If you look in any dictionary, you'll see some information on how to pronounce every word in there. Depending on the dictionary, it may be really easy to understand or it may be really hard. Also, it might be totally consistent, or it might be a big mess.
I don't like things to be inconsistent, so when I give pronunciation information, it may be a little hard to understand at first. That's why, before I go any further with this blog, I need to define the pronunciation symbols I'll be using (and already have used a little) when I talk about the sounds of English.
For starters, here's a table that covers almost all the basic sounds (which excludes diphthongs and clusters). If you're unfamiliar with the terminology I use here, check out my newly updated Working Glossary of Linguistics.
Keep in mind that this is a hypothetical dialect of English that attempts to maintain all the distinctions made by all the major dialects at once. Your particular dialect will almost certainly use fewer phonemes.
- pr, py, tS, tSr, kr, ky
- br, by, dZ, dZr, gr
- sp, spr, spy, st, str, sk, skr, sky
- y@w, yuw, yuwl, yuwr
- il, iyr, iy, iyl, @l, @w, @r, @y, ul, uw, uwr, uwl
- ol, or, oy, oy@l, el, eyr, ey, eyl, Ul, Uw, Uwl Ur, Url
- Ol, al, aw, aw@l, aw@r, Al, Ar, Ar@l, Ay, Ay@l, Ay@r
- Superfluous phonemes are optional, either because they are obligatory (as in /tents/) or systematically excluded from normal pronunciation (as in /hwAyt/).
- The nucleus of a stressless syllable is always represented by the ambiguous vowel /@/, possibly in combination with a glide. /@/ might serve to syllabify the following consonant or be phonetically realized as something like /i/ (for a syllable with a non-labial coda) or /U/ (for any other syllable). The combination /@ng/ is generally realized as something like /iyng/ or (more colloquially) a syllabic version of /n/. /@w/ typically sounds more like /Uw/ than /uw/, especially at the end of a word.
- A few words (such as "shh" and "hmm") contain /@/ as a superfluous vowel to show that its normal function is to syllabify the following consonant.
- Two vowels may not occur in a row; a consonant or morpheme boundary must intercede. In very few words (such as "uh-oh"), the glottal stop is phonemic and must therefore be represented (as /?/).
- The listed blends (with the exception of /@w/) prefer to stay intact within a single syllable. A given vowel-blend combination can only occur within a syllable if it is listed. /uwr/ is often converted to /or/ (as in "tour"), and /yuwr/ is usually something like a stressed /y@r/ (as in "pure").
- Consonants prefer primarily to attach to stressed syllables and secondarily to be onsets. This means that a single consonant would rather be a stressed coda than a stressless onset. If two consonants occur between vowels, one should typically be a coda and one an onset, unless the first syllable is stressless and the second stressed. Exceptions are /t/, which often becomes a coda in order to weaken, /d/, which does so before a possible syllabic /n/, and /g/, which often does so after /n/. If a syllable ends in /ng/, the /g/ is only released when the following syllable of the word begins with a glide. Consonantal blends may be broken up only between a stressed and a stressless syllable. In such a case, the blend's initial sound may separate from the rest to serve as a coda for the stressed syllable.
- Aspiration occurs on voiceless stops or phonemically as /h/ but only in the onsets of stressed syllables. /h/ can also serve as a coda but only in a few words (such as "ugh") and a handful of words of Scots origin (such as "loch") for some speakers.
- An additional consonant /N/ serves to nasalize the vowel it follows. It is not listed as a normal phoneme because it only occurs in a few words (such as "uh-huh") and a handful of words of French origin (such as "restaurant") for some speakers.
I should note that I did all this analysis a couple years ago, and I'm posting this a bit half-heartedly. Still, it's good information to use in combination with a linguistics course in phonology as a sort of second opinion. And I hope this sort of thing will be useful to people working on constructed languages, especially those based on English.
Anyway, I really feel that this whole practice of segmenting speech into discrete phonemes that need a bunch of messy phonotactic rules to explain how the speech actually sounds is just a little off target. I think speech should be thought of as something more like a sequence of actions (not unlike distinctive features) whose effects necessarily overlap. I'll be posting more on this developing theory soon. (Just letting you know in case you're a total nerd who found this paragraph at all interesting.)