Talk:Marshallese phonology

Nice article, but much of the phonetic detail seems unsupported by the sources. For example, this article says the difference between p and b is the position of the lips, when the ref suggests it's the position of the tongue. I'm removing such detail until it's sourced. — kwami (talk) 01:15, 3 June 2011 (UTC)

If possible, this article needs to have in-line citations added. Mo-Al (talk) 03:15, 3 June 2011 (UTC)

Original research?

I'm concerned about the possible original research in this article. Only one academic source is cited (A Brief Introduction to Marshallese Phonology by Heather Willson), and it does not make the unusual claim that this article makes that there are three approximants, all of which are variants of a voiced glottal fricative /ɦ/. Rather, it describes these three sounds as more normal glides /w/, /j/, /ɰ/. This article posits an extremely abstract analysis of "horizontal" approximants and "vertical" vowels. In general I'm quite wary of such analyses. I think there's a definite fashion among certain phonologists to prefer such analyses because they make a language seem weird and exotic. Unusual analyses like this are strong claims, much stronger than claiming a less abstract, more familiar-looking analysis, and strong claims require strong sources. I don't see any sources cited for these claims, much less a clear consensus of sources as we'd require to posit such an unusual analysis as the one we present as "preferred". Benwing (talk) 22:18, 2 April 2012 (UTC)

Some other issues I've noticed (not exhaustive!):

• I seriously doubt that Marshallese has only closed syllables. AFAIK no such languages exist. In fact, the first two sentences in the section on syllables contradict each other, as one says that open syllables exist and the other says they don't exist. Both quote the same source, and only the first statement (that open syllables do exist) is found in the source.
• The claim that Romanian has non-high glides of the sort postulated in this article is false. What Romanian does have is diphthongs like /oa/ and /ea/, where the primary length and dynamic stress is on the second half. This is quite different from claiming that you can have a non-syllabic /e/, /ɛ/ or /æ/ between two vowels. Possibly you can, but I've never heard of it, and the single source does not appear to support such a claim. The source also notes that there is disagreement about whether a glide like /ɰ/ exists at all on the surface -- i.e. instead you simply hear a long vowel. This suggests that the allophonic transcriptions given in this article may be largely OR. Benwing (talk) 22:34, 2 April 2012 (UTC)

I believe Spanish can have non-high semivowels as part of how it treats vowel hiatus. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 01:44, 3 April 2012 (UTC)

Again, this is AFAIK next to a consonant, in which case it's really a quickly-passed-over vowel. It's a totally differently story when one of these sounds occurs between vowels and claims to be a true glide. Benwing (talk) 09:34, 5 April 2012 (UTC)

I grew up in the Marshall Islands, and heard Marshallese spoken every day. Though I never became fluent (my first language is English), I am familiar with the sounds of the language. Marshallese has the disadvantage of being rather poorly documented online -- for years, the most comprehensive introduction to Marshallese phonology you could find online was a PDF tutorial by the Peace Corps, that kept mentioning to ask your instructor for help. With the more comprehensive introduction to Marshallese phonology, and the Marshallese-English Dictionary (which has full phonemic references for each word), it finally became possible to start writing a real phonology article with sources.

Marshallese approximants are indeed undefined by height, taking their height entirely from the vowels they neighbor. There's a more comprehensive example of this in action in the pronunciation guide for the lyrics of Ij Yokwe Lok, a song I heard practically every day growing up. The problem with /j ɰ w/ is that the specifications for the IPA describe them as closed in height, and are equivalent to asyllabic /i̯ ɯ̯ u̯/. Since most languages that have such approximants do not have more open approximants as you note, there is a lot of room for allophony. But in Marshallese, with its vertical vowel system, the approximants have to remain at the same height as the vowels they neighbor, or you risk producing additional asyllabic morae into the word. This is why the greeting Io̧kwe {{IPAc-mh|(y)|yuy'|(y)|yaw|.|KW|wey|ey}} is analyzed in the Marshallese-English Dictionary as {yi'yaqey} (CV'CVCVC) — Eo̧kwe [æ‿ʲ, æ‿ˠ, ɛ̯ɑ‿ˠ, æ‿ʷ, ɛ̯ɒ‿ʷ] would be {yaqey} (CVCVC) and pronounced differently. As for the use of /ɦʲ ɦˠ ɦˠʷ/ to describe these phonemes, that was actually a compromise from an earlier dispute over ad-hoc IPA /j̆ ɰ̆ w̆/ that were used in the past. I didn't want to use /ɦʲ ɦˠ ɦˠʷ/ initially, because the source material did in fact use /j ɰ w/ as phonemic radicals, but these particular IPA symbols are defined as closed in height when the actual phonemes are undefined by height. /ɦʲ/ has possible allophones /j e̯ ɛ̯ æ̯/, and the articulation [j] is specifically the phoneme /ɦʲ/ that neighbors the horizontally undefined /ɨ/ on either side. The IPA symbol /ɦ/ was selected as a replacement phonemic representation because it's the simplest, most flavorless voiced approximant, which all other voiced approximants implicitly co-articulate by virtue of having the glottis open at all. This ended up working out well, because /ɦʲ/ is palatal-flavored but not necessarily defined by vowel height. But if this too is inadequate, it means the IPA does not have the means to adequately represent Marshallese phonemes without the use of ad-hoc transcriptions. It should be noted that this has plagued attempts to orthographize Marshallese since first Western contact — Western writing systems tend to assume two-dimensional vowel phonemes as a given, and the result was the infamously variable Marshallese spelling system that exists today.

So was original research involved? I wouldn't have thought so, but I suppose I can't be entirely certain. I grew up in the islands afterall, and heard Marshallese spoken on the street practically every day. I suppose it is possible that, if not original research, there is a systemic bias where the references were triangulated with my own memories. Like I said, this stuff hasn't been extremely well documented online until recently. Through most of the 90s and 00s there wasn't much of a Marshallese presence online at all — people have more immediate day-to-day concerns, as they are generally poor people without a great deal of access to technology and media. It is not a bad life, but historically it has been a very disconnected life. And there is a significant Marshallese diaspora in the United States, but they mostly learn and speak English. Actual useful online materials in a language with barely five-digit numbers of speakers have not been plentiful until late. And now, you can find a lot more of it, especially on sites like YouTube.

However, growing up, Marshallese was the "second language in the world" for me. It's like native English speakers in the United States who hear Spanish spoken all over the place — it's obviously a language of significance, even if you haven't become fluent in it yourself. There was a tectonic paradigm shift between my childhood when the language was everywhere, and being transplanted to another continent and gaining internet access, just to find how scarce the language was in the world. Writing these articles has been my service to increase awareness of that language. - Gilgamesh (talk) 09:07, 3 April 2012 (UTC)

Even if there was original research involved, if you can back up the article's claims with inline citations, I think the article will be improved. The sources don't have to be available online. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 16:56, 3 April 2012 (UTC)

Okay. Like for which parts? Note that I don't necessarily have offline book references. But I can reference entries in the online MED, as well as YouTube videos (particularly language examples). - Gilgamesh (talk) 05:33, 4 April 2012 (UTC)

Any claim in the article that isn't explicitly referenced inline is uncited. If you don't have access to materials, what was the basis of all the contributions you've made here and at Marshallese language last year? — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 16:47, 4 April 2012 (UTC)

I said I didn't have offline texts. I have online ones. The various PDFs already referenced, and the bulk of my lexical references are from the MED. I suppose the individual words can be sourced. But for place names, a lot of their meanings are remembered. I knew as a child, for instance, that Ebeye was originally Epjā, but today people who live there call it Ibae. One weakness of the MED is that it only gives place names in unfiltered Marshallese, and seldom gives their more familiar anglicized names. But I've noticed the MED has been gradually improving over time. - Gilgamesh (talk) 07:47, 5 April 2012 (UTC)

Alright, I added inline sources for the words cited. Note that the MED uses its own phonological guide, but it has a page that tries to explain it. Between that and the PDFs, Marshallese phonology is documented. It does indeed have a vertical vowel system and heightless approximants. But if /j ɰ w/ is too specifically close for heightless approximants and both /j̆ ɰ̆ w̆/ and /ɦʲ ɦˠ ɦˠʷ/ are to be seen as inadequate ad hoc transcriptions, what should we use? What is the best way to describe approximants in IPA that are phonemically undefined by height? - Gilgamesh (talk) 08:16, 5 April 2012 (UTC)

You didn't address the contradictory statements about only closed syllables existing or not. But I really want to see a source that claims that Marshallese has heightless approximants. It's not in the academic source you use. I'm not so concerned about the particular words, and I'd imagine Aeusoes isn't either; it's more in the general claims made about the structure of the phonology and the particular ways that specific abstract phonemes are realized on the surface. If there's no source for this, then it is OR -- regardless of whether it's true or not.

As for /ɦʲ ɦˠ ɦˠʷ/, if they're your own invention, they're OR. I'd consider being willing to accept them anyway, but only if you (1) find sources that actually agree that these approximants are heightless; (2) find a lack of agreement on any other solution; and (3) find clear evidence (ideally, explicit comments to the effect) that phonologists have been searching around for symbols for the glides and aren't happy with any they've found. So far I see one academic source, which doesn't confirm any of these: It happily uses regular glides, gives (AFAICT) no indication that these are problematic, and doesn't indicate (AFAICT) that the glides are heightless. The basic idea here is that we should use what the sources use -- even if the sources disagree, better to use one of the existing conventions than make up a new one, either the one that seems most used or the one that seems least problematic. Making up your own symbols is only a last resort, in rare cases where there is no consensus among phonologists along with general agreement that all solutions proposed so far are bad. If the glides are indeed heightless, I could easily imagine exactly such a situation arising among the phonologist experts, but this needs to be shown. It may sound like I'm being a stickler for rules, but it's just because I've seen so much OR appearing in phonology-related and other linguistics-related articles in Wikipedia.

This is not at all to knock the great work you've done on this article -- it's a fascinating language and one I never would have encountered it had it not been for your work. I just want to make sure everything here is actually correct!

BTW if the PDF references you've given are the only ones you have, then you have a severe shortage of reliable sources. I'd suggest you go find more sources; they will probably back up most of your statements and contradict some of them (in the process revealing a great deal about the current state of linguistics research on Marshallese), and in any case make it possible to properly source the controversial statements. Feel free to contact the relevant professors working in the area and ask them both for references and questions about the actual analysis of the phonology. If you don't have access to JSTOR, I might be able to help. E.g. Byron Bender's "Marshallese Phonology" (1968), Oceanic Linguistics, Vol. 7, No. 1, pp. 16–35 is on JSTOR and you should almost certainly cite it unless there's a more recent comprehensive paper. This paper clearly mentions the height-only vowels but says nothing about heightless glides, choosing to note them as /j/, /w/, and /h/ (noting that the last one barely exists at all, mostly just for theoretical purposes).

BTW the fact that numerous authors comment that these glides often don't exist at all on the surface suggests that you need to discuss this more in the text, and your surface allophonic transcriptions need to reflect this.

And thanks again for all the work on this page! Benwing (talk) 09:34, 5 April 2012 (UTC)

I'm not entirely sure I understand what you mean about there only existing closed syllables. Where in the article was this mentioned?
And right now, I'm reviewing the sources again. Unfortunately, one of the main problems with linguistic resources is that many of them are unavailable online. The whole "academic shelfware" problem, when some of us don't have access to that kind of material. I hope I can find the reference for heightless approximants, because I know this to be very true from my memories, which I probably triangulated references with more heavily than I may like to admit. I know that approximants in Marshallese can have very open realizations, and that this is also reflected in the spelling conventions and shown in oft-encountered spelling variations. Actually, come to think of it, I know a page that describes this in detail. Finding Words in the Dictionary at the MED. It mentions the interchangeable spelling patterns. A good contrasting example of this is the expanded greeting io̧kwe eok. The dictionary's phonological guide gives these as {yi'yaqey yẹq}, which in our IPA templating is {{IPAc-mh|(y)|yuy'|(y)|yaw|.|KW|wey|ey|_|(y)|yow|kw}}. The {yi'y-} part is used by the dictionary where the specifically closed articulation [i̯] is used, analyzed as an asyllabic vowel sandwiched between two approximants rather than a single approximant. But the simpler notation {y} has no such stipulation, and words can begin with the approximant at the same height as the vowel it precedes. In this sense, what I termed the heightless approximant is probably more like a phonemic coefficient coloring the vowel for frontness but not height, while the vowel phoneme itself colors the articulation for height but not frontness. To Western ears (including mine, since I grew up there), no matter how many time a word like Em̧m̧an is articulated by a native Marshallese speaker, the pronunciation is [ɛ‿ʲ, ɛ‿ˠ, ɛ̯ʌ‿ˠ, ɛ‿ʷ, ɛ̯ɔ‿ʷ], never {{IPAc-mh|(y)|yuy'|(y)|yeh|mh|mh|hay|n}}. In earlier incarnations of these articles, if I recall correctly, I believe we were using a system like /j/ for phonemes, and [i̯ e̯ ɛ̯ æ̯] for articulations. But the problem is that the IPA specifically assigns [j] as an exact equivalent of [i̯] — an asyllabic [i]. And not only is the phoneme's onset vary more in height than that, but [i̯] before a non-high vowel is specifically what the dictionary describes as {yi'y-} — that {{IPAc-mh|(y)|yuy'|(y)|-}} at the beginning of io̧kwe that is not shared by the beginning of eok. - Gilgamesh (talk) 14:20, 5 April 2012 (UTC)

No further comment/response? - Gilgamesh (talk) 13:34, 7 April 2012 (UTC)

Would you like us to tag uncited claims? — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 22:35, 7 April 2012 (UTC)

Please, that would be a big help. Just try not to horizontally fatten the tables too much. - Gilgamesh (talk) 13:01, 8 April 2012 (UTC)

Why has no one tagged the claims? It's difficult to improve this article when there are disputes without clearly pointing them out. - Gilgamesh (talk) 14:45, 10 May 2012 (UTC)

Done. I've also marked some other citation issues. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 16:24, 12 May 2012 (UTC)

Alright, thank you. From the edits it looks like there's a lot to investigate. I'll try to get to it all when I can. - Gilgamesh (talk) 06:55, 15 May 2012 (UTC)

Alright, I just investigated the referenced tagged as a possibly unreliable resource. The PDF says that its author, Peter Rudiak-Gould, learned the language from the people of Ujae Atoll. To be honest, it's not going to be easy to get better credentials floating around than that. Marshallese doesn't have an enormous online presence, and most of the American personel who live there, never learn the language, at least not conversationally past several phrases. (One of my greatest regrets.) It seems rather remarkable when someone takes the time to fluently learn Marshallese language from its native speakers, and publish in-depth documentation about it online. Along with the MED and A Brief Introduction, I would consider it a crucial online reference. So what do we need to satisfy the criteria of a reliable source? - Gilgamesh (talk) 07:06, 15 May 2012 (UTC)

The source seemed to have a rather short, vague, even prescriptive coverage of Marshallese phonology. Do you not have access to JSTOR articles, Gilgamesh? — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 17:23, 15 May 2012 (UTC)

No, I don't. What is JSTOR? And if you mean the "good enough" notes, I saw them more as helpful learner's aids than hard and fast limits. There is considerable non-approximant consonant allophony permitted. And by the way, did you notice the two references I added from the Marshallese-English Dictionary? It helps explain some of the spelling variation reflecting the underlying approximant phonology. Particularly the ā- vs. ea- vs. eo̧-. - Gilgamesh (talk) 23:27, 15 May 2012 (UTC)

JSTOR is an online article database. If you email me through my user page, I can send you a number of articles that will help you with the article.

Those sources are useful, though neither source explicitly backs up the claim of the existence of such phonemes. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 23:57, 15 May 2012 (UTC)

Alright, this is rather embarrassing, considering I've been on Wikipedia since 2003. ...how do I do that? How do I email you through your user page? I couldn't see an email address, or determine any special feature to contact users through their user pages. I'd heard it exists, but don't think I ever used it. Also, if the phonemes don't exist, then why do are they referenced extensively in the entries of the Marshallese-English Dictionary? If this can't be used as a primary reference, we're in a lot of trouble. - Gilgamesh (talk) 05:05, 16 May 2012 (UTC)

There should be an option at the bottom of my userpage/talkpage that says "E-mail this user." Once I receive an email, I can respond with pdf attachments.

It seems as though that dictionary sees some sort of "y h w" set as underlyingly present, though I'm not sure if this is with the same analysis present in this article. I'm also not sure how representative this is in the rest of scholarship on Marshallese. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 12:08, 16 May 2012 (UTC)

Well, I do know from my own experience that, if the approximants are indeed there, they are only as high as the vowels they neighbor. Rather than going from understanding of underlying phonology to samples of its pronunciation, I went historically from being exposed to its pronunciation to later learning about its underlying phonology from sites like these. The whole reason we're not using /j ɰ w/ for the phonemes is because the IPA defines them as strictly equivalent to /i̯ ɯ̯ u̯/. Marshallese does have [i̯ ɯ̯ u̯], but as asyllabic pronunciations of the vowel phoneme /ɨ/. I know well that the common greeting io̧kwe eok (which I'd known most my life) is two syllables, not four, even though it's phonologically CV'CVCVC CVC as {{IPAc-mh|(y)|yuy'|(y)|yaw|KW|wey|ey|_|yo|yow|kw}}. The [j] is an elided vowel of specific height between two consonants, but the [e̯] is actually the realization of an approximant given height by its neighboring vowel. Also, I found a YouTube clip of various common phrases, and their consonants and vowels can be listened to. The phrases being said are io̧kwe, kom̧m̧ool tata, kōn jouj, ej et am̧ mour, em̧m̧an, enana, aaet and jaab. I understand that ultimately you are trying to hold these articles to a higher standard of hard references and verifiability, and I think that's a commendable goal. But when it comes to Marshallese, the occasional meager haphazard references are the best we can usually do. Often the websites and documents involved, assume everyday access to Marshallese speakers by people spending time in the Marshall Islands, because Marshallese is a robust non-endangered indigenous language that also has practically zero usefulness outside its natively spoken area. Articles like these are largely a labor of love by those of us who fondly remember more about it in a world where most people will likely never hear it be spoken. - Gilgamesh (talk) 19:26, 16 May 2012 (UTC)

Actually, in phonemic transcription those descriptions are much less strict. That's why we can use <r> for the English rhotic even though that symbol is technically used for the alveolar trill. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 19:34, 16 May 2012 (UTC)

I browsed all the links in the user page (and also did a Ctrl-F text search), and I couldn't find Wikipedia features to email (or "e-mail") a user. Anyway, yeah, I'm aware about how phonemic descriptions aren't required to be as strict. But isn't it also generally good wisdom not to use symbols like /j/ instead of /ɦʲ/ when the articulation [j] (actually [i̯]) is also often an allophone of /(Cʲ)ɨ̯(Cʲ)/? It doesn't seem right to use /j/ as an abstract when the articulation [j] is actually an allophone of two separate phonemes depending on context. If approximants specified by height can often be allophones of other phonemes, then what is a good abstract way of describing a phoneme unspecified by height? Almost any adequate solution would seem to have to be ad hoc. Earlier, we used /j̆ ɰ̆ w̆/, but someone complained about that. I initially hadn't wanted to use anything like /ɦʲ ɦˠ ɦˠʷ/, but there didn't seem to be much choice left. It's the closest I could come to phonemically describing heightless approximants; /ɦ/ represents a "zero"-type voiced approximant (by virtue of air going through the throat), with the diacritics "flavoring" it to "lean" towards the front, the back, or the rounded back. There are many situations where ad hoc IPA transcriptions are acceptable because of the limitations of the IPA specification. IPA is, frankly, a legacy eurocentric linguistic descriptor that is often poorly suited for phonemic linguistic features found in languages like Marshallese (with its vertical vowels and heightless approximants) or Korean (with fortis consonants for which no standard IPA symbol or diacritic exists). - Gilgamesh (talk) 19:47, 16 May 2012 (UTC)

I bet the link isn't there because you don't have an email address set up for it. My gmail addy is the same as my Wikipedia username.

I haven't taken a deep look at the sources available on JSTOR. If any of them talk about these three phonemes, they may provide a (sourced) alternative.

Personally, I don't mind using representations of phonemes that are original when there are difficulties like the one you've outlined above. It's really the analysis that comes with it that we should concentrate on. — Ƶ§œš¹ _{[ãːɱ ˈfɹ̠ˤʷɪ̃ə̃nlɪ]} 21:29, 16 May 2012 (UTC)

Email sent. Well, the analysis is still the more complicated issue. But as for the phonemic representations, which do you think is better for approximants unspecified by height? /j̆ ɰ̆ w̆/ or /ɦʲ ɦˠ ɦˠʷ/? Or maybe you could suggest a third option? - Gilgamesh (talk) 09:35, 17 May 2012 (UTC)

Alright, I've skimmed most of the PDFs (actually did a Ctrl-F search for "Marshallese" and looked at adjacent notes). I'm not sure I learned anything I didn't already know, but I also understand I was supposed to locate more references to back up claims. ...to be honest, I'm not entirely sure I found any or not.
However, something did come to mind -- it's demonstrable that approximant phonemes (as documented at the MED and in currently-referenced sources, and in the YouTube clips) assume the same height as the vowels they neighbor. I'm still not comfortable with transcribing the phonemes with bare approximant symbols that prescribe a more specific closeness than the phonemes do. But I think I can understand why the approximants are heightless — there is a tendency for height-specific vowels to have height-specific semivowels as asyllabic allophones, as demonstrated earlier in the different approaches of each of the two words io̧kwe eok. In retrospect, I might have even transcribed the approximant IPA articulations as simple hiatus, except for the strange phenomenon of heightless approximant gemination...which I admit I still don't fully understand as a matter of articulation. Though phonetic analysis shows a tightly-bound and smoothly-transitioning relationship of vowels with their neighboring consonant phonemes, Marshallese orthography seems content to leave most approximants unwritten, though this is also a legacy of historic religious orthography before Marshallese phonology was as well understood as it is now.
At the moment, I'm not comfortable (as a matter of original research) changing the articulation transcriptions to indicate hiatus instead of its current flurry of approximants-as-semivowels, until we have a better understanding of how geminated approximants are distinguished from plain approximants between vowels. I mean, I know it has been suggested I have engaged in original research, but that becomes a very muddled question when one has grown up hearing the language on a daily basis. It's all too easy to triangulate references with memories, and none of it feels new or original, but just basic deduction. Though that suggests I cannot completely be neutral on the subject when my memories and associated subconscious assumptions color my edits — I cannot tell if I have a bias. It seems to me that there has to be a line of reasonable doubt though — or else no one could be neutral enough to edit any topics they feel a personal connection to. - Gilgamesh (talk) 13:50, 20 May 2012 (UTC)

Labiovelarization

Does Marshallese have both labiovelarization and nonvelarized labialization? If so, that would just support my impression that Marshallese has the most perverse phonology on the planet. (Marshallese is to phonology what Old Irish is to verbal morphology.) But if not, there's no need to transcribe labiovelarization with Cˠʷ all the time. Cʷ by itself can be assumed to mean labiovelarization. Angr (talk) 23:42, 5 October 2012 (UTC)

No, there is no labialization without velarization, except in the labial consonants which are themselves already labial. It's not as simple to ask whether palatalized labial consonants both palatalize and labialize the vowels they neighbor, because vowels are not phonemically differentiated except by height alone. And while I would not say Marshallese phonology is perverse, I will concede it is fascinatingly bizarre. - Gilgamesh (talk) 20:35, 2 December 2012 (UTC)

I don't see how your second sentence is an answer to my questions, but is there any objection to my simplifying all instances of ˠʷ to ʷ? Angr (talk) 21:03, 2 December 2012 (UTC)