Tuesday, 10 May 2011

what's this l like?

Felix Chan had a question about clear and dark l. You’ll recall that the standard rule taught to NNSs is that l should be clear before a following vowel, dark elsewhere. This rule is fine for words in isolation, but what about connected speech? Felix said,
I don't think Debbie will agree.

I am wondering whether the 'l' in the word 'will' should be pronounced
as a dark l (since the preceding sound is a vowel?) or a clear l
(since the following sound is a vowel?).

Let me say right away that the nature of the preceding sound is entirely irrelevant. (You get clear l in black, but dark l in tables. The preceding consonant is b in both cases.)

Felix continues

I am even more puzzled when I listen to the examples in Longman
Dictionary of Contemporary English for the word 'capital':

'Washington D.C., the capital of the United States' I heard a dark l
in the word 'capital'.

'Hollywood is the capital of the movie industry' Here, I heard a clear
l in the same word 'capital'.

So how can this be? In reply I said
You can get either clear or dark l in this position (word-final before a vowel in the next word). Partly it depends on syntax (how closely are the words linked together?), partly on speech rate, partly on personal differences. Generally speaking I would use a clear l unless there was a major syntactic boundary between the words, or unless I paused at that point.

I must confess that that reply is not based on any evidence beyond tradition and my own introspection. The account in the current (7th) edition of Gimson’s Pronunciation of English, edited by Alan Cruttenden, has this to say (p. 216):
When an affix beginning with a vowel is added or the next word begins with a vowel (fiddling, fiddle it, finally, parcel of books), the lateral may remain as dark and may remain syllabic or become non-syllabic; alternatively the lateral may become clear , in which cases it is usually non-syllabic. The lateral is less likely to become clear in those cases where the following word begins with an accented syllable, where a [ʔ] may intervene, as in real ale [riːɫ ˋʔeɪɫ], cf. real estate [ˋriːl esteɪt].

I think it would have been better to deal separately with the question of loss of syllabicity (which I call ‘compression’), because this is sensitive to the strength of the following vowel — it is an option available only if the following vowel is weak. The possible insertion of ʔ (‘hard attack’) is also a separate issue.

And personally I wouldn’t be caught dead saying riːl instead of rɪəl for real — but now I’m showing my age, and I’m well aware that hardly anyone nowadays distinguishes real from reel in the way that I do.

I expect someone somewhere has carried out experimental measurements of clear vs. dark l in a variety of contexts, including unscripted connected speech, but I can’t refer you to any such research. Nor can I lay my finger on Abercrombie’s interesting observation, made half a century or more ago, that we use a clear l in feel in the sentence I feel ill, but in the sentence I may not look ill, but I do feel ill we use a dark one.

I also suspect that Abercrombie used, and I use, a clear lateral in some contexts where Cruttenden has a dark one. (Although not sounding in general like northerners, Abercrombie and I grew up in the northwest of England, whereas Cruttenden grew up in London.)

Felix was delighted with my answer, but still came up with a further poser.
In the phrase 'annual leave', I should pronounce a dark l then a clear l at the boundary between the two words?

We might also consider such examples as tell lots, full length, table lamp, feel lonely. I said
Theoretically, yes. In practice, in fast speech at least, you get assimilation making both parts of the lateral clear.

You may not all agree with this, and there are obviously social and regional differences. Again, has anyone ever made systematic observations relating to this point?
_ _ _

I’m sorry there was no posting to this blog yesterday. This was for reasons beyond my control.


  1. In Derry, everyone distinguishes monosyllabic reel ril from disyllabic real riəl.

    I remember, during a teenage discussion of Marillion, someone mentioning the live album Real to Reel but calling it "Real to Real". Everyone immediately turned on the offender, jeering "It's called Real to Reel you eejit, not Real to Real!"

    Non-pre-vocalic Ls in Derry (and indeed in Northern Ireland in general) are only very slightly darker than clear Ls, if at all.

  2. I get the impression that dark l is used (by some) prevocally in stressed syllables when a word is given great emphasis e.g. I like it! Has anyone else noticed this?

    And what about Ed Miliband's prevocalic dark l? Quite striking.

  3. @Pete: I doubt whether your "real" is disyllabic instead of having a diphthong.

  4. The people who came up with the title of the TV programme 'The Real Deal' seemed to think it's a good rhyme. It's the host's catchphrase too: "Now that's the Real Deal!"

  5. @Kilian: I perceive real 'riəl as two syllables: the prefix re- meaning "thing" and the adjective-forming suffix -al.

    You can analyse it as a rising diphthong if you prefer (although actually I think this would multiply the number of vowels unnecessarily), but the important point is that it's distinct from reel ril.

    @Paul: yes, for me The Real Deal doesn't rhyme.

    Also I should have said earlier that clear Ls are found non-prevocalically in Ireland in general, not just in Northern Ireland.

  6. Pete, the etymology is irrelevant. Do you reely think that most of your compatriots perceive real 'riəl as two syllables because of it? (even if they absentmindedly identify the first with the prefix re-!) I don't dispute that the two syllables can be real, as in 'idea' in your rhotic dialect.

  7. @mallamb - yes, I reely reely believe it's perceived as two syllables, although I take your point that the etymology doesn't come into it.

    The English of that part of Ireland hasn't got a phonemic diphthong . The NEAR words have iər, but that's because /i/ as an allophone [iə] before r. This allophone doesn't appear before l as it does in RP; if it did the contrast between reel ril and real 'riəl would collapse as it has in other dialects.

    The same applies to monosyllabic cool kʉl and disyllabic cruel 'krʉəl.

  8. Yes Pete, I was aware of all that, and was already trying to explain myself better. I meant of course that 'real' can be two syllables in the same way as the 'dea' of 'idea' can, i.e. both phonologically and phonetically. Phonetically, rhoticity doesn't of itself exclude the possibility of a realizational centring diphthong any more than in AmE, though I don't think that diphthong has any phonological status such as makes [iə] graduate to [ɪə] in my BrE, giving [rɪəl], as my ignorant disregard of the etymology permits, which no more rhymes with my [dɪiᵊl] than yours does. I don't even think I really have any epenthetic ᵊ worth notating there – it's just that my dark l is not so mighty dark as to need notating either, but I'm very far from being a Michael Howard sound-alike, and don't want to give that impression by leaving out the ᵊ. On the other hand my punctilio in transcribing my /iː/ as a rising diphthong should not be taken to imply any more of a rhyme than for you – like John I wouldn’t be caught dead saying riːl instead of rɪəl for 'real' – it's the closeness of the vowel in 'reel' and 'deal' and the fact that any epenthesis that there is does not give rise to a centring diphthong that makes the non-rhyme more striking, whereas for you it's the syllable count.

    Lazar Taxon on "non-speech" a fortnight ago admitted to being a rhotic who uses the centring diphthong [ɪə] in "idea", and in reply I said "I'm glad we agree to call [ɪə] and [ɛə] centring diphthongs, but again they are phonetic ones. And there is nothing marginal about the sequence of phonemes /aidiə/ having [ɪə] in AmE, though some of the orthography is a bit peculiar: onomatopoeia, sangria (the missing acute in OED certifies it naturalized)…"

    And John Cowan pronounced on this:
    «I don't actually say meh, as I've never watched The Simpsons and have no real idea (a trisyllabic word, no centering diphthong involved) how to pronounce it.»

    Once again I can only draw attention to my use of brackets in "there is nothing marginal about the sequence of phonemes /aidiə/ having [ɪə] in AmE". And Lazar himself had [ɪə] in square brackets. Nobody is saying that the sequence of phonemes /aidiə/ is anything but trisyllabic in terms of phonotactic syllable structure in rhotic dialects, but if anyone does have the phonetic centring diphthong [ɪə] in their realization of it, then it's jolly well "involved", and inasmuch as it's a diphthong, it's monosyllabic, giving us a count of two phonetic syllables, each with a diphthong.

    I don't know why we persist in bothering to attempt any intersubjective use of the word syllable at all.

  9. BTW I should also have offered a less ambiguous phraseology for my question: I was asking if you reely thought that it was because of the etymology that most of your compatriots perceive 'real' ˈriəl as two syllables, i.e. whether there were not perhaps a few who perceive it as two syllables for the phonological reasons that you have now made clear and that I have tried to. I did say "I don't dispute that the two syllables can be real, as in 'idea' in your rhotic dialect." By which I meant that they can be "real" in any dialect, rhotic or otherwise, for any speaker of it, even for real or imagined etymological reasons as well as for the sort of reasons we have now explicitly discussed, but that 'idea' in your dialect was a clear-cut case, as I hope my argument from an earlier thread has shown.

    An anecdote from a very distant memory: "The Listener" ran a competition for the longest piece of plausible writing in English using nothing but monosyllables, and entries using "real" were retrospectively disallowed, this being "obviously" disyllabic. I wouldn't have said it was even then, as I'm sure /rɪəl/ was a routine pronunciation for it even then, and the OED's disyllable was itself an etymologizing pronunciation. I suspect John would agree with me.

  10. Probably relevant would be "Allophonic Variation in English /l/ and its Implications for Phonetic Implementation" by Richard Sproat and Osamu Fujimura, Journal of Phonetics 21:291-311.

  11. As usual, I was speaking only for myself when saying "a trisyllabic word, no centering diphthong". There may well be rhotic AmE speakers who do have a non-rhoticized centering diphthong in idea, but I'm not one of them; I say it with the same prosody as I see ya. (As far as I know, the rhoticized pronunciation idear, fully assimilating the word to rhotic NEAR, is not part of any AmE regional accent but is idiolectal: in my time I have heard at least two rhotic politicians speaking of "new idears".)

    As for /l/, all of mine are dark without exception, so real and reel are homophones and have centering diphthongs. For me, both this centering diphthong and the one in NEAR words have a tense onset [i], which is a matter of regional accent. The same is true in SQUARE and words like quail, which have centering diphthongs beginning with [e]. Per contra, rill [rɪl] has no diphthong, nor does really ([rili] in isolation or with emphasis, allegro [rɪli]), though a nonce adverb real-ly 'in a real way' would have one, with two /l/s at the syllable boundary.

  12. That all seems fair enough. You don't explicitly pick up on my point that the darkness of the l seems to vary with the epenthetic ə in these words, but correlation or causality is implied by "so" in your statement "As for /l/, all of mine are dark without exception, so real and reel are homophones and have centering diphthongs". So is it that although l for you is dark without exception, you have degrees of darkness, and it is because the l in your rill [rɪl] is less dark that it has no diphthong? Or is it because of the lax vowel? Cockney and even EE actually throw it in with 'real' and 'reel' with or without L-vocalizing. For me the centring diphthong in 'real' is a sine qua non but I feel riːl only sounds a bit more epenthesis-prone than rɪl because of the greater height before the transition to l. I don't really think I have anything that amounts to epenthesis in either, whereas I'm sure [rɪᵊɫ] is something I often hear from AmE speakers for your [rɪl].

    I always think Am [rili] and [rɪli] and Br riːli are pure delight, but the only distinction I make between rɪəlɪ and a nonce adverb 'real-ly' 'in a real way' is the two /l/s at the syllable boundary. Now 'reeling' is riːlɪŋ pure and simple.

  13. As a child I had a speech impediment of /l/-vocalization (along with issues with /ɹ/). I have never felt completely comfortable with lateral consonants, and have recently noticed that my always-dark-L* quite often has a component I can best describe as a dorsal tap. I cannot find any good example of the phenomenon, except my pronunciation at this Forvo page.

    *I'm American, so that's not too weird, but my /l/ is still a bit lacking in alveolar-ness.

  14. mallamb: To be clear, I pretty much only use this [ɪə] in "idea", and "idea" (being disyllabic) doesn't rhyme with "sangria" or "onomatopoeia" for me, which use [iːə]. For this reason, I do tend to consider /ɪə/ a marginal phoneme in my speech. Since I live in New England, this might be a regional compromise (many of the rhotics here do say "idear"), but from hearing so many instances of "idea" as an iamb even in American songs, and seeing disyllabic "idea" attested in American dictionaries (for example, at m-w.com), I suspect that this word can be a pretty peculiar case.

  15. @Lazar Taxon: I agree with you on "idea" vs. "sangria" and "onomatopoeia": the latter two have FLEECE+schwa sequences. So does "IKEA", which therefore differs from "idea" in more than just /d/ vs. /k/.

    As for "real" and "reel", I think that although they both have centring diphthongs, that in "reel" starts with [i] and that in "real" starts with [ɪ], so that the distinction is just about still there. I might be fooling myself, though.

    The diphthongisation of FLEECE (and FACE) before word-final /l/ certainly disappears as soon as a vowel-initial suffix is added, so I'm with mallamb on "reeling" [riːlɪŋ], except that I tend to have a [g] on the end.

  16. Mallamb: I think it's the combination of a [+tense] nucleus and a coda /l/ or /r/ that triggers my centering diphthongs. Thus peel, pail, pool, peer, pair, poor have one, but pill, pell, pol, pawl, pull etc. do not.

  17. I think those that don't recognize a syllable boundary between re- and -al need a reality check.

  18. Dirck, I also produce only velarized l's here in Manitoba and I have no speech impediment. The "exclusive dark-l" rule sounds like yet another interesting phenomenon stretching the wide plains of North America.

    In fact, I reserve all of my clear l's for French. So in effect this produces a tidy black-and-white "l-dichotomy" for me between the two languages.

  19. Steve, some English dialects handle syllabicity differently and that's the true reality. For me, 'real' is pronounced as one syllable /ɹilˠ/ because if it were two, I'd say */'ɹi(j)əlˠ/ or */'ɹil̩ˠ/ but I'm pretty sure I never say that.

    When I think on it, I suppose "pre-velarization" of /i/ here is likely, creating a tiny [+back] glide that gives it an air of two syllables.

  20. Lazar,
    Surely you must use this [ɪə] in "ideal" as well? And I was scraping the barrel with "sangria" and "onomatopoeia" for me, which are not exactly rimes riches for my "idea" either, but I did say "there is nothing marginal about the sequence of phonemes /aidiə/ having [ɪə] in AmE" and because that is a realizational statement, free variants exhibiting a cline towards [ɪə] are the most we could expect. Certainly I have [ɪə] in it, and there is no problem in saying that my centring diphthong there is phonological as well as phonetic. Whereas I can only say that I am quite capable of saying [sæŋˈgrɪə] etc, although those variants are not canonic for me, as opposed to [sæŋˈgriə] etc. I wouldn't say [sæŋˈgriːə] etc, but phonologically, being stressed it must be an /iː/ in [sæŋˈgriə]. I'm not about to say I have an opposition /ɪə/ ~ /iə/ or an incongruous /ˈi.ə/.

    I like the idea of a regional compromise for the "idear" also mentioned by John Cowan, but I too suspect that with 'real', 'idea' etc we are dealing with a pretty restricted set.

    John Cowan,
    I take you to be confirming the impression given by your earlier post that
    'really' [rili] in isolation or with emphasis has no diphthong because the l is not in the coda, as opposed to 'real' which does have one, as in the nonce adverb 'real-ly'. I mentioned treating 'feeling' like your ˈrili, but not 'really' (ˈrɪəlɪ for me, still with one l, not dark because it goes with the following ɪ, as opposed to 'real-ly', which has a dark one after ɪə before the clear one.)

    I would still have this ɪə in 'mania', and unlike my earlier examples with stressed ɪə etc it presents no phonological problems. Would you be any more likely to say you have a centring diphthong in these unstressed cases?

  21. For me mania, like idea, has three syllables.

  22. So you never merge the last two into a rising centring diphthong i̯ə, let alone jə?

  23. John said...

    For me mania, like idea, has three syllables.
    12 May 2011 14:30
    mallamb said...

    So you never compress the last two into a rising centring diphthong i̯ə, let alone jə?
    12 May 2011 15:15

  24. Some experimental evidence for clear/dark l in:

    "F2 variation in Newcastle and Leeds
    English liquid systems"

    Journal of the International Phonetic Association (2007) 37/2

  25. M.A.L. Lamb:

    Only if I were reciting poetry or singing a song that allowed only two syllables for the word. But in ordinary speech, however allegro it might be: never. I would hear [meɪnjək] as a very allegro main yuk, hardly maniac.

    Providing these additional syllables is very AmE, although it is a matter of degree: I say extraordinary with five syllables (omitting the first a), though I know there are others who use six.

  26. Well we were talking about 'mania' not 'maniac'. I too would hear [meɪnjək] as a very allegro 'main yuk'. Here the vowel in question is strong, which means that the only way I see of analyzing it is as the nucleus of a separate phonotagm, and in such cases there is nothing particularly AmE about the additional syllable. I might conceivably say ˈmeɪnjæk, but probably ˈmeɪnĭæk is as allegro as I would get. What I would see as AmE is the tendency to give that æ secondary stress.

    We are closer brethren than you think: if I'm being ponderous enough, I too say extraordinary with five syllables (omitting the first a), though never more. But I hope we can agree on not having five vowels in it! My five syllables would be ɪksˈtrɔːdn̩r̩ɪ. Normally I would have four: ɪksˈtrɔːdn̩ri.

  27. I don't know if this site is mucking up the diacritics for you, but it certainly is for me, so five syllables ɪksˈtrɔːdn.r.rɪ or I suppose possibly ɪksˈtrɔːdn.nᵊrɪ, four ɪksˈtrɔːdn.rɪ.

  28. Very late response for the record: I say [ɪksˈtrɔrdn̩eri]