Note: I am on my phone, so linguistic symbols are not intuitive to type, so I typed out phone names in prose

Hello, apologies if this is a dumb question, but I'm reading through Bruce Hayes' "Introductory Phonology", and he provides an example of complementary distribution of the allophones of English /l/; he provides four allophones for analysis: l, velarized l, [voiced l followed by voiceless release] and [velarized l with dental articulation]. He posits that these allophones, being in complementary distribution, are realized in unique environments: [voiced l followed by voiceless release]: after voiceless C [velarized l with dental articulation]: before interdental C Im a footnote, he adds that word-final and after voiceless C are environments that DO NOT overlap since English has no words which end in a voiceless consonant followed by a word-final /l/. However, what of words like "apple" or "tackle", which I understand to be [æp(velarized syllabic l)] and [t^hæk(velarized syllabic l)]?

These words seem to put /l/ word finally AND after a voiceless consonant. Is this due to a light unstressed [ə] prior to the /l/ which prevents the environments from overlapping? Is Hayes' claim an oversimplification? Does it have anything to do with the syllabic nature of /l/ in these environments? Or perhaps rule ordering? Any assistance appreciated!

  • Wow, thank you all so much! I'm learning so much already from the couple of responses I have gotten, this is great info for a budding linguist. Never tried stack exchange before, but I will certainly be posting more questions on here from now on :) Jul 30, 2018 at 18:18
  • So much trouble with phones caused by a phone! :-) Feb 9, 2020 at 6:31

2 Answers 2


The transcription of English words in the profession is extremely variable, because we don't agree on how narrow to transcribe. If you look at the transcriptions of English that he uses, he does not transcribe syllabic sonorants for reduced VR syllables in English, and in "dreadful, crucible" on p. 44 he writes syllabic l with [əl], not [l̩]. With respect to his transcriptions, the claim is probably not an oversimplification, and it is due to the supposed schwa in words like apple, crucible. I would not transcribe those words with [əl], but I speak a mildly different dialect, so I would accept that as a correct statement about his pronunciation.

An alternative is that he also actually says [æpl̩], and this illustrates the point that a phonetic transcription is always a work in progress, if you are trying to provide a narrow transcription. I assume (as many people do) that syllabic sonorants in English derive from reduction of əR sequences, so the proffered generalization could be correct at that point, but then there is further reduction, which messes up the generalization. The underlying question is, or should be, what are his generalizations generalizations about? Actual pronunciation?

What Hayes does not talk about (and at the level of introductory textbooks, this is universal) is that there are two things mushed together under the rubric "allophone". One is where you have a true phonological rule which categorically changes one sound ("set of distinctive features") into another categorically distinct one (flapping is an example). And, there is also phonetic implementation, where a physical feature of a segment is realized continuously within a range. For example, the amount of nasal air flow in English words like "rant" varies from speaker to speaker and context to context. In that case, we are filling in values of a physical parameter for a single unit, not selecting between units. There is reason to believe that, in fact, pronunciation of /l/ is so variable that some of it should be treated in continuous phonetic terms. Specifically, ostensive devoicing of /l/ is actually the result of overlap of glottis spreading from the preceding consonant and a following sonorant (it's just more salient in the case of /l/ because of the spit that [ɬ] generates). L-darkening is a well-known rule that applies "to a degree".

  • Thank you so much! This answered a number of questions I didn't even ask in my question, and now I have some new knowledge that clears up some of the fuzziness around allophones for me! I wasn't aware of "allophone" being an umbrella term for discrete AND continuous phoneme variations, that's really cool. When you say "I assume that syllabic sonorants in English derive from reduction of əR sequences, so the generalization could be correct at that point, but then there's further reduction, which messes up the generalization" which further reduction after əR --> R(syllabic) are you referring to? Jul 30, 2018 at 18:39
  • That is, the rule as stated could apply when the word "apple" has the form æpəl (it would not devoice l, because there is a vowel between p and l), but later on, syllables with schwa plus sonorant delete the schwa, giving you a syllabic sonorant as in [æpl̩]. Another option is simply to restate the rule so that it only applies to a non-syllabic l.
    – user6726
    Jul 30, 2018 at 19:16

There are two theories of how words like "apple" work in English. One says that there's a syllabic consonant: /æ.pl̩/. The other says that there's an underlying schwa, /æ.pəl/, and a phonological rule that turns resonants syllabic between schwa and a syllable boundary.

In this case, Hayes is using the theory with the schwa, and saying that the syllabic [l̩] is a phonetic detail, not a phonological one. In other words, "apple" is /æ.pəl/ [æ.pl̩]. With these assumptions in place, it is indeed true that you can never have a voiced consonant immediately preceding /l/.

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.