An interesting query from reader M.Y.:
I read your article on the alphabet olympics yesterday and followed one of the links, and then one of its links, and so on. I was merrily traipsing thru the internet when I came upon a page that threw me: "The Rules and Misrules of English Spelling".
The note on "th" (note (f)) gives a list of words with the "this" sound (what I'd call "voiced th" — ð rather than θ) that includes the word "with". I was surprised — I have always used unvoiced as the pronunciation of that word, and had never noticed anyone doing otherwise. Sure, voicing gets *added* sometimes due to context, but surely unvoiced is the target — right? Apparently wrong. My Pocket Oxford gives only the voiced pronunciation, and my Houghton Mifflin Canadian gives the voiced version first, as does my New Lexicon Websters. The two pronunciation sites I found online also gave voiced pronunciations.
I asked my wife to pronounce the word slowly and carefully, and she likewise gave an unvoiced pronunciation, and was surprised that anyone aimed for the other (tho' she did point out that Bono has a buzzy version when he sings "with or without you"). (I grew up in Nova Scotia, and my wife grew up in southern Ontario.) OK, so I've got a non-standard (or less standard) pronunciation — it's not the only one I have. I'm interested in what the distribution of this variant is, but I'm having a hard time finding it online. The word "with" is no help since it appears in so many descriptions of dialect differences (with pin/pen merger, with cot/caught merger, with raising, …). I tried just scanning a few dialect maps, but I couldn't find anything. Is it part of a more general dialectal difference? Is it a regional thing at all, or is it just the case that the pronunciation is so variable (given context) that different people home in on one of the two variants more-or-less at random (with a greater tendency toward the buzzy version)?
Short answer: I don't know. I've never heard a discussion of this point of pronunciation variation, except with respect to the varieties of English that have [wɪf] or [wɪv].
My own intuitions agree with those of M.Y. and his wife — I've got [θ] not [ð] in isolation. And so does the LDC American English Spoken Lexicon:
And also the pronunciation at the online Merriam-Webster site:
On the other hand, the audio pronunciation at the American Heritage dictionary site clearly has [wɪð]:
There are obviously going to be contextual effects, which themselves are surely subject to variation.
So I don't know what's going on. From the evidence so far, this is one of those variable phenomena that people don't pay much (conscious?) attention to. Are there significant geographical, social, or temporal dimensions of variation? Readers may be able to help us figure it out.