The vowels in North American "mare" and "near" are usually transcribed as [mEr] and [nIr]. "Mary/merry/marry" and "nearer" are likewise usually transcribed as [mEri] and [nIr@`]. Obviously, after the pre-rhotic vowel mergers have occurred, there is no *phonemic* need to employ [E@] and [I@] when you could just use [E] or [I] in these contexts. But do you think that phonetically, [E@] and [I@] might be better representations of the North American merged vowels? I lack the pre-rhotic vowel mergers, and when [Er] and [Ir] occur in my speech (as in "merry" or "mirror"), they are very short, very lax, and completely monophthongal, and they can never occur word-finally. Just from my own observations, it seems that when most North Americans say "mare", it's closer to my own "mare" [mE@`] than to my "merry" minus the [i]. Any thoughts?
Pre-rhotic AmEng vowels
Lazar, well, I think such should only be used for dialects in which merger is not complete, and then at that only for the unmerged prerhotic phonemes. However, I would be tempted to not use /E@/ or /I@/ but rather just /e/ and /i/, treating [E@] and [I@] as prerhotic *allophones* of such, especially considering that they formerly were [e] and [i], and [E@] and [I@] cannot be found anywhere else in any rhotic NAE dialects (at least any I know of) except for possibly some NCVS dialects (even though I normally near of [e@] rather than [E@] in that case).
The reason for this is that the distributions of [e] and [E@], and [i] and [I@], are complementary, and cannot be used to form any minimal pairs, besides possible [E@] for historical /{/ in some NCVS-affected dialects. Therefore, there is no reason to introduce separate phonemes /E@/ and /I@/ in such cases, and doing so would only result in an unnecessarily more complex phoneme system which is not substantiated phonologically or historically. Remember that phonemes should *not* represent anything that necessarily at all directly represents what is actually realized, but rather should be able to represent *only* any possible minimal pairs and anything that might be in the lexicon or produced morphologically at the *logical* level. Hence, at this level, there is no reason to separate [e] from [E@] or [i] from [I@], as long as they are clearly in complementary distribution.
The reason for this is that the distributions of [e] and [E@], and [i] and [I@], are complementary, and cannot be used to form any minimal pairs, besides possible [E@] for historical /{/ in some NCVS-affected dialects. Therefore, there is no reason to introduce separate phonemes /E@/ and /I@/ in such cases, and doing so would only result in an unnecessarily more complex phoneme system which is not substantiated phonologically or historically. Remember that phonemes should *not* represent anything that necessarily at all directly represents what is actually realized, but rather should be able to represent *only* any possible minimal pairs and anything that might be in the lexicon or produced morphologically at the *logical* level. Hence, at this level, there is no reason to separate [e] from [E@] or [i] from [I@], as long as they are clearly in complementary distribution.
Well the main reason why I prefer to use separate /E@/ and /I@/ for unmerged speech, rather than just /e/ and /i/, is because sequences like /er/ and /ir/ *could* possibly occur in my speech, and sometimes do occur across word boundaries. Compare:
"he rows" (emphasizing that it is *he*) - ["hir\oz]
"heroes" - ["hI@`oz]
"he rows" (emphasizing that it is *he*) - ["hir\oz]
"heroes" - ["hI@`oz]
<<Lazar, well, I think such should only be used for dialects in which merger is not complete, and then at that only for the unmerged prerhotic phonemes.>>
But as I said above, I'm speaking in purely *phonetic* terms - I'm not proposing any changes in phonemes. I just think that phonetic transcriptions like [nIr@`] for merged "nearer", for example, can be misleading because that word may in reality be closer to unmerged "nearer" [nI@`@`] than to unmerged [mIr@`]. Mind you, this would only really apply in a very narrow phonetic context in which different English dialects are being compared, but I still think that [E@`] and [I@`] might be more accurate for merged NAE than the current [Er] and [Ir].
But as I said above, I'm speaking in purely *phonetic* terms - I'm not proposing any changes in phonemes. I just think that phonetic transcriptions like [nIr@`] for merged "nearer", for example, can be misleading because that word may in reality be closer to unmerged "nearer" [nI@`@`] than to unmerged [mIr@`]. Mind you, this would only really apply in a very narrow phonetic context in which different English dialects are being compared, but I still think that [E@`] and [I@`] might be more accurate for merged NAE than the current [Er] and [Ir].