My new comprehensible input project for Latin: Latine

Hi all, just for anyone interested, I’ve started a side project to produce comprehensible input for Latin: Latine. It will take a long time but I’ve got the bare bones of the structure in place. It will be free (of course) and I’ve tried to switch off ads as far as possible by selecting ‘made for kids’.

I’ve tested it out: by the fourth video (only about 3 or 4 minutes each), kids were able to understand the use of an indirect question in the subjunctive without any explanation. (I’m focusing the first lessons on the copula, including its usage in indirect constructions. Next will be acc. + inf. construction with esse).

Initial videos here:

https://www.youtube.com/watch?v=Kg9IlGKeFOc&list=PLUgmCopQRnVtUojnC5Ru-A33MXoZywc6n

My explanatory document here:

https://tinyurl.com/Kalliagros23

Cheers, Chad

Hi,

It’s a great deed you have undertaken. As you point, there isn’t that much comprehensible input out there.

What worries me, though, is the question of the quantity of vowels.

On the other hand, if I understand the literature on comprehensible input correctly, it could be problematic providing input in a (macronised) form, which does not correspond to the (unmacronised) form in (modern editions of) authentic texts.

I don’t agree with that. Reading without macrons should not be an issue if the words are mastered enough. Otherwise, when will a student learn quantities if not at the beginning?

And that brings me to the audio. The generated speech is a big problem: the quantities aren’t observed. For example, we have, instead of ‘Rōma (e)st in Italiā’, ‘Rōma est in Itālja’. Same with ‘Macedoniā’ (5 syllabes ˘˘˘˘¯) which becomes ‘Macedōnja’ (4 syllabes ˘˘¯˘). It’s also clearly audible that the people who provided their voice to the engine are Italian, as we can hear a schwa after final ‘s’, for example (like “It’s-a me, Mario!”)

I suggest you to reconsider your way of dealing with quantities because you may miss your point.

I hope I’m not too dismissive given the amount of work it represents.

Many thanks, both good points! And quite easily fixed.

I’ve already put macrons in the notes, and on reflection I agree with you; I’ll copy these over to the videos too.

Also agreed on the audio - I’ve now figured out a way to get lengths into the generated voices (it took some experimenting).

I could record it myself (I use lengths) but then it would not be scalable, and I want to get to 1m to 5m words and so want to be able to use generated voices if possible (or maybe a mix of that with my voice), based on the studies as to how much comprehensible input is required to get to an intermediate level.

Now that I’ve worked this out I’ll go back and re-issue the first videos. It always takes a bit of experimenting at the beginning!

You’ll see also my content is quite different to other videos out there. Often they present beginners with groups of nouns (English and modern languages rely more on nouns), whereas I am focusing on the copula and pronouns at the beginning (the most frequent words - no nouns figure in the most frequent words in Latin!) and have already introduced subjunctive and infinitival constructions, which kids were able to understand just by watching without explanation. It’s just an experiment, will see where it gets to!

Thanks again! Cheers, Chad

I’m glad that you found my remarks useful.

If you can produce consistent content, it may be useful for students.

By the way, if you found a working method for speech generation in Latin, I’m interested. For now I found the Microsoft version unconvincing (the Italian accent problem), and the Linux one seems faithful but the rendering itself is terrible.

Hi, after some experimentation I’ve realised the best approach is to remove voice altogether (it isn’t scalable for me to make recordings). I will narrate the voice parts live for those with whom I watch the videos, and others can do the same if they like.

I’ve added macrons, removed voice and added music (for those who want it; those who don’t can mute). Here is v2 of the first video:

https://www.youtube.com/watch?v=VWsID7aqAYY

I’ve also added pronunciation notes in the info doc:

https://tinyurl.com/Kalliagros23

Here are the notes for lesson 1:

Rōma est in Italiā.

Scholars debate how to pronounce the long vowel ō in the first syllable of Rōma. (The same debate rages relating to the pronunciation of ē, ī, and ū.) Two positions currently proposed are as follows (there are many scholars other than Calabrese and Allen who support one or other of these approaches; I name them for convenience only):

The ‘Calabrese’ approach: Pronounce ō with the SAME QUALITY as the corresponding short vowel: see sec. 1 (The Latin vowel system and its development) in Andrea Calabrese (2005), On the Feature [ATR] and the Evolution of the Short High Vowels of Latin into Romance, in A View from Arjona. University of Connecticut Working Papers in Linguistics, Vol. 13, pp. 33-78).

The ‘Allen’ approach: Pronounce ō with a CLOSER QUALITY than the corresponding short vowel, e.g. long ‘ē’ as in French ‘gai’, short ‘e’ as in Eng. ‘pet’; long ‘ī’ as in Eng. ‘feet’, short ‘i’ as in Eng. ‘pit’; long ‘ō’ as in French ‘beau’, short ‘o’ as in Eng. ‘pot’; long ‘ū’ as in Eng. ‘fool’, short ‘u’ as in Eng. ‘put’: see Allen (1989), Vox Latina, pp. 47-50.

The actual position is likely to be more complex than either of these approaches taken alone, and will depend on the particular vowel, whether it is accented, whether it falls at word-end, and the specific time period used as the benchmark. See e.g. Adams 2013, Social variation and the Latin language, p. 69.

Rōma est in Italiā.

In connected speech, the e of est would be dropped, and the remaining -st would be joined onto the preceding word (‘prodelision’), i.e. [Rōmast]: Allen 1989 p. 123. However, I separate the words in careful speech, to ensure that the input is comprehensible.

Neāpolis est in Italiā.

Take care to pronounce the s at word-end as [s], and not as voiced [z], unlike word-final ‘s’ in English (e.g. ‘roses’ [rouziz]): Allen 1989 p. 35.

Tarentum est in Italiā.

The final syllable of Tarentum would be pronounced without the m, and with the vowel before it (u) becoming nasalised and lengthened: Allen 1989 pp. 30-31.

Tarentum est in Italiā.

In connected speech, the same phenomenon of prodelision described above for Rōma est would also apply here: the e of est would be dropped, and the remaining -st would be joined onto the preceding word, i.e. [Tarentumst]: Allen 1989 p. 123.

Londīnium est in Britanniā.

The letter n of the preposition in before labial ‘b’ is assimilated to [m], and so in would be pronounced here as [im] rather than [in], similar to e.g. in balneum [im balneum]: Allen 1989 p. 28.

Londīnium est in Britanniā.

Take care to pronounce both consonants, just as in English we pronounce both ‘t’ consonants in ‘rat-tail’ but we do not do so in ‘bitter’: Allen 1989 p. 11.

Mīlētus est in Asiā.

Take care to pronounce the word-internal s between vowels here as [s], and not as voiced [z] as in English (e.g. causae like ‘cow sigh’, not ‘cow’s eye’): Allen 1989 p. 35.

Pella est in Macedoniā.

The letter n of the preposition in before labial ‘m’ is assimilated to [m], and so in would be pronounced here as [im] rather than [in], similar to e.g. in muro [im muro]: Allen 1989 p. 28.

Italia est in Eurōpā.

Pronounce eu as a diphthong (combine a short ‘e’ with a ‘u’); do NOT pronounce as [yu] like in English ‘neuter’: Allen 1989 p. 63.


Incidentally, it’s probably a good thing that I don’t include voice in the end, as I use a pitch accent for Latin: in my view, the scholarship arguing for a stress accent is rather weak, and does not overturn the clear description of a pitch accent in the ancient sources: it is likely a hangover from modern classroom practices. Here are my notes in the same linked document on the point:


One area where I differ from leading modern sources is on whether to render the Latin accent via stress or pitch. The ancient sources make clear that the Latin accent was one of pitch (see e.g. Allen 1989 p. 83), and this is what I follow. ‘Roman grammarians, down to the 4th cent AD, describe L accent in terms appropriate only for a pitch accent’ (Sihler 1995 sec. 248). Bennett summarises the rules (sec. 55.6):

https://archive.org/details/latinlanguagehis00bennuoft/page/76/mode/1up?view=theater

The arguments for a stress-based approach are weak, e.g.:

The pitch accent is too similar to the Greek pitch accent: ‘The very similarity of the Latin statements to those which apply to Greek is therefore an embarrassment rather than a support to the idea of a pitch accent for Latin.’ (Allen 1989 p. 84) According to Allen, then, the evidence for a pitch accent is evidence against it. This argument has little weight. What if the Latin authorities were right? How else could they have described the pitch accent? I note also that Quintilian discusses the pitch system at 1.5.29–31. Does Allen really think that Quintilian applied the Greek system without thought to Latin, in a section of his work on speaking dedicated to accentuation?

It would mean that the Latin accent changed twice: ‘It seems unlikely that the prehistoric stress accent would have been replaced by a pitch accent and this quite soon again replaced by a stress accent.’ (Allen 1989 p. 84) Again, this argument has little weight. The individual changes of accent must be assessed, not their collective history. It would have been helpful if Allen considered, for instance, whether (1) the original stress accent of Latin changed to pitch along with the change of the placement of the accent from word-beginning to the ‘penultimate law’ (which is not disputed), and (2) the pitch accent changed over time to a stress accent as occurred in the Greek language (which is not disputed).

It does not take account of the vowel length in the accented syllable: Allen argues (1989 p. 85):

There is also a strong general reason for believing the Latin accent to be different in type from that of Greek. In Greek, as befits a pitch accent, its location and variety depend only upon those elements of the syllable which can carry variations of pitch … i.e., primarily upon the vowels and diphthongs. … In Latin, on the other hand, it is syllabic quantity alone that is relevant; it makes no difference whether the heaviness of the syllable results from a long vowel or diphthong, or from a consonantal closure. Thus > re-lic-tus > is accented in the same way as > re-lā-tus > (and differently from > re-li-go> ); the fact that the c, unlike the second part of the long ā, cannot carry variations of pitch is irrelevant. The contrast with the Greek system could hardly be greater, and speaks strongly in favour of a syllabic stress, rather than a vocalic pitch as in Greek.

But Allen is simply wrong here. According to the Latin rules of pitch described by the ancient sources (summarised in Bennett sec. 55.6, linked above), re-lic-tus is NOT accented in the same way as re-lā-tus: re-lā-tus would bear the circumflex on the penult (since it has a long vowel), whereas re-lic-tus would bear the acute on the penult (since it does not).

Ictus: Allen next argues (1989 p. 86):

A further significant contrast between Latin and Greek lies in the fact that in the last two feet of the Latin hexameter poets increasingly aim at agreement between the verse ictus and the linguistic accent, whereas there is no such correlation between the Greek ictus and accent. Which suggests that there is something in common between ictus and accent in Latin, but not in Greek; and the most probable common factor is stress.

Unfortunately, this is, again, extremely weak (if it is an argument at all). Allen does not explain why any relationship between ictus and accent suggests stress patterns which Latin hexameter poets preferred instead of melodic patterns which they preferred.

Iambic shortening: Allen next argues that the phenomenon of iambic shortening is ‘suggestive’ of a stress accent (1989 p. 86). However, subsequent work on the phenomenon itself questions the underlying assumption of this suggestion: for instance, Fattori argues that ‘the accent has no role in inducing IS [Iambic Shortening] regardless of its position’ (2021, p. 112).

I will leave this here, ending with a question: who bears the burden of proof? Should we start from an assumption of a stress-based Latin accent, and require ‘pitch-supporters’ to refute it? Or vice versa? I believe that we should start from what the ancient sources suggest (i.e. a pitch accent), and ask ‘stress-supporters’ to refute this assumption, if they can.


Cheers, Chad

I don’t have a definite opinion about the precise nature of the vowels, given the lack of consensus. I prefer the calabrese pronunciation just because of personal convenience.

Much better with the macrons. but I fear we lose something without the voice. BY the way, if there are no voices, I suggest you to put the stressed syllables in bold or else to ease learning.

Hi, it’s fine without voice — there are plenty of other videos out there of comprehensible input with voice. There just aren’t many videos of comprehensible input which focus on high-frequency words (including subjunctival and infinitival constructions using those words) which is what I want to focus on. These need to be dyed into the wool from the beginning to make reading golden age Latin easier.

It’s a shame that there are (as far as I’ve seen) no videos out there yet which render the Latin pitch accent described by the ancient sources. Stefan Hagel and others have done this for Greek, but Latin is still a ways behind. It’s interesting that, not too long ago, scholars speculated over what the Greek accents might mean — see the first preface to Chandler’s accentuation text here, with the sentence beginning ‘Kreuser’:

https://archive.org/details/accentuationgree00chanuoft/page/vi/mode/1up?view=theater

Nowadays, we follow the ancient Greek authorities’ description of a pitch accent. Perhaps the day will come when we follow the Latin authorities’ description of a pitch accent!

There’s quite a lot of music that opens up in Vergil and other authors when one renders e.g. the acute vs circumflex accents in heavy syllables (depending on the length of the vowel/diphthong in the heavy syllable). I haven’t heard any renditions of the Aeneid that access any of this yet.

I won’t, therefore, mark accented syllables with bold (which would obscure the different types of pitch accents), and I don’t think there’s a burning need to mark them out — unlike Greek, the Latin accentual system is very easy to learn as it is so regular, and exceptions to e.g. the penultimate law (in final-accented words) can be covered in the notes.

Cheers, Chad