cursed english phonemes

(from april's american english)

There are a lot of phonemes used in (General American) English that I'd expect an English speaker off the street to already pretty much understand as concepts, at least for native speakers. These include the long a /ej/, b, ch /tʃ/, d, short e /ε/, long e /ij/, f, (hard) g, h, short i /ɪ/, long i /aj/, j /dʒ/, k, l /l/, m, n, short o /ɑ/, long o /ow/, oo /uw/, oy /ɔj/, p, r /r/, s, sh /ʃ/, t, short u /ə/¹¹ Lots of people like using /ə/ for unstressed syllables and /ʌ/ for stressed ones, but I tend to be not super convinced that this distinction is phonemic (for dialects close to mine).

Part of the issue is that unstressed syllables can get pretty difficult to analyze. For a lot of words with a reduced vowel that one might write as /ə/ or <ᵻ>, I feel like I can’t conclusively narrow down the valid pronunciations to a range much tighter than “a short monophthong somewhere between [ɐ] and [ɪ] and [ʊ]”. I think in practice what it comes out as probably depends mostly on factors like adjacent consonants, but the spelling of the vowel can maybe contribute too., v, w, y /j/, and z sounds.

There are a few others that have a nuance or two which people might need explained, or at least need a moment to think through. People know there’s a th sound, and they know it’s /θ~ð/, but they might not realize that there are distinct soft th /θ/ and hard th /ð/ sounds—the distinction between the soft th in thigh and the hard th in thy. And while people are mostly comfortable with the /aw/ sound in now or out, words like know and you make it rather difficult to unambiguously refer to; they might not have a name for the sound.

There’s no great generally accessible way to talk about /ʒ/—most people won’t have any idea that “the zh sound” is supposed to refer to a normal English sound—but “that weird soft g–ish sound in genre, treasure, and zhuzh” basically gets the point across without further fuss.

I think most people will identify “short a” as /æ/, but they probably also think of the /ɑ/ in father as a sound that is both short and a—while it would be perfectly normal to say that “the y in happy makes a long e sound”, I would be a little surprised to hear “the a in father makes a short o sound” from a layman. And the term “long u” can refer to either /juw/²² Which doesn’t make a ton of sense to call a “phoneme”, even when compared to the other diphthongs (which you could probably pretty much get away with describing as a checked vowel + a semivowel if you wanted to). or /uw/, because jod dropping (and jod coalescence) can cause the pronunciation of an orthographic long u to depend on the preceding consonant.

This is about sounds weirder than those. We will start with more comprehensible sounds and get progressively more esoteric.

/ʊ/, the centralized u

You will not believe how long it took me to convince Celene that bus and bush had different vowels. (Maybe it would've helped if I pointed to tuck and took as a minimal pair—or even better, lucks, looks, and Luke’s.) I’m pretty sure this isn’t a Canadian English thing, my other Canadian friends do not somehow lack the ꜰᴏᴏᴛ-ꜱᴛʀᴜᴛ split. I don’t think Celene ever really did either, but she didn’t make the distinction sufficiently consistently to avoid driving me mad with the occasional [bəʃ].

I think most people are not especially confused by the concept of “the vowel in put, would, or book”. But they might have issues tracking this sound because there’s no good way to spell it. Like, the “ow sound” isn’t an unambiguous way to refer to /aw/, but people will usually interpret it as /aw/ rather than /ow/. I don’t think any of “u sound”, “ou sound”, or “oo” sound will work—people will guess /ə/, /juw/, /ow/, or /uw/ before they guess /ʊ/.

/ŋ/, the velar nasal

I think most native³³ ESL might actually be an advantage here, because the ng sound is sometimes heavily emphasized in English classes. Pronouncing -ing as -in’ can be “low status”, so it’s “important” to get it “right”. English speakers understand the difference between running and runnin’, in the sense that they can tell the difference and can articulate each of them correctly. But they might well give a nonsensical (or at least dubious) description of how the pronunciation differs. Probably people will agree that runnin’ ends with an n sound—but it’s easy to imagine someone claiming that running ends with a g sound, which seems to me like a pretty weird way of looking at it.

There’s a few things going on here. Part of it is that there’s a sense in which /ŋ/ is only borderline phonemic. Sure, there are minimal pairs like sin and sing, but I expect most laymen would assert that sink simply has an n sound in it, and not think to remark that it isn’t quite the normal n sound.

You could almost argue that there’s an underlying /ng/ → [ŋ] pattern happening—an n assimilates before k or g, and then a g is dropped but a k isn’t—except that finger and anger seem to suggest this doesn't happen in the middle of words. But then, it still does in singer, so maybe you need to still do it in the middle of a word if it’s at the end of the lemma?

Personally, even though I don’t think I ever really contrast between /nk/ and /ŋk/, finger vs singer is enough to convince me that /ŋ/ can’t just be /ng/. In some words, like incredible, I’m not totally sure whether I tend to use /n/ or /ŋ/.

Untrained native English speakers will generally find it very easy to mmmm or nnnn and hold it indefinitely, but won’t really be able to ŋŋŋŋ or even ŋ at all. They usually won’t be able to say the surnames Ng, Ngo, or Nguyen with an actual /ŋ/ at the start, because English never⁴⁴ Unless you say [ŋ̩krɛdibɫ̩] or something, but that doesn’t count. starts a word with /ŋ/. This is dumb, and people should totally learn to appreciate the /ŋ/ sound more.

/ɚ/, the rhotic vowel

This is the vowel in bird, girl, or letter. I think people maybe overstate how weird this one is sometimes. Like, yeah, it doesn't fit into the IPA’s outdated vowel quadrilateral very well. But the distinction between /i/ and /y/ (ü) doesn’t fit into it great either, and for very similar reasons.

Vowels are (mostly) acoustically distinguished by things called formants. Lots of vowels can be told apart by looking at the first and second formants, written F1 and F2. Vowel “closeness”, “openness”, or “height” mostly tracks F1, while vowel “frontness” or “backness” mostly tracks F2. The vowel triangle is bounded by /a/, /i/, and /u/: /a/ has very high F1, /i/ has low F1 and high F2, while /u/ has low F1 and low F2.

The issue is that /ɚ/ can’t really be identified in terms of F1 and F2. The defining characteristic of /ɚ/ is an exceptionally low F3. Which is a little weird! But the difference between (German) i and ü or e and ö is also largely due to rounding’s lowering F3 (though a lower F2 contributes too). In a sense, the weird part is that American English only has the one weird low F3 vowel. Hence why some people treat Gödel and girdle like homophones.

People like to write this as /ər/, which is at least a little dumb. Am I allowed to call a ring and erring a minimal pair even though er is mostly just the British spelling of uh? Some people totally pronounce er as /ɚ/, that’s gotta be a thing. How about to ring and Turing, is that good enough for me to call /ɚ/ a phoneme?

/ɔ/, the open o

This is the sound in caught, as opposed to the short o /ɑ/ in cot, if you don’t have the cot-caught merger. I have /ɔ/ in August, autumn, awful, and often coffee. The word on is a little harder to place one way or the other, though off leans more solidly /ɔ/. Other minimal pairs are wrought vs rot⁵⁵ I might maybe have an allophonic [ʷɹ] in wrought. It’s funny to imagine a type of guy for whom wrought vs rot constitutes a /ʷr/ vs /r/ minimal pair.

Of course, wrap and rap would be a better example. I think there’s a decent chance that if you really look close I tend to articulate a sort of rounding in wrap somewhat more often than I do in rap, but they still feel too much like homophones to me for me to start actually arguing that I have a /ʷr/ phoneme.

Nonetheless, if you wanted an English sound even more cursed than the others I’m covering, consider /ʷr/. , thought vs thot, raw vs Ra, and pawed vs pod⁶⁶ Am I allowed to count paw and pa as a minimal pair even though they can be synonyms? I think probably yes, because even though /pɔ/ could be either, /pɑ/ can’t be a pet’s foot..

The idea of someone using a sound like my [ɑ] in a word like awful is really weird to me. Surely people don’t think that’s just a silent w? But maybe if you have the merger you use a sort of in between vowel—maybe it even leans slightly towards the [ɔ]-ish end in some words and the [ɑ]-ish end in others, even if you’d never be able to consciously sort them without being told the orthographic conventions.

Even for people with the cot-caught merger, they can’t totally escape the wrath of. It appears in, for example, the word awwww ([ɔ̃ːːː]). But it’s also often used to transcribe the diphthongs in boy or bore⁷⁷ Yes, that’s a diphthong. If you don’t accept /ɔr/ as a diphthong because “r isn’t a semivowel” or some other such nonsense, I will simply call it /ɔɚ/.. Personally, I think those all start with a higher vowel than my /ɔ/, and I'd be inclined to write them as /oj/ and /or/ since they're at least reasonably close to the vowel that starts my /ow/. I also think they’re the same vowel as in bowl for me, but that's usually written /bowl/⁸⁸ I might prefer /bol/—is bow lure vs bowler an /owl/ vs /ol/ minimal pair? I think it is. “Hey, what the heck is up with English approximants in codas” is going on the list of ideas. I find them confusing even before I start worrying about fire or boil.. I think /ɔr/ just doesn’t work, something like pouring vs paw ring forms an /or/ vs /ɔr/ minimal pair.

But I haven't really tried to like, really objectively measure the vowel heights, so I could be off here in some way—maybe my “/ɔ/” is lower than a real [ɔ] is supposed to be, or something.

/əj/, raised long i

I’ve decided that my stance is that this is phonemic for me. It occurs in ice; ice cream and eye scream is a minimal pair. The standard theory is that this is an allophone for /aj/ when it undergoes prefortis clipping, and so whiter and wider aren’t a good enough minimal pair because really the distinction is still the t’s fault. You can analyze ice scream and eye scream similarly—the ce induces prefortis clipping in ice, but the sc in scream is in a separate word from eye and therefore doesn’t.

Also, why is it in spider, if it’s just prefortis clipping? Is spidering and spy during (or spider and spied ‘er) a good enough minimal pair? Or can you argue that I prefortis clip before tapped d—even though I don’t in wider? Perhaps I prefortis clip before tapped d only when the d is tapped in the root word, I guess.

/ẽə̃/, tensed “/æ/”

Some people will try to tell you that am and at or hand and hacked have the same vowel. I do not understand how anyone would come to this conclusion if they were looking at General American from scratch, though I will grant that it could make total sense in some other dialect.

I grant that I don’t have a dialect where the distinction is “phonemic” due to the introduction of a minimal pair, though some such dialects do exist. But I definitely say hang with a flat out /ej/, and it’s not like am and aim are homophones but am is much closer to aim than to /æm/, surely it is possible for two vowels to eventually be far enough apart that it doesn't make any sense to group them as one phoneme even if there aren’t strictly speaking any minimal pairs per se.

If someone introduced themself to me as Manning with an /æ/, I would probably continue to make that distinction even if my first guess at the pronunciation of that word would’ve had an /ẽə̃/! I don’t have a minimal pair merely because no words happen to contain something like /æn/, not because such a word wouldn’t be distinct from one with /ẽə̃n/ if it did!

Someone please coin a minimal pair. I am very distressed to lack an indisputable argument that this isn’t an allophone. Maybe I should start pronouncing anime as /ænɪmej/. Help me.

I will grant that /ʔ/ and /x/ are me being weird. Probably the single word /krəsɔ̃nt/ isn’t enough that I should be allowed to count /ɔ̃/ as a phoneme, even if I want to. The wh sound /ʍ/ was, I think, legitimately distinguished by my grandma, but if I ever use /ʍ/ I'm doing it entirely on purpose for fun—and really it comes out more like a /hw/ anyways.