Danish vs. English

[Viktor77]

Quick question really, which do you think is more unphonetic, Danish or English? They're both pretty bad but after analyzing their phonologies I can't determine which is more unphonetic.

*Edit* Clarify: Which language has the more phonemic orthography?

[Åge Kruger]

Quick question really, which do you think is more unphonetic, Danish or English? They're both pretty bad but after analyzing their phonologies I can't determine which is more unphonetic.

What does this question even mean?

[Viktor77]

Quick question really, which do you think is more unphonetic, Danish or English? They're both pretty bad but after analyzing their phonologies I can't determine which is more unphonetic.

What does this question even mean?

From an unbiased pov, which language is worse when it comes to matching one sound with one letter? It doesn't matter what the sound is, but that each letter corresponds to only one sound?

[finlay]

"Which has the more phonemic orthography?"

[Viktor77]

"Which has the more phonemic orthography?"

Yes, my apologies. My linguistic terminology is not up to snuff.

[Rory]

This isn't a "what do you think" question, it's an empirical question. We can find out. However, there are a couple of things to consider.

First of all, we need to establish a way to measure how phonemic an orthography is. We could use a binary system, and say that for every word that has a 1-to-1 correspondence between the letters and phonemes, that's "phonemic", and every word that doesn't is "non-phonemic". In that case, though, do we care about consistency in representation? Does it matter that c is used for different phonemes in English?
Another way of doing it is making a model that predicts the pronunciation of a given word based on the spelling, and then compare the modelled pronunciation with the actual pronunciation. We can use edit distance as a measure of divergence. But in English orthography (maybe Danish too), often there are competing principles for pronunciation - could we make a non-deterministic model that predicts several outcomes, and gives a probability to each possible pronunciation? The probability would then have to scale the edit distance score in order for us to make an accurate comparison.
When we move to generalize our measure from single words to across the whole lexicon, other considerations arise. Should I just grab an English dictionary and compute the values for every word, and then average it? Or should I weight each word's score by its lexical frequency? What about derived or inflected words? American English spelling is less phonemic if you take into account morphological derivations. Do we want to exclude or include them?

So, this is a complex question, although it's one that we can solve through the use of online dictionaries and corpora. Not a matter of opinion.

(Another problem is we don't even really know what vowel phonemes Danish has, but that's a different issue.)

[alice]

More generally, is it possible to establish some sort of measurement of the correlation between phonology and orthography for any given language? Obviously, it would need to run in both directions.

[finlay]

"Which has the more phonemic orthography?"

Yes, my apologies. My linguistic terminology is not up to snuff.

Also, be careful in the sense that no language has a "phonetic" orthography, otherwise we'd have to have three different letters for each of p/t/k - one for the beginning of a word, one after s, one at the end of a word (etc). I think the only possible exception are those languages in the Caucasus that have only two or three vowel phonemes, assuming you subscribe to that theory in the first place, of course.

[Åge Kruger]

(Another problem is we don't even really know what vowel phonemes Danish has, but that's a different issue.)

We don't even know what vowels English has, for that matter. What is the canonical pronunciation of English?

[finlay]

That's a good point, actually, because most accents have little differences between which groups of words are lumped together as one phoneme, and sometimes big ones.

We can take a really simple common, regular and wide-reaching example: American and Scottish accents tend to be rhotic and have no TRAP-BATH split, whereas RP and other southern accents are non-rhotic and have the TRAP-BATH split. This means that for RP, the sets BATH, START and PALM are the same and TRAP has its own set, but in GA/SSE, TRAP and BATH are the same set and START is another. PALM is yet another in GA (merging with LOT and possibly THOUGHT), but merges with TRAP and BATH is SSE.

Now, which has the most phonemic orthography? I mean, they all have the same orthography but it represents different phonemes, right? So RP has things like ‹ar› giving /ɑ:/, and you can't predict from the spelling 100% whether something's /ɑ:/ or /a/ if it's ‹a› (you have to know what set it's in, BATH/PALM or TRAP). In GA, you don't know from the phoneme /ɑ/ whether the spelling is ‹a› (PALM) or ‹o› (LOT), but at least you don't have ambiguity with START, which is pretty consistently ‹ar›. It's probably some kind of personal bias, but I'd give it to Scottish English just because at least you pretty much know that if something's spelt with ‹a› it's /a/ and if it's spelt with ‹ar› it's /ar/.

And that's just the regular ones... never mind the irregular spellings which are spelt with ‹a› but just belong to another set entirely. Like 'warm' or 'walk' or 'talk'...

[Io]

How are p/t/k different after /s/? :S

[Radagast]

I don't think it can be made up - both languages have bastardized orthographies where foreign loanwords are accepted in their original spelling. This means that there are theoretically infinite spellings for every segment in both languages. Otherwise english would win by default because it has a larger number of loanwords with weird spellings than danish.

[Legion]

How are p/t/k different after /s/? :S

Loss of aspiration.

[Viktor77]

This isn't a "what do you think" question, it's an empirical question. We can find out. However, there are a couple of things to consider.

First of all, we need to establish a way to measure how phonemic an orthography is. We could use a binary system, and say that for every word that has a 1-to-1 correspondence between the letters and phonemes, that's "phonemic", and every word that doesn't is "non-phonemic". In that case, though, do we care about consistency in representation? Does it matter that c is used for different phonemes in English?
Another way of doing it is making a model that predicts the pronunciation of a given word based on the spelling, and then compare the modelled pronunciation with the actual pronunciation. We can use edit distance as a measure of divergence. But in English orthography (maybe Danish too), often there are competing principles for pronunciation - could we make a non-deterministic model that predicts several outcomes, and gives a probability to each possible pronunciation? The probability would then have to scale the edit distance score in order for us to make an accurate comparison.
When we move to generalize our measure from single words to across the whole lexicon, other considerations arise. Should I just grab an English dictionary and compute the values for every word, and then average it? Or should I weight each word's score by its lexical frequency? What about derived or inflected words? American English spelling is less phonemic if you take into account morphological derivations. Do we want to exclude or include them?

So, this is a complex question, although it's one that we can solve through the use of online dictionaries and corpora. Not a matter of opinion.

(Another problem is we don't even really know what vowel phonemes Danish has, but that's a different issue.)

You are completely right of course. From my perspective these issues are rarely brought up as I've grown up with media which just assumes some equation for calculating phonemic value. For example, if you read literature on dyslexia rates, it often makes reference to the phonemic value of English and how that may be a cause of increased dyslexia rates in the USA (a question I'd like to have compared with the rates in Denmark, but that's a different subject altogether). It hardly even fazes me that phonemic values as a whole are a complex question. I ought not to have asked this question for its simply too vague and too simple. I appreciate all of your responses.

[Io]

Legion, they're not aspirated word-finally either, i think.

[Legion]

Legion, they're not aspirated word-finally either, i think.

Yes, but you were asking how they were different after /s/, and that's how. English p/t/k are sapirated syllable initially, so not after s nor in coda position. But in coda position they are not pronounced the same than in onset after s, due to being… in coda position. This difference is true for any language though.

[finlay]

This isn't a "what do you think" question, it's an empirical question. We can find out. However, there are a couple of things to consider.

First of all, we need to establish a way to measure how phonemic an orthography is. We could use a binary system, and say that for every word that has a 1-to-1 correspondence between the letters and phonemes, that's "phonemic", and every word that doesn't is "non-phonemic". In that case, though, do we care about consistency in representation? Does it matter that c is used for different phonemes in English?
Another way of doing it is making a model that predicts the pronunciation of a given word based on the spelling, and then compare the modelled pronunciation with the actual pronunciation. We can use edit distance as a measure of divergence. But in English orthography (maybe Danish too), often there are competing principles for pronunciation - could we make a non-deterministic model that predicts several outcomes, and gives a probability to each possible pronunciation? The probability would then have to scale the edit distance score in order for us to make an accurate comparison.
When we move to generalize our measure from single words to across the whole lexicon, other considerations arise. Should I just grab an English dictionary and compute the values for every word, and then average it? Or should I weight each word's score by its lexical frequency? What about derived or inflected words? American English spelling is less phonemic if you take into account morphological derivations. Do we want to exclude or include them?

So, this is a complex question, although it's one that we can solve through the use of online dictionaries and corpora. Not a matter of opinion.

(Another problem is we don't even really know what vowel phonemes Danish has, but that's a different issue.)

You are completely right of course. From my perspective these issues are rarely brought up as I've grown up with media which just assumes some equation for calculating phonemic value. For example, if you read literature on dyslexia rates, it often makes reference to the phonemic value of English and how that may be a cause of increased dyslexia rates in the USA (a question I'd like to have compared with the rates in Denmark, but that's a different subject altogether). It hardly even fazes me that phonemic values as a whole are a complex question. I ought not to have asked this question for its simply too vague and too simple. I appreciate all of your responses.

The interesting thing here is that I have at least one very dyslexic friend who says that if he'd been Chinese he thinks there'd have been less of a problem, because what he has had to do with English words is learn each one as a unit. Like, it's difficult for him to imagine each word as a series of sounds and break it down that way. Another said once that he has to read every letter individually, though, so I think the moral of the story is that dyslexia's not monolithic and is technically a bunch of related disorders, IIRC.

[Io]

Legion, I know they're aspirated syllable-initially, I was asking how they're different after s, and when I'm told they just aren't aspirated I normally wonder why after s needs special mention in addition to word-final or any other position lack of aspiration.

[sangi39]

For some English speakers, in addition to /p t k/ being aspirated (beginning of words or stressed syllables) or unaspirated (after /s/ or word-finally), they have these three sounds as unreleased voiceless plosives in word-final position. I'm not sure such a distinction exists for /p/ and /k/ in my variety of English but there is a 3-way distinction for /t/, i.e. [t_h]~[t]~[?] appearing word-initially in the case of [t_h], after /s/ for [t] and word-finally for [?]. There's a bit more going on as well since I've also got [r\]~[4] as an intervocalic allophone of /t/ (even across word boundaries) which turns up most often when talking to family (thank you Yorkshire, although the [4] variation seems to be an extension of that feature with US influence) otherwise it'll be, more often than not, [?].

But yeah, I may be wrong but Legion seems to be talking about the unreleased final voiceless plosive thing.

[Richard W]

... I was asking how they're different after s, and when I'm told they just aren't aspirated I normally wonder why after s needs special mention in addition to word-final or any other position lack of aspiration.

They get special mention because the voicing contrast /b/ v. /p/ is neutralised after them syllable-initially. Life gets complicated with plastic Aztec ASDIC.

[Nortaneous]

They get special mention because the voicing contrast /b/ v. /p/ is neutralised after them syllable-initially. Life gets complicated with plastic Aztec ASDIC.

I don't think it's entirely neutralized. IMD those would be something like [ˈpʰɫæsd̥ɪk̚ ˈæzˌtʰɛk̚ ˈæsˌd̥ɪk̚].

[Richard W]

They get special mention because the voicing contrast /b/ v. /p/ is neutralised after them syllable-initially. Life gets complicated with plastic Aztec ASDIC.

I don't think it's entirely neutralized. IMD those would be something like [ˈpʰɫæsd̥ɪk̚ ˈæzˌtʰɛk̚ ˈæsˌd̥ɪk̚].

That's where it's complicated. Note that in these cases the cluster is not syllable-initial.

[finlay]

Legion, I know they're aspirated syllable-initially, I was asking how they're different after s, and when I'm told they just aren't aspirated I normally wonder why after s needs special mention in addition to word-final or any other position lack of aspiration.

They're usually glottalised and/or ejective in word-final position. So many English speakers have ejective finals it's unbelievable. Technically I can't speak for America right off (IIRC they're more likely to have unreleased stops), but you just try watching "The Weakest Link" and count the number of [pæŋkʼ]s you hear (my phonetics teacher knew a guy who did a study on this). In connected speech it's more difficult to tell, and I don't have a sample right here to point any out. I'm fairly sure my final plosives are not usually plain unaspirated.

Then there's that whole thing where English /b d g/ tend to be unaspirated voiceless. They do. It calls into question whether it's really /spɪn/ in the word 'spin' or if it's actually /sbɪn/, which would end up as [spɪn]. One thing that annoys me a bit is that [b̥] notation. Why they can't just write [p] is beyond me. I've only heard weak justification for it.

Oh yeah, and I have something similar to Nortaneous at least for the relevant part - [ztʰ] in Aztec. I've no idea what an ASDIC is, although it's making me thing of Asda (our version of Walmart, kinda - at least it's owned by Walmart), so I'm thinking [zd]. I also have [spʰ] in Passport. I might be thinking about that one too much, but it might be to do with it being two morphemes so I have a mental boundary between the two... i dunno.

[Nortaneous]

I also have [spʰ] in Passport. I might be thinking about that one too much, but it might be to do with it being two morphemes so I have a mental boundary between the two... i dunno.

I think I have that also. And [stʰ] (well, more like [st̚ tʰ]) in "sixteen".

[Io]

Thanks Finlay.

[pæŋkʼ] should be 'bank' right? I've been paying attention to English /b/ ever since you brought that up a while ago and I really don't hear it as [p], maybe there is some devoicing word-initially but it certainly isn't [p].