Were initial voiced stops voiceless in early Latin?

Question

The pronunciation of the stops b, d, g / p, t, k is—it seems—unstable. In German, the voiced stops are unvoiced at the end of words, such as in lieb mir, das Lied, Guten Tag (/liːp mi:r, das liːt, guːt(ə)n taːk/); in Korean, they are unvoiced word-initially, such as in 보 (bo: /po/), 단군 (Dan-gun: /tangun/), 관 (gwan: /kwan/); in Norwegian, Swedish, Danish and English, a -p- is voiced e.g. post initial s before a vowel, such as in spade (no: /²sbaːde/; en: /sbeɪd/).¹ Without going to deeply into linguistics territory: It is a well-established phenomenon.

Latin, it would seem, has the same phenomenon represented from older times, with the etruscan C serving both as /k/ and /g/; hence ‘C.’ is short for Gaius. Gian Biagio Conte writes that

What is left shows the existence of a melting pot of peoples and languages. Only gradually does the use of Latin and the Latin alphabet assert itself. The Latin alphabet itself, in fact, gives clear testimony of the situation in early Rome. In substance, it is derived from a particular West Greek alphabet, the one used in the powerful Campanian city of Cumae, but it is also somewhat influenced by Etruria. (This explains, for example, how the letter C serves as the abbreviation for the name Gaius: Etruscan had a single sign for the two velar consonants, voiceless and voiced.

I have read this and similar explanations many places, but it seems no-one entertains the notion that the early speakers might have had a devoicing of initial /g/ to /k/. My question, thus, is did the early Latin speakers devoice initial stops?

As a final note, it is probably worthy of note to mention the perhaps unstable spelling of words with an initial /k/: If borrowed from Greek and followed by α, it might get the spelling ka, such as in kalendae (e.g. ‘Martis Romani festae uenere kalendae | —exoriens nostris hic fuit annus auis—’ Tib.3.1); if followed by consonants or the vowels a, e, i, it is always c; but if followed by v—pronounced /uː/, /ʊ/ or /w/—it is spelt either q or c. This article by Sturtevant provides a useful comment on this.

Notes

¹ Though this might be contended. Is it merely an unaspiration?

Answer 1

To elaborate on jknappen's excellent answer:

The Etruscan language, originally spoken across Tuscany, had no voicing distinctions. They had an aspiration distinction, with /tʰ/ and /t/ contrasting, but no contrastive /d/. (The phonetics are unknown, but it may have actually been quite close to American English /t/ [tʰ] and /d/ [t]; all we really have to go from is their spelling.)

When the Etruscans borrowed an early form of the Greek alphabet, then, they found a whole lot of letters they didn't need. Theta and tau were used for /tʰ/ and /t/, for example, but delta was unnecessary. So somewhere around the sixth century BCE, they started purging the unneeded letters: delta, beta, omicron, and xi were unceremoniously disposed of (as Etruscan had no /d/, /b/, or /o/, and used chi to write /ks/).

But by that time, a strange convention had arisen around /k/. The early Greek alphabet had two separate letters for /k/, named kappa and qoppa; the latter was originally something like /q/ or /kʼ/ in Phoenician, but no dialect of Greek had either of these sounds, so it became just "kappa 2, in case we get tired of the first one".

Eventually the Greeks got rid of qoppa, but the Etruscans didn't. Since they didn't have a /g/, gamma became another redundant version of /k/. And since there were already "spare" letters for /k/, why not keep one more? Eventually, the convention became: use kappa before /a/, qoppa before /u/, and gamma before /i/ and /e/. Presumably there were some phonetic differences here, with qoppa being a bit farther back and gamma a bit farther forward, but there isn't enough evidence to say for certain.

When the Latins then borrowed the alphabet from the Etruscans, and applied it to their own language, they ran into a problem: they did have voicing distinctions, and needed some way to represent /b d g/. For /b/ and /d/ they re-borrowed beta and delta from Greek—but gamma was already being used for something different. So a new convention arose:

• Qoppa was used for /k/ before /u/ and /w/; /kʷ/ is written as if it were /kw/
• Kappa was used for /k/ before /a/
• Gamma was used for /k/ anywhere else, or for /g/ anywhere

In time, this got simplified somewhat, to better fit Latin phonology:

• QU was used for /kʷ/
• K was pretty much never used
• C was used for /k/ or /g/

And eventually, someone had the bright idea to stick an extra line on the side of C (possibly inspired by Q), to clarify its voicing; at the same time, they threw away the letter Z, since /z/ had merged with /r/, so they had an empty spot in the alphabet to put their new letter into:

• QU was used for /kʷ/ (*)
• K was pretty much never used
• C was used for /k/
• G was used for /g/

And this is the version of the alphabet used up through the Classical period. The only relics of the earlier version are a few words with kappa-alpha that stuck (like Kalendae), and the abbreviations C and Cn for Gaius and Gnaeus.

But there's solid evidence that the voicing itself never went away. As soon as the letter G was invented to write the voiced /g/, we see it used in some words but not others, in a consistent way; this usage lines up pretty much exactly with what we'd expect from other Indo-European languages (**). If the distinction between /g/ and /k/ were actually lost at the beginnings of words, we wouldn't expect to see it come back unaltered like this.

---

(*) In a few places, like the word equus, we also see QU used for plain /k/. But that's just force of analogy. In general, QU can be taken consistently as /kʷ/, and CU as /ku/.

(**) For example, initial /g/ in Latin consistently corresponds to initial /k/ in Germanic, while initial /k/ in Latin consistently corresponds to initial /h/ in Germanic: gelū ~ cold, but canis ~ hound. If Latin had actually lost the distinction between word-initial /g/ and /k/, we'd expect this pattern to be totally lost, since the last contact between Italic and Germanic was many centuries earlier. For the pattern to have survived through the G-less years, the voicing distinction must also have survived.

Answer 2

Some of the phenomena are in fact an interference of Etruscan: As far as we know, the Etruscan language made no phonemic distinction between voiced and voiceless stops. Because the Romans adapted the Latin alphabet from Etruscan, some pecularities of the Etruscan alphabet carried over to early Latin writing: Three different letters for the sound /k/: K before A, Q before U, C otherwise, no letter at all for /g/—it was introduced later (ca. 230 BCE) and placed in the alphabet at the place of the than abolished letter Z that was added again later to write loan words from Greek.

These pecularities have nothing to do with the pronunciation of Latin, it always maintained a contrast between voiced and voiceless stops, as it can be seen by comparison with Proto-Indogermanic and other Indogermanic languages.