The Index Diachronica

[bradrn]

Some problems I’ve encountered while trying to compile Polynesian sound changes:

• What kind of transcription should we use? When the phonetics are known, IPA seems like obviously the best choice, but of course the phonetics often are not known. When that occurs, I suggest using the same transcription as the source, preceded by an asterisk.
• Often a sound change seems to be shared between multiple related languages, without them necessarily forming a valid clade (cf. the wave model). How should we transcribe this? The current ID records the same change multiple times in the section for each language, which keeps each language in one section but results in a lot of duplicate entries. On the other hand, putting each shared sound change in its own section could become unmanageable.
• Many sources describe sound changes informally in words, e.g.:

  Nothing has been said about the development, in the Tuvalu dialects of Ellicean and in most, perhaps all of the Outliers development … of a contrast between long and short consonants through the loss of short, unstressed vowels between identical consonants.

  How should this be managed? In this particular case, not only is a precise condition not specified, but the languages it applies to are not delineated clearly.

[Man in Space]

Some problems I’ve encountered while trying to compile Polynesian sound changes:

• What kind of transcription should we use? When the phonetics are known, IPA seems like obviously the best choice, but of course the phonetics often are not known. When that occurs, I suggest using the same transcription as the source, preceded by an asterisk.
• Often a sound change seems to be shared between multiple related languages, without them necessarily forming a valid clade (cf. the wave model). How should we transcribe this? The current ID records the same change multiple times in the section for each language, which keeps each language in one section but results in a lot of duplicate entries. On the other hand, putting each shared sound change in its own section could become unmanageable.
• Many sources describe sound changes informally in words, e.g.:

  Nothing has been said about the development, in the Tuvalu dialects of Ellicean and in most, perhaps all of the Outliers development … of a contrast between long and short consonants through the loss of short, unstressed vowels between identical consonants.

  How should this be managed? In this particular case, not only is a precise condition not specified, but the languages it applies to are not delineated clearly.

I thought I responded to this, but I guess I didn’t.

- Honestly…I think it might be best to use whatever the customary orthography is and then provide some sort of explanation or “key” for “translating” into IPA. Or at least have some sort of dual-boot going on.
- Vis-à-vis the wave model…I’d say group the families as according to custom and then just note when the changes occur (maybe with some explanatory text, like “This was a shared areal innovation in X area at Y time”.
- If they use words, we do our best. The current ID resorts to prose appreciably often.

[Man in Space]

For the record, here is chridd's ID on Austronesian. Am I correct in thinking you want to go narrower in scope, to just the Polynesian langs?

[bradrn]

- Honestly…I think it might be best to use whatever the customary orthography is and then provide some sort of explanation or “key” for “translating” into IPA. Or at least have some sort of dual-boot going on.

This is also a possibility, to be sure.

One of the things I keep on thinking about here is automated processing… which includes things as simple as filtering to a single phoneme (like on chridd’s page). If we want that — and I think it is a very desirable thing to have — we’ll need at least some way to map everything to a consistent representation. And if we have that, we might as well show everything in that format in general.

(Or, on the other hand we could standardise everything to Americanist notation… yeah, that’s a somewhat tongue-in-cheek proposal, but given that lots of linguists use something vaguely Americanist anyway, it might actually be worth serious consideration.)

- Vis-à-vis the wave model…I’d say group the families as according to custom and then just note when the changes occur (maybe with some explanatory text, like “This was a shared areal innovation in X area at Y time”.

‘According to custom’ is tricky, though… for instance, should we have an ‘Italo–Celtic’ branch? There’s often multiple competing subgroup structures, each with their own selection of sound changes, so it could get quite difficult to have one single grouping method. In CS terms, maybe we should be structuring this as a DAG, not a tree.

- If they use words, we do our best. The current ID resorts to prose appreciably often.

We want to be better than the current ID, though. On the other hand, I guess a big part of better quality involves not adding our own interpretations to things which aren’t in the original source.

EDIT: After some thought, perhaps it might be best to include the original words, and then also a more formal interpretation, keeping it clear that the latter wasn’t literally in the original source. That way we can get the best of both worlds.

For the record, here is chridd's ID on Austronesian. Am I correct in thinking you want to go narrower in scope, to just the Polynesian langs?

Correct. Austronesian is huge!

[bradrn]

Something I’ve been pondering lately: I can see two approaches to this project, and I’m not entirely sure which is best.

One is to aim as much as possible to produce a curated and authoritative archive of sound changes. The appeal of such a project is obvious. But it will require a lot of effort, editorialisation and potentially original research, in areas such as choosing between contradictory sources, synthesising diverse descriptions of a language family, deciding on subgroupings, and so on. Most of this is subjective, too, which just makes it more difficult.

The other option is to simply act as a repository for sound changes. This would involve simply transcribing the sound changes from each source, with only minor alterations for consistency and to remove ambiguity. Such an approach would be a lot easier for us, but making minimal judgements on reliability exposes us to poor-quality data.

I feel part of the problem of the original ID is that it takes the worst of both approaches. It makes opinionated decisions on subgrouping, and reports only one or two sources for each set of sound changes. Yet the decisions are often in conflict with diachronic consensus (insofar as such a thing exists), and the sources are often poor ones. No matter which approach we choose, the next version of the ID will benefit from us being consistent.

When I compiled my Middle English sound changes, I essentially was attempting the first approach here: using as much information as I can to publish an authoritative list of sound changes. Yet I found that took a significant effort, and required a lot of domain-specific knowledge to do properly, which I didn’t have. I also found myself very uncertain about what ‘authoritative’ can even mean in the context of historical linguistics.

Thus, for the Polynesian changes, I’m planning to try the opposite: focus on transcribing each source consistently and cleanly, while not worrying too much about whether it has gaps or not. (If we want, it’s still very possible to add a modicum of moderation by e.g. placing a star rating next to each source, as discussed earlier.)

[Darren]

This reminds me, I should get back to working on the LP sound changes. It's a pity that a) the sources are so sparse and b) I can't get my hands on the Fayu grammar which is the only full grammar of a normal LP language rather than the fascinating but somewhat useless outliers of Abawiri and Iau. (On that topic, if anyone could obtain said grammar for me, that would be amazing)

Something I’ve been pondering lately: I can see two approaches to this project, and I’m not entirely sure which is best.

One is to aim as much as possible to produce a curated and authoritative archive of sound changes. The appeal of such a project is obvious. But it will require a lot of effort, editorialisation and potentially original research, in areas such as choosing between contradictory sources, synthesising diverse descriptions of a language family, deciding on subgroupings, and so on. Most of this is subjective, too, which just makes it more difficult.

The other option is to simply act as a repository for sound changes. This would involve simply transcribing the sound changes from each source, with only minor alterations for consistency and to remove ambiguity. Such an approach would be a lot easier for us, but making minimal judgements on reliability exposes us to poor-quality data.

I reckon that approach 1 is good, but it's better to have approach 2 rather than nothing at all for a given family. Plus 1 is only really possible for well studied languages, and is limited by skills. Personally I don't mind a bit of inconsistency in exchange for improved coverage.

[bradrn]

This reminds me, I should get back to working on the LP sound changes. It's a pity that a) the sources are so sparse and b) I can't get my hands on the Fayu grammar which is the only full grammar of a normal LP language rather than the fascinating but somewhat useless outliers of Abawiri and Iau.

Hmm… what does ‘normal’ even mean, when it comes to LP?

(On that topic, if anyone could obtain said grammar for me, that would be amazing)

I’ll see if I can do anything.

Something I’ve been pondering lately: I can see two approaches to this project, and I’m not entirely sure which is best.

One is to aim as much as possible to produce a curated and authoritative archive of sound changes. The appeal of such a project is obvious. But it will require a lot of effort, editorialisation and potentially original research, in areas such as choosing between contradictory sources, synthesising diverse descriptions of a language family, deciding on subgroupings, and so on. Most of this is subjective, too, which just makes it more difficult.

The other option is to simply act as a repository for sound changes. This would involve simply transcribing the sound changes from each source, with only minor alterations for consistency and to remove ambiguity. Such an approach would be a lot easier for us, but making minimal judgements on reliability exposes us to poor-quality data.

I reckon that approach 1 is good, but it's better to have approach 2 rather than nothing at all for a given family. Plus 1 is only really possible for well studied languages, and is limited by skills. Personally I don't mind a bit of inconsistency in exchange for improved coverage.

I disagree: inconsistency is probably the worst thing we can do when it comes to a project like this. And so, given the limitations you list, if I had to pick one approach, it would probably be 2.

That being said, now that I think of it… I guess we could report all the original sources, as well as an additional ‘moderated’ set of sound coming from our synthesis. It’s hardly a priority, though. (And in any case, I guess that would itself be significant research, and publishable in its own right — in which case we could just cite it like any other source!)

[linguistcat]

I'd appreciate course 2 as well. I've been able to scrape together some sound changes for the Japonic family (although mostly mainland Japanese), but unless I stick to a specific reconstruction of Proto-Japonic and the resulting sound changes, it might be difficult to get anywhere. So it's probably best to list separate reconstructions and the assumed sound changes that go with each. After about Early Middle Japanese they tend to agree so that'll be easier but different people will cite slightly different orders for certain changes.

[bradrn]

I'd appreciate course 2 as well. I've been able to scrape together some sound changes for the Japonic family (although mostly mainland Japanese)

Ooh… could you write them up somewhere and make a Japonic thread please? The other stuff can be dealt with as long as we’re reasonably consistent about how we treat it. (I’m literally just about to post some sound changes in the Polynesian thread, which show show some strategies I’ve been trying to deal with such tricky cases.) [EDIT: posted]

[linguistcat]

I'd appreciate course 2 as well. I've been able to scrape together some sound changes for the Japonic family (although mostly mainland Japanese)

Ooh… could you write them up somewhere and make a Japonic thread please? The other stuff can be dealt with as long as we’re reasonably consistent about how we treat it. (I’m literally just about to post some sound changes in the Polynesian thread, which show show some strategies I’ve been trying to deal with such tricky cases.) [EDIT: posted]

I'll have to find a book I stored and sort through some PDFs but I'll try to get to it in the next week or so. I'll try to make a post for each version I can track down.

[Darren]

This reminds me, I should get back to working on the LP sound changes. It's a pity that a) the sources are so sparse and b) I can't get my hands on the Fayu grammar which is the only full grammar of a normal LP language rather than the fascinating but somewhat useless outliers of Abawiri and Iau.

Hmm… what does ‘normal’ even mean, when it comes to LP?

Basically East or West Tariku languages, which all have roughly similar phonologies and from what little I can see morphology. The Far West languages (Saponi/Rasawa/Awera) are too understudied to tell; Central Tariku (Edopi + Iau) is just batshit (although Iau is far crazier than Edopi). Duvle doesn't seem too weird although again understudied as always; then East Lakes Plains (Abawiri + Taburta + Dabra) are poorly documented (what a shocker) apart from Abawiri, which has innovated 240% more consonants than PLP, so is very confusing when it comes to sound changes.

(On that topic, if anyone could obtain said grammar for me, that would be amazing)

I’ll see if I can do anything.

I disagree: inconsistency is probably the worst thing we can do when it comes to a project like this. And so, given the limitations you list, if I had to pick one approach, it would probably be 2.

That being said, now that I think of it… I guess we could report all the original sources, as well as an additional ‘moderated’ set of sound coming from our synthesis. It’s hardly a priority, though. (And in any case, I guess that would itself be significant research, and publishable in its own right — in which case we could just cite it like any other source!)

What's so bad about inconsistency? I feel like that would just bring down the quality of the "good" (option 1) changes, or exclude outright all the "bad" ones. Would it not be enough to make it obvious to the reader where each set changes have been compiled from?

[bradrn]

(On that topic, if anyone could obtain said grammar for me, that would be amazing)

I’ll see if I can do anything.

I mean, don’t get your hopes up too much. If you can’t access it, probably I won’t be able to either.

(By the way, what’s the full citation for the grammar?)

I disagree: inconsistency is probably the worst thing we can do when it comes to a project like this. And so, given the limitations you list, if I had to pick one approach, it would probably be 2.

That being said, now that I think of it… I guess we could report all the original sources, as well as an additional ‘moderated’ set of sound coming from our synthesis. It’s hardly a priority, though. (And in any case, I guess that would itself be significant research, and publishable in its own right — in which case we could just cite it like any other source!)

What's so bad about inconsistency? I feel like that would just bring down the quality of the "good" (option 1) changes, or exclude outright all the "bad" ones. Would it not be enough to make it obvious to the reader where each set changes have been compiled from?

What’s bad about it is that it’s not clear which changes have been stitched together from multiple sources and further refined by us (option 1), or which ones have been faithfully reproduced from the original source (option 2). What makes this a particular issue is that taking option 1 doesn’t merely involve generating sound changes: it also enforces a particular subgrouping, and requires ignoring other sources. Neither of these are especially compatible with option 2.

But then again, if you include both versions and do make the difference sufficiently obvious, that’s probably enough to make it clear that you’re being consistent, just in a different way. (Meta-consistent?) But that’s basically what I suggested in the bit you quoted.

[Darren]

I mean, don’t get your hopes up too much. If you can’t access it, probably I won’t be able to either.

(By the way, what’s the full citation for the grammar?)

Page Maitland 2020, "A grammatical description of Fayu: a Western Lakes Plain language of West Papua". It's on researchgate but no full text and no luck requesting it.

But then again, if you include both versions and do make the difference sufficiently obvious, that’s probably enough to make it clear that you’re being consistent, just in a different way. (Meta-consistent?) But that’s basically what I suggested in the bit you quoted.

I can get behind this.

[bradrn]

I mean, don’t get your hopes up too much. If you can’t access it, probably I won’t be able to either.

(By the way, what’s the full citation for the grammar?)

Page Maitland 2020, "A grammatical description of Fayu: a Western Lakes Plain language of West Papua". It's on researchgate but no full text and no luck requesting it.

Hmm, I’ve had no luck with this. It looks like she’s at the University of Newcastle; they have a thesis collection, but I can’t find Maitland there. I saw one place which listed Bill Palmer as her supervisor, but she isn’t listed on his ‘Supervisors’ page — perhaps because it was a Bachelor’s thesis, and not Masters or PhD. However, her email is listed on the website, so you could well contact her and ask for her thesis directly. (In fact, if you don’t, I may well… I’d be interested too!)

(Incidentally, Newcastle seems remarkably strong in linguistic research. Even if I didn’t find what I was looking for, I still found a bunch of other interesting theses in the process. It also looks like they have both Bill Palmer and Catriona Malau née Hyslop, which I hadn’t known.)

[Darren]

Hmm, I’ve had no luck with this. It looks like she’s at the University of Newcastle; they have a thesis collection, but I can’t find Maitland there. I saw one place which listed Bill Palmer as her supervisor, but she isn’t listed on his ‘Supervisors’ page — perhaps because it was a Bachelor’s thesis, and not Masters or PhD. However, her email is listed on the website, so you could well contact her and ask for her thesis directly. (In fact, if you don’t, I may well… I’d be interested too!)

Thanks, I'll try emailing.

[linguistcat]

I'll have to find a book I stored and sort through some PDFs but I'll try to get to it in the next week or so. I'll try to make a post for each version I can track down.

This will likely take longer than I thought. I seem to have misremembered where I found a table of sound changes, or hallucinated it entirely. I do have one reference at the ready (Samuel E. Martin, 1987), but that only gives when evidence for certain sound changes showed up in writing, when we now have evidence from other sources now that these sound changes occurred earlier, sometimes a lot earlier.

[bradrn]

I'll have to find a book I stored and sort through some PDFs but I'll try to get to it in the next week or so. I'll try to make a post for each version I can track down.

This will likely take longer than I thought. I seem to have misremembered where I found a table of sound changes, or hallucinated it entirely. I do have one reference at the ready (Samuel E. Martin, 1987), but that only gives when evidence for certain sound changes showed up in writing, when we now have evidence from other sources now that these sound changes occurred earlier, sometimes a lot earlier.

I’m not sure the exact dating of sound changes is really within the scope of this project. All we really need for now is the sound changes themselves, in as much detail as possible. So, since we seem to agree that it’s best to transcribe each source independently, perhaps just start with Martin 1987 for now, and we’ll figure out how to integrate multiple sources later.

(For another example, my current Polynesian source barely even gives information about the ordering of sound changes. So I’ve simply transcribed what I could, and in the HTML version I added a note to the effect that there’s little ordering information. It should be clearer once I’ve published the HTML version, which should be ready in the next few days.)

[bradrn]

Maybe we should try and come up with an actual list of aims (cringe, I know) in the main thread so we all know what we're doing.

On the basis of Darren’s salutatory (and not cringe at all) suggestion from the Lakes Plains thread, I think my own aim is:

To create a highly reliable and comprehensive repository of sound change information from the world’s language families.

I like this idea because it has the potential to be very useful to multiple different groups, at least if it gets popular enough:

• For conlangers such as ourselves, it provides a way to quickly gain confidence as to which sound changes are attested or not.
• For historical linguists, it could act as reliable database with which to test phonological generalisations (as opposed to the unsupported claims I regularly see in the literature).
• For linguists more generally, it could be a place to make their diachronic work more widely known.

But I’m happy to hear other suggestions about what we should aim for!

[linguistcat]

I’m not sure the exact dating of sound changes is really within the scope of this project. All we really need for now is the sound changes themselves, in as much detail as possible. So, since we seem to agree that it’s best to transcribe each source independently, perhaps just start with Martin 1987 for now, and we’ll figure out how to integrate multiple sources later.

I'll just post the screenshot of the table I got along with the sound changes so you'll see what I mean, but I'm already mostly done with this at least. I'll go through my other main source, which I thought had a nice concise table of changes but seems to not, after I finish that.

[bradrn]

I’ve finished my mockup of some Polynesian changes: https://bradrn.com/files/polynesian-biggs-mockup.html. This is compiled mostly from Biggs 1978, plus the changes for Takuu, Nukumanu and Nukeria from Davletshin 2015. Please let me know your thoughts! (Comments on the sound changes themselves can go in the Polynesian thread, general comments can go here.)