Accent in Modern Compounds: Speculations

Ancient Greek had mostly recessive accentuation in compounds, as we saw last post; there are exceptions, some nice, some messy, and a major group of exceptions with verbal nominals.

If an Ancient Greek compound survived into Modern Greek, it had no reason to change accent location; we saw that for simple words as well, two posts ago. The default in Modern Greek is also recessive; and again, recessive accent is defined by what was acceptable in Ancient Greek.

That's because Modern Greek speakers could work out which recessive accents were acceptable, from the compounds surviving in the language. Modern Greek speakers never saw any compound noun ending in -ης change from penult to antepenult stress—because in Ancient Greek, antepenult stress with -ης was impossible. So recessive stress for -ης still means stress on the penult.

For example, given "tailor" is ράφτης < ῥάπτης, a Western-style tailor is φραγκοράφτης, "Frankish tailor". There's nothing in Modern phonology to prevent *φραγκόραφτης; but with no precedent for an accent like ˈ$.$.ης from Ancient Greek, noone would stick their neck out to coin such a novel accent.

On the other hand, -ας was a novel inflection in Modern Greek, which allowed antepenult stress inherited from the third declension: γέρων > γέροντας. That means that if a noun ending in -ας entered into a compound, its recessive accent could go back to the antepenult.

Take μάγκας, which is what I started off with. Someone pretending to be a mangas, but too soft to live up to the needed machismo, is a "butter-mangas", a βουτυρο-μαγκας. The compound is recessively accented; and because γέροντας is accented three syllables back, so is butter-mangas: βουτυρό-μαγκας.

Here's another instance of Ancient Greek accentuation hanging over Modern Greek recessives. Ancient Greek diminutives ending in -ιον developed into Modern Greek -ι neuters. So Ancient παιδίον "boy" is Modern παιδί. "Altar-boy" in Modern Greek is "priest-boy", παπαδο-παιδι. The compound needs to be recessively accented. Again, recessive accent in Modern Greek should result in *παπαδόπαιδι. But it does not: it results in παπαδοπαίδι. παπαδοπαίδι obeys the mora restrictions of Ancient Greek, in which it would have been *παπαδοπαίδιον.

The -ον suffix has not been there for 1500 years; so it's not like any Modern Greek speaker is mentally adding the -ον suffix back in, and counting the missing mora. Instead, Modern Greek speakers are noticing the -ι(ον) compounds that have survived from Ancient Greek:

περι-βόλιον > περιβόλι "garden", εἰκονοστάσιον > κονοστάσι "iconostasis", ὀκταπόδιον > χταπόδι "octopus", ὡρολόγιον > ρολόι "clock", ἐλαιοτριβεῖον > λιοτρίβι "oil press". (The last one has moved up its accent—but -εῖον and -ιον often alternate as locative suffixes.)

None of them is accented any further back than a syllable before -ι-. So no new compound is accented further back than a syllable before -ι- either. And that includes novelties like αραποσίτι "Arab wheat = maize" and καφεκούτι "coffee box".

There's a second way of making sure accent is recessive: changing the inflection of the second noun. Instead of X-Y with the original inflection of Y, you strip the inflection of Y, and add a new inflection instead. The usual inflection to add is -ο; for instance, αστραπή "lightning" + βροντή "thunder" > αστραπό-βροντ-ο, with the final stress of βροντή wiped out into recessiveness. The possibility of switching inflection in compounds is ancient, e.g. ἄκρα "edge" + πρῷρα "prow of ship" > ἀκρό-πρῳρ-ον "edge of prow".

The reason why such compounds are recessive is, stripping the original inflection makes this a new word, which is accented from scratch. That's "recessive" as in "recessive by Ancient Greek norms", of course; when -ι rather than -ο is used as the new inflection, the accent is only one syllable back, like with παπαδοπαίδι: έξω "out" + εκκλησία "church" > ξω-κλήσ-ι "chapel" (εξωκκλήσιον).

Modern Greek particularly likes switching to -ο as a new inflection when compounding neuters ending in -ι. (Yes, I just said that -ι is also used as a new inflection. This is language, don't expect it to be economical.) The -ι suffix is stripped from the noun, and the old neuter inflection -ο brought back in. Then the whole compound is accented recessively. παπαδο-παιδι cannot be accented as *παπαδόπαιδι. But βούτυρο + παιδί "butter boy, weakling" can be compounded as βουτυρο-παιδ-ο, and accented as βουτυρόπαιδο.

The suffix -ι- was an Ancient diminutive, which is why historically it makes sense to strip it out. Of course, the full Ancient diminutive was -ι-ον, so the switch inflection is merely restoring the deleted -ον suffix. But because Greek speakers are not historians, -ι is also stripped when it has nothing to do with the Ancient diminutive: τροχό-σπιτ-ο "wheel house = caravan" < σπίτ-ι < ὁσπίτιον: Latin hospitium; Μακλαουντόσο-γ-ο "MacLeod clan" < σό-ι < Turkish soy.

I'm going to resist the temptation to get sidetracked into discussing the mechanics of how words are attached in Modern Greek, and work through where the linking consonants in παπα-δ-οπαίδι or Μακλαουντόσο-γ-ο come from. Maybe another post.

So when *do* Modern compounds preserve accent? I'm going to use some diagnostic prefixes. παλιο- "old; damned", as a generic derogatory term, can be applied to any noun—just as you could apply pseudo- to any Ancient noun (and, either as colloquial ψευτο- or learnèd ψευδο-, to any Modern noun as well). More vulgarly, σκατο- "shitty" is just as generic in its application.

The prefixes are recessive: παλιόκοσμος < κόσμος "damned world", παλιόκαιρος < καιρός "damned weather". The prefixes also force the -ι neuter suffix to be dropped—which again results in recessives: παλιόπαιδο "damned boy = scoundrel" < παιδί, like καμποχώραφο "valley field" < χωράφι. But feminine nouns are left with their accent intact when compounded with those suffixes, even if they're stressed on the ultima.

It's been too long since I've used a YouTube song to illustrate a linguistic point. Take Παλιοζωή, παλιόκοσμε, και παλιοκοινωνία, music by Iosif Ritsiardis, lyrics by Mimis Traiforos, ca. 1950:

The song lyrics would do service in another post, along with Έ ντε λα μαγκέ ντε Βοτανίκ and Εφτά νομά σ' ένα δωμά, on how truncating the final syllable of Greek words still leaves them intelligible. In this post, though, let's concentrate on the title words:

• παλιόκοσμε < κόσμε "damned world", recessive
  
• παλιοκοινωνία < κοινωνία "damned society", recessive in Ancient Greek, because the suffix was -ιᾱ, with a long final syllable
  
• παλιοζωή < ζωή "damned life". In no way whatsoever recessive.

The feminine παλιοζωή avoids becoming the recessive *παλιοζώη, whereas the masculine καιρός does become the recessive παλιόκαιρος. This time, we can't blame it on Ancient Greek phonology: Ancient Greek wouldn't have a problem with *παλαιοζώη. In fact, Ancient Greek took ὀπή "hole", and created the recessively accented μετόπη "metope".

What I think has happened here is: Ancient Greek had relatively few instances of compounds ending in a simple feminine noun, like μετόπη—which would be recessively accented; but it had abundant instances of compounds ending in a feminine verbal noun—which kept its accent: μεταβολή "change", καταστροφή "disaster", ἐντροπή "shame", προκοπή "progress". The relation between -βολή, -στροφή, -τροπή, -κοπή and the verbs βάλλω, στρέφω, τρέπω, κόπτω was no longer apparent, and the verbs no longer looked like that anyway. So Modern Greek speakers had no clear notion that these were verbal nominals at all. Their conclusion was, instead, that feminine compounds keep their accent.

Except that "wild duck" is a feminine compound, and it does not keep its accent: άγρια + πάπια > αγριόπαπια. For that matter, κουφός + βράση > κουφόβραση "muggy weather". Yes, βράση is ancient βράσις "boiling", and -ις nouns as I explained are recessive. But how did people work that out in Modern Greek, with the -ις suffix long gone?

Some verbal nominal suffixes remained productive in Greek; some did not. Verbal nominals had different accentuation in compounds than simple nouns did; but Modern Greek speakers would not have a clear notion which was which; we saw them already conflated for with feminines. Modern Greek speakers shouldn't have had a clear notion of what nouns had long endings in Ancient Greek, to base their decision on how to accent their compounds.

What it does look like, is that Modern Greek inherited a lot of verbal nominal compounds, with non-recessive accent; and they also got a lot of compounds with default recessive accent. Modern Greek couldn't tell from the morphology which compounds were built on verbal nominals. But they could see that there was *some* sort of pattern, and they guessed what the pattern was, with the evidence they had before them: the endings they could see.

I pulled off the shelf Konstantinos Minas' commentary on Modern Greek grammar—

Μηνάς, Κωνσταντίνος. 2008. Παρατηρήσεις στη γραμματική στης νεοελληνικής. Athens: Νεφέλη.

—and Minas lets on what kind of guesswork is going on. Let me quote at some length: (pp. 39).

For nominals that are either compounds or derived from compounds (parasyntheta): if their endings are characteristically permanently oxytone (stressed on the ultima), then the stress does not move: e.g. υπαρχηγός, υπολοχαγός, προσανατολισμός, αλληλοθαυμασμός, παλιοσπιτάρα, διαβλητός, προσχολικός "deputy leader, under-captain (= lieutenant), orientation, mutual admiration, damned huge house, reproached, pre-school". But with alpha-privative, stress moves up in adjectives: e.g. αφύσικος, αδιάκριτος "unnatural, indiscrete". But: αρχιστράτηγος, αντιστράτηγος, υποστράτηγος "generalissimo, deputy general, major general".

Unpacking:

• If a modern derivational suffix, like the augmentative -άρα, is stressed, then sticking it onto a suffix will not affect its stress. That makes sense as a layering of rules: σπίτι, παλιό-σπιτ-ο (recessive accent), παλιοσπιτ-άρα (new suffix on the compound, and that suffix is now stressed). Same goes for -μός, which has survived from Ancient Greek.
  
• If a suffix is an ancient derivational suffix (verbal nominal), which is stressed, then the accent does not move. This happens even when the suffixes are no longer productive. Because Modern Greek speakers don't know Ancient Greek derivational morphology, they guess what those suffixes are—they see suffixes like -τής and -ία keep their stress on existing compounds, so they maintain that pattern. That's the only meaning "characteristically permanently oxytone" can have for such ossified suffixes.
  
• The alpha privative (α-, corresponding to un-) is always recessive in Ancient Greek (pace ἀϝεργός > ἀργός, the exception Tom Recht unearthed in comments.) So it stayed recessive in Modern Greek, whose speakers worked out the pattern easily.

But the patterns work by analogy: Modern Greek speakers trying to reverse engineer the Ancient Greek distinction between recessive normal compounds, and verbal nominals preserving their accent (and with obvious-looking suffixes).

There's an oddity in Minas' list: a deputy-captain (υπολοχαγός) is not recessive, but a deputy general (υποστράτηγος) is recessive. Now, the general-compounds are Ancient: ἀρχιστράτηγος is in the Septuagint and Josephus, ἀντιστράτηγος in Thucydides, ὑποστράτηγος in Xenophon. Back then, accentuation was consistently recessive.

Yes, -ηγός/-αγός is a active psychopomp-type compound (στρατ-ηγός "army leader", λοχ-αγός "company leader"), which is accent-preserving. But preposition + "army leader" is treated as a new compound, and is accented from scratch, ignoring the existing accent of the compound. So Minas has got the ancient rule wrong here.

On the other hand, ὑπαρχηγός is a Modern coinage—it's not in TLG. And "under-captain" *is* in Xenophon... as the recessive ὑπολόχαγος. So Modern learnèd Greek is getting the recessive accentuation wrong: it's reaccenting the word as accent-preserving.

I haven't researched the history of why υπολοχαγός is accented where it is; but I'm reminded of the ugliness of contemporary coinages like μετααποικιακός "post-colonial", which ignore hiatus—something both Ancient Greek and Vernacular Modern Greek respect. I think such learnèd modern coinages did not have the scholarship of Puristic to get the Ancient rules right, and aren't colloquial enough to get the Modern rules right. The result was not to apply the phonological rules that smooth compounds over—eliminating hiatus, reaccenting the word recessively. There is a fearfulness from the coiners of such words about getting them right, which leaves the compound components undigested and intact, violating the phonology of both languages.

So it's no surprise that when the modern fad of using para- as a productive prefix with nouns arrived in Greek (modelled after paramilitary), the word was applied without recessive accent: παρα-στρατός "para-army", παρα-κράτος "para-regime". Of course, these compounds are endocentric (a para-army is a kind of army), and have no precedent in older Greek: ἀπό-κρατος is an exocentric adjective, "without strength" ("de-strength-ed"), and ἀπόστρατος "de-armied" is a Modern but Classically-correct exocentric adjective, meaning "demobbed". When the novel usage arrived, word coiners just weren't as scrupulous about accent: the words were being coined in a linguistic Neutral Zone, beyond the reach of Ancient or Modern rules.

Puristic Greek has a lot to answer for.

I'm going to propose out of whole cloth the following as ways Greek speakers would guess what's going on with compound accent in the modern language. My guess is, if people have the choice between preserving the accent on Y, or accenting X–Y as a new word (recessively), there will be factors like these influencing them:

• The more the X–Y compound is like Y in isolation, the likelier that Y will preserve its stress. That's other things being equal, and is assuming that stress is iconic—it reflects Y being somehow unchanged in the compound. In particular:
  
  • If X–Y is exocentric, it's not describing a Y, and it's likelier to be recessive. If X–Y is endocentric, it's likelier to preserve accent. (I think that's part of the reason behind παραστρατός.)
    
  • If X–Y switches inflection, e.g. replacing -ι with -ο, then the second half of the word no longer looks like Y. So it will be recessive.
  
  
• If a derivational suffix is added to a compound, then its accent takes priority: it shouldn't matter whether it is added to a compound or a simple word, the derivational suffix is the last thing to affect the word.
  
• If there is an obvious precedent of a verbal nominal suffix preserving its accent in compounds, and the word looks like ending in that suffix (even if it etymologically doesn't), the accept will stay put. By analogy.
  
• If the accent would obviously violate the precedent of the Ancient three mora rule, because words with those inflections have not switched accent in Modern Greek, then the accent does not move. (So -η or -ης would not normally have an accent three syllables back.)
  
• If the inflection is new to Modern Greek, there is no precedent from Ancient Greek about whether to have recessive or non-recessive accent—leave the accent in place. (E.g. loanwords from Romance or Turkish in -ες.) In that case, whether to move the accent back to the antepenult depends on whether there is precent in simple nouns for accent on the antepenult: again, analogy with precedent.
  As it happens, there are no nouns ending in -ες and accented on the antepenult. A compound of a such a noun isn't going to be the first such noun to be accented on the antepenult: compounds will not invent new accentuation in the paradigm, when Greek speakers are so desperately looking for precedent. Just as I'd argued for *φραγκοράφτης "Western tailor"; the accentuation *ψευτόκαφες "bogus coffee" should also be impossible, for the same reason. (Unsurprisingly, Google only finds 3 instances of ψευτοκαφές, and none of ψευτόκαφες.)
  
• On the other hand, if a noun is two-syllables long and recessively accented, it's going to look like all the other two-syllable nouns which are recessive in compounds, and it will be under pressure to follow suit.

There will be contradictory outcomes because of the contrary pressures. The word "priest", παπάς, has an ending that's post-Classical (it's Koine, not Early Modern); but unlike -ες, you can point to -ας words accented in the antepenult; so recessive accent is possible for -παπάς compunds. "Damned priest" is the accent-preserving παλιοπαπάς; there would have been little precedent for compounds either way, and the outcome seems to have been to leave the accent alone. On the other hand, one of Kazantzakis' favourite invectives is τραγόπαπας, "goat-priest", with recessive accent.

I admit, I'm not sure how to explain this, except to speculate that παλιο- has less semantic weight than "goat", so was less likely to affect the accent of the compound—and that, because of recessive -ας nouns in Modern Greek, the accent of -άς compounds could go either way. The fact that there are such inconsistencies in accentuation shows that Modern Greek has not come up with a consistent rule, to deal with the twin system it inherited from Ancient Greek.

I'm not happy with where I've got to; so I'm going to do a quantitative survey of Greek compounding next post, and see what that tells me.