English vowel systems and lexical sets

[Richard W]

I personally don't really like 'cute' analyses such as these, which try to solve so many phonological problems in an analysis, yet open up so many other problems that are just as bad as the problems they attempt to solve. I prefer more conventional analyses in this regard, even if they have open questions like why does intrusive-r get inserted after some vowels and not others.

I was hoping to see some assertions about syllable structure, and whether it is phonemic.

I'm not convinced that linking-r has to be phonological - I remember my mother being somewhat surprised that her (English, non-rhotic) primary school pupils insisted on pronouncing drawing with word-internal intrusive-r. And of course, the indefinite article a is usually an before vowels, though it is occasionally uttered with intrusive-r instead.

[Nortaneous]

What KIT/FLEECE merger? These varieties just violate the proposed universal that consonants are preferentially syllabified into onsets, which most English varieties violate anyway unless you bring in ambisyllabicity. FLEECE has a tense realization of /i/ and a [j] offglide; NEAR has a tense realization of /i/ and a [r] offglide. -jl is a prohibited coda just like -jr. Some American varieties also prohibit -wl, so "school" is disyllabic, although this is complicated by the emergence of a monophthongal /o/ phoneme distinct from THOUGHT in words like "both", sometimes misspelled "bolth" for this reason.

"few" can't have /iw/ unless you posit that word-initial NEW is actually /jiw/, because e.g. a useful thing (not *an useful thing).

I would precisely analyze most NAE varieties at least as having ambisyllabicity, for the reason that flapping of /t d/, and in applicable varieties /n nt/, patterns with where ambisyllabicity would be expected.

Not intervocalically except in a stressed syllable onset? (with secondary stress for words like "autism")

I have some ideas about analyzing English high vowels.
<snip>
"fit" /fit/
"feat" /fijt/
"few" /fiw/
"fear" /fir/

English is not a monosyllabic language, so please explain how you plan to analyse words like virulent and cirrus. Also give a thought to the difference between the verb tarry (rimes with carry) and the adjective tarry (rimes with starry). Are you planning to allow a distinction between the pronunciations of Sirius and serious?

If not /vir.jə.lənt/, /vir.ə.lənt/. The clouds are /sir.əs/.

Leaving aside -y, which could just as well be /-əj/ but let's write /-ij/, /ter.ij/ vs. /tar.ij/.

Sirius and serious are merged.

I can't think of any reason in the NAE variety I'm most familiar with to prefer a diphthongal analysis to a VC one. AFAIK the only trouble is that it leaves unexplained why FATHER, THOUGHT, and NURSE can appear without a coda consonant. (NURSE, unlike Vr sequences, doesn't condition schwa epenthesis when followed by /l/ - the surnames Carroll and Gerl only differ in the voicing of their onset, and I remember being confused as an elementary schooler about why Carl was pronounced with two syllables when it was written with one. /ar/ requires the least tongue movement of any of the Vr sequences, so it's likely the least prone to schwa epenthesis on articulatory grounds, but I don't think there are other words ending in /-arl/ in common spoken use, so monosyllabic pronunciations of e.g. "snarl" could be spelling pronunciations or loans from a higher register.)

I was hoping to see some assertions about syllable structure, and whether it is phonemic.

It might be possible to argue this using allophony of /æ/ after nasals.

Assume [ŋ] is /n/ before velars, with assimilation of /ng/. This is historically true and certain NAE varieties often have at least phonetic postplosion in -ing. /æ/ is [eə] before [m n], but [æj] before [ŋ]. Now consider the near-minimal pair [bæjŋkɚ] and [peəŋko̽w] - /bænk.ər/ ~ /pæn.kəw/? I've heard the surname Bernanke pronounced both [bɚnæjŋki̽j] and [bɚneəŋki̽] - /bər.nænk.ij/ ~ /bər.næn.kij/?

I don't know how to deal with this without either syllable structure or morphemes that have to be imputed when an analysis is unavailable.

[Travis B.]

What KIT/FLEECE merger? These varieties just violate the proposed universal that consonants are preferentially syllabified into onsets, which most English varieties violate anyway unless you bring in ambisyllabicity. FLEECE has a tense realization of /i/ and a [j] offglide; NEAR has a tense realization of /i/ and a [r] offglide. -jl is a prohibited coda just like -jr. Some American varieties also prohibit -wl, so "school" is disyllabic, although this is complicated by the emergence of a monophthongal /o/ phoneme distinct from THOUGHT in words like "both", sometimes misspelled "bolth" for this reason.

"few" can't have /iw/ unless you posit that word-initial NEW is actually /jiw/, because e.g. a useful thing (not *an useful thing).

I would precisely analyze most NAE varieties at least as having ambisyllabicity, for the reason that flapping of /t d/, and in applicable varieties /n nt/, patterns with where ambisyllabicity would be expected.

Not intervocalically except in a stressed syllable onset? (with secondary stress for words like "autism")

I analyze most NAE varieties as having ambisyllabic consonants in all intervocalic positions except where the following syllable is stressed (unless the consonant is in a final position in the morpheme it belongs to and the following morpheme starts with a vowel, including across word boundaries). It neatly explains why flapping occurs where it does, and in the case of my own lect it fits my intuition that intervocalic consonants in such positions belong to both the preceding and following syllables.

I have some ideas about analyzing English high vowels.
<snip>
"fit" /fit/
"feat" /fijt/
"few" /fiw/
"fear" /fir/

English is not a monosyllabic language, so please explain how you plan to analyse words like virulent and cirrus. Also give a thought to the difference between the verb tarry (rimes with carry) and the adjective tarry (rimes with starry). Are you planning to allow a distinction between the pronunciations of Sirius and serious?

If not /vir.jə.lənt/, /vir.ə.lənt/. The clouds are /sir.əs/.

Leaving aside -y, which could just as well be /-əj/ but let's write /i/, /ter.i/ vs. /tar.i/.

Sirius and serious are merged.

I can't think of any reason in the NAE variety I'm most familiar with to prefer a diphthongal analysis to a VC one. AFAIK the only trouble is that it leaves unexplained why FATHER, THOUGHT, and NURSE can appear without a coda consonant. (NURSE, unlike Vr sequences, doesn't condition schwa epenthesis when followed by /l/ - the surnames Carroll and Gerl only differ in the voicing of their onset, and I remember being confused as an elementary schooler about why Carl was pronounced with two syllables when it was written with one. /ar/ requires the least tongue movement of any of the Vr sequences, so it's likely the least prone to schwa epenthesis on articulatory grounds, but I don't think there are other words ending in /-arl/ in common spoken use, so monosyllabic pronunciations of e.g. "snarl" could be spelling pronunciations or loans from a higher register.)

I don't like VC analyses for the English I am familiar with in part because the "short" vowels are equal in length to the "long" vowels analyzed as VC, and this would be true for almost all NAE varieties, whereas unambiguous VC such as /Vl/ and /Vr/ are together longer than the vowel in isolation. (Edit: There is the exception here of what I would synchronically analyze as /ər/, which finally and before a consonant is equal to a normal vowel in length, but is longer, and ends closer, when followed by a vowel.)

As for /rVl/ and /rl/, Carroll for me is unambiguously disyllabic while Carl is unambiguously monosyllabic. Likewise, barrel is disyllabic while snarl is monosyllabic. Trying to make Carl and snarl disyllabic sounds wrong to me, as if I were unnaturally drawing out the words.

[Darren]

I've mentioned my AusEng VC analysis several times before and I still think it works. It works very well since there's pervasive intrusive /r/, a complete ban on vowel hiatus, and genuine long vowels and diphthongs. The only sticky point is things like the Sirius/serious, courier/scoria, merry/Mary and Surrey/sari contrasts – I'd analyse them as /ˈsirəjəs~ˈsirrəjəs, ˈkʊrəjər~ˈskʊrrəjər, ˈmerəj~ˈmerrəj, ˈsarəj~ˈsarrəj/, but they do require geminates which don't otherwise crop up outside of compounds.

As for ambisyllabicity, I don't think there's any convincing argument for which syllable an intervocalic consonant begins. Sure there's t-flapping, but then compare [bæt ~ bæɾɜ] "bat/batter" vs. [hɐʊ̯ɫʷ ~ hɐʊ̯ɫʷɘj] "hole/holey", [staː ~ staːɻɘj] "star/starry" (not *[ˈhɐy̯ɫʷɘj] or *[ˈstaɻɘj]) - there's a tendency to treat morpheme-final intervocalic consonants as belonging to the previous syllable, yet this doesn't block flapping. Really the analysis is more "any nucleus must be followed by a consonant" which is true.

[Travis B.]

I've mentioned my AusEng VC analysis several times before and I still think it works. It works very well since there's pervasive intrusive /r/, a complete ban on vowel hiatus, and genuine long vowels and diphthongs. The only sticky point is things like the Sirius/serious, courier/scoria, merry/Mary and Surrey/sari contrasts – I'd analyse them as /ˈsirəjəs~ˈsirrəjəs, ˈkʊrəjər~ˈskʊrrəjər, ˈmerəj~ˈmerrəj, ˈsarəj~ˈsarrəj/, but they do require geminates which don't otherwise crop up outside of compounds.

As for ambisyllabicity, I don't think there's any convincing argument for which syllable an intervocalic consonant begins. Sure there's t-flapping, but then compare [bæt ~ bæɾɜ] "bat/batter" vs. [hɐʊ̯ɫʷ ~ hɐʊ̯ɫʷɘj] "hole/holey", [staː ~ staːɻɘj] "star/starry" (not *[ˈhɐy̯ɫʷɘj] or *[ˈstaɻɘj]) - there's a tendency to treat morpheme-final intervocalic consonants as belonging to the previous syllable, yet this doesn't block flapping. Really the analysis is more "any nucleus must be followed by a consonant" which is true.

Does your analysis of BAD versus LAD involve /r/?

[Richard W]

I have some ideas about analyzing English high vowels.
<snip>
"fit" /fit/
"feat" /fijt/
"few" /fiw/
"fear" /fir/

English is not a monosyllabic language, so please explain how you plan to analyse words like virulent and cirrus. Also give a thought to the difference between the verb tarry (rimes with carry) and the adjective tarry (rimes with starry). Are you planning to allow a distinction between the pronunciations of Sirius and serious?

If not /vir.jə.lənt/, /vir.ə.lənt/. The clouds are /sir.əs/.

Leaving aside -y, which could just as well be /-əj/ but let's write /-ij/, /ter.ij/ vs. /tar.ij/.

Sirius and serious are merged.

OK, so for some alien version of English, rather than the standard lects in general.

I was hoping to see some assertions about syllable structure, and whether it is phonemic.

It might be possible to argue this using allophony of /æ/ after nasals.

The initial analysis I quoted above would have relied on it to phonetically contrast Sirius and Serious.

[Nortaneous]

English is not a monosyllabic language, so please explain how you plan to analyse words like virulent and cirrus. Also give a thought to the difference between the verb tarry (rimes with carry) and the adjective tarry (rimes with starry). Are you planning to allow a distinction between the pronunciations of Sirius and serious?

If not /vir.jə.lənt/, /vir.ə.lənt/. The clouds are /sir.əs/.

Leaving aside -y, which could just as well be /-əj/ but let's write /-ij/, /ter.ij/ vs. /tar.ij/.

Sirius and serious are merged.

OK, so for some alien version of English, rather than the standard lects in general.

Unmarked GenAm as spoken by the capital-area professional-managerial class - what more of a standard is there? (Schwa epenthesis in /arl/ is substandard but common - don't they do that in Aqua Teen Hunger Force?) The Sirius-serious merger is widespread. The schwa-schwi distinction can probably be analyzed away with reference to morpheme boundaries, and HAPPY often has a phonetically perceptible offglide, so there's no reason not to analyze it as an unstressed vowel + /j/, i.e. /əj/.

[Travis B.]

English is not a monosyllabic language, so please explain how you plan to analyse words like virulent and cirrus. Also give a thought to the difference between the verb tarry (rimes with carry) and the adjective tarry (rimes with starry). Are you planning to allow a distinction between the pronunciations of Sirius and serious?

If not /vir.jə.lənt/, /vir.ə.lənt/. The clouds are /sir.əs/.

Leaving aside -y, which could just as well be /-əj/ but let's write /-ij/, /ter.ij/ vs. /tar.ij/.

Sirius and serious are merged.

OK, so for some alien version of English, rather than the standard lects in general.

Aside from that I question whether GA has a glide in HAPPY, this is bog-standard GA. To you "standard lects" of English must be SSBE, it seems.

[Darren]

Does your analysis of BAD versus LAD involve /r/?

Ah yes, you've hit on the other snag with this system. Analysing "bad" as /bærd/ is slightly problematic but on the other hand it doesn't bring up any functional problems.

[anteallach]

It might be possible to argue this using allophony of /æ/ after nasals.

Assume [ŋ] is /n/ before velars, with assimilation of /ng/. This is historically true and certain NAE varieties often have at least phonetic postplosion in -ing. /æ/ is [eə] before [m n], but [æj] before [ŋ]. Now consider the near-minimal pair [bæjŋkɚ] and [peəŋko̽w] - /bænk.ər/ ~ /pæn.kəw/? I've heard the surname Bernanke pronounced both [bɚnæjŋki̽j] and [bɚneəŋki̽] - /bər.nænk.ij/ ~ /bər.næn.kij/?

I don't know how to deal with this without either syllable structure or morphemes that have to be imputed when an analysis is unavailable.

Easy, in a VC analysis: /bæjnkər/ vs. /pænkəw/. This also fits with my suspicion that what is going on here is that newer loanwords aren't being subjected to an older sound change, thus phomemicising it.

I am personally sceptical about attempts to get rid of /ŋ/; whatever the rules are, they just seem to have too many irregularities. And that's coming from someone who quite often has word-final [ŋg].

[Travis B.]

It might be possible to argue this using allophony of /æ/ after nasals.

Assume [ŋ] is /n/ before velars, with assimilation of /ng/. This is historically true and certain NAE varieties often have at least phonetic postplosion in -ing. /æ/ is [eə] before [m n], but [æj] before [ŋ]. Now consider the near-minimal pair [bæjŋkɚ] and [peəŋko̽w] - /bænk.ər/ ~ /pæn.kəw/? I've heard the surname Bernanke pronounced both [bɚnæjŋki̽j] and [bɚneəŋki̽] - /bər.nænk.ij/ ~ /bər.næn.kij/?

I don't know how to deal with this without either syllable structure or morphemes that have to be imputed when an analysis is unavailable.

Easy, in a VC analysis: /bæjnkər/ vs. /pænkəw/. This also fits with my suspicion that what is going on here is that newer loanwords aren't being subjected to an older sound change, thus phomemicising it.

I am personally sceptical about attempts to get rid of /ŋ/; whatever the rules are, they just seem to have too many irregularities. And that's coming from someone who quite often has word-final [ŋg].

To avoid recent loanwords and names, there is contrast in my own dialect between banker /ˈbeɪnkər/ [ˈb̥ẽ̞ŋkʁ̩ˤ(ː)] and pancake /ˈpænkeɪk/ [ˈpʰɛ̃ŋke̞ʔk/; I also have Pankow /ˈpæŋkoʊ/ [ˈpʰɛ̃ŋko̞(ː)] (which can also be /ˈpɑŋkoʊ/, or shall I say /ˈpaŋkoʊ/, [ˈpʰãŋko̞(ː)]) and Bernanke /ˌbərˈnænki/ [ˌb̥ʁ̩̃ˤːˈnɛ̃ŋci(ː)].

[Nortaneous]

It might be possible to argue this using allophony of /æ/ after nasals.

Assume [ŋ] is /n/ before velars, with assimilation of /ng/. This is historically true and certain NAE varieties often have at least phonetic postplosion in -ing. /æ/ is [eə] before [m n], but [æj] before [ŋ]. Now consider the near-minimal pair [bæjŋkɚ] and [peəŋko̽w] - /bænk.ər/ ~ /pæn.kəw/? I've heard the surname Bernanke pronounced both [bɚnæjŋki̽j] and [bɚneəŋki̽] - /bər.nænk.ij/ ~ /bər.næn.kij/?

I don't know how to deal with this without either syllable structure or morphemes that have to be imputed when an analysis is unavailable.

Easy, in a VC analysis: /bæjnkər/ vs. /pænkəw/. This also fits with my suspicion that what is going on here is that newer loanwords aren't being subjected to an older sound change, thus phomemicising it.

I am personally sceptical about attempts to get rid of /ŋ/; whatever the rules are, they just seem to have too many irregularities. And that's coming from someone who quite often has word-final [ŋg].

Then you have /æj/ appearing only before /ŋ g/. But maybe this is right - merger of /æj/ into /ej/ is common, and there's also e > ej in the same environment, which has been phonemicized by newer loanwords, e.g. egg /ejg/ vs. Genghis /gengis/.

There's an alternate syllabification-based explanation, but I think it can be ruled out. Based on egg vs. Genghis, it could be that the pre-raising forms were /eg/ and /gen.gᵻs/, and the environment for raising is actually / _(n)g$ - that is, the voiced velar has to appear in the coda. But doesn't penguin have FACE in these varieties?

To avoid recent loanwords and names, there is contrast in my own dialect between banker /ˈbeɪnkər/ [ˈb̥ẽ̞ŋkʁ̩ˤ(ː)] and pancake /ˈpænkeɪk/ [ˈpʰɛ̃ŋke̞ʔk/; I also have Pankow /ˈpæŋkoʊ/ [ˈpʰɛ̃ŋko̞(ː)] (which can also be /ˈpɑŋkoʊ/, or shall I say /ˈpaŋkoʊ/, [ˈpʰãŋko̞(ː)]) and Bernanke /ˌbərˈnænki/ [ˌb̥ʁ̩̃ˤːˈnɛ̃ŋci(ː)].

"Pancake" also works, but nasal assimilation isn't obligatory in noun-noun compounds.

[Richard W]

Aside from that I question whether GA has a glide in HAPPY, this is bog-standard GA. To you "standard lects" of English must be SSBE, it seems.

No, but "standard lects" would include both GA and SSBE. Note that Wells' lexical sets aimed to cover both, and some monosyllables failed to be included. Some fail because of incomplete mergers, and literacy wreaks havoc.

[Travis B.]

Aside from that I question whether GA has a glide in HAPPY, this is bog-standard GA. To you "standard lects" of English must be SSBE, it seems.

No, but "standard lects" would include both GA and SSBE. Note that Wells' lexical sets aimed to cover both, and some monosyllables failed to be included. Some fail because of incomplete mergers, and literacy wreaks havoc.

Aside from Nort's description of HAPPY, what he was describing was plain old GA, which for a majority of people has the serious-Sirius merger, the full Mary-merry-marry merger, and the weak vowel merger. You seem to be forgetting that GA generally maintains fewer vowel distinctions before historical /r/ than SSBE (aside from that many GA speakers don't merge tour with tore).