Can the stress pattern of "uroboros/ouroboros" be explained by any principle, or is it random? [duplicate]

I have been studying the pronunciation of Greek-derived words in English, and I've found an odd anomaly. There are (at least) two possible pronunciation patterns for plural word-forms that end in -⁠mata (which correspond to singular word-forms that end in -⁠ma).

Antepenult stress (expected)

It appears that it is always possible to place the stress on the antepenult (third-to-last) syllable, such as:

• lémma, lémmata /ˈlɛmə/, /ˈlɛmətə/
• schéma, schémata /ˈskiːmə/, /ˈskiːmətə/
• teratóma, teratómata /ˌtɛrəˈtoʊmə/, /ˌtɛrəˈtoʊmətə/~/ˌtɛrəˈtɒmətə/
• stígma, stígmata /ˈstɪɡmə/, /ˈstɪɡmətə/

This is what would be expected from the stress rules of Latin (I believe most or all of these words came through Latin before entering English), since the penultimate "a" in the original Greek words is short.

Maybe Latin vowel quality and stress rules are not very relevant; certainly, few English speakers know about them nowadays. And even with antepenult stress, some of these pronunciations do not seem to follow all of the traditional rules for pronouncing Latin-derived terms in English. I was reminded by some comments in this Languagehat blog post that in antepenult stressed syllables, we would expect tense ("long") vowels to be shortened due to trisyllabic laxing (TSL), yielding pronunciations like /ˌtɛrəˈtɒmətə/ (which I just noticed is in fact listed in the Oxford English dictionary) and ?/ˈskɛmətə/ (which is not). (Compare to genus~genera /dʒiːnəs/~/dʒɛnərə/, and perhaps opus~opera /oʊpəs/~/ɒpərə/). However, there are other Latinate plural forms that typically don't have trisyllabic laxing (matrices, also I think helices), or where both the lax- and tense-voweled pronunciations coexist (codices, apices), so I guess the apparent lack of it in schemata should not be surprising. (There are even some exceptions to TSL in singular nouns like obesity.)

Penult stress (unexpected)

The other pronunciation seems to exist only for a couple of words, in which the penultimate "a" is commonly stressed and broadened: stigma ~ stigmáta /stɪgˈmɑˑtɐ/ (listed by both the Merriam-Webster dictionary and the Oxford Learner's Dictionaries), and also sometimes schéma ~ schemáta /skiːˈmɑːtə/ (listed only by the Oxford Learner's Dictionaries). (Neither the Merriam-Webster dictionary nor the Oxford Learner's Dictionaries list *"lemmáta"). I'm wondering why this pronunciation exists. It's an irregular plural either way, and putting the stress on the penultimate syllable creates a stress change between the singular and the plural forms.

(It's true that all of these words are rare, and some people might pronounce some of their plurals differently. The pronunciations I give here are simply the ones listed in dictionaries.)

The plural form "stigmata" is probably most common in religious or theological contexts (where it refers to the wounds of Christ), so at first I wondered if the penultimate stress was supposed to be closer to the original Greek pronunciation, but according to Wiktionary, the stress is on the antepenult in Greek as well.

So, what's up with this? Did the stress perhaps change in Latin over time, is this a tradition of Greek accentuation that I'm not familiar with, is this a simple matter of ignorance of the original position of the stress that has now become entrenched with time and usage, or is this somehow explainable by established English stress patterns? (For example, do English speakers have some natural preference for a penultimate stress on polysyllabic words?)

If nobody knows of a reason, I suppose I can just put it down to people encountering the word mainly in text, not knowing which syllable to stress, and choosing the penult because it sounds better to them.

(I asked this question a while back on Wordreference, but nobody knew of any reason.)

Generally in English the addition of an inflectional ending does not affect stress. That is, the ending itself is not stressed, and stress remains on the same syllable as it is in the uninflected form. For instance, even though in morphologically simple forms, we don't get 3 unstressed syllables at the end of a word in English, the addition of the -ing inflection of the progressive aspect does not affect the stress, even when it is added to a verb ending with 2 unstressed syllables: interest/interesting, amnesty/amnestying.

Likewise, the regular English plural ending is inflectional, and its addition never causes stress to shift to the right, nearer the end of the word: parentage/parentages, Cavendish/Cavendishes.

Therefore, if the -ta is a plural ending and is counted as an inflectional ending, like the regular plural ending, one would predict that stress would be unaffected by adding it, and you would get stigmata ending in 2 unstressed syllables. That is a big "if", though, since I don't think stigmata is actually the plural of stigma in current English, and I never heard it pronounced on anything but the penult. So I think the stress you are asking about is archaic in English.

This is a beautiful question that should have garnered a few hundred upvotes by now.

If I may venture a theory (since no official theory seems to exist at this point).

"For example, do English speakers have some natural preference for a penultimate stress on polysyllabic words?"

You may be onto something here. Speakers of languages that lend themselves to versification ... let me rephrase this: ... that are conducive to composing rhyming poetry ... poetry that rhymes ... have an affinity for ... well, rhyming.

The same people's fondness of meter is less pronounced, but isn't far behind. We're all poets in our hearts.

Today, when professional poetry is hibernation, everyday, mundane, if you will, rhyming is alive and well in everyday, mundane speech. Itsy-Bitsy spider and all.

If your boss' name is Matt, he'll never be just Matt: he's Matt the Rat. Coppers always gorge on Whoppers. A particularly intelligent woman is Wits with Tits. The CEO of a think tank is Main Brain. And so forth.

Now STIG-ma-ta, with the stress on the first syllable, rhymes with absolutely nothing at all, whilst stig-MA-ta rhymes with sonata, cantata, regatta, persona non grata, and terra cotta.

In conclusion I'd like to point out that in this our epoch of habitual unmitigated hypocrisy and crass doubletalk, the truth oftentimes strikes people as comical.

A few years later, it seems to me that the most likely influence on the English pronunciation of stigmata was the group of polysyllabic words ending in the letter A that come from languages like Italian or Spanish, where penult stress is common (and familiar to many English speakers). This fits in with Ricky's observation that there are a number of words that rhyme with stigmata when it is pronounced with /ˈɑːtə/.

Greg Brooks' Dictionary of the British English Spelling System discusses this class of words in §10.3, and gives a list that includes stigmata, but seems to be composed mostly of words from Italian or Spanish:

<a> [is] pronounced /ɑː/ in RP [...]

• often when <a> is the penultimate vowel letter and there is at least one earlier vowel letter in the word separated from it by at least one consonant letter and the relevant <a> is followed by a single consonant letter followed by word-final <a, i, o>: armada, avocado, balaclava, banana, bastinado, cantata, cascara, cassava, cicada, cinerama, cyclorama, desiderata, desperado, farrago, Gestapo, gymkhana, iguana, incommunicado, karate, legato, liana, literati, marijuana, mascara, meccano, pajama, palaver, panorama, pastrami, pyjama, safari, salami, schemata, sonata, soprano, staccato, stigmata, sultana, svengali, tiara, toccata, tomato, tsunami, virago. [...] Exceptions: alpaca, piano with /æ/; dado, data, halo, lumbago, potato, sago, tornado, volcano with /eɪ/

It may also be relevant that due to vowel reductions that occurred in the history of Latin, native Latin words of more than two syllables usually don't have short <a> in a penult open syllable. There are a number of polysyllabic English words from Latin that end in the letter A and that have antepenultimate stress (e.g. nebula, replica, retina, viscera) but Latin words ending in the spelling pattern -aCa specifically tend to have penult stress; e.g. errata or arcana. In fact, -ata (with a long, stressed A) is a common ending for past participles in Latin, and a number of English words that are spelled with -ata and pronounced with /ˈɑːtə/ originate from this Latin ending; e.g. sonata (which the OED says is from " Italian sonata [...] use as noun of feminine past participle of sonare") and cantata.