Marginal distinctions

[Travis B.]

I have observed that at least in the English here there are complications with regard to the traditional NAE vowel distinctions. For instance, many of the distinctions merged in the Mary-merry-marry merger have been resurrected in a marginal fashion through consonant elisions and resulting vowel cluster reductions. Take the following:

airy: "/ˈɛri/" [ˈɛ̝ːʁˤi(ː)]
every: "/ˈɛri/" [ˈɜːʁˤi(ː)]
rather: "/ˈrær/" [ˈʁʷˤɛːːʁˤ]
matter: "/ˈmær/" [ˈmɛːʁˤ]
Saturday: "/ˈsærde/" [ˈsɛːʁˤɾe̞(ː)]

other: "/ˈʌr/" [ˈʌːːʁˤ]
star: "/stɑr/" [sʲtʲɑ(ː)ʁˤ]
stutter: "/stʌr/" [sʲtʲʌːʁˤ]
alright: "/ɔˈrəɪt/" [ɒːˈʁˤəe̯ʔ(t)], "/aˈrəɪt/" [aːˈʁˤəe̯ʔ(t)], "/ɔˈrəɪt/" [ɔːˈʁˤəe̯ʔ(t)]

may: "/ˈme/" [ˈme̞(ː)]
maybe: "/ˈmei/" [ˈme̞ːi(ː)]~[ˈme̞ːj]

I put the phonemes in quotes because I do not think they are accurate analyses, but at the same time the traditional analyses are inadequate too (for instance, why does the traditional /ð/ get dropped in rather and other but not father, brother and mother?).

Here we see the traditional phonemic system breaking down, but no new phonemic system can replace it, as many of these changes are not productive (e.g. the /t/ and /d/ elision is productive, but the /ð/, /v/, or /b/ elision is not). You cannot create new words that reflect the non-productive changes, indicating that they are not integrated into the underlying phonemic system, but at the same time that they are not regular products of the old phonemic system.

[abahot]

What dialect of NAE realizes the rhotic as uvular?

[Travis B.]

What dialect of NAE realizes the rhotic as uvular?

This is Milwaukee dialect. Note that it's a pharyngealized (except after coronals) uvular (except after coronals, where then it also has postalveolar coarticulation) approximant (also note that it is also labialized word-initially).

[Travis B.]

One note - word-final /ɑr/ and /ɔr/ often lose their uvular character and become primarily pharyngeal in articulation. But this is off-topic.

[Zju]

What does a Milwaukee rhotic sound like? Is it close to any of these pronunciations? Searching for "Milwaukee uvular rhotic" turned up nothing of interest.

[Travis B.]

What does a Milwaukee rhotic sound like? Is it close to any of these pronunciations? Searching for "Milwaukee uvular rhotic" turned up nothing of interest.

Some of them sounded closer than others, but none were spot-on.

[Estav]

I have observed that at least in the English here there are complications with regard to the traditional NAE vowel distinctions. For instance, many of the distinctions merged in the Mary-merry-marry merger have been resurrected in a marginal fashion through consonant elisions and resulting vowel cluster reductions. Take the following:

airy: "/ˈɛri/" [ˈɛ̝ːʁˤi(ː)]
every: "/ˈɛri/" [ˈɜːʁˤi(ː)]
rather: "/ˈrær/" [ˈʁʷˤɛːːʁˤ]
matter: "/ˈmær/" [ˈmɛːʁˤ]
Saturday: "/ˈsærde/" [ˈsɛːʁˤɾe̞(ː)]

other: "/ˈʌr/" [ˈʌːːʁˤ]
star: "/stɑr/" [sʲtʲɑ(ː)ʁˤ]
stutter: "/stʌr/" [sʲtʲʌːʁˤ]
alright: "/ɔˈrəɪt/" [ɒːˈʁˤəe̯ʔ(t)], "/aˈrəɪt/" [aːˈʁˤəe̯ʔ(t)], "/ɔˈrəɪt/" [ɔːˈʁˤəe̯ʔ(t)]

may: "/ˈme/" [ˈme̞(ː)]
maybe: "/ˈmei/" [ˈme̞ːi(ː)]~[ˈme̞ːj]

I put the phonemes in quotes because I do not think they are accurate analyses, but at the same time the traditional analyses are inadequate too (for instance, why does the traditional /ð/ get dropped in rather and other but not father, brother and mother?).

Here we see the traditional phonemic system breaking down, but no new phonemic system can replace it, as many of these changes are not productive (e.g. the /t/ and /d/ elision is productive, but the /ð/, /v/, or /b/ elision is not). You cannot create new words that reflect the non-productive changes, indicating that they are not integrated into the underlying phonemic system, but at the same time that they are not regular products of the old phonemic system.

The simplest explanation to me seems to be that all of these are simply phonetically reduced pronunciations without phonemic restructuring (in the form of consonant loss); I believe similar elisions can be observed in some accents of Spanish.

Therefore, I'm skeptical of "many of these changes are not productive (e.g. the /t/ and /d/ elision is productive, but the /ð/, /v/, or /b/ elision is not)": can /ð/, /v/, or /b/ really only be totally elided (optionally) in a lexically fixed set of words, and never in any others? By introspection, it doesn't seem implausible to me that /ð/ might be phonetically elided sometimes in brother, mother, father; I haven't looked for measureable evidence yet. Similarly, I feel unsure whether /v/ might be elided for me in beverage in some fast speech context. It's hardly a pronunciation I would be conscious of.

[Travis B.]

The simplest explanation to me seems to be that all of these are simply phonetically reduced pronunciations without phonemic restructuring (in the form of consonant loss); I believe similar elisions can be observed in some accents of Spanish.

Therefore, I'm skeptical of "many of these changes are not productive (e.g. the /t/ and /d/ elision is productive, but the /ð/, /v/, or /b/ elision is not)": can /ð/, /v/, or /b/ really only be totally elided (optionally) in a lexically fixed set of words, and never in any others? By introspection, it doesn't seem implausible to me that /ð/ might be phonetically elided sometimes in brother, mother, father; I haven't looked for measureable evidence yet. Similarly, I feel unsure whether /v/ might be elided for me in beverage in some fast speech context. It's hardly a pronunciation I would be conscious of.

The key thing is that phonetically reduced pronunciations without phonemic restructuring implies that they are productive, i.e. they are surface phonological rules that can produce new pronunciations. For instance, post-tonic intervocalic /t d n nt nd/ elision in the dialect here does not entail any phonemic restructuring by itself, because they not only can apply to any applicable word (there are some environments in which they are disfavored which I won't discuss here, but again this is regular), but they also apply to new words (borrowings or coinages). But the big issue is with /ð v b/ elision because in the dialect here they it only applies to certain words and cannot apply to new words. The reason why this is a problem from a phonemic analysis standpoint is that if one sticks with the conventional analysis one runs into the fact that there are lexicalized rules which cannot be derived from the phonemic layer combined with things such as stress and sandhi. Consequently there are words that just do not fit in the conventional phonemic analysis.

[Darren]

Consequently there are words that just do not fit in the conventional phonemic analysis.

This rarely seems to bother linguists – for instance the analysis of Wichita phonology is entirely content to note that /m/ occurs in two native roots but isn't accepted as phonemic, with no further explanation. Most likely all languages work like this. As I mentioned a while ago, in my own dialect there's no interplay between the TRAP and DRESS vowels except in the roots "thank" and "hello" where they seem to have swapped places; "gday" is subject to elision of /ɾ/ resulting in an unprecedented hiatus [əæi̯] through a change that's resisted pretty strongly everywhere else.

Your dialect seems to be in a state of flux where it's rapidly starting to drift away from the standard, and phonemicise distinctions like consonant and vowel length and a loss of the obstruent voicing contrast. I'd be interested in hearing a recording of your speech if you're happy to share it.

[Travis B.]

Your dialect seems to be in a state of flux where it's rapidly starting to drift away from the standard, and phonemicise distinctions like consonant and vowel length and a loss of the obstruent voicing contrast. I'd be interested in hearing a recording of your speech if you're happy to share it.

It is more complex than just phonemicizing consonant and vowel length and losing obstruent voicing distinction, because underlyingly the obstruent voicing distinction is fully preserved, but it is often transformed into other things. For instance, stop/affricate voicing contrast is reflected not just in vowel length but also in homorganic nasal elision (e.g. homorganic nasals are elided before fortis coda plosives/affricates while lenis coda plosives/affricates do not trigger this), (pre)glottalization, and sandhi (e.g. voicing reappears word-finally when the following word in an utterance starts with a vowel or semivowel). However, this is complicated by intervocalic elisions, because in intervocalic elisions the phonetic length of the preceding vowel is preserved, and if it cannot stand before the following vowel (e.g. it is LOT, TRAP, BATH, DRESS, KIT, STRUT, or FOOT) or in some cases if the following vowel is /i/ or /ə/ the two merge, with the resulting length being the preceding length plus one (i.e. short becomes long and longs becomes overlong).

Note that there are very clear register differences in my dialect, e.g. in high registers speech here is pretty much GA plus the NCVS with different /r l/ realization and a few other minor differences (such as utterance-final and initial (plosive/affricate-only) devoicing and light palatalization in some environments), while in everyday speech it can be quite different from GA. For instance, when I am on the phone or I am speaking with non-native English-speakers who I perceive as having trouble with my speech I automatically switch into my high register (e.g. I even try to clearly voice initial lenis plosives/affricates), but in everyday speech I aggressively elide and assimilate.

Also note that I have found that I do not have many problems communicating with other Americans from elsewhere or even with many foreign-born individuals here in the US who have been here for a while; I do not think that my dialect is as different from other NAE varieties as it may seem like from posts like these. Case in point, flap elision seems to be a very common feature in NAE overall, as is /n/ and /nt/ flapping (which opens them up to elision). Also note that in NAE rounded vowel fronting is extremely common (despite other people's transcriptions); my idiolect is actually kind of odd in that it preserves mid and high rounded back vowels when not preceded by coronals or palatals (and this may be dying out to a degree, e.g. my daughter tends to centralize or front mid and high rounded vowels not followed by /l/ even when preceded by labials). A very good way to sound foreign here is in fact to preserve [u ʊ o̞~o̞ʊ] consistently; this will make you clearly sound like you are not a native NAE-speaker.

[Darren]

Your dialect seems to be in a state of flux where it's rapidly starting to drift away from the standard, and phonemicise distinctions like consonant and vowel length and a loss of the obstruent voicing contrast. I'd be interested in hearing a recording of your speech if you're happy to share it.

It is more complex than just phonemicizing consonant and vowel length and losing obstruent voicing distinction, because underlyingly the obstruent voicing distinction is fully preserved, but it is often transformed into other things. For instance, stop/affricate voicing contrast is reflected not just in vowel length but also in homorganic nasal elision (e.g. homorganic nasals are elided before fortis coda plosives/affricates while lenis coda plosives/affricates do not trigger this), (pre)glottalization, and sandhi (e.g. voicing reappears word-finally when the following word in an utterance starts with a vowel or semivowel). However, this is complicated by intervocalic elisions, because in intervocalic elisions the phonetic length of the preceding vowel is preserved, and if it cannot stand before the following vowel (e.g. it is LOT, TRAP, BATH, DRESS, KIT, STRUT, or FOOT) or in some cases if the following vowel is /i/ or /ə/ the two merge, with the resulting length being the preceding length plus one (i.e. short becomes long and longs becomes overlong).

That suggests to me that stop voicing isn't going to last very much longer. Not that predicting sound changes is really legit, but it seems like enough information has been offloaded onto other segments that the voicing is fairly redundant. Didn't you make a future English conlang which did that?

Note that there are very clear register differences in my dialect, e.g. in high registers speech here is pretty much GA plus the NCVS with different /r l/ realization and a few other minor differences (such as utterance-final and initial (plosive/affricate-only) devoicing and light palatalization in some environments), while in everyday speech it can be quite different from GA. For instance, when I am on the phone or I am speaking with non-native English-speakers who I perceive as having trouble with my speech I automatically switch into my high register (e.g. I even try to clearly voice initial lenis plosives/affricates), but in everyday speech I aggressively elide and assimilate.

Do you try and voice initial lenis plosives only because of your linguistic awareness, or is this something everyone there does? Are people aware that they pronounce these sounds differently from the "standard"?

Also note that I have found that I do not have many problems communicating with other Americans from elsewhere or even with many foreign-born individuals here in the US who have been here for a while; I do not think that my dialect is as different from other NAE varieties as it may seem like from posts like these.

That's why I'd be keen in hearing you speak. From your transcriptions it sounds like you're speaking a completely different language, but I imagine in connected speech it would be easier. Plus I'm probably unconsciously partly accustomed to GenAm accents from TV anyway.

Also note that in NAE rounded vowel fronting is extremely common (despite other people's transcriptions); my idiolect is actually kind of odd in that it preserves mid and high rounded back vowels when not preceded by coronals or palatals (and this may be dying out to a degree, e.g. my daughter tends to centralize or front mid and high rounded vowels not followed by /l/ even when preceded by labials). A very good way to sound foreign here is in fact to preserve [u ʊ o̞~o̞ʊ] consistently; this will make you clearly sound like you are not a native NAE-speaker.

That I can well believe. In fact it seems fairly universal in English dialects to front non-low back vowels; my GOOSE and GOAT vowels have [y̯] offglides

[Richard W]

...For instance, many of the distinctions merged in the Mary-merry-marry merger have been resurrected in a marginal fashion through consonant elisions and resulting vowel cluster reductions...

Surely this is only an issue if the merger is still active. Is it?

You cannot create new words that reflect the non-productive changes, indicating that they are not integrated into the underlying phonemic system, but at the same time that they are not regular products of the old phonemic system.

There's always borrowing. For example, I would pronounce the first word of 'Sudeiri Seven' as /sʊdeɪriː/ and not be bothered by the sound change that gives /ɛː/ instead of /eɪ/ before /r/.

[Travis B.]

That suggests to me that stop voicing isn't going to last very much longer. Not that predicting sound changes is really legit, but it seems like enough information has been offloaded onto other segments that the voicing is fairly redundant. Didn't you make a future English conlang which did that?

The key thing is that this is all fully productive except that some elisions are not productive, and one cannot intentionally produce words with a certain vowel length, nasal elision, or (pre)glottalization, indicating that in this regard the old phonemic system is still in place.

Do you try and voice initial lenis plosives only because of your linguistic awareness, or is this something everyone there does? Are people aware that they pronounce these sounds differently from the "standard"?

I am not sure; I do this in particular on the phone, and I have long been aware that unaspirated voiceless versus aspirated voiceless initial plosives/affricates are not distinguished well on the phone. When saying my last name on the phone I even always say "B as in 'boy'" to make it clear that my last name starts with a /b/ and not a /p/.

That's why I'd be keen in hearing you speak. From your transcriptions it sounds like you're speaking a completely different language, but I imagine in connected speech it would be easier. Plus I'm probably unconsciously partly accustomed to GenAm accents from TV anyway.

That I can well believe. In fact it seems fairly universal in English dialects to front non-low back vowels; my GOOSE and GOAT vowels have [y̯] offglides

To me honestly I think a lot of present-day Americans are somewhat deluded as to what their own phonological system is. This is a big part of why I am interested in NAE phonology ─ I years back came to the conclusion that most Americans simply do not understand the phonology of their own native language. For instance, most people mark rounded non-low vowels as back vowels, when I so frequently hear /u/ as [y] from Americans and when I hear real [u] from someone when it follows a coronal it is almost always from a foreigner.

[Travis B.]

...For instance, many of the distinctions merged in the Mary-merry-marry merger have been resurrected in a marginal fashion through consonant elisions and resulting vowel cluster reductions...

Surely this is only an issue if the merger is still active. Is it?

The merger is still active in that new words and loanwords still obey the Mary-merry-marry merger except where productive elisions such as the post-tonic intervocalic /t d n nt nd/ elisions apply (i.e. allowing non-merged vowels to come into contact with /r/).

You cannot create new words that reflect the non-productive changes, indicating that they are not integrated into the underlying phonemic system, but at the same time that they are not regular products of the old phonemic system.

There's always borrowing. For example, I would pronounce the first word of 'Sudeiri Seven' as /sʊdeɪriː/ and not be bothered by the sound change that gives /ɛː/ instead of /eɪ/ before /r/.

See, elisions aside, loans still obey the Mary-merry-marry merger.

[Estav]

/ð v b/ elision [...] cannot apply to new words

I'm saying I somewhat doubt that this is true, but obviously I can't be certain of that. Of course, you have a better chance of accurately describing your own accent than I do, but I think it's possible that you might not have perfect awareness of every non-phonemic elision that may occur in your speech or that of those around you. I assume that it is possible to optionally include [ð] in other, rather or [v] in every; am I right?

There's always borrowing. For example, I would pronounce the first word of 'Sudeiri Seven' as /sʊdeɪriː/ and not be bothered by the sound change that gives /ɛː/ instead of /eɪ/ before /r/.

I don't have an issue with pronouncing /eɪr/ in that kind of context: it also occurs in words like payroll and playroom.

[Travis B.]

/ð v b/ elision [...] cannot apply to new words

I'm saying I somewhat doubt that this is true, but obviously I can't be certain of that. Of course, you have a better chance of accurately describing your own accent than I do, but I think it's possible that you might not have perfect awareness of every non-phonemic elision that may occur in your speech or that of those around you. I assume that it is possible to optionally include [ð] in other, rather or [v] in every; am I right?

Yes, the elisions are optional in those words, and in higher registers they are absent.

There's always borrowing. For example, I would pronounce the first word of 'Sudeiri Seven' as /sʊdeɪriː/ and not be bothered by the sound change that gives /ɛː/ instead of /eɪ/ before /r/.

I don't have an issue with pronouncing /eɪr/ in that kind of context: it also occurs in words like payroll and playroom.

Actually, I do have /e/ in these word, probably because of the morpheme boundary.