Talk:Diaphoneme

Lead rewrite

@J. 'mach' wust: I'm having trouble seeing how the lede glosses over the distinction between an allophone and diaphonic variant. The former is a contextual variant within the speech of a single speaker or dialect, contrasting with other pronunciations of the same phoneme within that same lect. The latter is a dialect-specific pronunciation, contrasting with other lects' pronunciations in the same context. How does the lede miss this? What wording do you recommend? — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 16:13, 9 March 2016 (UTC)

A better example of a diaphoneme should show how the assignation of the diaphoneme to the phonemes can differ between varieties. An example might be the BATH vowel. The current example with different realizations of the PRICE vowel can just be explained as allophonic variation of the phoneme /aɪ/. That is, unless we would adopt your restrictive definition of allophones “within the same lect”. But that is beside the point (and would complicate the lead even more, because as of now the lead misses it completely and so do the articles phoneme and allophone). The point is that the BATH vowel is a more illustrative example than the PRICE vowel. --mach 🙈🙉🙊 16:33, 9 March 2016 (UTC)

But it can't be explained that way. What you call my "restrictive definition" of allophones is the typical restrictiveness given to the meaning of the term allophone. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 16:42, 9 March 2016 (UTC)

I agree that something isn't quite right about the lead. The PRICE example would be suitable as an illustration of the original Daniel Jones 'diaphone', in which various allophones (square brackets) are grouped for a variety of speakers. But in that case all the allophones are allophones of the same phoneme. When we get to diaphonemes we surely need an example which shows how a diaphoneme (as used in WP transcriptions) may subsume more than one phoneme (slant brackets) under a single symbol, either for distributional reasons (e.g. the use of the HAPPY vowel i) or dialectal (possibly the BATH vowel, though I think that example may not be ideal). There's a brief but useful discussion of the problem in Wells 'Accents of English', pp 69-70. RoachPeter (talk) 17:09, 9 March 2016 (UTC)

Diaphonemes do also include dialectal variation of the same phoneme, so the example isn't wrong, exactly. Would the New York example further down in the article be a better choice for the lead? — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 17:40, 9 March 2016 (UTC)

It certainly is not wrong, but it is not helpful either, because readers may confuse it with mere allophonic variation. I do not have Wells’s Accents of English at hand. Since he explicitly discusses the problem, I assume he provides good examples that are easy and to the point. I suggest, then, that we take one of his examples. That will allow us to add another citation to the article.

I wonder what “may not be ideal” about the BATH vowel as an example. Is it too predictable diachronically? --mach 🙈🙉🙊 00:05, 10 March 2016 (UTC)

Looking at my notes on Wells, I don't see mention of any examples. The selection is talking about the theoretical problems of polylectal grammars, which is already covered in the article (there is even a blockquote from that very page range).

No example will be perfect, though I can probably drag up a sourced one if we feel like it would be helpful.— Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 02:57, 10 March 2016 (UTC)

I could not believe Wells would not use any examples. So I went to the library and it turns out he used two examples:

1. FORCE and GOAT vowels: “for many speakers [...] /fors, got/ [...] [f]or others [...] /fɔːs, goʊt/”. No explicit diaphonemic transcription.
2. Finer points on the FACE vowel, taking into account the Pane–pain merger and the Wait–weight merger:

Words	Diaphoneme	“those north-of-England accents which make this distinction”	“other accents”	“large majority of speakers of English”
late, mate, mane	“underlying monophthong” (|eː|)	/eː/	/eː/	/ɛɪ/
main, wait	“underlying [...] diphthong” (|ɛɪ|)		/ɛɪ/
eight, straight	“presence of underlying /x/”: (|eːx|)	/ɛɪ/

I think the latter makes for a better example. Wells relates it with more detail, and we already have Pane–pain merger and Wait–weight merger that we can link to, and the relative complexity illustrates the intricacies of diaphonemes better than the FORCE example (or the BATH example). --mach 🙈🙉🙊 12:29, 10 March 2016 (UTC)

If we are going to use an example that involves mergers or splits, it seems less clear to me to use ones where only a handful of speakers speak one way or another. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 14:40, 10 March 2016 (UTC)

My point is that we must use an example that involves mergers or splits. Otherwise, we fail to illustrate how the concept of diaphonemes is different from mere allophonic variation.

I think if the late/main/eight example is good enough for Wells, then it is certainly good enough for Wikipedia. It is better than the New York example because we can cite a source and, I think, because it is easier to illustrate – the New York example involving two unrelated diaphonemes, it is more difficult to get to the point. Also, the New York example could mislead readers into believing that diaphonemes are only about transcribing dropped sounds. Wells’s late/main/eight example may not concern many speakers, but it is readily understandable all the same. Of course, if you know a better example, by all means don’t hold back. --mach 🙈🙉🙊 16:13, 10 March 2016 (UTC)

The examples Wells gives there aren't designed to clearly illustrate what a diaphoneme is, but rather to point out problems in the theory behind constructing a panlectal grammar. So appealing to his authority as a rationale for these examples isn't convincing. I've found a source for the BATH vowel, but I am also curious as to why Roachpeter believes that would be less than ideal. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 16:48, 11 March 2016 (UTC)

Sorry for any confusion - my remark about the BATH vowel was just based on my recollection of the (rather ancient) literature on diaphonemes, archiphonemes and the like. I don't remember BATH being used as an illustration, but it would take me a while to find what examples *were* used, and I'm not regularly going in and out of a university library as I used to. I just thought an example backed up by a literature citation would be best, and if this exists for BATH, that's fine. RoachPeter (talk) 17:33, 11 March 2016 (UTC)

I do not understand what you mean by asserting that Wells’s examples “aren’t designed to clearly illustrate what a diaphoneme is, but rather to point out problems in the theory behind constructing a panlectal grammar”. Sure, Wells is critical of the term, but how does that make his examples any less illustrative?

Anyway, I am fine with using the BATH example. Care to tell us where you found it? --mach 🙈🙉🙊 11:10, 12 March 2016 (UTC)

If nobody provide the source for the BATH example, I am going to put the late/main/eight example into the article because it has been used in a relevant source (Wells). --mach 🙈🙉🙊 07:54, 15 April 2016 (UTC)

Having just revisited this page, I am concerned that the lead paragraph uses a very complex example to give the reader an idea of what a diaphoneme is. I'm particularly concerned about the representation of 'eight' with //ex//. This convention only works if you accept that a phonological representation may contain underlying segments that are modified or deleted by phonological rules. There seems to be no explanation given of how this is done in the case of 'eight'. While having underlying representations and segment-changing rules can lead to neat (or intellectually satisfying) solutions such as the Spanish example cited in the article, it is not (as far as I know) a necessary component of diaphonemics and it certainly can't be assumed that the average reader will know about the mechanisms of generative phonology. The fact that Wells, in the cited source, mentions the //ex// analysis does not mean that he endorses it. I think it would be better to have a simpler example. In looking for a simple example, there is a basic issue that ought to be brought out here. In the case of phonemes and archiphonemes, we can see that the relation between them is one of many to few. However, in the case of most of the examples given for diaphonemes, the relation between phonemes and diaphonemes is one of few to many, just as the relation of phonemes to allophones is one of few to many. The diaphonemic system implied in the 'late-wait-eight' example, and in the table in Help:IPA for English, contains more entities than the phonemic inventory of most individual varieties of English. RoachPeter (talk) 15:03, 25 April 2017 (UTC)

I fully agree: The generative grammar assumption of underlying segments that never surface is rather awkward. I’d be happy to go for a more accessible example – in fact, that is why I originally proposed a bath example. However, Ƶ§œš¹ was reluctant to change the original example. It used different pronunciations of the word eye which in my opinion could mislead readers into assuming that this article were about allophonic variation among different varieties of English. I was happy that Ƶ§œš¹ implicitly consented to the eight example in the end (see [1] and [2]). That is why I kept it even though I think the bath example would be easier to understand. --mach 🙈🙉🙊 17:55, 25 April 2017 (UTC)

Minor correction. I didn't oppose using the BATH vowel. I just never found the source I thought I had access to that backed it up. If someone else wants to add it, I'm fine with that. I echo Roachpeter's concerns with it. — Ƶ§œš¹ _{[lɛts b̥iː pʰəˈlaɪˀt]} 20:50, 25 April 2017 (UTC)