Akana and the comparative method

[bradrn]

Mention of a conlang conference led me to this interesting paper by one of its organising committee members, in which language construction is used to test assumptions about language. And that in turn reminded me of an idea which I’ve had floating around in my brain for a while, but never actually got around to investigating.

The idea is this: we now have ~15 years of reconstruction relays from the Akana project. Most of these (if not all) consist of a protolanguage, its descendants, and an independently reconstructed duplicate of the protolang. That means this is one of the very rare instances where we can directly compare the output of the comparative method to the original protolang. Furthermore, the reconstruction is happening under ‘perfect’ conditions, where we have complete data for all the languages. And we’ve done this for several different unrelated protolangs.

So: might this give us a unique opportunity to directly validate the comparative method? It would at least be interesting to see how similar the reconstruction is to the original protolang. If there’s any systematicities in what gets reconstructed and what doesn’t, that would be well worth knowing too.

Even if no-one else is interested, I’d at least like to explore this idea a bit. The first step is, of course, simply getting the data — but that has its own challenges, now that the Akana website has disintegrated. Does anyone know where I might be able to find the previous protolangs and reconstructions?

[Moose-tache]

The problem is that the same logic is used for both directions: making and interpreting the daughter languages. Usually it's the exact same people doing both jobs at different times. It would be circular to use the data to validate the linguistic technique used to create the data.
That said, I'm sure I have a spreadsheet or two kicking around. I'll check...

[bradrn]

The problem is that the same logic is used for both directions: making and interpreting the daughter languages.

In some ways this is part of the appeal: this is a case where all the assumptions of the comparative method hold perfectly. So the question is — when all the assumptions hold, can everything be reconstructed correctly? Or, conversely, are some things simply unreconstructable?

Usually it's the exact same people doing both jobs at different times. It would be circular to use the data to validate the linguistic technique used to create the data.

I thought in a reconstruction relay, the whole point was for the other team to do the reconstruction independently? (At least, that was how it worked last year for the one I participated in.)

That said, I'm sure I have a spreadsheet or two kicking around. I'll check...

Thanks!

[Nortaneous]

The problem is that the same logic is used for both directions: making and interpreting the daughter languages.

In some ways this is part of the appeal: this is a case where all the assumptions of the comparative method hold perfectly. So the question is — when all the assumptions hold, can everything be reconstructed correctly? Or, conversely, are some things simply unreconstructable?

Don't teams generally try to ensure that everything is reconstructible? This is not a pressure that would exist in natural language.

[bradrn]

The problem is that the same logic is used for both directions: making and interpreting the daughter languages.

In some ways this is part of the appeal: this is a case where all the assumptions of the comparative method hold perfectly. So the question is — when all the assumptions hold, can everything be reconstructed correctly? Or, conversely, are some things simply unreconstructable?

Don't teams generally try to ensure that everything is reconstructible? This is not a pressure that would exist in natural language.

Interesting point. I don’t recall ever discussing this point explicitly when I participated in the relay last year, but it’s certainly possible we were thinking along these lines.

But, in any case, reconstruction is difficult and I imagine no-one has ever reconstructed the protolanguage entirely correctly. So it’s still an interesting test, I think.

[Zju]

Even if no-one else is interested, I’d at least like to explore this idea a bit. The first step is, of course, simply getting the data — but that has its own challenges, now that the Akana website has disintegrated. Does anyone know where I might be able to find the previous protolangs and reconstructions?

I dare to say that my reconstruction of 2014 relay's Proto-Ronquian is one of the best, if not the best of the Akanarian reconstructions. I still have the wordlist lying somewhere - maybe even with backups and bits of reconstructed grammar. Regarding the protolang itself though, I don't know where and if it could be obtained.

[zompist]

The idea is this: we now have ~15 years of reconstruction relays from the Akana project.

15 years?!

Does anyone know where I might be able to find the previous protolangs and reconstructions?

No idea, though my Faralo page and Proto-Ranic pages are still there. A lot of the people from that time are long gone so I don't know where you'd find the data. Learn to plumb the Wayback Machine, I guess...

It would be interesting to compare results and methods.... I hope you can track down the data.

[bradrn]

The idea is this: we now have ~15 years of reconstruction relays from the Akana project.

15 years?!

Of course I wouldn’t really know, but I got the impression it started around 2010.

Does anyone know where I might be able to find the previous protolangs and reconstructions?

No idea, though my Faralo page and Proto-Ranic pages are still there.

Thanks!

Question: what are the blue markings on the Faralo page?

EDIT: and I found the original Proto-Isles Lexicon. Plus its grammar, of course, but you didn’t attempt to reconstruct that. Is your page all that existed of the reconstruction, or did other people work on it too?

[Zju]

Woot! As luck would have it, the wayback machine preserves both of these:
Proto-Ronquian reconstruction: https://web.archive.org/web/20180305084 ... o-Ronquian
Proto-Ronquian reconstructed lexicon: https://web.archive.org/web/20211020032 ... an_Lexicon

[bradrn]

Woot! As luck would have it, the wayback machine preserves both of these:
Proto-Ronquian reconstruction: https://web.archive.org/web/20180305084 ... o-Ronquian
Proto-Ronquian reconstructed lexicon: https://web.archive.org/web/20211020032 ... an_Lexicon

Thanks! You don’t happen to know where I could find the original (non-reconstructed) Proto-Ronquian, do you?

[Zju]

Nope. I guess your best try is to contact somebody from that team (https://www.frathwiki.com/Akana#The_Thi ... .282014.29) and/or dhok or Serafín. Though the game forum is down (along with its host), so somebody better has a copy of the original protolang.

[xxx]

The problem is that the same logic is used for both directions: making and interpreting the daughter languages. Usually it's the exact same people doing both jobs at different times. It would be circular to use the data to validate the linguistic technique used to create the data.

I agree with that...
using linguistics to create a language is, in my opinion, already a short-circuit of thought...
so using this construction to prove anything about linguistics is an impossible challenge...

the only point of a conlang in linguistics is that you get a language without culture...
it's not for nothing that the Sapir Whorf hypothesis is an aspect that comes up again and again in conlanging...
even if it's often to justify the conculture of artlangs...

but even in this case the reasoning is biased, even for an engelang, where neither the construction nor the use of the language that could be questioned, can do without the L1 of the conlanger...

the only true use of linguistics in conlanging is to create a conlinguistics whose rules would be valid only in the conmonde that uses it, in other words a linguistics branch of literature (like magic linguistics or Extaterrestrial linguistics for instance...)

the stakes would be very low, not to say irrelevant, when compared with "natural" linguistics
(like conlangs, who limit their playground to a conworld; in fact, only the real world can give them the status of a real language, which is in their hands...).

[zompist]

Question: what are the blue markings on the Faralo page?

Anything related to the parent language, Ndak Ta.

EDIT: and I found the original Proto-Isles Lexicon. Plus its grammar, of course, but you didn’t attempt to reconstruct that. Is your page all that existed of the reconstruction, or did other people work on it too?

I did that reconstruction alone. I forget if anyone else tried it!

[bradrn]

the game forum is down (along with its host)

After some searching… no it isn’t! And, astonishingly enough, you don’t even need the Archive to access it.

That being said, the last post there seems to be a call for the protolangs to be released. And it looks like they never were. So perhaps I will have to contact dhok (or Serafín if I can find them).

Or, who else is listed? Cedh, vec, and… Pogostick Man, who if I’m not misremembering is our very own Man in Space, who of course is still active here. Might he still have the protolang lying around?

Also, while I’m at this: it looks like the Archive has preserved the old AkanaForum. Perhaps there’s some useful materials there. I’d have to dig around for them a bit, though.

[Nortaneous]

Woot! As luck would have it, the wayback machine preserves both of these:
Proto-Ronquian reconstruction: https://web.archive.org/web/20180305084 ... o-Ronquian
Proto-Ronquian reconstructed lexicon: https://web.archive.org/web/20211020032 ... an_Lexicon

Thanks! You don’t happen to know where I could find the original (non-reconstructed) Proto-Ronquian, do you?

here

[bradrn]

Woot! As luck would have it, the wayback machine preserves both of these:
Proto-Ronquian reconstruction: https://web.archive.org/web/20180305084 ... o-Ronquian
Proto-Ronquian reconstructed lexicon: https://web.archive.org/web/20211020032 ... an_Lexicon

Thanks! You don’t happen to know where I could find the original (non-reconstructed) Proto-Ronquian, do you?

here

Amazing, thanks! And amazingly detailed, too… I don’t think any of my conlangs have ever been this well worked-out, let alone a protolang intended to be kept secret.

[bradrn]

On closer inspection it looks like we don’t have access to the lexicon; could you allow public access to that, please?

[Nortaneous]

On closer inspection it looks like we don’t have access to the lexicon; could you allow public access to that, please?

here

[WeepingElf]

Is there any reason why this thread is here rather than in the Akana section?

[bradrn]

OK, so this is quite interesting. Taking a preliminary look over the actual Proto-Ronquian, and comparing it to Zju’s and KathAveara’s reconstructions… well, it’s quite different, let’s put it that way. For a start, both reconstructions have completely missed the voiceless nasals, the prenasalised stops, and the ejectives. Looking a bit closer, the latter have been reconstructed as Cʔ clusters, and presumably something similar happened to the prenasalised stops too. Still, this is enough to refute one of my preliminary hypotheses — that reconstructions will generally recover the correct number of consonant series, even if they don’t get their realisations quite right.

EDIT: looks like the voiceless nasals are reconstructed as *hN sequences. There’s also a series of reconstructed onsets *bʔ *dʔ *gʔ, which look like they could be the prenasalised series — especially since the grammar mentions a connection between the prenasalised and ejective consonants. In that case, Zju’s reconstruction does correctly find all the consonant series… but without their correct realisations, just as I predicted.

There’s also extra series of consonants. KathAveara in particular reconstructs a nonexistent labiovelar series. Curiously, both reconstructions agree on the presence of *ɴ, even though nothing of the sort actually exists!

The vowels didn’t fare much better. Both central vowels got merged into one, and the same happened to the two low vowels. Admittedly, there is a suggestion of oddness in *ə: ‘a vowel that shows up variously as e or a in RT and RTJ, with or without palatalising effect’. The article also mentions an inconsistency at the start of the ‘Important correspondences’ section, which (consulting Nortaneous’s lexicon) turns out to be due to *y. So I’ll chalk that one up as a half-pass, which might have been recognised eventually.

What else? The numbers are mostly correct (modulo the reconstructed phonology), but eighteen, five and especially two are a bit off. Similar story for the pronouns. In general, it seems that most word-shapes are reconstructed correctly, even if the precise phonology isn’t.

Interestingly — though not unexpectedly — when the two reconstructions agree on a numeral, it’s generally correct. The numerals for which the two constructions disagree seem more likely to have been reconstructed incorrectly. I can’t confirm that for the pronouns, though, since only Zju’s reconstruction of those is shown.

That’s enough for tonight, I think. I’ll see if I can do any more investigation tomorrow. Perhaps I’ll look into zompist’s Proto-Isles reconstruction too, though that one doesn’t seem to have as much information.

Is there any reason why this thread is here rather than in the Akana section?

I thought it would be best suited to here, since it’s really about studying the comparative method itself. The Akana aspect is incidental. But presumably it can be moved if it would work better there.