Accent on compounds with inflection switched to -as

In the survey of compound accentuation on slang.gr that I'm about to start on, I've left out compounds switching inflection. These inflections, I had reasoned, bring along their own accentuation; and since the whole word was being reaccented from scratch, that new accentuation is recessive. Which is clearly the case when the switch in inflection is to -o or -i: παιδί > παλιό-παιδ-ο, εκκλησιά > ξω-κλήσ-ι.

But slang.gr captures slang coinages; and slang is not restricted to -o and -i for its inflection switching on compounds. There are several other inflections that can be used as derivational morphology—to indicate a person associated with the preceding stem or stems.

(The language advisory, as always, should be taken with any post relying on source material from slang.gr, and Yr Obt Svt should not be taken as condoning the stereotypes or attitudes borne in the invective.)

Three such inflections that get a lot of use on slang.gr are the feminines -α and -ω, and the masculine -ας. For examples of the -ω, see τρελοκαμπέρω "madwoman" (ultimately from τρελο-Καμπέρος "Mad Kamberos", an aviator: details at Sarantakos'), αεραγκίστρω "air hook" (Kaliarda for "magnet"), αρπαχτοτσιμπούκω "grab-and-fellater, slut", καμαρωπαπούτσω "woman who (looks down, like she) is admiring her own shoes: shy woman".

For examples of -α, see for example κοντοκλώτσα "short-kicker, woman with short legs", νεροφίδα "watersnake, one who drinks lots of water", or any of the dozens of compounds ending in -μούνα, -πουστα, or -πουτσα "cunt, faggot, dick" (μουνί, πούστης, πούτσος). There's a difference of accentuation right there, between non-recessive -μούνα and recessive -πουστα/-πουτσα, which could be because of the referent being feminine or masculine, or with the original stress of μουνί not being recessive, and πούστης, πούτσος being recessive. That's not the affix I'm going to consider though.

I'm instead looking at -ας. We've already established that -a(s) is where all confusion breaks loose as to where to accent compounds recessively. There is such seeming confusion about -ας compounds as well: though the switch in inflection is meant to cause recessive accentuation in -ας compounds—as it does in -ος compounds—there is a lot of -ας compounds stressed on the penult.

This list of inflection-switching compounds ending in -ας, I swear on the bones of Psichari, was objectively pulled out of slang.gr without ulterior selection. I have to say that, because it results in some quite nice distinctions. Here's the list of -ας compounds stressed on the penult, and on the antepenult:

Antepenult	αγριόπουστας	πούστης	angry faggot
	ανθυποτίποτας	τίποτα	deputy vice nobody (bottom of the food chain)
	γαμαωδέρνουλας	δέρνω	fuck-and-basher (someone in command of situation; a subject of an extensive post previously on this blog)
	δε(ν)παλεύουρας	παλεύω	an "I can't deal with you" kind of guy
	διαβολόπουστας	πούστης	devil's own faggot
	θρασόπουστας	πούστης	impudent faggot
	καβατζόπουστας	πούστης	selfish faggot
	καραφλοχαίτουλας	χαίτη	baldie with a mane (going bald up top, but leaving their hair long)
	λαδοπόντικας	ποντικός	oil rat (someone greasy; someone who seeks bribes—"greasing")
	μαλακοπέρδουλας	πέρδομαι	shlemiel farter
	μαλακοπίτουρας	πίτουρα	bran shlemiel (probably, meddling shlemiel, from the proverb: "if you mess with bran, the hens will eat you")
	ντόμψωλας	ψωλή	shlemiel ("thump-dick"? English "dumb dick"?)
	ντουμανόψειρας	ψείρα	smoke-flea'er (getting high on someone else's marijuana smoke, from ψειρίζω "to flea" = "to steal" damned gypsy
	παλιόγυφτας	γύφτος	damned gypsy
	τρικάτσουλας	κατσούλι (?)	idler ("thrice cat?")
Penult	βαβουροπατάτας	πατάτα	noisy potato guy: fat spoiled person who pretends to be active
	γιαταμπάζας	μπάζα	a "fit for the landfill" guy, ugly, inexcusable
	γλειψαρχίδας	αρχίδι	ball-licker, flatterer
	δροσοπεζούλας	πεζούλι	a "cool bench" guy, idler (who hangs around cool windowsills in summer instead of working)
	εξωφυλαρούχας	ρούχα	a "watch over the clothes outside" guy (benchwarmer)
	καλοχαιρέτας	χαιρετώ	a "greet it fondly" guy, gladhander
	καψοκαλύβας	καλύβα	hut-burner, self-sabotaging
	λαμπογέλας	γελώ	shining-smiler, baldie
	μαμκακανανύστας	νάνι/νύστα	a "nom-nom, poo, beddie-byes" guy, idler
	μουνοτρέχας	τρέχω	a "run after cunts" guy, pussy-whipped
	μπορόλας	όλα	a can-do guy
	μπρατσορακέτας	ρακέτα	an arm racket guy, someone showing off his biceps
	ξερόλας	όλα	know-it-all
	οτινάνας	ό,τι να 'ναι	a "whatever" guy (someone random, inexplicable incoherent)
	πανταόλας	όλα	an "anything, any time" guy, overcommitting oneself
	παράκμας	ακμή (this is incorrect, but I'll explain why later)	decadent
	παρταόλας	όλα	a "take-it-all" guy, pimp, moneysucker
	πορδορούφας	ρουφώ	saliva-face, dribbling idiot
	σαλιαμούτρας	μούτρο	fart-sucker, flatterer
	φυσαρούφας	ρουφώ	a blow-and-suck: bong; someone who goes back on their word
Ultima	βαψομαλλιάς	μαλλιά	hair-dyer, older man concealing his age
	καικαλάς	καλά	a "yeah right" guy
	καραφλομαλλιάς	μαλλιά	bald-hair guy (going bald up top, but leaving their hair long)
	καραφωτιάς	φωτιά	a "real fire" guy, spitfire, sexually active man
	καφρομεταλλάς	μέταλλο	boor-metal guy (death metal fan)
	κρυφοσκυλάς	σκύλος	hidden dog-guy (secret fan of "dogpound" music—skyladiko, disreputable pop)
	μαστακουνάς	κουνώ	You'reShakingThemForUs (mock surname of someone who has said something stupid, Captain Obvious; the "them" are testicles)
	μπακαυτιάς	αυτί	a Bakatsias-ears guy, Big Ears
	πατσοκοιλιάς	κοιλιά	tripe-belly guy, fatso (with a belly full of tripe)
	σαπιοκοιλιάς	κοιλιά	rotten-belly guy, fatso (has let himself go)
	τραχανοπλαγιάς	πλαγιά	frumenty mountainside-guy, hillbilly, useless football player
	χοντρομπαλάς	μπάλα	fat ball guy, fatso
	χωριατλαμάς	Turkish atlama	village leaper (?), uncouth

See the patterns? Yes? No? To my surprise, the patterns actually fall out quite nicely, in all but a couple of cases. Let's go through slowly.

First, you will have noticed a disproportionate number of compounds which are actually just phrases with a masculine inflection tacked on. γιαταμπάζας for example is merely the phrase για τα μπάζα "for the landfill", with the -ας ending appended. In a normal compound, there is no conceivable place for definite articles: compounds link bare noun or verb stems. But this is not a normal compound.

We could argue that slang.gr is exaggerating the trend to such phrase-based compounds; but slang.gr did not invent them. ξερόλας "know-it-all" < (τα) ξέρω όλα certainly predates the site, and φαταούλας "eat-it-all, avaricious" < φα τα ούλα predates it yet further. As "know-it-all" shows, this is a known pattern of English; it's just innovative in Greek.

So these are compounds which quote a phrase in Greek. Like the quotative compounds we saw last post, ξεμπράβο and τι άμμος και ξε-άμμος, if a compounds quotes a word rather than treat it as a normal element, the stress is left alone. This is particularly needed when what is quoted is an entire phrase; it would make no sense to reaccent this like a normal noun–noun compound.

That rule takes care of a lot of our non-recessive compounds.

• γιαταμπάζας < για τα μπάζα
  
• εξωφυλαρούχας < έξω φύλα ρούχα
  
• καικαλάς < και καλά
  
• καλοχαιρέτας < καλοχαιρέτα
  
• μπορόλας < μπορώ όλα
  
• ξερόλας < ξέρω όλα
  
• οτινάνας < ότι να 'ναι
  
• πανταόλας <πάντα όλα
  
• παρταόλας <πάρ' τα όλα
  
• φυσαρούφας < φύσα ρούφα
  
• μαστακουνάς < μας τα κουνάς
  

In fact, it also takes care of a lot of inflection-switching compounds which are accented on the ultima, but don't end with -ας:

ελαμωρές	έλα μωρέ	an "oh come on" guy
ξερωγώς	ξέρω γω;	a "dunno" guy
πουθενάς	πουθενά	a "nowhere" guy, someone nowhere to be found

Note that where the phrase ends in a vowel, the -a- of -as is elided: Modern Greek has no interest in the hiatus that would result from */e.la.mo.ˈre.as/ or */kse.ro.ˈɣo.as/.

Our rule does *not* take care of μαμκακανανύστας < μαμ, κακά, νάνι: the baby talk words have clear accents, and should have been μαμ, κακά, νάνι-ας /ˈmam kaˈka ˈnani as/ > μαμκακανάνης /mamkakaˈnani-s/. I'll explain what's happened here at the end.

That leaves us with:

Antepenult	αγριόπουστας ανθυποτίποτας γαμαωδέρνουλας δε(ν)παλεύουρας διαβολόπουστας θρασόπουστας καβατζόπουστας καραφλοχαίτουλας λαδοπόντικας μαλακοπέρδουλας μαλακοπίτουρας ντόμψωλας ντουμανόψειρας παλιόγυφτας τρικάτσουλας
Penult	βαβουροπατάτας γλειψαρχίδας δροσοπεζούλας καψοκαλύβας λαμπογέλας μαμκακανανύστας μουνοτρέχας μπρατσορακέτας παράκμας πορδορούφας σαλιαμούτρας
Ultima	βαψομαλλιάς καραφλομαλλιάς καραφωτιάς καφρομεταλλάς κρυφοσκυλάς μπακαυτιάς πατσοκοιλιάς σαπιοκοιλιάς τραχανοπλαγιάς χοντρομπαλάς χωριατλαμάς

Hold on to φυσαρούφας "blow-and-suck = bong, equivocator", btw. We'll come back to it because it explains some seeming exceptions.

For our next cull of forms, recall that some -as words (those inherited from Ancient Greek) are recessively accented on the antepenult, while other -as words (those new to Modern Greek) are recessively accented on the penult. That means that there will be uncertainty about where to recessively accent -as accents in general. But:

• If the right-hand word in the compound is stressed on the antepenult, the stress stays on the antepenult in the compound.
  
• If the right-hand word is stressed on the penult, and is three syllables long (–´–), then the word is clearly not going to be accented on the antepenult no matter how it gets compounded: it will count like one of those Ancient Greek nouns with the long ultima. So Ancient Attic καλύβη "hut" has survived in Modern Greek in its Doric variant, καλύβα. In Ancient Greek, both καλύβη and καλύβα have long ultimas; so they can only be accented on the penult. Modern Greek gets that καλύβα is one of those penult-no-matter-what words, and doesn't try to accent it any further back. That accentuation extends by analogy to -ας endings on the same stem.
  
• If on the other hand the right-hand word is stressed on the penult, but is two syllables long (´–), then the word is treated as recessive. And compounding the word has the word still treated as recessive; now there's more syllables to the word, the accent can go three syllables back. Recall πάπια > αγριόπαπια "wild duck".
  
• If finally the original word is accented on the ultima, its accent is preserved.

So: accent is preserved in switching inflections to -as; but two-syllable–long ´– words are treated as recessive, so they end up accented three syllables back in compounds.

Let's put that to the test:

• τίποτα (´––) > ανθυπο-τίποτα-ς (´––)
  
• πίτουρ-α (´––) > μαλακο-πίτουρ-ας (´––)
  
• πούστης (´–) > αγριό-πουστ-ας (´––): two-syllables, recessive
  
• ψείρα (´–) > ντουμανό-ψειρ-ας (´––): two-syllables, recessive
  
• αρχίδι (–´–) > γλειψ-αρχίδ-ας (–´–): three-syllables, non-recessive
  
• πατάτα (–´–) > βαβουρο-πατά-τας (–´–): three-syllables, non-recessive
  
• μαλλιά (–´) > καραφλο-μαλλι-άς (–´)
  
• κοιλιά (–´) > σαπιο-κοιλι-άς (–´)

This pattern extends to non-compound use of the inflection switch as well: σαχλαμάρα "balderash" > σαχλαμάρας "self-promoting fool".

That deals with a lot of forms, though not all. Let's do the cull:

Antepenult	γαμαωδέρνουλας δε(ν)παλεύουρας καραφλοχαίτουλας λαδοπόντικας μαλακοπέρδουλας ντόμψωλας τρικάτσουλας
Penult	λαμπογέλας μαμκακανανύστας μουνοτρέχας παράκμας πορδορούφας
Ultima	καφρομεταλλάς κρυφοσκυλάς μπακαυτιάς χοντρομπαλάς

We're down to the exceptions now. For the first class of exceptions, recall φυσαρούφας, "blow-suck". This is not a normal, noun–noun compound, because there is no such noun as *ρούφα. But *ρούφα clearly is not a fluke, since it shows up again in πορδορούφας.

These are an innovative class of compound—like a lot of slang compounds are; the right-hand of the compound is the bare verb root, without the affixes that would turn the verb into a noun. In fact, the simplest thing to call this in Modern Greek is an imperative, ρούφα "suck!"; and φυσαρούφας is indeed a compound of two imperatives, φύσα! ρούφα! (Modern Greek doesn't have an infinitive, so the imperative is as good a place as any to get a bare verb form from.) So πορδορούφας is formed from πορδ-ή "fart" and the imperative ρούφα "suck".

Now, there are two major conjugations in Modern Greek. The old non-contracted verbs still have their present imperative end in -ε: Ancient λύ-ε, Modern λύν-ε. The old contracted verbs have merged into the αω-conjugation for the imperative, which ends in -ᾱ: ἀγάπα. Imperatives are recessively accented: δέν-ε "tie up", κάλλιο γαϊδουρό-δενε παρά γαϊδουρογύρευε, "better to tie up a donkey than to go looking for a donkey". But the -ᾱ ending was long, so it has stayed accented on the penult: αγάπα. This carries across to verbs which now have an -α imperative, even if historically they shouldn't: Ancient τρέχε, Modern τρέχα "run".

The long imperative endings in γέλα "laugh!", τρέχα "run!", ρούφα "suck!" explain the penult accents of λαμπογέλας, μουνοτρέχας, πορδορούφας. Precedent also counts for words ending up in this class: I'm reasonably sure that μουνοτρέχας "cunt-runner, pussywhipped" is modelled after the much older παπατρέχας "Priest Runaround, someone acting in haste".

There's a further nastiness to this class that I won't try to resolve: there is variation for inflection in these imperatives-turned-nouns, between -ας and -ης (where -ης is always penult-accented).

• Non-contracted verbs end up as nouns in -ης: φά(γ)ε "eat!" > χαραμο-φά-ης "someone eating in vain, a waste of space".
• Contracted verbs can have either -ης or -ας: κέντα "poke!" > πισω-κέντ-ης "back-poker, gay top" (possibly by analogy with πισω-γλέντ-ης "back-feaster, gay bottom"); γάμα "fuck!" > γιδο-γάμ-ης "goat-fucker". (You won't be surprised that -γάμης was the majority of such compounds in slang.gr .)

In fact, the choice between -ας, -ης, and -ος is generally a free choice in forming such compounds in Greek, and the accent rules for -ας, -ης, and -ος are all different. I don't think the choice is particularly predictable. In fact, I was surprised to see πούστης and γύφτος compound as -πουστας and -γυφτας, ignoring their original masculine inflections, and I think that's slang deliberately being innovative. But this article presupposes that -ας has been chosen as the inflection.

Culling those imperatives turned nouns, we have:

Antepenult	γαμαωδέρνουλας δε(ν)παλεύουρας καραφλοχαίτουλας λαδοπόντικας μαλακοπέρδουλας ντόμψωλας τρικάτσουλας
Penult	μαμκακανανύστας παράκμας
Ultima	καφρομεταλλάς κρυφοσκυλάς μπακαυτιάς χοντρομπαλάς

For most of the rest, the accent is caused by completely different -as suffixes, which have distinct functions, and happen to have different accent rules. Admittedly, that sounds like cheating; but those distinct functions of -as do exist, and the accent of a derivational affix takes priority over the accent of a compound.

• So -ουλας/ουρας is a distinct suffix for "person characterised by X", and its accent is consistently antepenult: γαμάω (και) δέρνω > γαμαωδέρν-ουλας, δεν παλεύω > δε(ν)παλεύουρας, καραφλός + χαίτη > καραφλοχαίτουλας, μαλάκας + πέρδομαι > μαλακοπέρδουλας. Cf. νυσταλέος "sleepy" > νυσταλέουρας "sleepyhead"
  
• παρά + ακμή > παρακμή "decline" should have switched inflection as *παρακμάς. But the -ας of παράκμας is not just a switch of inflection, and παράκμας is not in fact derived from παρακμή "decline". This instance of -ας is used to form truncated forms, and παράκμ-ας is a truncation of παρακμ-ιακός "decadent". That truncating -ας suffix is accented on the penult; cf. Παναθηναϊκός "Athens United Football Club" > Πανάθας, ανθυπ-ολογαχός "deputy vice captain = second lieutenant" > ανθύπ-ας.
  
• There is a longstanding accented -άς suffix for a professional of X, or more generally someone characterised by X: ψωμ-ί "bread" > ψωμ-άς "baker", ροκ "rock music" > ροκ-άς "rocker", φουστανέλ-α "fustanella, kilt" > φουστανελ-άς "kilt-wearer". Which means that a compound with *any* accent can end in -άς; but if the compound ends in unaccented -ας, then the original right-hand word probably wasn't accented on the ultima.
  That explains καφρομεταλλάς κρυφοσκυλάς μπακαυτιάς χοντρομπαλάς. This explanation is less arbitrary than it sounds, because the -άς suffix has already been added to the right-hand nouns: μεταλλ-άς "metal fan", σκυλ-άς "skyladiko fan", αυτιάς "big ears" already exist (the first two on slang.gr, the last on standard dictionaries). I don't think μπαλάς exists independently as ball-player; it does exist as the Blackspot seabream (Pagellus bogaraveo), but I'm happy for one out four to be a formation with -άς from scratch.

We're now down to:

Antepenult	λαδοπόντικας ντόμψωλας τρικάτσουλας
Penult	μαμκακανανύστας

Of these,

• λαδοπόντικας does look inconsistent in its accent with ποντίκι, ποντικός "mouse"; but this recessive accent is an established pattern with "rat"-compounds (τυφλοπόντικας "blind rat = mole"). My surmise is that τυφλοπόντικας is far far earlier than the current slew of -ας inflection-switching compounds, and its inflection was by analogy with recessive -os compounds: τυφλός + ποντικός > τυφλο-πόντικ-ος > τυφλοπόντικας.
  
• The etymology of τρικάτσουλας is obscure; I'm guessing τρία + κατσούλι "three-cats", where κατσούλι is an old dialectal word; if so, it may well predate the current pattern for -ας. But this is obscure enough for me to ignore.
  
• For ντόμψωλας, the meaning of ντομ is likewise obscure, but the right-hand word is clearly ψωλή "dick". The other compounds ending in stressed -άς were derived from nouns ending in /-a/ not /-i/; maybe accent retention only happens if they share inflection class (/ma.ˈlja/ > /va.pso.ma.ˈljas/), and the switch in inflection class throws it back to being recessive. (The new ending doesn't look like the old ending, so we're back to "accent the word from scratch, as a new word".) I'd need another few examples to confirm, but we have a more interesting word to look at.

Our more interesting word is μαμκακανανύστας. The word is derived from the babytalk words μαμ, κακά, νάνι. The coiner decided to put an -as ending on it. He couldn't.

He couldn't, because vernacular phonology would not accept the hiatus between νάνι and -ας (*/mamkakaˈnani.as/). Normally, there would be no hiatus, because the final /-i/ would be stripped off from the word as an inflection: for example, αρχίδ-ι > γλειψ-αρχίδ-ας. But the /-i/ is not an inflection; νάνι is babytalk, and is therefore treated as indeclinable and undecomposable.

The approach the coiner *should* have taken was instead to elide the vowel of /-as/, and produce *μαμκακανάνης /mamkakanani-s/, just like έλα μωρέ /elamoˈre/ > ελαμωρές /elamoˈre-s/. But the coiner did not do that.

Instead, the coiner switched to another masculine suffix that happens to end in /-as/: /-ˈistas/. This corresponds to English -ist. In fact, there are two -ist suffixes in Greek: the indigenous -ιστής /-iˈstis/, inherited from Ancient Greek, and -ίστας /-ˈistas/, which came back into Modern Greek via Latin and then Italian -ista. There's a clear register difference between the two, and the Italianised form is what you'd use with foreign stems (cf. ποδοσφαιρ-ιστής /poðosferiˈstis/ "football player", straight from Puristic, with μπασκετμπολίστας /basketboˈlistas/ "basketball player".)

Our coiner has decided to tack /-ˈistas/ onto /mam kaka nani/; the deletion of the second /i/ in a row (*mamkakananiˈistas) was easier to defend, and is established behaviour for the suffix. (γκαλερί > γκαλερίστας "gallery owner", χόμπι > χομπίστας "hobbyist", not *χομπιίστας.)

The suffix had an added advantage, which the coiner took advantage of—and may have intended all along. /mamkakaˈnistas/ ends in /-nista/, which is the word for sleepiness. (We already saw the related νυσταλέουρας.) "Sleepiness" is spelled with an upsilon, and that's how the coiner has spelled /mamkakaˈnistas/: μαμκακανύστας.

The avoidance of hiatus is a driving force for compounding in Greek in general: compounds have to avoid a vowel next to another vowel. The need is less frantic if it's /i/ before any other vowel; but the undecomposability of /nani/ meant that the colloquial pronunciation *[mamkakaˈnanjas] of */mamkakaˈnani.as/ would have been unacceptable: /nani/ would have become unrecognisable.

Switching to /istas/ is not the most extreme way to make sure there's a buffer consonant or two between the stem and the inflection. I dangled παπα-δ-οπαίδι and μακλαουντόσο-γ-ο before you in a previous post. In παπαδ-οπαίδι "priest boy = altarboy", "priest" is παπά-ς, παπά; the extra -δ- comes from the plural παπά-δ-ες.

For compounds of σόι /soj/ "extended family", the extra -γ- of μακλαουντόσο-γ-ο "Macleod clan" takes a little more explaining. The Turkish soy was easily analysed as an -ι neuter, with a -ιού genitive and a -ια plural. That means genitive [soˈju], plural [ˈsoja]; but [j] between vowels in Greek is normally the result of underlying /ɣi/, which means that the genitive and plural were analysed as /soɣiˈu/, /soˈɣia/. That would mean that the singular nominative would underlyingly have to be /ˈsoɣi/. It isn't actually pronounced as /ˈsoɣi/ [ˈsoʝi]; but when a compound like */maklauˈdoso.o/ called for an extra stem consonant, the /ɣ/ was available to be stuck on.

It's rare that a stem-final consonant has had to be invented from scratch; but that has been argued to have happened with κυριλέ "snobbish". The word is formed with the faux-French suffix -é; to talk about something being fancy, it made sense to add it to the word for "gentleman".

• That would have resulted in the hiatus */kiri.ˈe/, which would violate the rules of vernacular phonology; that would be fine for Puristic (/ˈki.rios/ itself has hiatus), but people were nervous about coining a new slang word with a Puristic hiatus.
• Reducing */kiri.ˈe/ to the vernacular */kirˈje/ would have been even less of a sociolinguistic plausibility.
• Everything would be so much simpler if /kiri-os/ had a consonant at the end of its stem; and a consonant was grudgingly invented: */kiri.ˈe/ > */kiri.-l-ˈe/.

Maybe because κυριλέ would sound like Κύριλος "Cyril", which happens to be etymologically related to κύριος (and thereby shares an upsilon with it). If the coiners of κυριλέ had not noticed the parallel with Cyril, the slang.gr folk certainly did; and inveitably, slang.grist "Jesus" paired κυριλέ /kiriˈle/ with μεθοδέ /meθoˈðe/, just as St Cyril was paired with St Methodius. (Making this, as slang.gr's in-jokey definition puts it, a "Jesuitism", and pas du tout slangue.)

That's /meθoˈðe/, not /meθoðiˈe/. It's an analogy that coined /meθoˈðe/ to match /kiriˈle/; and having the same number of syllables is pretty important in analogies. But I have ranged far from accentuation of compounds by now.

The story is not really as complicated as it may have sounded in leisurely exposition. If you're switching inflection to /-as/ in a compound ending in a noun, you keep the accent of the right-hand word—with ´– words considered recessive, rather than penult. But the accent is preserved if the compound does not end in a noun: if the compound ends in an imperative-turned-noun, or it involves a nominalised phrase. And if /-as/ has a distinct function as derivational morphology, with an associated intrinsic accent, then that accent takes over.

Even more briefly: there's an accent-preserving default for /-as/ compounds—with recessive taking priority over penult accent in the ambiguous ´– case. But that default only applies to default-type compounds, ending in a noun. Any disturbance to that pattern, and the compound becomes strictly accent-preserving. Any derivational meaning in the suffix takes priority in accentuation—as it always does; again, the compound is no longer the default ending in a noun.

This all means I have something extra to look for with my survey of slang.gr compounds that *don't* switch accent—of which there are well over 1300. That will take a little while to go through: I'm behind in other commitments. I'm hoping for surprises, though.