What is the origin of “long long i” before voiced consonants (the [ai] of wide, while, & tribe) versus “short long i” before unvoiced consonants (the [ʌi] of white, wife, & wipe)? When did this difference appear? Is it just a secondary distinction, or are there minimal pairs for these two diphthongs?
Obviously the distinction is not present in all accents—though its presence in most leads me to believe it's not a late arrival—and yet I've never found a dictionary that distinguishes them.
I believe this is called Canadian Raising.
Canadian raising is a phonetic phenomenon that occurs in varieties of the English language, especially Canadian English, in which certain diphthongs are "raised" before voiceless consonants (e.g., /p/, /t/, /k/, /s/, /f/).
/aɪ/ (the vowel of "eye") becomes [ʌɪ] or [ɐɪ], while the outcome of /aʊ/ (the vowel of "loud") varies by dialect, with [ʌu] more common in the west and a fronted variant [ɛʉ] commonly heard in Central Canada. In any case, the /a/-component of the diphthong changes from a low vowel to a mid-low vowel ([ʌ], [ɐ] or [ɛ]).
Below is an excerpt of the possible origin paragraph in the Wikipedia article.
Some have hypothesized that Canadian raising may be related historically to a similar phenomenon that exists in Scots and Scottish English. The Scottish Vowel Length Rule lengthens a wide variety of vowel sounds in several environments, and shortens them in others; "long" environments include when the vowel precedes a number of voiced consonant sounds. This rule also conditions /aɪ/ in the long environments and /əɪ/ in the short environments. Significantly, though, the Scots Vowel Length Rule applies only before voiced fricatives and /r/, whereas Canadian raising is not limited in this fashion; thus, it may represent a sort of merging of the Scots Vowel Length Rule with the general English rule lengthening vowels before voiced consonants of any sort.
The most common understanding of the Great Vowel Shift is that the Middle English vowels [iː, uː] passed through a stage [əɪ, əʊ] on the way to their modern pronunciations [aɪ, aʊ]. Thus it is difficult to say whether Canadian raising reflects an innovation or the preservation of an older vowel quality in a restricted environment.
In most dialects of American English, vowels are shorter before unvoiced consonants. If this is what you're hearing, then in your dialect the length difference seems to have induced a change in the quality of the diphthong as well. I believe that in some dialects, this determines whether long o and long a are diphthongs or not.
To find a minimal pair, you'd have to find two words with VCV, where the first vowel is long i, the consonant is unvoiced and the division into syllables differs. For example, I think bright-eyed and high tide form a near-minimal pair.
As a native of East Yorkshire, England, this difference has always interested me. In that area, we have the same difference in diphthong-length between, say, 'site' and 'side' as in RP English. But in WEST Yorkshire, they're both pronounced long. I'd always thought that it was semantically unimportant until I realized that in some varieties of North American English, the length really matters. Oddly, it was the writer/rider pair that I noticed. This difference seems to be maintained even when a non-English [in that position, at least] consonant is involved: ask most English speakers to say "Third Reich", and they'll say "Reich" with a 'short long-I' /ɑɪ/, because the 'ch' /ç/ is an unvoiced consonant.
I'd always learned it was allophonic and based on voicing, but I know of one minimal pair: -sider (e.g. "this DVD is a two-sider") and cider: Realized the old pun "Dixon Cider" doesn't work in my (western Montana) dialect any more than "I scream for ice cream". I'm not going to transcribe them here; it seems like anyone who could explain them, should know which is which!
