March 20, 2005

Raising and lowering those tighty whities

We're talking about vowels here, of course. At least, we will be, once some preliminaries are out of the way.

In his post on tighty/tidy whities, Arnold Zwicky mentioned that tighty and tidy are pronounced almost but not quite the same, due to the flapping and voicing of the intervocalic (i.e. between-vowel) /t/ and /d/. This is worth discussing a bit, at least for those who are interested in English pronunciation.

First issue: flapping and voicing. In nearly all varieties of North American English, /t/, /d/ and /n/ turn into a short tongue-tip tap when they precede a vowel and are not in the onset of a stressed syllable. In the same contextss, /t/ also becomes "voiced", i.e. it loses the feature that distinguishes it from /d/. Here are some examples of such contexts for /t/:

Flapping/voicing happens: Flapping/voicing doesn't happen:
at all
a tall

As a result, words that are historically distinguished by /t/ vs. /d/ can become essentially homophonous. In my own pronunciation, for example, latter and ladder are homophones, unless I'm trying hard to convey the distinction (which I would do by artificially suppressing the voicing process).

Second issue: vowel length before voiceless vs. voiced consonants. In most dialects of English, the voicing of syllable-final consonants makes a big difference to the duration of the previous vowel, especially in phrase-final stressed syllables. For example, in a fully-stressed syllable at the end of a phrase, the vowel in mad might be half again as long as the vowel in mat. This effect is much smaller in syllables that are medial and/or less stressed.

For some speakers, there can be a sort of residue of this difference in vowel length in cases where the consonant voicing has disappeared, so that the first vowel in bedding may be a little bit longer, on average, than the first vowel in betting. But the effect, where it exists, is typically very small relative to the normal variation in vowel duration for other reasons. In fact, it's hard to find clear evidence for this effect, outside of cases where facultative disambiguation may be involved. In my own speech, I'm pretty sure that betting and bedding are indistinguishable, absent a special effort to convey the difference.

Third issue: vowel quality change in /aj/ before voiceless vs. voiced. In some North American dialects, certain vowels are different in quality when voiceless vs. voiced consonants follow. The vowel most commonly affected is "long i" -- the vowel in five vs. fife, tide vs. tight, bide vs. bite, etc. In the affected dialects, this vowel is raised and fronted in the pre-voiceless cases. For these speakers, the diphthong in fife starts out near the vowel of bud, and ends near the vowel of bade; while the diphthong in five starts near the vowel of hod, and ends nears the vowel of hed. In IPA, this is something like the difference between [fʌef] and [fɑɛv].

(If you want to see what this means in terms of acoustic measurements,here was a discussion of this on phonoloblog last summer (here and here), in which Bob Kennedy and I exchanged plots of our formant trajectories (time functions of resonance frequencies) in fife vs. five.)

This vowel quality distinction is generally not affected by flapping/voicing of /t/. As a result, in my own speech, the same quality difference that appears in write vs. ride also appears in writer vs. rider. As a result, while latter and ladder are homophones for me, writer and rider are definitely not. The original /t/ and /d/ are both neutralized to a voiced tap [ɾ], but the initital-syllable diphthongs are very different.

(This change in "long i" before voiceless consonants is sometimes called "Canadian raising", but a similar change is found in many U.S. dialects as well. Back in 1942, Martin Joos claimed that Ontario speakers divided into two groups, one of which maintained the vowel quality distinction in spite of flapping and voicing, while the other group didn't; however, I've never been able to find any speakers of his second type, who would raise the diphthong in write but not in writer.)

Now we're in a position to discuss what Arnold said about the pronunciation of tighty and tidy:

On the pronunciation front: tighty and tidy get to be (almost) the same in pronunciation in American English via intervocalic flapping, which plays a role in a large number of reinterpretations, and plain spelling errors too. Interestingly, the two words aren't necessarily pronounced exactly the same, even if they both have an intervocalic flap. Full neutralization at the word level turns out to be rarer than people used to think; often there's some cue as to the "real" nature of the neutralized segment. For tighty vs. tidy, this would be in the length of the vowel preceding the flap -- shorter in tighty than in tidy, at least on the average.

For speakers who don't raise "long i" before voiceless stops, and who do retain the vowel length difference in a reliable way, this is accurate. For speakers who have the "long i" raising pronunciation, the two words tighty and tidy are clearly distinguished by vowel quality.

Although this change might seem to help prevent the tighty tidy substitution by preserving the distinction in pronunciation between the two words, life is more complicated than that. When a non-i-raiser whose speech is otherwise like mine says tighty, what I hear is equivalent to my own pronunciation of tidy. Thus in the cross-dialect situation, this change can actually create a confusion.

Finally, let's note that flapping and voicing apply to the /t/ in whitey too -- which opens up the possibility to misunderstand that word as something other than a diminutive of white. It could be related to wide, or perhaps to Y, or could be interpreted just as a cutesy reduplication of tidy. And Google indeed finds some forms with appropriate spellings:

It's Viewtiful Joe, but with Resident Evil thrown in, and the whole running around in your Tidy Widy's (Y fronts(pants)) when you get an ass full of lead (shot) sounds like the best thing since ?
It is your chance to break into the San Franciscan art scene without any of the hot-headed, tidy-widy high-nosers looking down on you. What more can you ask for?
So IMO I wouldn't say you would need a 1-hit-kill-blade to make everything perfectly tidy-widy, just some dodging skillz.

Is that all perfectly cleary-weary?


Posted by Mark Liberman at March 20, 2005 01:05 PM