The LINGUIST List is dedicated to providing information on language and language analysis, and to providing the discipline of linguistics with the infrastructure necessary to function in the digital world. LINGUIST is a free resource, run by linguistics students and faculty, and supported primarily by your donations. Please support LINGUIST List during the 2018 Fund Drive (http://funddrive.linguistlist.org)
Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

I am not sure I understand the question. You may want to submit the question again, telling us why you want this information.

You might want to start by reading up on phonetics a bit. Ladefoged's course on phonetics is available online, with sounds (http://www.phonetics.ucla.edu/course/contents.html).

You need to:

  1. Understand the difference between a phoneme and an allophone. For example, /p/ sounds quite different in 'pin', 'spin', and 'nip',

  2. Understand allomorphs. For example, the plural (written as (e)s) is pronounced differently in 'cats', 'dogs', and 'horses'. This concept applies to the regular past tense.

  3. Understand that there is dialectal variation in both phonemes and allophones.

Hope this helps. I am not sure what you mean by 'frequency' in this question.

Anthea Fraser Gupta


Hi, Kevin,

I can't answer your question with references either, off the top of my head, though as Dr. Fischer says there are such references available. In fact, with a little (OK, a lot of) study of the sound system of American English (btw, note also that there's no such thing, really: there is a vast array of different variants of 'American English', from 'Yankee' to 'Southern' to 'Midwest' to 'Western', and each one includes a wide variety of different 'accents', without even mentioning isolated variants like Ozark, Appalachian, etc. and others like 'Hispanic', 'Yiddish (English)', etc. Likewise, each of the varieties (mostly including a large geographic area each one) I have mentioned is itself divided into subvarieties. Indeed, in the focal, original areas of American English (for historical reasons, mostly found on or near the East Coast), even small areas can have a great variety of different ways of pronouncing the same words. One example: if we try to describe 'Philadelphia English', we need to take into account many different subvarieties: I remember one article about a kid who moved from one western (I think it was) suburb of Philadelphia to King of Prussia (which I'm not quite sure exactly where it is, but it is a suburb of Philadelphia) when she was six years old. Years later, natives of KofP could still tell she was not a native speaker of KofP English, although most linguists would consider six years old as plenty young enough to master natively a new language or dialect. In any case, the pronunciation differences, even between variants very close to one another indeed, is often detectable and therefore describable; and there are huge numbers of such differences. Now, an Aussie, for example, would be unlikely to detect all these subtle differences and would mostly notice an 'American accent' only. So a separate issue would be the audience you are directing your study to.)

To return to the previous, interrupted sentence: you could take a speech synthesizer for American English (ie, a program that converts written text to speech, a much easier task than speech recognition), but it would have to be a very good one to capture all (or at least some) of the subtleties in that enterprise, and turn it loose on lots of text and then transcribe phonetically the result (unless you could tell it was incorrect or unrealistic). I'm no expert, but I believe there are at least some good programs of that nature. In the good old days, I had a BASIC program that did that, though the result sounded a bit unnatural, since it was done without a pronouncing lexicon as a component. Nowadays, there's not much except the expense to keep anyone from including the entire OED, say, in a pronouncing dictionary (American pronunciation, of course, unless you wanted to have a British option). Very much of this dictionary, btw, could be derived automatically by rule and then later corrected in the cases where it were necessary. In fact, you could do the whole thing in one step automatically, and then do the corrections. However, if you want to avoid years of corrections, you would need a very good understanding of most of the subtleties of deriving pronunciations of English from the orthography (including the lexical ones, of course, which would involve checking against a pronunciation dictionary of irregular cases; but also including syntactic influences on pronunciation, differences of the pronunciation of a word in many set phrases, etc.) I don't know of any such program myself, but it just might be worthwhile googling around for one; with modern computing power, lots of things are now possible that were 'undreamed of in [my] philosophy, Horatio'. At any rate, I hope these comments have been of some help, especially in helping you understand the complexity (although, basically, an underlying simplicity) of the problem.

Good luck.

Jim

James L. Fidelholtz Graduate Program in Language Sciences Instituto de Ciencias Sociales y Humanidades Benem'erita Universidad Aut'onoma de Puebla, M'EXICO


Perhaps one of my other colleagues can give you the information you're actually asking ford (the frequency of sounds in spoken American English), , but permit me to correct a misconception on your part: a phoneme is a class of sounds that count as the same within a language. Glottal stops are not phonemes in English; they are allophones, usually of /t/. That is, there are no two words in English that differ only in one having a glottal stop where the other one has a [t]. I believe there may be corpora that will give you frequency of phonemes; not sure about frequency of individual sounds.

Susan D Fischer


All I would add to what my colleagues have written has to do with the level of phonetic transcription you are considering, the transcription of naturally occurring connected speech. For most linguistic purposes, we transcribe words and short phrases that we elicit, and on the basis of an extensive body of such transcription we derive a phonological description. We can then use that description to transcribe longer texts, including folktales, conversations, etc. But those are not transcriptions of naturally occurring connected speech. They tend rather to be a phonemic or, at best, an allophonic transcription.

To learn about transcriptions of speech, you might look at Linda Shockey's Sound Patterns of Spoken English (Wiley-Blackwell, 2003), which gives many examples but also discusses the complexity of the task of such transcription. It cannot be done without close reference to sound spectrograms, what are popularly called "voiceprints." As Shockey's transcriptions indicate, the phonetics of connected speech goes well beyond what is normally considered allophonic variation, and in ways that linguists often do not consider. The relationship between what is evident from connected speech and the citation forms that we would use in a phonemic transcription is complex and subtle. An example of such a transcription, although not a fully detailed one would be

[ɑmnǝʹgoʊ]

This is a natural reduction from

I am going to go.

in the Inland Northern variety of American English that I speak. As others have already indicated, that transcription represents something close to a personal variant. Young speakers from the same area would have distinctively different forms.

Herbert Frederic Stahlke