How many phonemes are in the english language?

[scratch123]

http://www.antimoon.com/forum/2005/6356.htm

This was the best link I could find on this topic and there are a bunch of conflicting answers.

http://semarch.linguistics.fas.nyu.edu/ ... /index.txt

This is somewhat related but deals with syllables instead which is more what I had in mind when I was searching for information about this. Do you guys think that this list looks complete?

[Lazar]

Phonemes, not phenomes. As other sources indicate, the number depends on dialect, and even on interpretation, so there's no one solid number that we can point at.

[RebeccaRGB]

I've counted 22 consonants (p b t d k g m n ŋ f v θ ð s z ʃ ʒ h ɹ j l w) and 12 vowels (i ɪ e ɛ æ a ə ʊ u o ʌ ɔ), but again, dialects.

[Iulus Cofield]

Don't diphthongs usually get counted as separate phonemes, at least in English? I think analyses of Greek and Japanese usually don't, but those are somewhat unusual cases, yeah?

[Makri]

Huh? e vs. ɛ in (any kind of) standard English?

[eSOANEM]

In my English (RP), [e] only appears in diphthongs (or in the odd borrowed word). I've also yet to see a minimal pair contrasting [θ] and [ð] and, in many people's idiolects around where I live, the th phoneme isn't contrasted with [f] or [v]

[Lazar]

Don't diphthongs usually get counted as separate phonemes, at least in English?

Yes. In some languages (with simpler vowel inventories) diphthongs can be transparently analyzed as a sequence of two vowel phonemes, but in English they can't. (Rebecca, you also left out the affricates [tʃ] and [dʒ].)

Huh? e vs. ɛ in (any kind of) standard English?

In many varieties of British and North American English, FACE is a monophthongal [e(ː)], which contrasts with [ɛ] in DRESS.

I've also yet to see a minimal pair contrasting [θ] and [ð]

Teeth-teethe, sooth-soothe, loath-loathe. They're different phonemes.

[Derek]

In many varieties of British and North American English, FACE is a monophthongal [e(ː)], which contrasts with [ɛ] in DRESS.

Isn't this just an allophone of [ei] though? Assuming we modify his list to count diphthongs separately of course.

I've also yet to see a minimal pair contrasting [θ] and [ð]

Teeth-teethe, sooth-soothe, loath-loathe. They're different phonemes.

I'll add thigh-thy (archaic) and thin-then (where the pin-pen merger is in effect).

[Lazar]

Isn't this just an allophone of [ei] though? Assuming we modify his list to count diphthongs separately of course.

It's not an allophone, it's just the way that FACE is pronounced in some dialects. (Allophony happens within one dialect, not between dialects.) But yes, it's the same phoneme which is more commonly transcribed as /eɪ/. In my opinion, there's no significant distinction between monophthongs and diphthongs in English - /iː/ and /uː/ very commonly have some diphthongization, while /eɪ/ and /oʊ/, and even /aɪ/, tend to become monophthongal in many dialects. I think a more useful distinction to draw is between checked and free vowel phonemes - the former, such as /ɪ/, cannot appear at the end of a syllable, while the latter, such as /iː/, can.

While we're at it, I'll make up a list of the phonemes that I consider to exist in my own dialect. You might disagree with some of my interpretational choices, but hey.

Consonants: /m/, /n/, /ŋ/, /p/, /b/, /t/, /d/, /k/, /ɡ/, /f/, /v/, /θ/, /ð/, /s/, /z/, /ʃ/, /ʒ/, /tʃ/, /dʒ/, /h/, /ɹ/, /j/, /w/, /l/
Vowels: /ɑː/, /ɒː/, /eɪ/, /oʊ/, /iː/, /uː/, /aɪ/, /ɔɪ/, /aʊ/, /ɑɚ/, /ɛɚ/, /ɪɚ/, /ɔɚ/, /ʊɚ/, /ɝː/, /ɚ/, /æ/, /ɛ/, /ɪ/, /ʊ/, /ʌ/, /ə/

So that's 46 in total. By my interpretation, General American would have either 45 or 46, while RP would have 44. (Other sources may put the total for General American as low as 38 or 39, but I disagree with them.)

[Makri]

What's behind counting rhotacized vowels/diphthongs as phonemes?

the former, such as /ɪ/, cannot appear at the end of a syllable, while the latter, such as /iː/, can.

I take it you mean something like "at the end of a phonological domain". Surely /ɪ/ can occur in open syllables.

[Lazar]

What's behind counting rhotacized vowels/diphthongs as phonemes?

Well, it's complicated. For a non-rhotic dialect like RP, it's pretty clear that the "derhoticized" diphthongs /ɛə/, /ɪə/, /ʊə/ need to be considered separate phonemes; and for a rhotic speaker like me, who nonetheless maintains distinctions like "Mary-merry-marry", it follows that /ɛɚ/, /ɪɚ/, /ʊɚ/ need to be considered separate phonemes too. (The only alternative - analyzing "Mary", for example, as /meɪɹiː/ - seems a bit too reductionist, and it ruins useful interdialectal comparisons. Indeed, many speakers here in the northeastern United States present a continuum between rhotic and non-rhotic speech, although mine is consistently the former.) I likewise posit a phoneme /ɔɚ/, because the equivalents of RP /ɔː/ have basically split into two distinct sets in North American English - I can see this clearly in my own speech, where "Laurie" [ˈlɒːɹi] and "glory" [ˈɡlɔɚi] constitute a non-rhyme. (N.B.: this is not the same as the moribund "horse-hoarse" distinction; those two are both [hɔɚs] for me.) The remaining one, /ɑɚ/, is a bit of an indulgence, but it's consistent with the others, and allows us to posit a convenient model of non-rhoticism as a set of phonemic mergers.

This model becomes a bit more questionable when we apply it to a dialect, like modern General American, which has obscured distinctions like "Mary-merry-marry" - meaning that all three of them can be analyzed as /ˈmɛɹiː/, and "mare" as /mɛɹ/. But again, such an analysis obscures useful comparisons with other dialects - and the notion (propogated by many American dictionaries) that the "Mary-merry-marry", "serious-Sirius" and "hurry-furry" mergers result in a significant change in the pronunciation of "mare", "sear" and "fur", rather than that of "merry", "Sirius" and "hurry", has always bugged me, as the phonetic evidence consistently suggests the opposite. (Sorry, this is a recurring rant of mine.) Indeed, the utility of the rhotic diphthong model again becomes evident when we look to the west of Plains-style GA: low-back-merged Californian English preserves equivalents of RP /ɔː/ only when paired with a rhotic element, merging all the others into /ɑː/ - so the more suitable option, in my view, is to posit a rhotic phoneme /ɔɚ/, rather than a phoneme /ɔː/ which always happens to be followed by /ɹ/.

TLDR: They're needed for the analysis of some dialects, and they make comparisons among all dialects much easier. I remain convinced that English has the most godawful complicated set of interdialectal vowel phoneme relationships of any language, although it is the one that I've examined most closely.

I take it you mean something like "at the end of a phonological domain". Surely /ɪ/ can occur in open syllables.

Well, it depends on which model of syllabification we use - I'm going by John Wells' practice (he analyzes "bitter", for example, as /ˈbɪt.ə/), although various dictionaries disagree. In more simplistic terms, a checked vowel must be followed by a consonant. They exist in English, German, Dutch and French, but not in most of the other Romance languages, nor, as far as I know, in any Slavic ones.

[Makri]

For a non-rhotic dialect like RP, it's pretty clear that the "derhoticized" diphthongs /ɛə/, /ɪə/, /ʊə/ need to be considered separate phonemes

Is it? They don't contrast with /ɛr/...

Well, it depends on which model of syllabification we use - I'm going by John Wells' practice (he analyzes "bitter", for example, as /ˈbɪt.ə/), although various dictionaries disagree. In more simplistic terms, a checked vowel must be followed by a consonant.

"bitter" may well contain a geminate, so it's not the kind of example I had in mind. I was thinking mainly about unstressed /ɪ/, but also about latinate words. You'd have to argue that they were all phonologized into English with geminates... Well, maybe they were.

[Lazar]

Is it? They don't contrast with /ɛr/...

Yes they do. Both RP and my own northeastern US dialect have what's called the "Mary-merry-marry" distinction, so "Mary" (/ˈmɛəɹiː/ or /ˈmɛɚiː/) does indeed contrast with "merry" (/ˈmɛɹiː/).

"bitter" may well contain a geminate, so it's not the kind of example I had in mind.

No, "bitter" doesn't contain a geminate in any English dialect.

You'd have to argue that they were all phonologized into English with geminates... Well, maybe they were.

Huh? Again, there are no geminates in English, except at morpheme boundaries like in "unnamed". But what has this got to do with John Wells' model of syllabification? The fact that he puts an intervocalic consonant in the same syllable as a preceding checked vowel doesn't mean that he considers the consonant a geminate. And... I don't really understand what you're saying here.

To clarify, if you subscribe to John Wells' model (in which the first syllable of "bitter" is considered to be closed), then it's correct to say that checked vowels can only occur in closed syllables. If you don't, then you can just say that checked vowels must always be followed by consonants. This has nothing to do with geminates.

[Makri]

If you use the insane syllabification VC.V, then "at the end of a phonological domain" becomes equivalent to "in open syllables", because you've gotten rid of open syllables everywhere else.

Again, there are no geminates in English, except at morpheme boundaries like in "unnamed".

This is precisely not a geminate.

Yes they do. Both RP and my own northeastern US dialect have what's called the "Mary-merry-marry" distinction, so "Mary" (/ˈmɛəɹiː/ or /ˈmɛɚiː/) does indeed contrast with "merry" (/ˈmɛɹiː/).

Ah, I see what you mean. But then why would you have to postulate more than just one new phoneme /ɛə/ for the likes of "Mary"?

[Lazar]

If you use the insane syllabification VC.V, then "at the end of a phonological domain" becomes equivalent to "in open syllables", because you've gotten rid of open syllables everywhere else.

No you haven't. As I've already indicated, the model is VcC.V and Vf.CV, with Vc being a checked vowel and Vf being a free vowel. It's the model used by John Wells for the Longman Pronouncing Dictionary, and it's used in an only slightly different form in the Cambridge Advanced Learner's Dictionary which you can see online.

This is precisely not a geminate.

This precisely is a geminate, which makes you doubly wrong. The point is that there are no geminates in English, except sometimes at morpheme boundaries, as in "unnamed".

Ah, I see what you mean. But then why would you have to postulate more than just one new phoneme /ɛə/ for the likes of "Mary"?

Okay, so you clearly haven't been paying attention. As I mentioned in my previous post, there's also the "serious-Sirius" distinction (which necessitates /ɪə/ or /ɪɚ/) and the "hurry-furry" distinction (which necessitates /ɜː/ or /ɝː/), in addition to the two reasons that I gave why /ɔɚ/ is needed in North American English.

[gmalivuk]

Again, there are no geminates in English, except at morpheme boundaries like in "unnamed".

This is precisely not a geminate.

How do you figure? How exactly are you defining gemination if that one doesn't count?

the "hurry-furry" distinction

I'm still unclear on what distinction you're talking about here. Is there a minimal pair with the same distinction?

[Lazar]

I'm still unclear on what distinction you're talking about here. Is there a minimal pair with the same distinction?

Well to innovate a bit, we could posit "myrrhy" (made up from "myrrh") and "Murray". It's the distinction by which one set of words (e.g. "furry", "purring", "transferable") use /ɝːV/ or /ɜːɹV/, and another set (e.g. "hurry", "Murray", "Surrey", "current") use /ʌɹV/. Like the "Mary-merry-marry" and "serious-Sirius" distinctions (with which it's roughly coextensive), it's universal in British Isles* and Antipodean English, and common in northeastern US English, but rare throughout the rest of North America.

*With the exception of Scottish English, but that's a special case, because they lack what John Wells terms the "first NURSE merger"** - that between "fir" and "fur" - which gave rise to the "hurry-furry" distinction in the first place. I told y'all this stuff was convoluted.

**The "second NURSE merger", by the way, is his term for the "hurry-furry" merger.

[gmalivuk]

If it's not contrastive between actual words, I'd still hesitate to call them different phonemes.

In any case, they're definitely not different phonemes in my dialect. I use the same vowel+r phoneme in fur, fir, her, hurry, and furry, as well as Murray, Surrey, and current.

[Lazar]

If it's not contrastive between actual words, I'd still hesitate to call them different phonemes.

I suspected that you would argue this, and I think you're mistaken. "Hurry" and "furry" definitely do not rhyme in my dialect or in RP, and there's no allophonic condition from the initial consonants that would cause this. Compare "furry" [ˈfɝːi] with "furrow" [ˈfʌɹɤʊ], and "hurry" [ˈhʌɹi] with stressed "her" [hɝː]. Or even better, "occurring" [əˈkʰɝːɪŋ] with "ocurrance" [əˈkʰʌɹəns]. I would argue that they definitely are phonemically different, and every academic treatment of English phonemes that I've seen agrees with me (such as John Wells' Accents of English - you'd think I'm in love with the guy).

While an exact minimal pair is easy proof of a phonemic distinction, I think you'll have to back up your claim that the lack of an exact minimal pair is disproof of a phonemic distinction. (Of the few cases used here to prove a contrast between /ð/ and /θ/, for example, all involve a word that's either archaic or rare - "thy", "sooth", "teethe", "loath". Even if these words didn't exist, "this" and "thin" would still suffice.) If "hurry" and "furry" don't rhyme, and I've proven that the /f/ and /h/ are not providing allophonic reasons for them not to rhyme, then I believe I've proven a phonemic distinction. (If it helps, remember that this is a merger of some instances of one phoneme into another phoneme - under consistent conditions-, not a merger of one whole phoneme into another.)

EDIT: Hallelujah, I have found a perfect minimal pair attested in a dictionary! "Worry" [ˈwʌɹi] and "whirry" [ˈwɝːi]. Although I've admittedly not heard of the latter. In any case, what I said above still holds.

[Makri]

I apologize, you're right that I wasn't paying proper attention and totally misinterpreted the whole phonologization of /ɛə/, /ɪə/ and /ʊə/. I agree with you on this point; yes, for those variations that don't have the mergers there, the phonologization is necessary!
What is still don't get how this has anything in principle to do with rhoticity. Because you emphasized that they were clearly phonemes in non-rhotic dialects, I thought they were meant to represent the diphthongs that arose from Vr before consonants. That was the notion I was objecting to.

As I've already indicated, the model is VcC.V and Vf.CV, with Vc being a checked vowel and Vf being a free vowel.

Doesn't this make the definition of a checked vowel in English wholly circular?

As for geminates... I'm disinclined to consider double consonants with a morpheme boundary in between geminates, just as I wouldn't want to call a sequence of /t/ and /s/ with a domain or even morpheme boundary between them an affricate. This may be because I've been exposed too much to representational approaches to phonology.

[Lazar]

Doesn't this make the definition of a checked vowel in English wholly circular?

No, because /Vf/ can occur word-finally, while /Vc/ can't. Nothing can be to "bit" as "be" is to "beat", or as "go" is to "goat".

As for geminates... I'm disinclined to consider double consonants with a morpheme boundary in between geminates, just as I wouldn't want to call a sequence of /t/ and /s/ with a domain or even morpheme boundary between them an affricate.

I don't think the analogy to affricates is valid, because an affricate is a single phoneme (which just happens to be a little more phonetically complex) that must be in either one syllable or another, whereas "geminate consonant", as I understand it, is just a term for two of the same consonant phoneme in a row - that is, not one consonant. For example, the phonemic analyses of Italian that I've seen don't posit distinct geminate phonemes - so "gatto", for example, would be /ˈɡat.to/, with two instances of /t/, rather than /ˈɡatt.o/ or /ˈɡa.tto/, with one instance of /tt/. Once I've accepted this, then the only difference between an Italian geminate and a putative English geminate such as that in "unnamed", is the morphemic requirement, which in itself doesn't seem to be a phonological question. (Although I could be totally off on this.)

[gmalivuk]

(Of the few cases used here to prove a contrast between /ð/ and /θ/, for example, all involve a word that's either archaic or rare - "thy", "sooth", "teethe", "loath".)

Sure, but that was just a list of examples that are contrastive in (almost?) every dialect.

If a contrasting pair that exists in some dialects but not others works (which it must for you, based on examples you've already given), then I give you either/ether, which for huge swaths of American English is a minimal pair distinguished only by the voicing of the <th>, and then/thin is another in dialects with the pen/pin merger. And if a lack of any consistent rule for why one allophone would be used instead of the other suffices , then I suspect that also proves that the two are distinct phonemes.

I'm disinclined to consider double consonants with a morpheme boundary in between geminates

But again: why not? I don't see why a lack of gemination within morphemes precludes it between them.

Also, it doesn't *always* happen between them. Compare the contrast in the length of the /n/ between unnamed and unaimed, with the lack of such contrast (at least in my pronunciation) between inordinate and innumerable.

[Lazar]

True, my making the analogy to /ð/ and /θ/ was a bit of an overreach. I just meant that I don't think an exact minimal pair is necessary if a contrast can be proved in another way, so... we agree on that.

[Makri]

No, because /Vf/ can occur word-finally, while /Vc/ can't. Nothing can be to "bit" as "be" is to "beat", or as "go" is to "goat".

Right, but then you've made the concept of open and closed syllables useless again and you could just stick with "certain vowels can't appear word-finally" - which is in my view the correct generalization.

Once I've accepted this, then the only difference between an Italian geminate and a putative English geminate such as that in "unnamed", is the morphemic requirement, which in itself doesn't seem to be a phonological question.

It depends. As I said, some theories do have different representations here, but it's not important anyway. I can see how one can want not to draw that distinction.

[Lazar]

Right, but then you've made the concept of open and closed syllables useless again and you could just stick with "certain vowels can't appear word-finally" - which is in my view the correct generalization.

...right, you do have a point there. Sorry. So yeah, we should just say that they can't occur word-finally.

[Derek]

I have the Mary-marry-merry merger in my dialect so I can't say too much about it, but according to Wikipedia it is the merger of /æ/, /ɛ/, and /eɪ/ before /r/. What reason would you give for not analyzing the combinations in this way to avoid introducing new phonemes?

On the other hand, I do have the caught-cot merger and /ɔ:/ only remains before /r/, so I see where you're coming from when you analyse it as /ɔɚ/, which rather inclines me to agree with the rest of your analysis.

[kirkedal]

For the purpose of automatic speech recognition, the phoneme inventory is usually around 39. This is not tailored for any dialects.

-Andreas