Talk:Danish phonology/Archive 3

Archive 1

Archive 2

Archive 3

New tables

I've added two tables to the vowel section, one for getting from phonemes to morphophonemes (that Nardog didn't originally object to) and one for getting from phonemes to allophones; Nardog is now objecting to both, but only on the grounds that the original table does the job equally well. That has, to put it mildly, not been my experience when trying to learn the phonology of Danish vowels from the original table. On the contrary, I found I got a much better overview of the phonology by making these tables, and it's of course extremely unlikely that I would be the only one. If the new tables are not included, everyone like me will have to go through the painstaking work of making such tables themselves – or we could spare them all the work by adding these tables that I sorely missed when starting to work through the article. In other words, the usefulness of the new tables has been proved through experience, while the alleged equal usefulness of the original table has been disproved. The reason the sort feature doesn't work well enough is that in some cases, two different phonemes correspond to the same morphophoneme in the same environment, so that e.g. the two occurrences of the phoneme /oː/ occur 15 rows apart.

Nardog claims that a list of allophones without the information on where they occur is useless, but I beg to differ. A list of vowel allophones is a list of the vowels that actually occur in a language, while a list of vowel phonemes is a list of random symbols whose correspondences to actual sounds are unknowable. With most languages, the two are sufficiently similar that it's not a problem, but when it comes to Danish vowels, the correspondences between phoneme symbols and actual sounds is a complete chaos, partly because of the nature of the language itself and partly because of the choices of symbols for the phonemes. So in the case of Danish vowels specifically, it's very useful to have a table that alerts the reader to this chaos of correspondences, whether the distribution of the allophones is explained in the same table or only in the original table. The reader is bound to make certain assumptions when seeing the table of vowel phonemes, assumptions that would be correct in regard to most other languages, which is why it's important to immediately let them know that when it comes to Danish, these assumptions are often wrong. This point isn't made sufficiently well by the original table, which in its unsorted version is organised according to morphophonemes, not phonemes. (The assumptions I'm talking about are of course e.g. that, of /aː/, /ɛː/ and /eː/, the two last would correspond to [ɛː] and [eː], not the two first. Or that [a] would be an allophone of /a/, not only of /ɛ/. Or that the two phonemes /ø/ and /œ/ wouldn't both have [œ̝] among their allophones. Or that of the phonemes and phones /ɒ/, [ɒ̽], /ɔ/, [ɔ̞], the first two and the last two would correspond to each other, not /ɒ/ to [ɔ̞], and /ɔ/ to [ɒ̽].) Ardalazzagal (talk) 23:33, 3 July 2020 (UTC)

I wouldn't say there's no utility to the tables you inserted in conception, but the way you went about it is quite suboptimal. The correspondence table is rigorously sourced, so if you derive another table from it, then that table should also be referenced, especially if it's put before the existing table. (You also seem to have copied the narrow transcription for allophones, which makes little sense. A deliberately simplified transcription is used for phonetic transcriptions throughout the article, including the vowel charts. For instance, since the difference between [ɛ̝ː] and [ɛ̝j] is simply a free variation, we count it as one phone and stick to one notation, ⟨ɛ̝ː⟩.)

Thinking it over, I actually see more use in the list of allophones than in the table of phonemes that correspond to multiple morphophonemes. Regardless, wouldn't a table like this serve both purposes?

Phoneme	Phone	Narrow tran- scription	Morpho- phoneme	Example	Note
/i/	[i]	[i]	|i|	mis
/iː/	[iː]	[iː]	|iː|	mile
/e/	[e̝]	[e̝]	|e|	list
	[ɛ̝]	[ɛ̝]		brist
/eː/	[e̝ː]	[e̝ː]	|eː|	mele
/ɛ/	[e]	[e]	|ɛ|	læst
	[ɛ]	[ɛ]		bær
	[æ]	[a]		række	/ɑ/ in innovative varieties.
/ɛː/	[eː]	[eː]	|ɛː|	mæle
	[eː~ɛː]	[eː~ɛː]		bære	[ɛː] in innovative varieties.
	[ɛ̝ː]	[ɛ̝ː~ɛ̝j]		kræse
			|eː|	grene
/a/	[æ]	[æ]	|a|	malle
/aː/	[ɛː]	[ɛː]	|aː|	male
	[æː]	[æː]	|ɛː|	græde	/ɑː/ in innovative varieties.
/ɑ/	[ɑ]	[ɑ̈]	|a|	takke
			|ar|	var	Only in a handful of words.
			|ɛ|	kræft
				række	/ɛ/ in conservative varieties.
/ɑː/	[ɑː]	[ɑ̈ː]	|ar|	arne
			|aː|	trane
			|aːr|	har
			|ɛː|	græde	/aː/ in conservative varieties.
/y/	[y]	[y]	|y|	lyt
/yː/	[yː]	[yː]	|yː|	kyle
/ø/	[ø]	[ø]	|ø|	kys
	[œ̝~œ]	[œ̝~œ]		grynt	[œ] in innovative varieties.
/øː/	[øː]	[øː]	|øː|	køle
	[œ̝ː~œː]	[œ̝ː~œː]		røbe	[œː] in innovative varieties.
/œ/	[œ̝]	[œ̝]	|œ|	høns
	[œ]	[œ]		gør
	[ɶ]	[ɶ̝]		grøn
			|ø|	drøv
/œː/	[œ̝ː]	[œ̝ː]	|œː|	høne	Rare.
	[œː]	[œː]		gøre
/u/	[u]	[u]	|u|	guld
	[u~o]	[u~o̝]		brusk	[o] in innovative varieties.
/uː/	[uː]	[uː]	|uː|	mule
				ruse	/oː/ in innovative varieties.
/o/	[o]	[o̝]	|o|	sort ('black')
	[ɔ̝]	[ɔ̽]		ost ('cheese')
/oː/	[oː]	[o̝ː]	|oː|	mole
			|uː|	ruse	/uː/ in conservative varieties.
/ɔ/	[ʌ]	[ɒ̽]	|ɔ|	måtte
			|ø|	tøj
/ɔː/	[ɔ̝ː]	[ɔ̽ː]	|ɔː|	måle
/ɒ/	[ɒ]	[ɔ̞]	|ɔr|	vor
/ɒː/	[ɒː]	[ɔ̞ː]		morse
			|ɔːr|	tårne

Pinging @Austronesier and Kbb2: for input. Nardog (talk) 02:01, 5 July 2020 (UTC)

Danish phonology is complex, but not "a complete chaos". And FWIW, our description based on Basbøll and Grønnum still leaves out some quite noticeable details; if we were to follow Brink & Lund, we could add extra-short diphthongs and ð-colored vowels—listen to the unstressed example here[1] and tell me which vowel in the table is actually pronounced by the speaker ;)

I can see the point that taking the morphophonemes as a starting point is counterintuitive. Our current table is very dense, but at least it gives all the necessary context about when and why which (morpho-)phoneme is produced in what manner. Whereas a mere listing of allophones[2] without giving rules doesn't help the reader. The "tautosyllabic environment", especially with respect to morphophonemic |r|, is extremely important in Danish. E.g., vowels following /r/ and vowels not following /r/ represent two different subsystems; while the latter are increasingly pushed upwards in the vowel trapezoid (especially on the left side), the former are increasing pushed down. Maybe it could help if we disentangle the subsystems, or add tables that illustrate the different realiaztions, e.g. like this for short front vowels after /r/:

	default	after /r/
/i/	[i]	[i]
/e/	[e̝]	[ɛ̝]
/ɛ/	[e]	[a]
/a/	[æ]	[ɑ̈]
/ø/	[ø]	[œ̝]
/œ/	[œ̝]	[ɶ̝]

And FWIW, I agree with Nardog that the phoneme-phone interface is more important than morphophoneme-phoneme interface. –Austronesier (talk) 13:00, 6 July 2020 (UTC)

I've got no idea where to put this but THANK YOU for finally using proper IPA in this article. You have no idea how confusing it was (well, I know you do - but still). Sol505000 (talk) 19:17, 1 September 2020 (UTC)

Phonetic realization of the so-called "soft d" (Danish: blødt d) is totally different from description in article

The article says "/ð/ – the so-called "soft d" (Danish: blødt d) – is a velarized laminal alveolar approximant [ð̠˕ˠ].[20][21][22] It is acoustically similar to the cardinal vowels [ɯ] and [ɨ].[21] It is commonly perceived by non-native speakers of Danish as [l].[23][21] Very rarely, /ð/ can be realized as a fricative.[22]" There is a reason why - as the article correctly notes - this consonant is commonly perceived by non-native speakers as [l]: The "soft d" is not just an ordinary approximant with a loose laminal-alveolar oral stricture, as e.g. in Spanish, but in addition has two lateral escapes, which makes it something of an "incomplete" [l], where the tip/blade of the tongue fails to actually touch the alveolar ridge, which distinguishes it from the ordinary lateral approximant [l]. Phonetically the Danish "soft d" is a lateralized laminal-alveolar approximant [ð̠ˡ], or in other words, a lateral approximant with incomplete oral closure [l̞]. It is usually NOT velarized. This type of incomplete lateral approximant is not as uncommon in the world's languages as it may seem: It occurs e.g. as one phonetic realization of Dutch syllable-coda /l/ and is a common realization of Brazilian Portuguese syllable-coda /l/, where it is allophonic with the usual [[ʊ̯] pronunciation (similar to some Cockney realizations of /l/). What makes the Danish approximant so special is that it is not derived from /l/ but from /d/. This blogger arrived at the same conclusion as me: http://phonetic-blog.blogspot.com/2010/11/danish.html Soundshift (talk) 04:45, 8 August 2020 (UTC)

@Soundshift: Most sources describe the realization of "younger" speakers (= young in the 1990s or early 2000s) as "velarized". The velarization on Danish /ð/ is very different from the one observed e.g. in [ɬ]. The back of the tongue has a position quite similar to [ɯ˖], and Danish /-Vð/ sounds awfully similar to [Vɨ̯]-diphthongs in N.Welsh to my ears.

The lateral-like acoustic quality certainly comes from the fact that the whole blade of the tongue leaves only a very narrow space for the airstream, creating formants at the sides similar to the ones of /l/. But there is just as much air passing in the middle, so I'm not sure if one can call such approximants/vocoids "lateralized". But agree, lateral [l̻]/[ʎ] is certainly a good starting point for the soft d, and something like [l̻˕ˠ] could be a good approximation, but that's of course totally OR and not applicable in WP. An adequate description of [ð̠˕ˠ] is hampered by the fact that it actually blends into the preceding vowel. Den Store Danske Udtaleordbog describes a whole bunch of /ð/-colored vowels, which are totally unaccounted for in IPA. –Austronesier (talk) 10:24, 10 August 2020 (UTC)

@Austronesier: Among the various Danish sound files I've examined so far, only once have I heard the soft /d/ pronounced with velarization, which is a minor allophone at best or just idiosyncratic. I checked non-syllabic -/d/ as well as its syllabic realization (spelled -ed or -de), but could not detect even a trace of velarization. You write "is very different from the one observed e.g. in [ɬ]" (I guess you mean [ɫ], not [ɬ]). That's simply because it's in fact NOT velarized. Yes, the back of the tongue is in a [ɯ˖] position, but that is true of all consonants without accompanying secondary articulation. It's essentially the neutral unmodified shwa [ə] oral background. The secondary articulation called "velarization" (including uvularization and pharyngealization) is different from that neutral position in that it requires a depressed (concave) dorsum (as opposed to a convex dorsum, which produces palatalization). You're right, my "lateralized" is something of a misnomer because it suggests a secondary articulation, which it isn't at all. The soft /d/ is a lateral (as opposed to a central) approximant/vocoid with a fricative-type central stricture [l̞] (instead of the "stop" stricture of ordinary [l]). As it is essentially a lateral approximant, most of the air escapes laterally, which means only a tiny amount escapes centrally, too little to cause any friction, but enough to give the sound its characteristic acoustic quality, thus distinguishing it from its counterpart with a central "stop" stricture. The /-d/-colored vowels you mention are further proof of my "lateral" argument because they parallel Danish vocalization/r-coloring of syllable-coda /r/. It is no linguistic accident that both rhotics and laterals are grouped together as "liquids" and in many languages are just different realizations of one and the same "liquid" phoneme. Soundshift (talk) 04:12, 14 August 2020 (UTC)

In the end we need reliable sources to back up a statement in the page, and velarization is backed by sources from two eminent Danish phonologists, Grønnum and Basbøll. If we want replace that information or at least say that it is contested, we need reliable sources as well. And in such sources, the "lateral" character is always decribed as an auditory feature of the soft d, but not as an articulatory feature.

Note that phonologists and phoneticians have been looking a into different direction for many years: soft d structurally aligns with the semivowels/vowels and in fact might be right that. Basbøll groups non-syllabic [ð] among the vocoids, and stops short before calling syllabic [ð̩] a vowel for systematic reasons. Following SDU co-editor Steffen Heger, Thomas Olander (actually not a phonologist, but a linguist with awsome credentials in IE accentology) calls [ð] a semivowel, and [ð̩] a vowel [3]. And this new paper (doi:10.3765/plsa.v5i1.4739) gives further evidence for the vocalic nature of /ð/, and calls for rethinking the use of ⟨ð⟩ to transcribe this sound. –Austronesier (talk) 11:30, 14 August 2020 (UTC)

Thanks for the links to the new paper by Chloe Brotherton & Aleese Block and Thomas Olander's Overview of the Danish Sound System. I fully agree with Olander's analysis of the soft /d/ as a (semi)vowel. The corresponding approximant would be the normal [l], while the corresponding fricative would be [ɮ], not [ð]. The soft d is thus something of a true lateral vowel (not just approximant) thanks to the central aperture left over from the original [ð]. The vowel quality is also borne out by the aforementioned paper, where the high harmonics-to-noise ratio (HNR) clearly places the soft d in the same league as the central vowels in terms of sonority (the mean value is significantly lower for the approximant category). An interesting scientific comparison would be one between the Danish approximant [l] and the soft d (or vocalic ([l̞], my tentative notation for this lateral vowel).

Of course, I'm fully aware that all this is not sufficient to make it onto the page in the form of a revision of what is said there about the soft d. All I can hope for is that my comments on the Talk page will contribute to a rethink.Soundshift (talk) 09:02, 15 August 2020 (UTC)

@Soundshift: Well, our goal here is to present everything about Danish phonology that can be gleaned from reliable sources in an encyclopedic manner. And the function of the talk page is to discuss how it is best presented, and which subtopic (including "hot topics" like the instable vocalism or the soft d) can be covered here and to what degree of detail, without giving any aspect too much undue weight. I actually think the soft d does deserve a section of its own—of course without our WP:OR musings ;). @Nardog and Kbb2: What do you think? –Austronesier (talk) 13:14, 15 August 2020 (UTC)

If there's enough to write about, sure. FWIW I couldn't disagree more with Brotherton & Block though. ⟨ð⟩ is clearly the best (and only) option to symbolize a non-rhotic coronal vocoid and it totally does "align with the goals of the [IPA]". Nardog (talk) 11:17, 16 August 2020 (UTC)

Yeah, as long as its treated as a consonant, which it is in most sources that I've read. There's no better alternative, AFAICS. Sol505000 (talk) 08:21, 16 December 2020 (UTC)

Transcription and description of soft d do not match

The soft d is described as a velarised laminal alveolar approximant i.e. [ð̻̞ˠ] (I'm on mobile so can't easily do the zero width space to stop the stacking in accordance with style policies), but is currently transcribed with a retracted rather than a laminal diacritic

Is this a mistake, or is the mismatch present in the cited works? Tristanjlroberts (talk) 10:50, 15 December 2020 (UTC)

@Tristanjlroberts: The transcription as [ð̠˕ˠ] follows Basbøll and indeed does not clearly indicate tongue position apart from lowering [˕] (i.e. forming an approximant, not a fricative). The retraction diacritic is necessary to indicate alveolar position, because for historical reasons, [ð] is considered an adequate starting point the soft d. –Austronesier (talk) 12:43, 15 December 2020 (UTC)

@Austronesier: Thanks. Good point about the retraction diacritic (I was thinking of the common, but loose use of [ð̞] for non-rhotic alveolar approximants, but this isn't really appropriate for a narrow transcription like this). Specifying the laminal nature of the consonant in the words but not the transcription seems an odd choice here, but if it follows from the cited source it should be kept. Tristanjlroberts (talk) 18:28, 15 December 2020 (UTC)

@Tristanjlroberts: I was thinking of the common, but loose use of [ð̞] for non-rhotic alveolar approximants, but this isn't really appropriate for a narrow transcription like this. Is it really common though?

but if it follows from the cited source it should be kept The way we transcribe it should match its description. It doesn't matter how we transcribe it as long as it's in IPA. The sound can be retranscribed with ⟨ð̻ˠ˗˕⟩ to show all of its features. I'm gonna go ahead and apply the changes myself. Sol505000 (talk) 08:20, 16 December 2020 (UTC)

@Sol505000: I'm not sure how common it is, but I've seen it often enough that it seemed natural to me

Regardless, Wikipedia policy is generally to follow notable sources, even when they are contradictory or appear to be incorrect. Making such a change seems to violate the No Original Research policy, being original synthesis (albeit within one source, rather than between multiple sources as the term is usually used) and so is prohibited

Incidentally, the order of your diacritics also seems odd to me, I would expect the retracted one to come first, then laminal, then lowered, and only then velarised. I assume this is chosen to match the order given in the description, but as it is a bigger deviation from the order in the source and has some very unconventional orderings (e.g. not placing the velarised diacritic last, despite it being the only diacritic defined in the IPA standard as following the character) it also seems poorly justified Tristanjlroberts (talk) 12:43, 16 December 2020 (UTC)

As such I have reverted your change. Unless we get a source that explicitly includes the laminal diacritic in its transcription (which I would prefer), this change cannot be made whilst keeping to the policies of Wikipedia Tristanjlroberts (talk) 12:49, 16 December 2020 (UTC)

@Tristanjlroberts: IPA is a means of conveying information, not the information itself. The revert does not look justified to me, especially given the fact that the way we transcribe vowels and plosives does not follow any source. Rather, it is an in-house convention. Sol505000 (talk) 14:34, 16 December 2020 (UTC)

I wouldn't mark laminality in the transcription either, not because we have to use the exact same transcription as the sources (we don't) but because apicality/laminality are omitted even in rather arrow transcriptions more often than not. The IPA is not capable of denoting all phonetic details and that's where written descriptions and diagrams can help out. That said, we already have the narrow transcriptions of the alveolars [n̺, t̺s̺ʰ, t̺, s̺, l̺] which do not appear in the source, so I wouldn't strongly object to Sol505000's suggestion either. Nardog (talk) 20:17, 16 December 2020 (UTC)

Sorry if this is getting forum-ish, but isn't it that most occurrences of apical alveolar approximants are transcribed as rhotic [ɹ]? So somehow, laminality could be seen as a corrollary of not using [ɹ]. Just thinking... –Austronesier (talk) 12:20, 17 December 2020 (UTC)

⟨ɹ⟩ is seldom used for a non-rhotic because it was defined as a postalveolar approximant until 1989 and because the IPA is primarily a phonemic (or taxonomic phonetic) alphabet (IPA Handbook, pp. 27ff). I doubt apicality has much to do with it. Nardog (talk) 13:09, 17 December 2020 (UTC)

If we do include laminality, we should at least put the velarised diacritic last. The laminal, retracted, and lowered diacritics are supposed to appear below the symbol, but are displayed afterwards to prevent bad stacking issues, whilst the velarised diacritic is supposed to appear afterwards, and so should be appearing last of all Tristanjlroberts (talk) 20:12, 17 December 2020 (UTC)

g̊

According to https://ordnet.dk/ddo/artiklernes-opbygning/udtale?set_language=da [k] and [kʰ] should be [g̊] and [g̊ʰ]. --Espoo (talk) 06:17, 24 May 2021 (UTC)

See /Archive 2#Voiceless voiced plosives for how we arrived at the current set of symbols, particularly what I said to Fdom5997 below the section break. Nardog (talk) 06:34, 24 May 2021 (UTC)

Identifying "soft d" with L in non-native speakers

Surely this is restricted to speakers whose native language has a velarised [ɫ], right? As a German I do find this sound unsual, but I've never associated it with L. I can see how the velarisation could remind you of a "dark L", however. 188.96.64.40 (talk) 00:33, 2 October 2021 (UTC)

External link

I have found a video on YouTube about the phonological history of Danish (Why Danish sounds funny to Scandinavians), but I do not know whether Wikipedia will accept it. תיל"ם (talk) 10:22, 14 February 2022 (UTC)

New sources

The entire archive of the Annual Report of the Institute of Phonetics, University of Copenhagen—papers from which I've tried to get a hold of to no avail—is now available online (h/t Schwa dk). I made do with Goblirsch, who covered the topic citing others, but it may be worthwhile to consult the primary sources.

Speaking of which, a new textbook on Danish phonology came out about a year ago: Udtalt by Heegård Petersen, Juul, Pharao & Maegaard (2021). Based on the sample excerpt, they seem to have stuck with Basbøll's scheme (which I think we here agree is a shame), but they seem to cover innovations in a later chapter. There was also a study of modern young speakers' vowels published in 2016 (Juul, Pharao & Thøgersen, "Moderne danske vokaler", Danske Talesprog 16), which I suspect the chapter draws from. Heger's Sprog og lyd also received the first update in 40 years in 2021. They're all a tad too pricey for me and the closest libraries are thousands of kilometers away, so go for it if you can find them. I've been told by a Dane linguist that Grønnum's description of innovative features is inaccurate, and given that the speakers this article most prominently represents are now more than 90 years old, we could use an update. Nardog (talk) 21:50, 23 March 2022 (UTC)

Agree that Danish phonetics/phonolgy is in need of new approaches. I am working on a few things myself ;)

To push things forward I have created a free modern Danish pronunciation dictionary at https://udtaleordbog.dk/ . It currently has around 270.000 word entries with broad transcriptions of different speech styles. The IPA is updated to modern IPA standards, which isn't found in any published Danish text books yet. A brief overview is available here: https://en.udtaleordbog.dk/ipa.php ə.dk (talk) 10:18, 2 May 2022 (UTC)