from an upgrade: the smallest link target they have is 4.8 MB on a single web page? What are they thinking?)

Hierher gehören die neutralen Substantive auf ας, G. α-ος, auf ος, G. ε-ος, die Adjektive auf ης (St. ες), sowie die Eigennamen auf [35 different suffixes], welche substantivierte Adjektive auf ης sind, einige sonstige substantivierte Adjektive, wie ἡ τριήρης

But all that's not the point of this post. The point is this.

Name plurals: Greek proper names should be identifying an individual uniquely, especially as they had no surnames (so nothing like "the Joneses"). But Greek literature got into the habit of using plurals of proper names, with Plato and Aristophanes, to refer to "Famous Person and people of their ilk"; e.g. "the Platos of this world". Smyth is too small to tell you about it, but it's mentioned in Kühner-Blass, at the end of the same page. So Plato Symposium 218b mentions Aristophaneses:

καὶ ὁρῶν αὖ Φαίδρους, Ἀγάθωνας,Ἐρυξιμάχους, Παυσανίας, Ἀριστοδήμους τε καὶ Ἀριστοφάνας: Σωκράτη δὲ αὐτὸν τί δεῖ λέγειν, καὶ ὅσοι ἄλλοι;
I have only to look around me, and there are Phaedruses, Agathons, Eryximachuses, Pausaniases, Aristodemuses, and Aristophaneses--I need not mention Socrates himself--and all the rest of them.

Now you'll notice something odd about Aristophaneses. Its plural accusative is not /aristopʰáneːs/ but /aristopʰánaːs/. That's not Decl3, it's Decl1. ZOMG, this sounds like a pretty cool analogy. I'm going to tell you the story I'd worked out in my head. Then, why Kühner-Blass says no, you dolt, it's obviously wrong, and this is not an analogy at all. And in the end, I'm going to propose it kind of is after all.

If this was an analogy, this would be the story.

• I need a plural for Aristophanes, stat.
  
• What the hell *is* the plural of Aristophanes?
  
• Um, names don't have plurals, so I can't decline it in my head.
  
• Adjectives in -ης have plurals, and following that pattern, we should get Ἀριστοφάνεις.
  
• But come on, this isn't an adjective, like "Excellent-looking" or something. It's a noun. Right?
  
• And the only nouns I can think of right now, that I can get a plural from, are Decl1 nouns. Like θύτας.
  
• Because I've blanked out on triremes and takers of cities being nouns. Nah, they're adjectives too.
  
• So it's Ἀριστοφάνας.

Or more simply: I have to come up with a new plural. I don't model the plural after adjectives, because I no longer feel Aristophanes to be an adjective, but after nouns. And the only nouns I know of ending in -ης and taking plurals are Decl1 nouns.

Not so: A nice story about markedness and category shift, that is completely and utterly wrong. Kühner-Blass say why in notes 8 and 10 in the same section:

• Working from the proto-Greek for the Decl3, the accusative plural of the adjective should be /eas/. In fact that's the accusative ending used in Homer: ἀολλέας, ἐϋπλεκέας.
  
• (It's even messier, because vowel-stems like this should have had */ans/ not */as/, but I'm going to dodge that.)
  
• This /eas/ should contract in Attic to /ɛːs/.
• But Attic did not use the original accusative plural ending. It uses the nominative ending, /eːs/ < /ees/, for the accusative as well: nominative hairɛːsiteíkʰeːs, accusative hairɛːsiteíkʰeːs, and not hairɛːsiteíkʰɛːs (τοὺς αἱρεσιτείχεις not τοὺς αἱρεσιτείχης).
• Kühner-Blass present /ea/ > /eː/ in their list of contractions, "ε + α = ει (st. η) im Akk. Pl. der III. Dekl. auf εας". They then admit that this is analogy: "indem der kontrahierte Akk. Pl. sich gern nach der Form des Nominatives richtet; vergl. Choerob. in Bekk. Anecd. III. p. 1191: ὅτι ὁμοφωνία ἐστὶ τῆς αἰτιατικῆς τῶν πληθυντικῶν πρὸς τὴν εὐθεῖαν τῶν πληθυντικῶν"
  
• Why the analogy? For starters, this is as close as a nominative and an accusative masculine plural would ever have got in Attic Greek if the analogy didn't happen: /eːs/ vs. /ɛːs/. For second, an accusative plural in /ɛːs/ would sound identical to the nominative singular; that's not enough reason to prevent analogy, but it might reinforce people avoiding /ɛːs/.
  
• /ea/ should have gone to /ɛː/, but outside this accusative plural, Attic otherwise contracts /ea/ to /aː/ in inflections. Kühner-Blass give a range of reasons why: for /ostéa/ > /ostâː/, analogy with the uncontracted neuter, for /kléea/ > /kléaː/, the Attic rule of alpha after epsilon; for /kʰruséas/ > /kʰrusâːs/, avoiding the genitive singular /kʰrusɛ̂ːs/ (again, not a strong reason). The compelling parallel though was in the plural pronouns: nominative /hemées, humées/ > /hemêːs, humêːs/ "we, ye", accusative /heméas, huméas/ > /hemâːs, humâːs/ "us, you". Not /hemêːs/ vs. */hemɛ̂ːs/ (ἡμεῖς vs. *ἡμῆς): that kind of close similarity does not get to survive in pronouns, which is why when /hemêːs, humêːs/ became /eˈmis yˈmis/, analogy shunted /yˈmis/ across to /eˈsis/.
  
• So if you didn't want the accusative plural of Ἀριστοφάνης to be Ἀριστοφάνης, you had two choices. One, slip across to the nominative plural Ἀριστοφάνεις. Two, imitate Decl2 contract adjectives like χρυσᾶς—or more importantly, ἡμᾶς ὑμᾶς, and adopt Ἀριστοφάνας. Which lets you keep a distinct accusative plural.
  
• There is at least one Attic inscription which did just that with a normal Decl3 adjective, /pseudâːs/ ψευδᾶς. Otherwise though, the only place the disambiguating /aːs/ ending turns up in Decl3 is proper name plurals, and adjectives ending in -ετής "-year-old".
  
• Those "-year-old" are reason enough to establish that this is not Decl1 creeping in by analogy, as Kühner-Blass harrumph: "Anmerk. 10. Dagegen der Akk. Pl. Ἀριστοφάνας Plat. Symp. 218, b ist nicht als Übergang in die I. Deklination aufzufassen, vgl. Anm. 8 [on ψευδᾶς], und den Grammatikern nicht zu glauben, die ohne Belege auch andere Kasus nach der I. Deklination bilden: οἱ Δημοσθέναι, οἱ Ἀριστοφάναι (Herodian L. II, 697). Vgl. § 148, Anm. 7 τριακοντούτας u. dergl., bei welchen Wörtern übrigens (Hdn. I, 81) in der Femininbildung auf -ις sich wirklich eine Analogie mit denen auf -της I. Deklination zeigt."
  
• Let's unpack this. We only get to see these endings in the accusative plural: not just in names like Aristophaneses, but also in adjectives like τριακοντούτας "thirty-year-olds", which is also Plato. You don't see in Attic Greek a nominative Ἀριστοφάναι or τριακοντοῦται, or a dative Ἀριστοφάναις, τριακοντούταις. Ἀριστοφάνας only coincidentally looks like a Decl1 form: it's really the Decl3 ending, with exceptional phonology.
  
• The Roman-era grammarians like Herodian, always happy to establish a rule where there is none, came to the same conclusion I did: the plural of names like Aristophanes is Decl1. So they tell you that the plural nominative of Ἀριστοφάνης is Ἀριστοφάναι, and the plural nominative of Δημοσθένης is Δημοσθέναι. But they're just making that up.

True enough, Diogenes Laertius has the Decl3 plurals Ἀντισθένεις Ἀριστοτέλεις Ξενοκράτεις Διογένεις, Didymus the Grammarian has Ἀριστομήδεις, Aelian Ξενοφάνεις, and Plutarch Σωκράτεις. Once Attic Greek was just a written language, though, the epigones went along with what the grammarians told them. Plutarch already had Decl1 Σωφάναι, Synesius has Κλεισθέναι, Theophilus of Antioch Διογέναι, and Theophylact of Ochrid Ἀριστοφάναι.

Doch: OK, we'll concede that this is all hypercorrection and pedantry and artificial. But Kühner-Blass also note that those same "-year-old" compounds get routinely declined in Decl1 post-Classically: ὀκτωκαιδεκέτην (18yo), ἐνενηκοντούτην (90yo) in Dio Chrysostom, δωδεκαετῇ (12yo) in Josephus, ἑκατοντούταις (100yo) in Philostratus, τεσσαρακοντοῦται (40yo) in Eusebius. That's the same analogy as Ἀριστοφάναι, but it doesn't look as artificial any more. It's starting to look like Greek was already inclined to regard the "-year-old" compounds as Decl1.

There's a piece of older evidence that it did. The masculine "-year-old" compounds in -έτης have feminine counterparts ending in -έτις: τριακοντούτης "30yo man", τριακοντοῦτις "30yo woman". By analogy to πολίτης "townsman" πολῖτις "townswoman", Kühner-Blass notes. But feminines in -τις are not derived from masculine Decl3 adjectives in -της. They don't need to: the masculine adjective in -της already gets used as a feminine as well, so ἀνάντης is both masculine and feminine for "up-hill". No, feminines in -τις are derived from masculine Decl1 nouns. Like πολίτης. And forms like ἑπτέτις "7yo girl" and τριακοντοῦτις "30yo woman" don't show up as late as Plutarch and Josephus. They're already there in Thucydides and Aristophanes and Plato and Xenophon.

So where are we? We have what are basically adjectives and adjectival proper names in -ης in Decl3. We have nouns in -ης in Decl1. We have an accusative plural for the adjectives that would case trouble. It goes to the nominative plural by default, to deal with that trouble. For names and "-year-old" compounds though, it goes to a form that looks like the Decl1 nouns instead. It's not a Decl1 ending: it only occurs in the accusative, and there's reason enough within the Decl3 for that ending. Especially with the analogy with pronouns ἠμᾶς ὐμᾶς.

But the three instances where the accusative plural looks like a Decl1 noun... are the instances where the Decl3 stem doesn't look like an adjective. Personal pronouns are going to do desperate things to avoid ambiguity; but personal pronouns are like nouns, not like adjectives. Proper names may have started as descriptions, but now they identify individuals, like nouns do. And thirty-year-olds are people: if τριακοντούτης was an adjective, Plato would have no reason to coin a feminine adjective τριακοντοῦτις. Don't get me wrong, grammatically Plato is still declining τριακοντούτης in Decl3: you have to wait till Josephus for that to break down. But in his derivational morphology, he's treating it like a Decl1 noun.

I can accept that contracting /ea/ to /aː/ in τριακοντούτας is morphophonology that has nothing to do with Decl1. But isn't it convenient that this exceptional contraction happens only in the adjectives that look most like nouns anyway, proper names and "-year-old" compounds?

Kühner-Blass lays out all this evidence to say the Decl1 take on Ἀριστοφάνας is wrong—mentioning τριακοντοῦτις as an aside ("übrigens"). I don't think it is an aside, I think it proves that people were slowly starting to think of these noun-like adjectives in noun-like morphology. In Attic Greek, this happened only with the morphologically ambiguous accusative plural /aːs/; but every analogy has to begin somewhere.

Modern Greek reversal: As a coda, the Modern plural of Aristophanes has its own analogical story. The third declension is dead, though learnèd loans have half-heartedly revived some adjectives, which people are at a loss to decline. (Sarantakos in Greek has more on that.) Aristophanes and Demosthenes are treated as first declension nouns in the mainstream.

When it comes to plurals, the contemporary language has a choice. The modern default is to go from -ης to -ες: ψεύτης ψεύτες, μαθητής μαθητές. But for agentive nouns in -τής, that looks to be a recent, artificial levelling. The traditional plural of -τής was -τάδες, and that survives in colloquial use: μαθητής μαθητάδες.

But if a word was a late coinage or import into Greek and ended in -ής, its plural wasn't -ές or -άδες. Newfangled words, being unfamiliar, took a morphologically more transparent ending, and kept the thematic vowel: -ήδες. So καδής καδήδες (kadı, Ottoman judge), καφετζής καφετζήδες (café owner), πεταλωτής πεταλωτήδες (farrier); βαρκάρης βαρκάρηδες (boatman), μανάβης μανάβηδες (greengrocer). Modern Greek kinda-sorta has masculine adjectives ending in -ης (feminine -α), and they have the newfangled ending too: ζηλιάρης ζηλιάρηδες "jealous", ζηλιάρα ζηλιάρες.

So we have two old plurals, -ες and -άδες, of which the former is beating back the latter; and we have a newfangled-word plural, which ain't going nowhere: μαθητάδες has become μαθητές, but μανάβηδες is not becoming μανάβες. (Modern Greek speakers, stop guffawing at the very notion it could.)

What's the plural of Κωστής "Con" and Νικολής "Cole"? And Αριστοφάνης? You got it: Κωστήδες Νικολήδες Αριστοφάνηδες. New-fangled word plural. It's not the names that are new-fangled; it's the notion that they should have a plural. Ancient Greek took the plural of Aristophanes in the direction of a less marked inflection: the adjective was becoming a noun. Modern Greek takes the plural of Aristophanes in the direction of a more marked inflection: it's an unfamiliar kind of noun. And when it gets there, Αριστοφάνηδες patterns with ζηλιάρηδες. So it's back in the company of adjectives.

Um... yeah. God wot how many of you got this far.