It would be very helpful to have for a programming project I'm working on involving grapheme-to-phoneme translation. I've been able to find many rules for phonemes but not too many for allophones.

  • Allophonic variation is exceedingly complex and not subject to a single set of rules. Individuals vary a lot, depending on tiredness, emotion, etc. And they vary even more from each other, even in the same speech community. Change communities and you start all over. Allophones are the stuff of sociolinguistic surveys, and correlate more with socioeconomic factors than linguistic ones. Linguists wouldn't mind having a complete list, mind -- but we know how many billions it would cost to develop it.
    – jlawler
    May 15, 2015 at 17:18
  • 1
    Why do you need a list of allophones for grapheme-to-phoneme translation? Are you doing speech synthesis? May 15, 2015 at 19:31

2 Answers 2


You ask about allophonic rules for "the English language". There are no such rules. There are different allophonic rules for the various dialects of English. But even collecting a set of rules for some major, more or less standardized English dialect would be problematic, because of linguists' very strong tendency to disagree about the facts, to say nothing of their interpretation. Probably a fair number of linguists would not even agree that there are such things as allophonic rules (but I do).


If you are interested in grapheme to phoneme translation, you don't need to care at all about allophonic rules. Allophones are not at all reflected in orthography. However, you do need to care about what sense of "phoneme" you are interested in for any translation. People use "phoneme" in many different ways; sometimes they mean "underlying form", and sometimes they mean "phonetic form, but undoing any rules that introduce derived-only segments". The usual example of a derived-only segment for English is aspiration, and another example is the flap [ɾ]. But the flap can come unpredictably either from /t/ or /d/, which means that you can't translate from a surface flap to a specific underlying consonant (thus, under one definition, the flap is a phoneme). Whether or not a flap is a segment of underlying representations is completely unclear. Many people like to assert that there are no underlying flaps, but the reasoning for that is usually that you could predict a surface flap from some other phoneme, and therefore you can "avoid" having an underlying flap.

I also suggest that you look at the concept of "phonetic implementation", since a lot of what used to be thought of as coming from "allophonic rules" (such as "devoicing of /l/ after /s/") is not the result of a phonological rule, it's a mechanical reflection of how the glottis is opened in producing /s/ in English (which is itself in the realm of "phonetic implementation" and not "phonological rule").

  • David Stampe's view is that flap is not a phoneme of English, because language learners follow a principle of least effort, and it's easier for them to assume that every flap they encounter comes from t or d than it is to admit a phoneme flap into their system of phonemes. (The flapping process is not something that has to be learned -- it's in your realm of "phonetic implementation".)
    – Greg Lee
    May 15, 2015 at 22:11
  • How does he account for the difference between "capitalistic" and "militaristic"? Also, what is his underlying representation for "water"?
    – user6726
    May 15, 2015 at 22:24
  • @GregLee: what do you mean when you say the flapping process is not something that has to be learned? This seems impossible to me, if we're using an ordinary definition of "learning". Is voicing and flapping alveolar stops an innate process? If so, why does it only show up in specific dialects of English, rather than in all languages? May 15, 2015 at 22:34
  • In Natural Phonology, innate processes can be suppressed (analogous to constraint domination in OT).
    – user6726
    May 15, 2015 at 22:38
  • 1
    I can't speak for David on these issues of detail about phonemes, but I do know that he has argued against identifying phonemic form with a level of derivation, which you seem to be doing. Speaking only for myself, I can't follow your reasoning at all about "capitalistic" / "militaristic". Are you assuming that phonological rules and processes make no reference to morphological structure? I certainly don't assume that. And I have no difficulty making an arbitrary choice of underlying form for "water" -- easiest thing in the world. I'll make it a /t/.
    – Greg Lee
    May 16, 2015 at 18:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.