κατσούλι "kitten": where did the cutesy /ts/ come from?

Tom Recht had a simple question in comments the other day, which admits of an almost simple answer. There is a catch, in that there is no clear phonological reason for what has happened, and I offer an unconvincing guess at it.

Tom Recht's question:

I'm curious about the word κατσούλι, which is intriguingly similar to the Hebrew for 'cat', [xatul]. Of course they're both presumably related to cat, Byzantine κάττα/κάττος, Latin catta/cattus, etc. (ultimate origin unknown), but none of those has the -ul- part that you see in κατσούλι and xatul. Any idea where the Greek word comes from?

Anonymous delivered, in his response:

The "oul" in greek is part of the suffix -ouli.

The part that needs explanation, I think, is rather how kati gave katsi. Did a tsitakistic(sic) dialect give it to the standard?

This all makes sense, but let me unpack it a bit more slowly.

Ancient Greece was not familiar with cats; Greeks kept weasels (γαλῆ) as pets, and you'll occasionally see pedants calling cats γαλῆ, but the real exposure to cats was in the Roman era. There was a Classical word for cat, αἴλουρος, which Puristic has preserved: felines are αιλουροειδή, "aeluroids". But that's the name for an exotic animal from Egypt; once people routinely interacted with cats, they used the Roman name for cats, cattus.

(I could be completely wrong about this, but I'll keep going.)

In LSJ, cattus is reported as κάττα and κάττος, both from Scholia. This is a fairly common occurrence in LSJ: the scholia on Ancient text explain the old words using contemporary, mediaeval words. LSJ is strip-mining the scholia for snippets of antiquity, and adds these glosses to its coverage artificially, although properly they're out of its scope.

In Standard Modern Greek, the word was reborrowed from Italian gatto, as γάτα. In Greek, unlike Italian, cats are by default feminine, and the masculine γάτος is explicitly a tomcat. That's Standard Greek; but Greek dialect preserved the earlier, Latin form: the Cretan for "cat" is κάτης. Hence the Renaissance poem Ο Kάτης και ο Μποντικός, "The cat and the mouse", which has just come out in a new edition. (You can get an earlier edition online via Tassos Kaplanis' Cretan Lit class wiki.)

Now, once Greek has borrowed a foreign stem, it can play around with its inflections and derivational morphology. From γάτα "cat" (Standard Greek), you get:

• γάτος "male cat"
  
• γατί "neuter cat; (implicit) diminutive of cat"
  
• γατάκι "neuter diminutive of cat, kitten"
  
• γατούλα "different, feminine diminutive of cat, she-kitten"
  
• γατίλα "cat smell"

A "kitten" in the sex-kitten sense is thus going to be γατούλα, as memorably and annoying chanteused about by the Greek, infantilised equivalent of Brigitte Bardot (God help me), Aliki Vouyouklaki (0:29):

Νιάου νιάου βρε γατούλα / με τη ροζ μυτούλα / γατούλα μου μικρή—τσα τσα τσα
Νιάου. Σ' έχουνε μη στάξει / κι είναι από μετάξι / η γούνα σου η γκρι
"Meow meow, kitten, with your little pink nose, my little kitten. Cha cha cha.
Meow. They treat you with kid gloves, and your grey coat is made of silk."

Um, yeah. A younger, more innocent Greece. Glad that's over.

Modern Greek doesn't happen to have the normal neuter diminutive γατ-ούλι, just its feminine counterpart γατ-ούλα. As it turns out Cretan does has the equivalent neuter equivalent, and has had it at least since the Renaissance. The hyperlinks are the online abridged Kriaras dictionary; the citations are from the dead tree full version:

• κάτα "cat" (fem): Mass of the Beardless Man, Meursius' Dictionary
  
• κάτης "cat" (masc): Entertaining Tale of Quadrupeds (mainland, ca. 1364); Oracles of Leo the Wise; Teseide; Cat and Mouse
  
• κατσί "neuter cat; (implicit) diminutive of cat": To an Old Man Not To Marry, Chronicle of Morea
  
• κατσούλι "neuter diminutive of cat, kitten": Marco Antonio Foscolo: Fortunato (Crete, 1655)
  
• κατσίπαρδον "cat–pard, cheetah" (Italian gattopardo): Oracles of Leo the Wise
  
• κατσουλόπαρδος "cat–pard, cheetah": Erotokritos

So κατσούλι is derived from κατσί, and κατσί in turn is derived from κάτης.

The problem here is that the neuter κατσί has changed /kat-is/ to /kats-i/, for no obvious linguistic reason. Anonymous, in delivering, wondered whether this was Tsitacism (τσιτακισμός). Tsitacism is the onomatopoeic word for the process in a large number of Greek dialects, of affricating what was the palatal stop [c], to [tʃ] or [ts]. This affrication is pretty common across languages, since [c] is a very unstable sound to pronounce: lots of contact area between the roof of the tongue and the palate, easy to let some air through when trying to pronounce a plosive.

We see that kind of affrication all the time with renderings of Latin /ki, ke/ and /ti/, which must at once time have ended up pronounced as [ce, ci]:

• citatio [kitatio] > [citacjo] > Italian [tʃitatsione], German [tsitat], French [sitasjɔ̃], Spanish [θitaθjon], English [saɪtɛɪʃən]

The way to account for all these vaguely palatal modern pronunciations of what used to be /k/ and /t/ is that they were palatal stops as [c] (because of the following front vowel), and the two [c]'s then broke down, at different times in different languages, into various permutations of [tʃ, ts, s, ʃ, θ].

And Tsitacism would explain the τσ in κατσούλι as being from κατσί, because κατσί has a front vowel and κατσούλι doesn't. (Kriaras' etymology accordingly reads: "From the noun κατσί and the ending -ούλι, or less likely from Latin catulus".)

But Tsitacism doesn't explain κατσί, because in almost all dialects of Greece, the only phoneme to undergo that affrication is /k/: Κυριακή "Sunday" /kirjaki/ [cirjaci] ends up in dialect as [tʃirjatʃi] or [tsirjatsi], but Τρίτη [triti] does not end up as *[tritʃi] or *[tritsi].

There are three exceptions where /t/ does palatalise, and they don't account for κατσί in Crete. The two Hellenic exceptions are Lesbos and Tsakonia. Tsakonian is spectacular with its palatalisations: it palatalises not only /k/ and /t/ before front vowels, but also /p/. In fact, historical /k/ ends up pronounced further front than /p, t/: πίνω /pino/ > κίνου [cinu] "to drink", τιμώ [timo] > κιμού [cimu] "to honour", κήπος [cipos] > τχήπο [tɕipo] "garden". (Or, using a Tsakonian transcription other than the one I've invented, τζήπο [tsʰipo].)

The non-Hellenic exception is Aromanian, which is still a language of Greece: it likewise palatalises /t/ to [c]. The Aromanian specialist Nikos Katsanis has in fact claimed Tsakonian and Aromanian palatalise /t/ for the same reason.

• Katsanis, N. [Κατσάνης, Ν.] 1989. Κουτσοβλάχικα και Τσακώνικα (Arumanian and Tsakonian). Ελληνική Διαλεκτολογία 1: 41-54

No, not because there is a Romance substrate to Tsakonian; but because Tsakonian and Aromanian are both far enough removed from written Greek, that they would not have been subject to its conservative influence, making them pronounce <τ> as written.

But κατσί is unlikely to have wandered to the Morea from Lesbos, Tsakonian, or the mountain pastures of Thessaly. So affrication can't be the explanation.

Kriaras' dictionary has these references for κατσί:

• Triantaphyllides, Manolis. Collected Works I 358.
  
• Pernot, Humbert. Études linguistiques III 423.

Both are foundational works. Pernot's work is supposed to be his grammar of the dialect of Chios; it ended up being his historical grammar of Modern Greek.

• Pernot, H. 1907. Études de Linguistique Néo-Hellénique I: Phonétique des parlers de Chio. Fontenay-sous-Bois.
  
• Pernot, H. 1946. Études de Linguistique Néo-Hellénique II: Morphologie des parlers de Chio. Paris: Les Belles Lettres.
  
• Pernot, H. 1946. Études de Linguistique Néo-Hellénique III: Textes et Lexicologie des parlers de Chio. Paris: Les Belles Lettres.

Triantaphyllides vol. I, pp. 305–490 is his PhD thesis Die Lehnwörter der mittelgriechischen Vulgärliteratur, 1909, Trübner: Strassburg.

The Collected Works of Triantaphyllides are easy to find in University libraries—but I'm at home, and have not photocopied that text; Google Books is giving me one snippet for cattus, but not the right one. Pernot's Études are less easy to find in my garage, but I did find them; he does not however explain the [ts], and I suspect Triantaphyllides didn't either.

My suspicion is that the expected diminutive *κατίν—which we do see in Standard γατί—was modified to κατσί(ν) under the influence of the diminutive -ίτσιν, which was particularly widespread in Early Modern Greek. The accent is wrong for an analogy (*kaˈtin "cat" ~ piðimat-ˈitsin "little leap" < kaˈtsin). A haplology from *kat-ˈitsin "little cat" to kaˈtsin also is awkward, because the syllable being eliminated isn't precisely repeated. But that's my guess.

The Anastasiadis–Symeonidis reverse dictionary gives 10 words in Standard Modern Greek ending in stressed /ˈtsi/; there is a parallel to κατσί in βουτσί "barrel" < βουτίον (attested in the Hippiatrica) < Hellenistic βοῦτις "vessel in the shape of the frustum of a cone" (Hero of Alexandria) < Late Latin buttis. There's less of a cause to pronounce "barrel" in a cutesy way than there is for "kitten"; so my guess is probably wrong; but there's a large number of Ancient diminutives in -τίον that have stayed as -τί: αφτί, γατί, κουτί, πορτί, σκουτί, χαρτί. Whatever happened with βουτίον > βουτσί did not happen with κυτίον > κουτί "box".

So this is a dread irregular phonetic change, and linguists appeal to analogy when they don't have a better answer invoking phonological rules.

If someone has already solved this, I'll be happy to hear it.