Wednesday, 1 July 2009

Beth am y llall?

As promised in response to yesterday’s comments, here’s how I would teach the sound [ɬ], the Welsh ll as in Pontcysyllte.

This consonant is a voiceless alveolar lateral fricative. To make it you have to control the voicing (voiceless), the place of articulation (alveolar) and the manner of articulation (lateral fricative). I would teach each of these features separately, and then combine them.
To get awareness of voicing, pronounce [fvfvfv θðθðθð szszsz ʃʒʃʒʃʒ] and then [mm̥mm̥mm̥ nn̥nn̥nn̥ ll̥ll̥ll̥]. As you do so, check the voicing by covering your ears with your hands or by feeling the front of the larynx with your fingers. As you alternate between voiced and voiceless make sure that everything in the mouth remains unchanged: the only difference should be at the larynx.
Concentrate now on [ll̥ll̥ll̥]. Alveolar place should not be a problem, for native speakers of English at any rate.
(Resist any well-intentioned but misguided suggestions involving placing the tongue between or against the teeth.)
Try the English words please, plum, clear, claw. Observe that the l-sound in these words is actually devoiced, [], because of the effects of aspiration. Isolate this sound (the symbol is meant to be a letter l with a ring underneath — sorry about the shortcomings of the font). It is a voiceless lateral approximant. Note how the air escapes over the side rim(s) of the tongue. (Over both sides? Or just one, and if so which?)
(Starting from English subtlety is also misguided, since this word has lateral release of the preceding alveolar plosive, which introduces an entirely unnecessary complication. Its lateral is only slightly devoiced and is not fricative. Furthermore, it is dark, which will be inappropriate for most kinds of Welsh. Welsh ll, on the other hand, is a fricative. It does not have lateral release, just lateral escape.)
Your remaining task is to change the sound from a lateral approximant to a lateral fricative. This means that you have to narrow the gap between the side rims of the tongue and the side teeth. Try to do this while saying a long []. It should change into [ɬ].
Feel the air escaping turbulently over the side of the tongue. Keeping everything else constant, change from breathing out to breathing in. Check that you can feel a cold airstream at the side (not in the centre line). Go back to breathing out.
Lastly, we need to check the place of articulation. If you have carried out the above steps successfully, the place of articulation is still alveolar. The ll sound is a single phonetic segment, and should not have any accompanying elements. There should be no element of velar friction before, during, or after the lateral fricative. There should be no central friction (s or θ) before, during or after it. The pronunciations [xɬ, θɬ], often used by non-Welsh speakers, are not acceptable.
Try the words llaw (hand), lle (place), llo (calf), llwybr (path). Then try pell (far), twll (hole). Then allan (out), and felly (thus). Like the other Welsh voiceless consonants, [ɬ] is considerably longer intervocalically than would be the case in English: but there is no change in quality.
Make sure that you can hear and make the difference between dallu (to blind) and dathlu (to celebrate).
Practise some placenames: Llangollen, Llanelli, Machynlleth. (The last one is [maˈxənɬeθ].)
If you use [ɬ] in place of English [s], you can produce a common type of lisp (lateral sigmatism). Try [ɬɪkɬ piːɬɪɮ] (six pieces).

You are now equipped to pronounce not only Welsh but also Icelandic and Zulu words with this sound. Icelandic has a voiceless lateral affricate. Try fjall [fjatɬ] (mountain), Þingvellir [ˈθɪŋɡvetɬɪr] (the site of Iceland’s first parliament).

Zulu has voiceless and voiced alveolar lateral fricatives. Use your control of voicing to switch between them: [ɬɮɬɮɬɮ]. Then try -hlala [ˈɬaːla] (sit) and -dla [ɮa] (eat), not to mention amandla [aˈmaːndɮa] (power).

Lastly, do not fall foul of the “exotic sounds syndrome”. For speakers of languages that have it, [ɬ] is a perfectly ordinary sound that does not require any special effort. You, too, must use it effortlessly.
(Experts on Welsh, Icelandic and Zulu please correct me if I’m wrong anywhere.)


  1. mm̥mm̥mm̥ nn̥nn̥nn

    That seems a bit of a nasal détour.

    Not sure I understand why the naturally coming lateral release makes it more complicated than it eases it, but you certainly have a point about the darkness.

  2. @Lipman: I assume we're using the terms in the same way as one another. Release = the last stage of a plosive: the change from being obstructed to being unobstructed. Escape = the constant state of a fricative or approximant. Welsh ll does not have lateral release, since it isn't a plosive. So English examples of lateral release are irrelevant.

  3. Then I understood. Anyway, your ll̥ll̥ll̥ approach makes more sense in most cases.

  4. Thank you. I think this keep me busy for the rest of the week.

    And I did not know that Icelandic had ɬ, too.

    Intellectually I know that it's not an 'exotic sound', but that's just all the more reason to be annoyed that I can't produce it.

  5. nope - never had any problem pronouncing this sound - ever - never even thought about it

  6. I have been informed that the sound written as an L in the Mongolian capital's name "Ulan Bator" is also one of these funky little sounds.

  7. I'm pretty sure that when I was a kid, before the age of eight, I used to mispronounce the Serbian voiceless fricative (of the [s̪] type) as a lateral fricative. I was sent to speech therapy, but I quit after two different therapists gave up.
    It wasn't until I started going to school and had other kids mocking me for my lisp that I began consistently producing the right sound. There's nothing like peer pressure, as they say.
    Although in casual speech, when I've let my guard down, I still occasionally produce a lateral fricative.

  8. The proto-Hebrew שׂ, usually transcribed ś, is supposed to be a voiceless alveolar lateral fricative, too. It merged with /s/ in what became standard biblical Hebrew, and in another dialect with /ʃ/. This is the basis of the "shibboleth" episode.

  9. I just want to say that I love this blog.

  10. I've heard anecdotal reports that lateral fricatives (in Welsh and elsewhere) are produced unilaterally, with the air escaping just through one side. I've also heard some (often non-native Welsh speakers) describe the sound as more palatal or alveo-palatal (and they certainly do seem to make their /ɬ/ that way). James Crippen, who writes about Tlingit, has said that he's found that Tlingit speakers seem to vary between apical and laminal productions of their lateral fricatives. Has any study been made of the articulation of /ɬ/ in Welsh or other languages that has noticed these sorts of variations?

  11. Yes, the ll was pronounced unilaterally after negotiations about a participation of the English collapsed. In retaliation, the Welsh expelled the dark l from among the country's English.

  12. Re (i) unilateral laterals and (ii) apical/laminal variation.

    My experience of doing old-style chalk palatograms in phonetics classes with Swedish students is that both uni and bilateral articulations occur as personal variants. So I wouldn't be surprised to find similar variation among speakers of Welsh.

    Many years ago I checked the phonetics literature for reports on apical and laminal productions of dentals/alvoeolars, after reading that there might be language specific preferences. Around 10% of speakers of English were said to have laminals. I found the same proportion among the Swedish students of phonetics. So this could also be an individual thing. The exception is [s], which is mostly laminal, rarely apical. The EPG people should have statistics on this, including cross-linguistic data. The laminal dentals (alveolars) are effectively palatalized. With apical dentals, the vocal tract opens abruptly posteriorly from the constriction, suppressing pharyngeal resonances around 6-8kHz. With the laminal dentals (or palatalized dentals), the vocal tract opens gradually posteriorly to the constriction, and those pharyngeal resonances are active. So if there really is individual variation between apical and laminal laterals by speakers of Welsh, then the laminals will tend to sound "palatalized".

  13. I wonder where this development came from? This strange devoiced l and its relationship with its british-lenited counterpart /l/ is seemingly the reverse of what might be expected. Converting stops to be [+voice] when lenited is a familiar rule, but a fricativisation-like mechanism is a rule found in lenition in Scottish Gaelic and in Welsh voiced stops, and fortis -> lenis may be expected too.

    The various l sounds of conservative dialects of Scottish Gaelic and their relationship with their lenited counterparts make it seem to me all the more odd that Welsh underwent this development.

  14. I'm not expert on Islandic, but does it really have a voiceless alveolar lateral fricative? Before I have only seen the ll i words like fjall described as [fjatl]. The phoneme [ɬ] does however occur in some Norwegian dialects (Trøndelag in Central Norway, parts of Northern Norway), where other dialects have [tl] (Western Norway) or [sl] (Eastern Norway).

  15. I've been learning this sound in the Creek/Muskogee language, where it is represented by "r." This post is helpful, as there are no native Creek speakers within hailing distance.

    I had a semester of Old Icelandic but managed to remain innocent of this consonant, probably because we studied medieval texts instead of the living language.

  16. Okay, an old post but decided to comment on it anyway as this sound has always fascinated me – and as I know Icelandic.

    Here in Iceland, the tradition of phonetic transcription does not include [ɬ], even though I as a non-native speaker of the language hear clear noise here very often. My native tongue is Danish, by the way.
    In the tradition here, the sound is classified as an unvoiced lateral approximant, [l̥] (hope the ring went below there).

    The word "fjall" is then transcribed [fjatl̥], and "allt" as [al̥t] – no matter how much they sound like [fjatɬ] and [aɬt] to me. On the other hand, words that start on hl- tend, according to my ear, to be more truthful in having [l̥], e.g. "hlaupa" [l̥œyːpa] ‘run’ (some phoneticians stay strictly with [œi] because it fits the system of diphthongs ending in either [i] or [u], while others choose to show the rounding (that is always present) in transc) and "hljóðfræði" [l̥jouːðfraiːðɪ] meaning ‘phonetics’, but even here I sometimes hear a fricative. I would support the fricative being a free variant of the unvoiced l here, but traditions tend to be quite strong here (!)
    In "Þingvallavatn", the "ll" is simply [tl] because of the surrounding vowels, in older tradition transcribed as [d̥l].

    A lateral fricative obviously doesn't fit into the nice devoicing system of consonants in Icelandic, see this: In certain environments [m n ɲ ŋ l r] are devoiced to [m̥ n̥ ɲ̊ ŋ̊ l̥ r̥] – but there's already a problem here, the approximant [j] is being devoiced with so much force so even the Icelandic traditions changed it from the otherwise nicely fitting [j̊] to [ç], as in "hjól" [çouːl] ‘wheel’. Maybe the story of [l̥] will go the same way – indeed, it is not yet as convincing to change as it is the case with [ç].

  17. In response to Paul, when I first tried to produce the Welsh "ll" (native English speaker), I used one side of the tongue as you describe. It's only since reading descriptions of how to pronounce it through both sides that I've changed my method - redundantly, it seems.