Talk:Icelandic phonology

Content change

I’ve moved the phonological content from the main Icelandic language article to here, leaving copies of the tables and a brief summary there. I’ve moved the old content of this page to Icelandic orthography, where it is better suited. Max Naylor 18:16, 27 April 2007 (UTC)

Laterals

Is there a phonemic contrast between the "plain" laterals [l, l̥] and the velarized laterals [l̥ˠ, lˠ]? None of my (admittedly very scanty) resources indicates that L is ever velarized in Icelandic, let alone phonemically so. —Angr 15:43, 12 June 2007 (UTC)

Well, I was also surprised to see it but it seems that for some Icelanders "veldi" ((I would) choose) and "velgdi" ((I would) make warmer) are not homophones. I believe the reference comes from the book "Íslensk tunga" [1] but I don't have it. Perhaps someone who has the book close by will come and quote the relevant passage. Stefán 19:14, 12 June 2007 (UTC)

Voicing contrast

The current text says:

Icelandic has an aspiration contrast between plosives, rather than a voicing contrast, something relatively rare among European languages

Is it really? I'm pretty sure it's the exact same in Danish. The difference between Danish /p/ and /b/ is not one of voicing, it's one of aspiration. --Pinnerup 23:49, 29 July 2007 (UTC)

Well, it says relatively rare. Danish certainly is another example, Faroese probably another. Those are the only ones I am aware of. If there are more then we should consider rewording the statement. Stefán 03:52, 30 July 2007 (UTC)

Scottish Gaelic is another one; it's actually not rare in Northwestern Europe. —Angr 04:52, 30 July 2007 (UTC)

Right, I've removed it, and another paragraph which seemed to be confusing the written and the spoken language. We still say that a voicing contrast in nasals is rare among the languages of the world. Is that an accurate statement? Also, we say that length is contrastive for consonants, but not vowels. Am I correct in saying that length is only contrastive for [m], [n], [r] and [s]? I cannot think of any others. Also this may not be terribly accurate, the research of Pind suggests that it is the relative length of the vowel compared to the following consonant rather than the absolute length of either one which is the important factor. There is also eyddra vs eitra in which either the vowel length is contrastive or my statement above about which consonants can be contrastive is wrong. Comments? Stefán 16:30, 30 July 2007 (UTC)

Translation

This page has a lot of example words with no translation. It would be a good thing to translate the example words. Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 04:58, 30 August 2007 (UTC)

Agiesh hulumar Johannsson the great (talk) 14:36, 23 April 2017 (UTC)

J in the phonology

j is an approximant in the IPA. Why is it put in the fricative bit here? Also, j is included as being one of the fricatives which veer towards being an approximant here but in 'Icelandic language' it is not included. Munci (talk) 18:46, 3 May 2008 (UTC)

diphthongs

Are the unusual diphthongs rising or falling? That is, where should we put the <  ̯> diacritic, and how does that fit the length diacritic? kwami (talk) 20:57, 3 June 2008 (UTC)

Glottalization of non-aspirated stops?

This abstract seems to be saying that the difference between aspirated and unaspirated consonants in Icelandic involves the glottis being open (or closed) during articulation of the consonant. This would suggest that at least some unaspirated Icelandic stops are in fact glottalized or ejective. I don't speak Icelandic, but I occasionally see Icelandic shows on my cable TV (the SCOLA channel), and I can definitely hear some glottalized /k'/ sounds (which have a distinctive "click" or "pop" sound). Thoughts or comments on this from anyone who does speak Icelandic? Richwales (talk) 05:13, 18 January 2009 (UTC)

If the ejective quality were that strong and clear I'm sure it'd have been noticed already. I rather suspect that it is more like in Eastern Armenian, where the unaspirated voiceless stops have a quality that almost sounds like ejectives but not quite, and I've heard they're better described as tense. That would make them similar to the Korean tense stops, though phoneticians don't seem to be quite sure if tenseness, stiff voice or faucalized voice is the best way to describe their distinctive quality. --Florian Blaschke (talk) 15:13, 22 April 2010 (UTC)

Incorrect IPA

I have corrected a few important mistakes in the IPA of vowels, where it was previously written [øy]. This sound does not exist in Icelandic, and so I don't know who put it there but it's wrong. I think someone might have used AIPA thinking it meant [øj]. In any case, I've changed it to [øɪ], but if you like you can change it to [øi], though [øɪ] sounds better to me. Also, I changed a few [c] and [ç] to [cj] and [çj] (vekja was [veca], now it is [vecja] and hjá was [çau] but now it is [çjau]. There is definitively a difference in sound between "kja" and "ki" in Icelandic.

Anyways, I hope this helped. (Egein (talk) 14:59, 1 August 2009 (UTC))

You've left the 2nd change unfinish'd. You seem to have made the changes only in the consonant table, and even in the section of the table where it claims that eg. orthographic <kj> indeed is /c/ rather than /cj/! And then there's the paragraph under the phoneme table where it discusses this phenomenon, claiming that eg. góla vs. gjóla is a minimal pair.

Now, I'm sure there's a distinction between kja and ki - different vowels! - but is this one between /cʰja/ and /cʰi/, or /cʰa/ and /cʰi/? Try something like ké vs. ke: if you're right, these should be distinguishable as /cʰjɛ/ and /cʰɛ/.

Eliminating the [uiː] allophone of /u/, which you don't mention here, has a similar problem of leaving the table now presenting a non-distinction.--Trɔpʏliʊm • blah 11:33, 24 August 2009 (UTC)

Voiceless laterals

I see many sorces representing these as fricativs rather than voiceless approximants (eg. John Wells here: [2]). Which is right, or is there variation? --Trɔpʏliʊm • blah 11:54, 24 August 2009 (UTC)

Differing voiceless nasals

Is it really possible to distinguish [m̥ n̥ ɲ̊ ŋ̊] out of context? Wisapi (talk) 15:26, 26 January 2010 (UTC)

Do you mean to distinguish them from each other or to distinguish them from their voiced counterparts? Either way, I think the answer is yes, why wouldn't it be? +Angr 15:29, 26 January 2010 (UTC)

I meant distinguishing them from each other because in my attempts to pronounce them they only sounded like an air release through the nose. Or should air also come out through the mouth in [n̥ ɲ̊ ŋ̊]? Wisapi (talk) 19:56, 26 January 2010 (UTC)

No, like all nasal stops air should come out through the nose only. I guess the difference between them is really hard to hear if you just say them in isolation without any vowels next to them. +Angr 20:52, 26 January 2010 (UTC)

I tried with my wife (we're both native speakers) and she couldn't distinguish between my production of voiceless m and voiceless n out of context (I turned my face away from her). There are not many minimal pairs - I had to search for ten minutes before finding one - 'reimt' ("haunted") vs. 'reynt' ("tried"). Interesting question, I hadn't really thought about this before. Haukur (talk) 21:45, 26 January 2010 (UTC)

I tried a sentence pair with my wife - "Það hefur verið reimt þarna" vs. "Það hefur verið reynt þarna" - and we agreed that there was no particular problem in distinguishing them. I've thought of another minimal pair - 'lengt' ("extended") vs. 'leynt' ("secret") and I think the difference here is very slight. But that's not specific to the voiceless pairs, I think the distinction between 'lengd' ("length") and 'leynd' ("secrecy") is not made by all speakers and I'm not sure I make it myself. Haukur (talk) 22:09, 26 January 2010 (UTC)

I've tried the words "reimt" and "reynt" alone (not in a phrase) and I think I could tell them apart by the release after [m] (which sound as a very weak [b̥]). What do you think? Could saying them in a whole sentence give other clues? And what about [ɲ̊ ŋ̊]? Can you guys tell them apart (isolated or within a word)? From what I got in this article, 'lengt' and 'leynt' aren't exactly a minimal pair: 'leynt' has a <y> ([i(ː)]) which 'lengt' hasn't. Wisapi (talk) 23:17, 26 January 2010 (UTC)

You may be right about the release. The word 'lengt' is pronounced as if written 'leingt'- Icelandic orthography has a quirk about vowels before ng/nk ('banki' is pronounced as 'bánki', 'vinka' as 'vínka' etc.) I can't think of a minimal pair between [ɲ̊ ŋ̊] and I doubt one exists, but I think the same goes for the voiced variants. Haukur (talk) 00:14, 27 January 2010 (UTC)

I just wanted to add that we have a three way minimal pair since "Það hefur verið rengt þarna" (from the verb 'rengja') is a valid sentence with ŋ̊. I sometimes listen for the pronunciation of 'lygnt' and I think it is a bit off and on whether people have the ŋ̊ or whether it becomes n̥. In the latter case it sounds the same as 'lint', but I think there is an audible difference between the ŋ̊ and n̥ versions. Without any context I can believe that it is difficult to distinguish between these sounds. Stefán (talk) 11:55, 27 January 2010 (UTC)

Excellent point! Indeed, reynt-reimt-rengt is a three-way minimal pair (and of course clearly distinct from reynd-*reimd-rengd). Haukur (talk) 12:34, 27 January 2010 (UTC)

I think the term "minimal triplet" is more common than the oxymoron "three-way minimal pair". +Angr 12:44, 27 January 2010 (UTC)

Duh, yes, thanks. Haukur (talk) 13:37, 27 January 2010 (UTC)

I believe here too one can distinguish between 'lyngt' and 'lint' by the release of [ŋ̊], which sound as weak [ɡ̊] (or perhaps even a ligual egressive click) ^{No vowel-before-/ng/ quirk here?}. Should this be the case, I'd be interested in hearing of a minimal pair between [ɲ̊ n̥], for then, I imagine, the tongue would glide from [ɲ̊] to the alveolar ridge making no audible release, like [n̥]. Wisapi (talk) 16:03, 27 January 2010 (UTC)

The word is 'lygnt' (it is the nominative neuter form of the adjective 'lygn'), still the pronuciation is [lɪŋ̊t]. I could also believe that there is something at the onset of ŋ̊ which would help distinguish it from n̥ but that is speculation. As for a minimal pair between [ɲ̊ n̥] they don't exist as far as I believe. Certainly not in this environment. The highest peak in Iceland is called either Hvannadalshnjúkur og Hvannadalshnúkur but I think that the second component (hn(j)úkur) is [n̥(j)ukʏr]. Stefán (talk) 17:07, 27 January 2010 (UTC)

Do you mean that the pronunciation of 'Hvannadalshnjúkur' could be with a [ɲ̊] owing to the <j>? That's an important question because we will have to add this enviroment (<nj>) to the table then. It is also important to revise the vowels table because of the lyngt's pronunciation; I know it's not a big difference between [i] and [ɪ], but, as it stands now, it claims that the pronunciation should be [liŋ̊t]. Wisapi (talk) 00:47, 31 January 2010 (UTC)

Like I said, I think 'hnjúkur' is [n̥jukʏr], I was just wondering whether it could be [ɲ̊ukʏr] but I don't think it is. We don't need to modify the table because the word is 'lygnt'. It is *not* lyngt and the spelling does not contain the cluster ng, but rather the cluster gn. Compare 'hringt' [r̥iŋ̊t] from the verb 'hringja' (to call) with 'hrygnt' [r̥ɪŋ̊t] from the verb 'hrygna' (to spawn). Stefán (talk) 01:13, 31 January 2010 (UTC)

I understood that you were just wondering if the pronunciation of 'hnjúkur' could be something else, but as you left the question open, I was willing to draw attention to its importance, so more people may give their opinion. As to 'lygnt', I'm sincerely sorry that I bothered you again; I still hadn't noticed the different clusters. Wisapi (talk) 12:48, 31 January 2010 (UTC)

<ö> before <gi>, 'lögin' [ˈløːjɪn]

Examining the vowels table, I found something that may be a typo, or, if not, deserves more explanation: 'lögin' was transliterated as [ˈløːjɪn] without the expected <ɪ> after <ø>. Wisapi (talk) 00:20, 31 January 2010 (UTC)

It is a typo. Stefán (talk) 01:14, 31 January 2010 (UTC)

Fixed, it now stands as [ˈløɪːjɪn]. Wisapi (talk) 12:52, 31 January 2010 (UTC)

Initial <h> before consonant, aspiration or devoicing?

I think it would be clarifying to add a row for that in the consonants table. It seems to be pronounced in both ways, but further down in the vowels table hræsni was only given the pronunciation [ˈr̥aistnɪ]. Is there a prefered pronunciation? By the way, what about <hv>? I've read in a grammar (Colloquial Icelandic, p.7 http://books.google.com/books?id=MD7m_cV6KrIC&printsec=frontcover&dq=colloquial+icelandic&cd=1#v=onepage&q=&f=false) that some Icelanders consider the pronunciation [xv] to be better Icelandic. Is that outdated? Wisapi (talk) 01:08, 31 January 2010 (UTC)

I think it is fair to say that the consensus is that the initial clusters hr as in 'hrota', hn as in 'hneta' and hl as in 'hljóð' signify unvoiced sounds and not some cluster of an aspiration followed by a voiced sound. So, there are not two ways of pronouncing these words and the table should be modified accordingly. As for initial hv people pronounce it differently, either [kv] or [xv]. In the middle of the 20th century 3/4 of the population had the [kv] pronuciation and that fraction has been increasing rapidly. Some influencial people have claimed that [xv] is preferable. It is clear that they have lost this battle. Let me strongly recommend to anyone learning Icelandic to choose the [kv] since it is so much more popular. Stefán (talk) 01:34, 31 January 2010 (UTC)

Alternative transcriptions

Looking through an Icelandic grammar (Kennslubók í Nútíma Íslensku handa Ítölum by Riccardo Venturi (Rikarður V. Albertsson)), I came across a phonetic table and found some differences to the one we have here. I'll try to present all of them in the following subsections as coherent and faithful as I can, so you may understand as much as possible of this other concept. The author, his apparent precision notwithstanding, seems to have used some characters interchangeably as the ' (apostrophe) and ` (grave accent) to mark the words' initial stress (and some of them hadn't even that: "laut [löi:tʰ], háll [hau^¿d’l’]" (non-stressed words?)). Occasionally, he also presented a symbol for a phoneme but then used another way to transliterate it in that phoneme's example word. Assuming that the phones represented by the first symbols are but the implicit, theoretical phoneme's form, and not the form that exist within a word, I'll chiefly use the latter form and present the former only together with the example word. This one I'll call prototypical phoneme while reserving 'archetypical' to the phones derived from removing the diacritics of another ([t']->[t]). Wisapi (talk) 17:19, 3 February 2010 (UTC)

Voiceless or ejective

One of the things I found out was that the sounds covered here as voiceless stops [b̥ d̥ g̥] were represented there as ejective consonants [b' d' g'] and the letters <p t k, b d g> were assigned to either them or their aspirated counterparts [pʰ tʰ kʰ] in the following fashion:

---

Observations:

1. [g']/[kʰ] is followed by back vowels and [g'ʲ]/[kʲ] by front vowels, cf. next section. There's no [kʰʲ] within a word, but only as a prototypical phoneme.
2. Exception (which might be a typo): degi ['dei:ji].

That word is not an exception, the combination 'egi' is always pronunced like that. Haukur (talk) 12:04, 4 February 2010 (UTC)

Yes, the combination 'egi' wasn't mistranscribed, but the initial <d>, was. It should be [d']. Wisapi (talk) 00:57, 17 February 2010 (UTC)

---

• in initial position:

1. <b d g> were mapped to [b' d' g(ʲ)']
2. <p t k> were mapped to [pʰ tʰ kʰ/kʲ]

• in middle position:

• after a vowel

1. b -> not found
2. d -> not found
3. g -> never a stop

<gi> -> [jɪ]

before <t> or <s> -> [x]

otherwise -> [γ]

1. p -> only found in 'skipti' (see below)
2. t -> [tʰ] exception: gæta [`g’jai:t’a]

That word isn't an exception, an orthographic 't' between vowels is often aspirated in the north and usually not in the sourth. Haukur (talk) 12:04, 4 February 2010 (UTC)

1. k -> [kʲ] / (possible [kʰ]?)

before <t> or <s> -> [x]

• geminated

1. bb -> [b':]
2. dd -> [(^¿)d’(:)] : Edda -> [`ed’:a] , þíddur -> [`θi^¿d’ür] . The prototypical sound of <d> is [d’(:)].
3. gg -> ejective g "[g']", palatalised before front vowels

[g’:] -> vagga -> ['vag’:a]

[c:] -> leggja -> ['leʰg’ʲ:a]

The pre-aspiration doesn't belong there, this looks like a transcription of *lekkja. Haukur (talk) 12:04, 4 February 2010 (UTC)

1. pp -> [ʰb']
2. tt -> [ʰd']
3. kk -> [ʰg']

• after a consonant -> all six letter were mapped to [b' d' g'(ʲ)]

exception: skipti [`skʲɪb’d’ɪ]

Should be [`skʲɪfd’ɪ]. Haukur (talk) 12:04, 4 February 2010 (UTC)

note the pronunciation of <i> in: herti [`her’d’i]

• in final position (-> same rule as initial position?)

• after a vowel

1. b -> not found
2. d -> not found
3. g -> not found

1. p -> not found
2. t -> [tʰ]
3. k -> [kʰ]

• after a consonant -> presence of archetypically voiceless stops

1. b -> not found
2. d -> not found
3. g -> not found

1. p -> not found
2. t -> [t']
3. k -> [k']

I also think it's worth noting here the pronunciations I found for the words containing following clusters:

• <ll> allt [al’t’], ell [eʰ d’l’], olli [`o^¿d’lI]

núlli [`nu<sup>¿</sup>d’l:I] (archetypically: [(<sup>¿,h</sup>)d’l’], perhaps 'ell' should enter into this group), hóll [hou<sup>¿</sup>d’l’:], stóll [sd’ou¿d’l'], alla [`a^¿d’l’:a]ull [ü^¿d’l’], fýll [fi^¿d’l’], höll [hö^¿d’l’]

this should just be [`nul:I]; recent loanwords usually have [l:] rather than [tl]

Olli [`ol:I]

all the transcriptions here are ipsis litteris, including the one of 'ell' (space) and 'olli' (second 'l' is not ejective). These are probably typos.

• <nn> inni [`In:I], enn [en:], brenna [`b’ren:a], einn [ei^¿d’n’]

enginn [`éiη_jgʲIn] (archetypically: [η]; acute accent reprisenting tone?)

• <n>langa [`lauη'g’a], henti [`hen’d’I]

• <fn> svefn [sveb’n’]
• <m> [m(’)(:)] kempa [`kʲem’b’a]
• <x> [xs] lax [la:xs] (the pronunciation [ks] was not listed. Does it exist only in loanwords?)

--Wisapi (talk) 17:19, 3 February 2010 (UTC)

No, it exists all over the place. Most people born after ca. 1950 use [ks], most people

born before that use [xs]. My wife makes fun of me for using [xs]. Haukur (talk) 12:04, 4 February 2010 (UTC)

[kʲ g’ʲ] instead of [cʰ c]

Both archetypical pronunciations, amongst others, are given:

• for <g>

1. [gʲ] -> gef -> [gʲe:v]
2. [c:] -> leggja -> ['lehg’ʲ:a]

but only [gʲ] is used:

gæta [`g’ʲai:t’a]

gætti [`g’ʲaiʰd’ɪ]

• for <k>

1. [k’ʲ] -> kæla -> [`kʲai:la]
2. [c] -> reki -> [`re:kʲI]

but only [kʲ] is used:

kæla [`kʲai:la]

reki [`re:kʲI]

kempa [`kʲem’b’a]

poki [`pʰo:kʲI]

skipti [`skʲIb’d’I]

Further examples including the cluster <ng>/<nk>:

anga [`auηg’a]

Ingi [`iηjg’jI]

löng [löiη’k’]

enginn [`éiη_jgʲIn]

ungur [`uηg’ür]

þyngð [θiηɣð]

banka[`b’auη’g’a]

And another surprising phonetic value for <g> that I hadn't found here:

• [g’v] -> Guð -> [g’vüð]

--Wisapi (talk) 17:19, 3 February 2010 (UTC)

I am not sure if [v] is the correct transcription, but I definitely do recognise a pronunciation along those lines in this word ([k_wY:T]?), and I also seem to hear a different /l/ than the usual one between á and v in words like hálfur and álfur. I'd like some confirmation here. Oh, and the /a/ part of the á diphthong doesn't sound like the /a/ in a normal a to me, either. Skomakar'n (talk) 22:46, 23 July 2012 (UTC)

Vowels

• o -> [o] [wo:] kom [kho:m] [khwo:m]
• e -> [e(:)] [ei] er [e:r], ferð [ferð]
• é -> [je] él [je:l]

--Wisapi (talk) 17:19, 3 February 2010 (UTC)

Comments on the above

It looks like this book has some idiosyncratic transcriptions that don't follow IPA exactly. That's not unusual; most works describing the phonology of a single language have their own conventions that deviate in some way or another from IPA principles. For example, does the author ever explicitly call his "[b' d' g']" ejective? It's far more likely that he's just following a convention that the apostrophe indicates voicelessness in that book. The same goes for not indicating stress: many authors omit stress marks on monosyllabic words because it's obvious where the stress goes. Whether one uses the symbols for palatalized velars (kʲ etc.) or the symbols for palatals (c etc.) is also a matter of convention; as long as the language you're describing doesn't contrast the two, it doesn't much matter which set of symbols you use. (Over at Scottish Gaelic phonology, we use the symbol ç to stand for a voiceless postpalatal or prevelar fricative but the symbols ʎ and ɲ to stand for a postalveolar or prepalatal lateral and nasal, even though ç, ʎ and ɲ are all in the same "Palatal" column of the IPA.) +Angr 18:05, 3 February 2010 (UTC)

Finally having some time to read a bit further in that grammar, I found out that what was meant by these apostrophes after [b d g] is that they are desonorizzate (dessounded), i.e. the vocal cords do not cease their vibrations before the onset of the articulation, but after. The author mentioned that he could could also have reprisented them by [b̥ d̥ g̊], but preferred the apostrophy. I think it would be a good idea to have this last set of symbols used here with the proper explanation of how they differ to simply voiceless consonants. Wisapi (talk) 02:47, 6 June 2010 (UTC)

However, I just read, in final positions or after certain consonantal clusters, these desonorizzate consonants will be truly voiceless, i.e. the vocal cords will stop vibrating before their onset. These are marked in this grammar by [pʼ tʼ kʼ]. Wisapi (talk) 12:39, 6 June 2010 (UTC)

Dialectical variation

I highly recommend mallyskur.is for pronunciation samples. Haukur (talk) 12:07, 4 February 2010 (UTC)

Are the recordings freely licensed? If so, we can copy them to Commons and use them directly. +Angr 13:01, 4 February 2010 (UTC)

The site doesn't really say. I know the people behind it and I could try to brooch the subject but I would expect the usual hemming and hawing, discomfort and incomprehension. In any case, licencing would be complicated by the fact that some of the recordings are of people reading recent works of literature aloud. Some of it is just spontaneous talk which wouldn't be subject to copyright - but neighboring rights would come into play. Haukur (talk) 13:58, 4 February 2010 (UTC)

P.S. I assume you mean "dialectal variation", not "dialectical variation". +Angr 13:03, 4 February 2010 (UTC)

When I see Angr writing in <small></small> tags I know I've made a boo-boo. Haukur (talk) 13:58, 4 February 2010 (UTC)

Would you rather I point out your mistakes in great big letters? +Angr 17:39, 4 February 2010 (UTC)

problems with phonemes

Hi. I added a tag questioning the factual accuracy of the consonant section because it appears to confuse phonemes with allophones. Although I don't know Icelandic phonology very well, it appears that the following, at least, are not phonemes: velar laterals, velar nasals, palatal nasals, velar fricatives, glottal stop. It's also arguable whether either palatal stops or fricatives are phonemes (probably better analyzed as /k/, /g/ or /h/ followed by a front vowel or /j/); likewise for the voiceless resonants (predictably voiceless in some situations, sequences of /h/+resonant in other situations). The voiced dental fricative is likewise not a phoneme, as the article itself indicates.

It would be much better to clarify this situation, e.g. by enclosing the non-phonemes in brackets and the questionable phonemes in parens.

Benwing (talk) 04:05, 20 April 2010 (UTC)

Perhaps someone could quote the material from Scholten (2000). I wouldn't be surprised if there was, as the paragraph below the chart describes, some debate as to the phonemicity of certain sounds. — Ƶ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 06:53, 20 April 2010 (UTC)

I don't have Scholten but I have "The Germanic Languages", with a chapter on Icelandic. It has a table somewhat similar to the one here, described as "The Modern Icelandic consonant system". The important thing about the table in this book is that (a) it's never described as a table of phonemes, and (b) the book explicitly says:

The system represented here is not strictly phonemic in the classical sense nor is it the system of underlying segments in the generative sense since in contains a number of segments that are predictable (for the most part at least) in terms of their environment. This is true for the alternation between palatal and velar stops, on the one hand, and voiced and voiceless sonorants, on the other.

I wonder if the table in Scholten isn't similar, in that it's a table of major allophones, not of phonemes.

I also wonder if some of the reason for this unwillingness to commit to a phonemic representation is due to the fact that simplistic phonological theories have difficulty with a sound that functions as the allophone of two different phonemes. For example, the "voiceless r" allophone is an allophone of /r/ before /p,t,k,s/ and word-finally after voiceless consonants; but it's also the common pronunciation of the cluster /hr/.

Benwing (talk) 08:12, 20 April 2010 (UTC)

<u> in "-num"

Hi! I found in the Icelandic grammar "Kennslubók í Nútíma Íslensku handa Ítölum by Riccardo Venturi (Rikarður V. Albertsson)" that <u> is pronounced like an <o>. Unfortunately, the author uses the same character for both [o] and [ɔ], so I don't to which he's actually referring. The examples he gives are "dölum" ([ˈdʼöːlom]) and "himnunum" ([ˈhimno:nom]). Here, however, we have that this sound ([ɔ]) is only found in in the first <u> of the ending "-unum", as in "augunum" [ˈøɪːɣɔnʏm]. Wisapi (talk) 22:08, 5 June 2010 (UTC)

götva [ˈkœːtva]

Shouldn't this be pronounced [ˈcœːtva], since <k> is followed by a front vowel [œ]? Wisapi (talk) 23:44, 5 June 2010 (UTC)

No, that rule only applies to unrounded front vowels. – Krun (talk) 10:35, 20 July 2010 (UTC)

As it turns out, the "rounded front" vowels are actually rounded central vowels anyway. - Gilgamesh (talk) 06:36, 7 May 2021 (UTC)

hver and compounds

I've just hit upon this particularity: according to this Kennslubók, 'hver, 'einhver' and 'sérhver' "should actually be pronounced -[kʰvʏr] ([kwʏr])" (actually the author uses [ü] instead of [ʏ]). Suspecting this would be a prescripted —and not descriptive— pronunciation, I was also puzzled by the fact that the auther didn't then prescript the [xvʏr] pronunciation. Wisapi (talk) 00:37, 6 June 2010 (UTC)

aspirated consonants in non initial positions

Still according to this grammar there actually are aspirated [k p t] in middle and final positions. Examples vaka [ˈvaːkʰa], vík [viːkʰ], skapa [ˈsgʼaːpʰa], glöp [gʼlœːpʰ], gráta [ˈgʼraːtʰa] (I guess there was a typo here with the diphthong), net [neːtʰ]. There is even a whole topic about this in the grammar and before that the author had already used transcriptions with aspirated in other positions. Wisapi (talk) 02:19, 6 June 2010 (UTC)

IPA table

I've set up WP:IPA for Icelandic to go along with the {{IPA-is}} template. Needs to be reviewed, esp. the L's.

Also, I changed the transcription of long diphthongs. I assume what was meant was that the entire diphthong was lengthened, but [ai:] makes it look as though it's the off-glide that's lengthened. It's pretty standard to put the length mark on the peak of sonority. — kwami (talk) 05:27, 11 November 2010 (UTC)

Glottal stop?

The glottal stop is listed in the consonant table but is not commented in the text. I'd like to see a source on that and examples of word containing it. Icelandic being a Germanic language, my gut instinct is to be very doubtful about its phonemicity. - Dingbats (talk) 10:35, 14 December 2010 (UTC)

Vowel length

Why is there a separate section "Vowel length" now when the issue is already explained in a much more detailed way in the preceding section? The section "Vowel length" would seem redundant then. Not to mention that the detailed explanation names a source, while the "Vowel length" section is completely unsourced. --Florian Blaschke (talk) 03:08, 27 June 2011 (UTC)

This whole article is a mess. If you want to fix it, go for it. — Ƶ§œš¹ _{[aɪm ˈfɹ̠ˤʷɛ̃ɾ̃ˡi]} 03:40, 27 June 2011 (UTC)

Final devoicing, anyone? (re-post from orthography talk page)

I'm very surprised to find that the description in the article not only ignores but implicitly denies the obligatory phrase-final voiceless pronunciation of the fricatives ([v],[ð],[ɣ]), even though the recorded pronunciation examples clearly display it. Árnason (2011) certainly does mention final devoicing of fricatives, and, less consistenly, of sonorants (p.237), as a "pervasive postlexical effect that takes place before a pause", using the example of dag pronounced as [ta:x].--91.148.130.233 (talk) 02:13, 7 December 2014 (UTC)

Right, he also mentions final devoicing on p.107 and gives the examples of bauð pronounced as [pœyːθ] (cf. bjóða [pjouːða]) and gaf as [kaːf] (as opposed to gefa [cɛːva]) (p.107).--91.148.130.233 (talk) 20:37, 14 December 2014 (UTC)

[s]

It seems to me that this is most often realized as [s̠], even perhaps [ʂ]/[ʃ] when followed by [k] or [j] respectively:

• skulu [ˈs̠kʏ.lʏ] ~ [ˈʂkʏ.lʏ]
• sjálfur [ˈs̠jau̯ɬ.fʏr] ~ [ˈʃjau̯ɬ.fʏr]
• skilja [ˈs̠cɪlʲ.ja]
• sál [s̠au̯ɬ]
• lexía [ˈlɛk.s̠i.a]

Are there any thoughts on this? Thadh (talk) 14:02, 27 August 2020 (UTC)

Recent edits

I recently added a lot of new material to this article, sourced from at least two different newer references. But I am not an expert on the subject, so it only seems responsible to generally ask for my edits to be looked over and perhaps details corrected, etc. I did footnote the references by document, but not by page or using citation templates. (I am so out of practice with that.) Fortunately the references themselves aren't that massive, the longest being perhaps a couple dozen pages. - Gilgamesh (talk) 20:36, 23 April 2021 (UTC)

Problematic IPA for diphthongs, especially long diphthongs

I appreciate that many of the source references use this same notation, but using [aiː auː eiː œiː ouː] as-is has always seemed like a barrier to quick reading, as the symbol ⟨ː⟩ is generally used for only the immediately preceding symbol. We know that the notation actually means [a͡iː a͡uː e͡iː œ͡iː o͡uː], where a unitary diphthong is phonetically longer than it otherwise would be expected to be. But superficially, to those not familiar with Icelandic broad IPA, the mind is thinking the monosyllabic equivalent of a lopsided [aii auu eii œii ouu], or worse, parsing it as implied hiatus [a.iː a.uː e.iː œ.iː o.uː]. I've also seen the sporadic usage of [aːi aːu eːi œːi oːu] instead, but this isn't necessarily better because the mind will still tend to see it as either lopsided monosyllabic [aai aau eei œœi oou], or still imply hiatus [aː.i aː.u eː.i œː.i oː.u] because it's not common in IPA conventions to use ⟨ː⟩ between vowels where hiatus isn't implied. We know what it actually means is that the whole diphthong as a unitary phone is pronounced audibly longer than not, and where potentially ambiguous, it is conventional to indicate unitary articulations with a tie bar anyway. I know it is common on Wikipedia to do away with tie bars to simplify the transcription, but in this case omitting the tie bar only increases the potential for confusion. Compare conventions for Polish IPA, where cz /t͡ʂ/ and trz /tʂ/ are not strictly homophonous phoneme sequences and the tie bar (or lack thereof) makes a difference both in reading and in imparting the crucial phonological information. And though I appreciate that sequences of vowel syllables without a consonant between them appear to be relatively uncommon in Icelandic, they can still occur, like in Skógafoss [ˈskouːaˌfɔss]. Someone with a more well-read understanding of the phonology can still infer where the hiatus is and isn't in that transcription, but to someone less familiar with it, it can seem to imply [ˈsko.uː.aˌfɔss]. A hiatus symbol ⟨.⟩ could still be used [ˈskouː.aˌfɔss], but it just seems clearer to use a tie bar where adjacent vowel symbols form a diphthong, whether long or short [ˈsko͡uːaˌfɔss]. I propose making this tie bar for Icelandic diphthongs conventional practice on Wikipedia in the interests of clarity, especially for those who want to try to pronounce individual Icelandic terms at a glance with generic IPA knowledge, but haven't (at least yet) fully read any more in-depth articles on Icelandic phonology. This is not a simple matter of trivia, either, because Icelandic terms can be very difficult to pronounce untrained, and most articles with Icelandic IPA don't have corresponding English-specific pronunciations (IPA or otherwise), and the Icelandic IPA should ideally serve to aid rather than hinder attempts to pronounce it. - Gilgamesh (talk) 13:01, 24 April 2021 (UTC)

Three levels of phonological representation?

It occurs to me that what this article currently describes as phones and phonemes, are actually closer to phonemes and morphophonemes respectively. Otherwise, an appropriately phonetic transcription would use diphthongs like [ɪ͡ɛː] instead of [ɛː], and might use true central vowel symbols like [ɵ̞ ʉ̞] instead of front vowel symbols like [œ ʏ]. I know that can theoretically be numerous levels of phonetic nitpicking in IPA transcription and that it's unusual to recognize more than two levels (phonemic vs. phonetic), but Icelandic is already a relatively unusual language in this regard, and there is precedent in recognizing three levels where it makes phonological sense, like in Korean phonology with its ⫽morphophonemic⫽, /phonemic/ and [phonetic] levels of transcription. This is because, like Icelandic, Korean words can attempt to cram a lot of morphophonemic information that can end up simplified or otherwise modified in complex ways, meaning that there's a raw version (morphophonemic), the abstract articulation of sounds in practice (phonemic), and the way they actually sound in speech (phonetic). If this article's "phonemic" level of Icelandic phonology has as much of a gulf as its does between the language's "phones" (⟨ɡ⟩ vs. ⟨k ~ ɣ ~ c ~ j ~ ˠ⟩), and if the phones themselves as indicated still use fossilized abstract symbols that can stray significantly from how the language is actually spoken (⟨a, ai, au, ɛː, œ, ʏ⟩ vs. ⟨ä, a͡i, ɑ͡u, ɪ͡ɛː, ɵ̞, ʉ̞⟩, and one of the references did suggest ⟨ɐ⟩ instead of ⟨ä⟩), then the "phones", as not perfectly phonemic as they are, are not actually phonetic, either, are they? I know that, to some degree, it may be considered original research to refer to the phonemes and phones instead as morphophonemes and phonemes if there isn't accepted precedent for this in the reliable references. However, adopting the convention for Korean IPA on Wikipedia, we could still use the morphophonemic and phonemic IPA bracketing for ⫽phonemes⫽ and /phones/, respectively, and only use [square brackets] for more narrowly phonetic descriptions.

Also, I've found at least two references that specifically describe /œ ʏ/ as central vowels rather than front vowels. It would be interesting to know how long ago these symbols were adopted for Icelandic IPA convention. I know that many of the cardinal central vowel symbols (including ⟨ɞ ɵ⟩) weren't even part of the IPA at all until 1993, and it helped remedy the previous situation where fully front ⟨œ ø⟩ or fully back ⟨ɔ o⟩ vowel symbols kept being used for central vowels, perhaps with centralization diacritics ⟨œ̈ ɔ̈ ø̈ ö⟩ if clarity was needed, but not the most ideal situation when some languages could have all three backnesses of vowels in their inventory, as languages like Swedish (or indeed Icelandic) do. Considering that the true front vowels trigger the palatalization of velar consonants, while the true central vowels do not, it seems misleading to use front rounded symbols for central rounded vowels when frontness actually seems to actively matter in the language's living phonotactics. For the time being, conventions for transcribing the phones are what they are, and absent some appropriate references, we probably shouldn't consider changing the primary symbols for phones. But all that is my observation. - Gilgamesh (talk) 15:30, 24 April 2021 (UTC)

Well said. I noticed that under the Monophthongs table, one bullet point says that /ʏ/ is phonetically near-front [ʏ], which matches the chart, but another bullet point says that /ʏ/ is phonetically central [ʉ̞]. It's an example of the confusion you describe. As it is, those bullet points are in direct conflict with one another. A clarifying note at least would be helpful. Damangio (talk) 04:03, 29 April 2021 (UTC)

This article in general needs a lot of help and clarification, and it's obviously not getting enough attention or review. I'd try my hand at rewriting it with references, but I am not an expert. I don't know anyone who is. That seems to be a significant problem. And it seems like no trivial matter given how complicated Icelandic phonology and phonotactics appear to be.

There's also the apparent inconsistency that, from the references I have read, the phonetic onsets for mid short monophthongs vs. mid diphthongs appear to be opposite of their prescribed symbols. It's already established in the article that /ɛː/ is actually more like [ɪ͡ɛː]. But in Flego-Berskon (2020), short /ɛ/ (along with the other mid vowels) has a true-mid articulation [e̞], but /ei/ (along with the other mid diphthongs) as relatively more open-mid articulation [ɛi]. Though it doesn't, on the other hand, show the onsets of the open diphthongs /ai au/ as front [a͡i] (or [æ̞͡i]) and back [ɑ͡u] respectively, though we know to some degree this has to be the case because /ai/ triggers palatalization like an entirely front vowel while /au/ does not, so I'm not 100% certain how reliable that table is. It seems mostly truthy, though. Also, though I can't read Danish, it seems like the vowel chart in Volhardt (1988) (which the article uses for its own chart) is even less accurate, positioning the central vowels as more or less fully front, but also apparently illustrating them in comparison specifically with Danish vowels. Besides there being other references for Icelandic vowels that are actually in English, I'm not sure how verifiable a Danish reference can be if editors can't read it.

If it were me deciding the symbols (which it's not, but if it was), I would do vowel symbols like this:

	Front onset	Central onset	Back onset
Close onset	i		u
Near-close onset	ɪ	ᵿ • (ᵿ͡i)
Mid onset	e • e͡i	ɵ • ɵ͡i	o • (o͡i) • o͡u
Open onset	æ͡i	a • (a͡i)	ɑ͡u

Using common clear "default" mid symbols for all mid vowels, and clear front, central (or central-compatible) and back symbols for all vowels with those onsets. I know /ᵿ/ is not officially an IPA symbol (it's more an Oxford English Dictionary symbol for a variable archiphoneme whose neutral position is in that spot), so /ʉ/ could be used instead. Why separate /æ͡i a͡i/? Because one triggers palatalization, and the other does not, which at least is borne out by the noticeable differences in vowel onset quality with Google Translate's Icelandic text-to-speech function for gæi 'guy' vs. generic fragment gæ. Though the fact that no reference I've found actually describes separate /æ͡i a͡i/ onsets (just the preceding consonant being velar or palatal) would not only make this rather original research (more so than the rest of this paragraph), but also further imply that the distinction between /æ͡i a͡i/ (as with non-phonemes [ᵿ͡i o͡i]) is more strictly allophonic on the vowel level itself, with the true morphophonemic representation being closer to ⫽ʲa͡i a͡i⫽ (notice the double-slash morphophonemic bracketing), with the two only contrasting after velar consonants. And perhaps even ⫽ʲa͡i⫽ can be reduced to the two-phoneme sequence ⫽ja͡i⫽ with the conditional understanding that it only normally implicitly exists like this after velar phonemes ⫽kja͡i ɡja͡i⫽ in native words. After all, though there is certainly more than one source that describes this contrast on the quasi-phonetic level, I haven't yet seen specific notation describing it on the most basic morphophonemic level, but it would most likely seem to be ⫽kja͡i ɡja͡i⫽ vs. ⫽ka͡i ɡa͡i⫽. (All right, it's official, this paragraph is firmly in the territory of original research. But on to the next paragraph.)

As for consonants, there are non-OR alternative phone (= middle phoneme) symbols that are used in the references that I would probably prefer to have used, like purely approximant symbols /ʋ ð ʝ ɣ/, with the understanding that they are implicitly /ʋ ð̞ ʝ˕ ɣ˕/, and that they can allophonically be relaxed as approximants or constricted to full fricatives [v ð ʝ ɣ]. I don't suggest /j ɰ/ on the phone level (though maybe ⫽j⫽ is probably okay on the morphophoneme level) because not only does /ɰ/ never seem to appear in any of the references, but these symbols tend to imply specific levels of rounding and closeness, whereas the phones they describe don't seem to be strictly limited in those terms. This seems to be evidenced by Eiríkur Rögnvaldsson (2020) treating coda /ʋ ɣ/ as optional (rather than obligatorily deleted) after /u/, which would theoretically be possible if both /uʋ uɣ/ are labiovelarizing [uʋʷ uɣ˕ʷ] and coalescing into [u(w)], though actually saying that last part in the article might be a wholly unnecessary OR leap too far.

Okay, self-indulgent phonology moment over.

Ultimately, we have plenty of references, some of them quite recently published, and finding a way to clean up this article would take the usual consensus process. Such a thing is very difficult when it's so hard to find even one other active editor to work on it. Without it, this article seems fated to well-meaning but haphazard edits that may fail to accurately describe the language and worsen the situation. In my experience, that problem often arises when a language has so few actual speakers in the world. The scope of the problem was even more apparent when I was working on Marshallese language where the language is in vigorous use but most native speakers of the language don't have internet at all (geographically-constrained higher cost of internet connection vs. lower average income), but that can't be said of the language of such a relatively affluent, well-connected country like Iceland, yet even with Icelandic it's still proving surprisingly hard to find expert active editors on the subject. - Gilgamesh (talk) 07:01, 29 April 2021 (UTC)