English questions

[KathTheDragon]

Intervocalic /r/ might drop at a morpheme boundary.

Never does for me.

If saw and soar are homophonous, sawing and soaring need not be

They are for homophonous for me.

In a word like Myanmarese, as the 'r' was never intended to be sounded, so you might find commonwealth people who don't insert it. Likewise, one might find no insertion in Khmer[ic]ise - final 'r' is also silent in the Khmer language, but not in Northern Khmer.

This would sound really weird, and tmk nobody in England at least talks like that.

I thin there may be some phonetic simplification of -arian (e.g librarian or vegetarian) - [eəɹjn̥̩̩] doesn't fit the sound pattern of RPish speech well.

Not as far as I know - it's [ɛːrijn̩] for me

[Qwynegold]

Can someone give an example of a (common) English word with an intervocalic R, that's pronounced [˞˞ ] in GA or [∅] in RP, please?

Have mercy on an old man use your words. That just looks like a really tiny ㄴ to me.

I meant a word that has intervocalic rhotactization in GA, but which is unpronounced in RP. I wasn't sure how different varieties of English work like, so I didn't know if there is such a word. In that case, I'll take willm's suggestion of furry. Thanks for the replies!

[KathTheDragon]

I meant a word that has intervocalic rhotactization in GA, but which is unpronounced in RP. I wasn't sure how different varieties of English work like, so I didn't know if there is such a word. In that case, I'll take willm's suggestion of furry. Thanks for the replies!

But the /r/ in "furry" is pronounced in BrEng!

[Linguoboy]

Can someone give an example of a (common) English word with an intervocalic R, that's pronounced [˞˞ ] in GA or [∅] in RP, please?

Have mercy on an old man use your words. That just looks like a really tiny ㄴ to me.

I meant a word that has intervocalic rhotactization in GA, but which is unpronounced in RP.

I'm glad you clarified because that wasn't my takeaway at all. I thought you were looking for cases where an orthographic medial <r> was silent in RP or--alternatively--for cases where orthographic medial <r> was pronounced as some mysterious (rhotic?) phone in GA.

I'm not sure I agree that phonetically there's no approximant in GA furry. Hiatus in English is usually filled with a glide of some sort, otherwise you end up with a diphthong. Furry is unquestionably disyllabic, no matter how rapidly I said it. (Contrast, say, quiet, which, while usually disyllabic, in rapid speech can contract to [kʰwaɪ̯t̚], contrasting in my speech with quite due to my non-Canadian Canadian raising.)

[Travis B.]

I would be surprised if there were words that had intervocalic /r/ in GA but not in RP because I am used to the general tendency of RP being to insert extra /r/s intervocalically at hiatuses where the preceding vowel could be followed by /r/, e.g. draw/r/ing. Along these lines I would expect saw/r/ing and Myanma/r/ese as well.

[Travis B.]

I'm not sure I agree that phonetically there's no approximant in GA furry. Hiatus in English is usually filled with a glide of some sort, otherwise you end up with a diphthong. Furry is unquestionably disyllabic, no matter how rapidly I said it. (Contrast, say, quiet, which, while usually disyllabic, in rapid speech can contract to [kʰwaɪ̯t̚], contrasting in my speech with quite due to my non-Canadian Canadian raising.)

/ər/ before a vowel is always [ʁ̩ˤ(ː)ʁˤ] for me; it always has an intervocalic glide. Note that for me the only vowels that can exist before other vowels without glides, excluding the case of intervocalic elision, are /eɪ i/; of course intervocalic elision has a strong tendency to result in diphthongs and long or overlong vowels.

[Richard W]

I would be surprised if there were words that had intervocalic /r/ in GA but not in RP because I am used to the general tendency of RP being to insert extra /r/s intervocalically at hiatuses where the preceding vowel could be followed by /r/, e.g. draw/r/ing. Along these lines I would expect saw/r/ing and Myanma/r/ese as well.

I tried to dig out some literature on intrusive 'r' within words, but I could only access one article, at https://zenodo.org/record/894800#.YDLcAf6ny24 - Mompeán-Gonzalez, Jose, & Mompeán-Guillamón, Pilar. (2009). /r/-liaison in English: An empirical study. http://doi.org/10.1515/cogl.2009.031. It analysed news archives from the BBC World Service in 2004 and 2005. It found no intrusive 'r' in drawing (0 with, 1 without) but intrusive 'r' in withdrawal (4 with, 0 without). That's a paltry haul for a study of intrusive 'r' within words, but backs my claim about variability. I actually speak the other way round; it may be relevant that for me, withdrawal has two syllables.

For Estuarine English speakers who drop their glottal stops, does sorting have hiatus? I've seen the point made that hiatus may be avoided in English by using a glottal stop rather than 'r', and a glottal stop may be confounded with zero.

[Qwynegold]

But the /r/ in "furry" is pronounced in BrEng!

Ah, I meant, if there are no words that fit the requirements, then I will have to settle for something that doesn't fill all the criteria (the RP criteria).

[Qwynegold]

I'm glad you clarified because that wasn't my takeaway at all. I thought you were looking for cases where an orthographic medial <r> was silent in RP or--alternatively--for cases where orthographic medial <r> was pronounced as some mysterious (rhotic?) phone in GA.

I did mean that too, though. See below for a fuller explanation.

I'm not sure I agree that phonetically there's no approximant in GA furry. Hiatus in English is usually filled with a glide of some sort, otherwise you end up with a diphthong. Furry is unquestionably disyllabic, no matter how rapidly I said it. (Contrast, say, quiet, which, while usually disyllabic, in rapid speech can contract to [kʰwaɪ̯t̚], contrasting in my speech with quite due to my non-Canadian Canadian raising.)

Wiktionary says it's /ˈfɝi/. But phonetically it would be something different?

The reason why I was asking this is because I'm trying to write a short explanation about the orthography and recommended pronunciation for my IAL in layman's terms. And I needed an example word. When it comes to intervocalic /r/ in foreign languages, I thought I've perceived English speakers often pronouncing it in a way that's "too off", because they analyze the structure as Vr.V (instead of V.rV). But if there is an actual [ɹ] or some such there, then I don't need to write that much instructions.

[Travis B.]

I actually speak the other way round; it may be relevant that for me, withdrawal has two syllables.

I should note that withdrawal is a weird word for me in that it is phonemically trisyllabic but phonetically disyllabic with an overlong vowel in the second syllable, i.e. /wɪθˈdʒrɔəl/ but [ˌwɘθˈtʃɻ͡ʁɒːːo̯]

[Travis B.]

I'm not sure I agree that phonetically there's no approximant in GA furry. Hiatus in English is usually filled with a glide of some sort, otherwise you end up with a diphthong. Furry is unquestionably disyllabic, no matter how rapidly I said it. (Contrast, say, quiet, which, while usually disyllabic, in rapid speech can contract to [kʰwaɪ̯t̚], contrasting in my speech with quite due to my non-Canadian Canadian raising.)

Wiktionary says it's /ˈfɝi/. But phonetically it would be something different?

I would not necessarily trust dictionaries; after all I've seen supposedly reliable dictionaries give cot-caught-merged "American" pronunciations despite the fact that a majority of Americans are cot-caught unmerged.

In this case I think a better way to treat "furry" is as /ˈfɜri/, with /ɜ/ being a vowel phoneme that only exists before /r/ in rhotic varieties and with an intervocalic glide /r/ rather than having a rhotic vowel existing in hiatus.

The reason why I was asking this is because I'm trying to write a short explanation about the orthography and recommended pronunciation for my IAL in layman's terms. And I needed an example word. When it comes to intervocalic /r/ in foreign languages, I thought I've perceived English speakers often pronouncing it in a way that's "too off", because they analyze the structure as Vr.V (instead of V.rV). But if there is an actual [ɹ] or some such there, then I don't need to write that much instructions.

How I see things is that unstressed intervocalic consonants in English, at least in NAE, have to be treated as distinct from both onset consonants and coda consonants; in a way they belong to both the preceding syllable and the following syllable.

[Richard W]

And I needed an example word. When it comes to intervocalic /r/ in foreign languages, I thought I've perceived English speakers often pronouncing it in a way that's "too off", because they analyze the structure as Vr.V (instead of V.rV). But if there is an actual [ɹ] or some such there, then I don't need to write that much instructions.

There's also the issue of morpheme boundaries, that has led to 'tarry' being /ˈtærɪ/ as a monomorphemic verb but /ˈtɑːrɪ/ as a dimorphemic adjective meaning ‘to do with tar’. I feel the difference in morpheme structure stops ‘gory’ and ‘glory’ being perfect rimes, but this difference is very subtle.

[Richard W]

How I see things is that unstressed intervocalic consonants in English, at least in NAE, have to be treated as distinct from both onset consonants and coda consonants; in a way they belong to both the preceding syllable and the following syllable.

That applies to British English as well - probably to the whole language.

[Nortaneous]

I'm not sure I agree that phonetically there's no approximant in GA furry. Hiatus in English is usually filled with a glide of some sort, otherwise you end up with a diphthong. Furry is unquestionably disyllabic, no matter how rapidly I said it. (Contrast, say, quiet, which, while usually disyllabic, in rapid speech can contract to [kʰwaɪ̯t̚], contrasting in my speech with quite due to my non-Canadian Canadian raising.)

Wiktionary says it's /ˈfɝi/. But phonetically it would be something different?

I would not necessarily trust dictionaries; after all I've seen supposedly reliable dictionaries give cot-caught-merged "American" pronunciations despite the fact that a majority of Americans are cot-caught unmerged.

In this case I think a better way to treat "furry" is as /ˈfɜri/, with /ɜ/ being a vowel phoneme that only exists before /r/ in rhotic varieties and with an intervocalic glide /r/ rather than having a rhotic vowel existing in hiatus.

What's objectionable about a rhotic vowel existing in hiatus? I think it's just a case of the lax-tense system breaking down - we have contrasts between /i ij/ and /u uw/, but not /ɝ ɝr/. Like the "high tense vowels" FLEECE and GOOSE ( = /ij uw/), /ɝ/ can occur in hiatus; like the lax vowels KIT and FOOT, it can appear before coda /l/ without triggering breaking. (A lot of people pronounce "Carl" as a disyllable, but I've never heard anyone pronounce "curl" as one.)

It's not as if true hiatus doesn't exist elsewhere in English - you still have hiatus in THOUGHT and maybe FATHER. (cawing, ...spa-ish? dunno, there's probably a better example)

How I see things is that unstressed intervocalic consonants in English, at least in NAE, have to be treated as distinct from both onset consonants and coda consonants; in a way they belong to both the preceding syllable and the following syllable.

This is basically reasonable, but there's probably a case for NAE just totally violating the Maximal Onset Principle.

I was thinking the other day about the two pronunciations of Lancaster. People who don't know how it's pronounced say something like [ˈleəŋˌkæstər]; people who do know how it's pronounced say something like [ˈlæŋkɨstər]. This [eə]-[æ] contrast happens elsewhere as well - "banker" with [æ] vs. "panko" with [eə], for example - and given Lancaster, it seems like it might be possible to avoid having either phonemic /eə/ or a contrast between /nk/ and /ŋk/ by appealing to syllabification:

[ˈleəŋˌkæstər] /læn.kæstər/
[ˈlæŋkɨstər] /lænk.ɨstər/

[ˈbæŋkər] /bænk.ər/
[ˈpeəŋkəw] /pæn.kʌw/

(I don't know how to syllabify intervocalic -st- and at this point I'd even consider a unit geminate presigmatized stop)

This requires two different nasal assimilation steps, one before and one after æ-tensing, but I think that's actually correct - the later one can fail. I heard someone today consistently pronounce "Vanguard" (the investment company) as [ˈveənˌgɑrd], with no nasal assimilation. I could also believe [ˈleənˌkæstər] and [ˈpeənˌkʌw].

(The other problem is that "panko" needs /-ʌw/ rather than /-əw/, because shouldn't it syllabify as -nk.- if the following syllable is unstressed?)

[Travis B.]

Wiktionary says it's /ˈfɝi/. But phonetically it would be something different?

I would not necessarily trust dictionaries; after all I've seen supposedly reliable dictionaries give cot-caught-merged "American" pronunciations despite the fact that a majority of Americans are cot-caught unmerged.

In this case I think a better way to treat "furry" is as /ˈfɜri/, with /ɜ/ being a vowel phoneme that only exists before /r/ in rhotic varieties and with an intervocalic glide /r/ rather than having a rhotic vowel existing in hiatus.

What's objectionable about a rhotic vowel existing in hiatus? I think it's just a case of the lax-tense system breaking down - we have contrasts between /i ij/ and /u uw/, but not /ɝ ɝr/. Like the "high tense vowels" FLEECE and GOOSE ( = /ij uw/), /ɝ/ can occur in hiatus; like the lax vowels KIT and FOOT, it can appear before coda /l/ without triggering breaking. (A lot of people pronounce "Carl" as a disyllable, but I've never heard anyone pronounce "curl" as one.)

See I analyze NAE different from you, probably because my native dialect operates differently from yours. /ɪ i/ are both monophthongs to me, not a monophthong and a diphthong. Likewise /ʊ u/ and /oʊ/ are monophthongs to me, with the following [w] in hiatus after /u o/ (except if there was an elided intervocalic consonant!) being inserted; final /o/ can also have [w] added after it, but this is optional and may be affected by stress (e.g. I am much more likely to insert [w] after hello than after tomato). Of course, none of these vowels cause breaking of following /l/ for me, implying that underlyingly they are not diphthongs for me.

About /r/, the reason why I don't like "/ɝ/" is because it makes it special as rhotic vowels go, even though "/ɚ/", /ɔr/, and /jur/ (with a preceding [j]) or /ur/ (when the historical /j/ was lost) may all be realized as syllabic rhotics. So why does it get to be special and be considered its own phoneme without necessarily having a following /r/? (The reason to consider "/ɚ/" a phoneme pair is because initially it alternates with [ə(ː)ʁˤ] and it is realized with a glide as [ʁ̩ˤːʁˤ] before another vowel. You could argue that the following [ʁˤ] before a vowel is inserted, but then you would have to argue that it is inserted for all the other vowel+/r/ pairs.) Of course, I don't believe that /r/ being a glide triggers /l/ breaking for me, since I pronounce Carl as a monosyllable.

It's not as if true hiatus doesn't exist elsewhere in English - you still have hiatus in THOUGHT and maybe FATHER. (cawing, ...spa-ish? dunno, there's probably a better example)

I forgot about THOUGHT, which yes, I can have hiatus for. FATHER can have hiatus, but only in contrived examples (considering that morpheme-final FATHER can only exist in stressed morpheme-final syllables, and even then it is rare), and only across morpheme boundaries.

How I see things is that unstressed intervocalic consonants in English, at least in NAE, have to be treated as distinct from both onset consonants and coda consonants; in a way they belong to both the preceding syllable and the following syllable.

This is basically reasonable, but there's probably a case for NAE just totally violating the Maximal Onset Principle.

When I try to pretend that NAE violates the Maximum Onset Principle, I end up inserting glottal stops before each unstressed vowel...

To me, how I take this is that (ignoring things like palatalization) /p t tʃ k/ each have at least three different allophones (four in the case of /t/). At the start of stressed syllables they are [pʰ tʰ tʃʰ kʰ]. After a vowel or a sonorant in a coda, they are [ʔp ʔt~ʔ ʔtʃ ʔk]. Intervocalically they are [p ɾ tʃ k]. And otherwise they are [p t tʃ k]. If these were to validate the Maximal Onset Principle, you would expect [ʔp ʔt~ʔ ʔtʃ ʔk] to be found intervocalically, but they are not, at least in NAE. Likewise, in the case of /t/ if intervocalic [ɾ] were part of a coda, one would expect it for final /t/ (which rather is [ʔt~ʔ] except after another obstruent, where it is then [t]).

Going back to /r/, though, it just seems to me that the [ʁˤ] found between vowels here is intervocalic in the same way that [ɾ] for intervocalic /t/ is. It seems strange to me to argue that /r/ itself would be special in being syllabified with the preceding vowel (whether or not that vowel is a syllabic /r/ or is merely r-colored, or (thanks to elisions, as in the case of every) is not r-colored at all).

I was thinking the other day about the two pronunciations of Lancaster. People who don't know how it's pronounced say something like [ˈleəŋˌkæstər]; people who do know how it's pronounced say something like [ˈlæŋkɨstər]. This [eə]-[æ] contrast happens elsewhere as well - "banker" with [æ] vs. "panko" with [eə], for example - and given Lancaster, it seems like it might be possible to avoid having either phonemic /eə/ or a contrast between /nk/ and /ŋk/ by appealing to syllabification:

[ˈleəŋˌkæstər] /læn.kæstər/
[ˈlæŋkɨstər] /lænk.ɨstər/

[ˈbæŋkər] /bænk.ər/
[ˈpeəŋkəw] /pæn.kʌw/

(I don't know how to syllabify intervocalic -st- and at this point I'd even consider a unit geminate presigmatized stop)

Intervocalic /st/ for me seems to behave somewhat like intervocalic /nt/, which has to be treated as a single intervocalic unit that belongs to both syllables (of course, intervocalic /nt/ is realized as a single consonant, [ɾ̃], most of the time, when it isn't elided).

This requires two different nasal assimilation steps, one before and one after æ-tensing, but I think that's actually correct - the later one can fail. I heard someone today consistently pronounce "Vanguard" (the investment company) as [ˈveənˌgɑrd], with no nasal assimilation. I could also believe [ˈleənˌkæstər] and [ˈpeənˌkʌw].

(The other problem is that "panko" needs /-ʌw/ rather than /-əw/, because shouldn't it syllabify as -nk.- if the following syllable is unstressed?)

The variety here has strong nasal assimlation (as I kid I thought the Hoan Bridge in Milwaukee was the Home Bridge, because I had never seen its name in writing, and I naturally always heard the pronunciation where the /n/ in Hoan assimilated to the /b/ in bridge, but at the same time it has a clear distinction between /æ/ [ɛ~ɛə~eə] and /eɪ/ [e] before what is realized as [ŋ]. For instance, I have /eɪ/ in banker but /æ/ in Lancaster and vanguard. Note that this is not a matter of productiveness, because when I make up -ang words I naturally pronounce them with /eɪ/ but when I learn words from elsewhere, ones that I may have never heard spoken, I readily pronounce them with /æ/. Aside from being a learnèd versus non-learnèd distinction, it might have to do with stress and being polysyllabic; words that a perceive as having more than one stressed syllable seem to be more likely to be pronounced with /æ/, while short words with one stressed (typically initial) syllable are more likely to be pronounced with /eɪ/.

[Nortaneous]

See I analyze NAE different from you, probably because my native dialect operates differently from yours. /ɪ i/ are both monophthongs to me, not a monophthong and a diphthong.

OK, phew, I've used Mid-Atlantic vs. Inland North as an example of how small differences in realization can reflect large differences in underlying structure before... good to know I wasn't completely talking out of my ass

About /r/, the reason why I don't like "/ɝ/" is because it makes it special as rhotic vowels go, even though "/ɚ/", /ɔr/, and /jur/ (with a preceding [j]) or /ur/ (when the historical /j/ was lost) may all be realized as syllabic rhotics.

Can they?

I have (something like - my notation obv doesn't strictly work diachronically since Mid-Atlantic is innovative here) /i̯ur/ > /i̯or/ ~ (more commonly) /i̯ɝ/ as a diachronic change, as well as /ur/ > /or/ (that is, CURE > FORCE ~ NURSE and POOR > FORCE... "CURE" was a poor choice on Wells's part), but I can't merge FORCE and NURSE.

So why does it get to be special and be considered its own phoneme without necessarily having a following /r/? (The reason to consider "/ɚ/" a phoneme pair is because initially it alternates with [ə(ː)ʁˤ] and it is realized with a glide as [ʁ̩ˤːʁˤ] before another vowel.

Does it? I call it /ər/ for symmetry, but I don't have any principled reason for this - it never decomposes, and glide insertion is automatic. (cf. /tʌwkɨjəw/ "Tokyo" and /kəwala/ "koala")

In my analysis, the r-diphthongs are symmetrical to the j-diphthongs - there's PRICE FACE CHOICE FLEECE /aj ej oj ij/, and there's START SQUARE FORCE NEAR /ar er or ir/. The w-diphthongs aren't symmetrical, though, because only height is contrastive before /w/. So the only vowels that can occur in prevocalic position (i.e. in hiatus) are FATHER, THOUGHT, and NURSE.

This has a diachronic explanation, but probably not a synchronic one.

To me, how I take this is that (ignoring things like palatalization) /p t tʃ k/ each have at least three different allophones (four in the case of /t/). At the start of stressed syllables they are [pʰ tʰ tʃʰ kʰ]. After a vowel or a sonorant in a coda, they are [ʔp ʔt~ʔ ʔtʃ ʔk]. Intervocalically they are [p ɾ tʃ k]. And otherwise they are [p t tʃ k]. If these were to validate the Maximal Onset Principle, you would expect [ʔp ʔt~ʔ ʔtʃ ʔk] to be found intervocalically, but they are not, at least in NAE. Likewise, in the case of /t/ if intervocalic [ɾ] were part of a coda, one would expect it for final /t/ (which rather is [ʔt~ʔ] except after another obstruent, where it is then [t]).

That's an interesting perspective on BrEng t-glottalization!

I think the stops can be modeled with positional effects alone, without reference to syllables, although I'll admit I haven't thought about it much. Maybe something could be done with lenition of /g/, although I don't know what environments that can occur in, and it's both optional and not very salient.

Going back to /r/, though, it just seems to me that the [ʁˤ] found between vowels here is intervocalic in the same way that [ɾ] for intervocalic /t/ is. It seems strange to me to argue that /r/ itself would be special in being syllabified with the preceding vowel (whether or not that vowel is a syllabic /r/ or is merely r-colored, or (thanks to elisions, as in the case of every) is not r-colored at all).

Resyllabification explains things like the hurry-furry and mirror-nearer mergers, and /lo.jər/ > /loj.ər/ "lawyer".

This would take a lot of squinting at spectrograms, but I think when /v/ is lost in "every", you get something more like [ɛ.ɹɪj] than [ɛɹ.ɪj] or [ɛɹ̣ɪj]. (using dot below for ambisyllabicity, i.e. location on the syllable boundary)

The variety here has strong nasal assimlation (as I kid I thought the Hoan Bridge in Milwaukee was the Home Bridge, because I had never seen its name in writing, and I naturally always heard the pronunciation where the /n/ in Hoan assimilated to the /b/ in bridge, but at the same time it has a clear distinction between /æ/ [ɛ~ɛə~eə] and /eɪ/ [e] before what is realized as [ŋ]. For instance, I have /eɪ/ in banker but /æ/ in Lancaster and vanguard. Note that this is not a matter of productiveness, because when I make up -ang words I naturally pronounce them with /eɪ/ but when I learn words from elsewhere, ones that I may have never heard spoken, I readily pronounce them with /æ/. Aside from being a learnèd versus non-learnèd distinction, it might have to do with stress and being polysyllabic; words that a perceive as having more than one stressed syllable seem to be more likely to be pronounced with /æ/, while short words with one stressed (typically initial) syllable are more likely to be pronounced with /eɪ/.

I remember getting confused by intro linguistics stuff about nasal assimilation - there's coronal contact for the [n] in words like "invent", "input" isn't pronounced "imput" (except in very rapid speech), and so on.

I don't have any TRAP > FACE, but I do have æ > æj before /g ŋ/. (I'm not committed to /ŋ/ as a phoneme and would prefer to avoid it if possible, but it's the simplest statement of the conditioning factor.) Also DRESS > FACE in the same environment, but it's incomplete - "Genghis" has DRESS, and I'm pretty sure I pronounced "penguin" with FACE at one point but I'm not sure if I do anymore.

[anteallach]

This is basically reasonable, but there's probably a case for NAE just totally violating the Maximal Onset Principle.

I was thinking the other day about the two pronunciations of Lancaster. People who don't know how it's pronounced say something like [ˈleəŋˌkæstər]; people who do know how it's pronounced say something like [ˈlæŋkɨstər]. This [eə]-[æ] contrast happens elsewhere as well - "banker" with [æ] vs. "panko" with [eə], for example - and given Lancaster, it seems like it might be possible to avoid having either phonemic /eə/ or a contrast between /nk/ and /ŋk/ by appealing to syllabification:

[ˈleəŋˌkæstər] /læn.kæstər/
[ˈlæŋkɨstər] /lænk.ɨstər/

[ˈbæŋkər] /bænk.ər/
[ˈpeəŋkəw] /pæn.kʌw/

(I don't know how to syllabify intervocalic -st- and at this point I'd even consider a unit geminate presigmatized stop)

This requires two different nasal assimilation steps, one before and one after æ-tensing, but I think that's actually correct - the later one can fail. I heard someone today consistently pronounce "Vanguard" (the investment company) as [ˈveənˌgɑrd], with no nasal assimilation. I could also believe [ˈleənˌkæstər] and [ˈpeənˌkʌw].

(The other problem is that "panko" needs /-ʌw/ rather than /-əw/, because shouldn't it syllabify as -nk.- if the following syllable is unstressed?)

I must admit that it always seems to me when you describe this that a contrast between /n/ and /ŋ/ before velars is the simplest explanation for what's going on. Not being American, I don't have the TRAP allophony before nasals that you do, but I do have the possibility of unassimilated [n] before /k/ and /g/, and indeed vanguard is a good example.

For Lancaster (with the obvious caveat that my primary reference for this name is presumably a different place from yours) I definitely have [ŋ], but a full vowel [a] in the second syllable. If I understand Wells's syllabification rules correctly, that should put the /k/ in the first syllable (because it's more strongly stressed and /ŋk/ is a legal coda) unless -caster is treated synchronically as a morpheme.

[Travis B.]

See I analyze NAE different from you, probably because my native dialect operates differently from yours. /ɪ i/ are both monophthongs to me, not a monophthong and a diphthong.

OK, phew, I've used Mid-Atlantic vs. Inland North as an example of how small differences in realization can reflect large differences in underlying structure before... good to know I wasn't completely talking out of my ass

Whenever you talk about the underlying structure of English it always seems to me like we don't even speak the same primary branch thereof, even though I know that our lects can't be that far apart in realization.

About /r/, the reason why I don't like "/ɝ/" is because it makes it special as rhotic vowels go, even though "/ɚ/", /ɔr/, and /jur/ (with a preceding [j]) or /ur/ (when the historical /j/ was lost) may all be realized as syllabic rhotics.

Can they?

I have (something like - my notation obv doesn't strictly work diachronically since Mid-Atlantic is innovative here) /i̯ur/ > /i̯or/ ~ (more commonly) /i̯ɝ/ as a diachronic change, as well as /ur/ > /or/ (that is, CURE > FORCE ~ NURSE and POOR > FORCE... "CURE" was a poor choice on Wells's part), but I can't merge FORCE and NURSE.

TBH these alternations, aside from the allophony between initial [ə(ː)ʁˤ] and [ʁ̩ˤ] are probably best classed as allomorphy and are not productive. The affected words include your, you're, pure, cure, sure and like.

So why does it get to be special and be considered its own phoneme without necessarily having a following /r/? (The reason to consider "/ɚ/" a phoneme pair is because initially it alternates with [ə(ː)ʁˤ] and it is realized with a glide as [ʁ̩ˤːʁˤ] before another vowel.

Does it? I call it /ər/ for symmetry, but I don't have any principled reason for this - it never decomposes, and glide insertion is automatic. (cf. /tʌwkɨjəw/ "Tokyo" and /kəwala/ "koala")

I have [ˈtʰokiːo(ː)]~[ˈtʰokjo(ː)] but [kʰoːˈwaːɤ̯ə(ː)]~[kʰəːˈwaːɤ̯ə(ː)].

Will respond with more later.

[Travis B.]

In my analysis, the r-diphthongs are symmetrical to the j-diphthongs - there's PRICE FACE CHOICE FLEECE /aj ej oj ij/, and there's START SQUARE FORCE NEAR /ar er or ir/. The w-diphthongs aren't symmetrical, though, because only height is contrastive before /w/. So the only vowels that can occur in prevocalic position (i.e. in hiatus) are FATHER, THOUGHT, and NURSE.

This is complicated in the speech here because of intervocalic elision before /ər/ resulting in non-r-colored long or overlong vowels preceding /r/.

To me, how I take this is that (ignoring things like palatalization) /p t tʃ k/ each have at least three different allophones (four in the case of /t/). At the start of stressed syllables they are [pʰ tʰ tʃʰ kʰ]. After a vowel or a sonorant in a coda, they are [ʔp ʔt~ʔ ʔtʃ ʔk]. Intervocalically they are [p ɾ tʃ k]. And otherwise they are [p t tʃ k]. If these were to validate the Maximal Onset Principle, you would expect [ʔp ʔt~ʔ ʔtʃ ʔk] to be found intervocalically, but they are not, at least in NAE. Likewise, in the case of /t/ if intervocalic [ɾ] were part of a coda, one would expect it for final /t/ (which rather is [ʔt~ʔ] except after another obstruent, where it is then [t]).

That's an interesting perspective on BrEng t-glottalization!

I think the stops can be modeled with positional effects alone, without reference to syllables, although I'll admit I haven't thought about it much. Maybe something could be done with lenition of /g/, although I don't know what environments that can occur in, and it's both optional and not very salient.

Syllables provide good reason for effects such as atlas [ˈɛʔɰɘs] versus metro [ˈmɜtʃɻ͡ʁo(ː)]; the former has a glottal stop because /tl/ cannot form an onset outside of some loanwords, whereas the latter does not because /tr/ is a valid onset.

Going back to /r/, though, it just seems to me that the [ʁˤ] found between vowels here is intervocalic in the same way that [ɾ] for intervocalic /t/ is. It seems strange to me to argue that /r/ itself would be special in being syllabified with the preceding vowel (whether or not that vowel is a syllabic /r/ or is merely r-colored, or (thanks to elisions, as in the case of every) is not r-colored at all).

Resyllabification explains things like the hurry-furry and mirror-nearer mergers, and /lo.jər/ > /loj.ər/ "lawyer".

This would take a lot of squinting at spectrograms, but I think when /v/ is lost in "every", you get something more like [ɛ.ɹɪj] than [ɛɹ.ɪj] or [ɛɹ̣ɪj]. (using dot below for ambisyllabicity, i.e. location on the syllable boundary)

What about words like Saturday [ˈsɛːʁˤde(ː)] though, where the /r/ cannot belong to a following syllable? (Note that the vowel in Saturday is not the vowel in marry-merry-Mary as that for me is closer than it.)

The variety here has strong nasal assimlation (as I kid I thought the Hoan Bridge in Milwaukee was the Home Bridge, because I had never seen its name in writing, and I naturally always heard the pronunciation where the /n/ in Hoan assimilated to the /b/ in bridge, but at the same time it has a clear distinction between /æ/ [ɛ~ɛə~eə] and /eɪ/ [e] before what is realized as [ŋ]. For instance, I have /eɪ/ in banker but /æ/ in Lancaster and vanguard. Note that this is not a matter of productiveness, because when I make up -ang words I naturally pronounce them with /eɪ/ but when I learn words from elsewhere, ones that I may have never heard spoken, I readily pronounce them with /æ/. Aside from being a learnèd versus non-learnèd distinction, it might have to do with stress and being polysyllabic; words that a perceive as having more than one stressed syllable seem to be more likely to be pronounced with /æ/, while short words with one stressed (typically initial) syllable are more likely to be pronounced with /eɪ/.

I remember getting confused by intro linguistics stuff about nasal assimilation - there's coronal contact for the [n] in words like "invent", "input" isn't pronounced "imput" (except in very rapid speech), and so on.

Whereas here input definitely has an [m].

I don't have any TRAP > FACE, but I do have æ > æj before /g ŋ/. (I'm not committed to /ŋ/ as a phoneme and would prefer to avoid it if possible, but it's the simplest statement of the conditioning factor.) Also DRESS > FACE in the same environment, but it's incomplete - "Genghis" has DRESS, and I'm pretty sure I pronounced "penguin" with FACE at one point but I'm not sure if I do anymore.

I have TRAP > FACE before some but not all /ŋ/ (e.g. I have it in banker but not Lancaster). and I have DRESS > FACE in length and strength but not in Genghis or penguin.

[Richard W]

In this case I think a better way to treat "furry" is as /ˈfɜri/, with /ɜ/ being a vowel phoneme that only exists before /r/ in rhotic varieties and with an intervocalic glide /r/ rather than having a rhotic vowel existing in hiatus.

However, I just heard a US pronunciation of rural as something like [rɜ˞əl] on Episode 4 of Resident Alien; the lack of an intervocalic consonant did surprise me.