Thursday 24 March 2011

unpacking the conventions

On 22 March someone left a comment on this blog, under the cloak of anonymity, saying
It reminds me of how the dictionaries can be confusing. For example, one gives ˈsteɪʃ^ənəri, the other /ˈsteɪ.ʃ^ən.^ər.i/, where both are unpronounceable. As written, taken literally, without certain implications. The choice of sounds not to be pronounced is also so random, the prescriptivity arbitrary... Then you have the recorded actor's voice and it doesn't match the written.

I don’t usually react to anonymous negative criticism, but on this occasion I will.

The word stationary/stationery is a good example of how an English word can have several subtly different pronunciations all trivially different from one another. This is because of the interplay of two optional phonological rules in English. They are (i) syllabic consonant formation, which permits a sequence of schwa plus a sonorant to become a syllabic sonorant, and (ii) compression, which involves a reduction in the number of syllables in the word. As the author of a pronunciation dictionary, my dilemma is (a) whether to include all these variants, and (b) if so, how to avoid spelling them all out in detail, which would be too wasteful of space. (AmE is much simpler here, since Americans do not reduce the suffix vowel.)

In BrE our word can be pronounced in any of eight ways, which our seɡment-based transcription system enables us to distinɡuish as follows.
1. ˈsteɪʃənəri (four syllables, no syllabic consonants)
2. ˈsteɪʃn̩əri (four syllables, including a syllabic nasal)
3. ˈsteɪʃənr̩i (four syllables, including a syllabic liquid r̩ = ɚ)
4. ˈsteɪʃn̩r̩i (four syllables, including a syllabic nasal and a syllabic liquid)
5. ˈsteɪʃənri (three syllables, no syllabic consonant)
6. ˈsteɪʃn̩ri (three syllables, including a syllabic nasal).
7. ˈsteɪʃnəri (three syllables, no syllabic consonants)
8. ˈsteɪʃnr̩i (three syllables, including a syllabic liquid).

From the point of view of the EFL student, any one of these is fine. A similar but simpler combinatorial explosion affects words such as liberal and national. Then dictionary and missionary are like stationary.

If I were designing a pronunciation entry for an elementary or intermediate dictionary, I would select one variant and ignore the others. For LDOCE and the like, the entry ˈsteɪʃənəri is entirely adequate. On the other hand the COD’s entry ˈsteɪʃ(ə)n(ə)ri could be taken to imply, wrongly, that the word can be pronounced with two syllables only.

But LPD is meant to be a specialist dictionary. I do not want to dumb down by pretending that the other variants do not exist. I want to specify them unambiguously.

My solution is to use abbreviatory conventions. My entry reads
stationaryˈsteɪʃ ən ər_i -ən_ər i || -ə ner i
What is shown here as an underline should actually be a low breve, which for typographical reasons I cannot reliably reproduce here. This symbolizes the site of possible compression. As explained in the part of the dictionary that people tend not to read, a raised symbol indicates a segment that is usually absent (creating a consonant that will be syllabic unless compressed), though it may alternatively be included. An italic symbol indicates a segment that is usually present, but may be omitted (ditto).

To understand the entry you have to be able to ‘unpack’ the abbreviatory conventions. If you can do so, you will find that it covers all eight possibilities mentioned above.

The advice on page 149 of the third edition says that you can, if you choose, simplify the abbreviatory conventions by preserving italicized symbols, deleting raised ones, and ignoring the compression mark and syllable-division spaces. Applying this to our word we derive ˈsteɪʃnəri, variant number 7. Fine.

As for the claim that
the recorded actor's voice … doesn't match the written
— it is simply untrue. She says ˈsteɪʃn̩ri, version number 6. There is of course no way in which she could have produced all eight versions simultaneously.

Looking at the other pronunciation dictionaries, we see that variants 5 and 6 are not covered by the Cambridge EPD’s ˈsteɪ.ʃən.ər.i, ˈsteɪʃ.nər-. The ODP’s ˈsteɪʃn̩(ə)ri, ˈsteɪʃən(ə)ri does not cover versions 7 and 8.

The LPD entry may be complicated — but the facts are complicated. The optional segments are not ‘random’. Rather than ‘arbitrary prescriptivity’, you could claim that I go to an extreme of inclusiveness and non-prescriptivity. I certainly gave a very great deal of thought to the problem of how best to design an accurately inclusive entry for tricky words like stationary.


  1. As I think I've said here before, LPD is a specialist dictionary for advanced learners, so let it be as complicated as it needs to be! Simpler transcriptions are provided in every other dictionary.

    Learners are always keener to ask questions than to look things up for themselves.

    In this case it means accusing a lexicographer of incompetence when it is the learner who hasn't bothered to read and understand the front matter of the dictionary.

    And in the EFL setting I've often found myself saying that a word is rather tricky to explain, that I could waste a lot of lesson time unsuccessfully trying and that completely understanding the nuances of the word wasn't the focus of the lesson, so they learner should look it up in a good dictionary after the lesson. And sure enough, on asking the next day whether they have looked it up, the answer is always 'no'. My heavily ironic response 'So it can't have been that important' implying 'So you shouldn't have made such a fuss during the lesson and wasted everybody else's lesson time'.

    Sorry to turn this post into a rant on a pet hate of mine!

  2. I feel unrepresented. I say [ˈsteɪʃn̩ˌnɛɹi]. Oh snap, am I an outcast again?

  3. Am I deluding myself when I feel that I can sometimes say this word as two syllables
    with no syllabicity for the n or the r?

  4. No, JWL, I don't think you are. Syllabicity is on some sort of spectrum, and demarcation disputes are unavoidable so long as the concept is barely definable.

  5. JWL, you have a point. But like many of the extreme reductions to which you regularly draw attention, it seems to me not to belong to the prononciation familière ralentie that we traditionally record in pronunciation dictionaries. It also violates the phonotactics, having an illicit syllable-initial cluster nr.
    Would you feel happy if I, as a hymnwriter, say, were to treat this word as disyllabic? I thought not. But the kind of compressions I record are, as we have seen, amply attested in the work of well-known hymnodists.

  6. John, I'm sure you know JWL knows it violates the phonotactics, but to say it has an illicit syllable-initial cluster nr begs the question of whether one can regard the syllable as anything but a phonetic rather than a phonological unit (if indeed one can define it at all in any sort of intersubjective or uncontentious way). What you can define is the phonotagm, and in terms of what is licit and illicit for it, to boot. Circular, of course, but a virtuous circularity is all we've got.

    And you seem happy with monosyllabic hevn etc in hymnody, though you don't record that kind of compression in LPD: ˈhev ᵊn, never mind saying vn is a licit anything-final cluster anywhere but in the special conventions of hymnody.

  7. Actually both ˈsteɪʃnəri and ˈsteɪʃnr̩i (versions 7 & 8) feel wrong to me. I don't think you can compress the antepenultimate (-ion-) without also compressing the penultimate (-er-).

    Is that just me?

  8. No, Pete, it's not just you. That's why versions 7 and 8 are catered for in the ALTPRON field rather than in the MAINPRON.

  9. whether one can regard the syllable as anything but a phonetic rather than a phonological unit

    Surely the syllable can only be defined at the phonological level. For example, in order to determine how many syllables are contained in a particular utterance of the word "idea", it's necessary to know whether the speaker is rhotic (information not necessarily contained in that particular utterance).

  10. vp,
    An example of a language in which the syllable can relatively uncontentiously be defined at the phonological level is Chinese, in which the notions 'syllable' and 'phonotagm' are just different ways of looking at the same thing. With all the contending criteria for English I'm afraid you're on a hiding to nothing. But as far as JWL's comment is concerned, the syllabicity in question is obviously phonetic, and John deals with that point accordingly, using a pragmatic phonetic criterion. But his subsequent phonological criterion of the phonotactics of the sequence for the purpose of determining syllabicity seems to prejudge the issue of whether JWL can legitimately operate in terms of impressionistic phonetic criteria, and draw attention for example, as he may well do somewhere, to the pronunciation [ˈnreɪʃn̩] for 'narration'. Or even whether we may view that as a phonotagm with the outrageous initial /nr/, in which the realization may or may not include an epenthetic ə or a syllabic r.

  11. vp,
    At least I think nobody here would propose to introduce the etymological criteria that to this day influence some analyses! And at least the new online version of OED, while changing the aɪ of your 'idea' to the preposterous ʌɪ, has changed the iːə to ɪə.

  12. @mallamb:

    You imply that there are some arguments for locating the syllable at the phonetic rather than phonological level. Could you outline one? I had always thought that the syllable was uncontroversially a phonological notion.

  13. vp,
    All I meant to imply was that it's mightily difficult to locate it anywhere, but that JWL had introduced the question of phonetic syllables. Or do you think he was arguing for a phonological analysis of the syllable in which the initial cluster nr- is established phonotactically? And what I have outlined is the possibility of identifying realizational phonetic syllabicity within a distributionally determined phonological structure which I don't think anyone would call monosyllabic!

  14. vp,
    I do appreciate your usually giving some thought to my posts.

  15. All I meant to imply was that it's mightily difficult to locate it anywhere

    Until today, I would have said without much reflection that the syllable resides at the same level of analysis as the phoneme. I would still be inclined to that view, but am listening if anyone wants to argue that it lives somewhere else.

    I do appreciate your usually giving some thought to my posts

    Thanks! Or is "usually" contrastive here? :)

  16. It sounds to me that mallamb might be dancing around the notion of "semisyllables".

    So under such an analysis, a sequence /nre/ that some may consider "un-English" and automatically absurd can nonetheless surface in a language. One would label the initial /n/ of the syllable as a "semisyllable", a subset of the larger syllable which lacks a *moraic* vowel. Such semisyllables may contain non-moraic vowels that surface phonetically but are not phonemic (ie. they fail to carry the mora of the syllable). The pronunciation [ˈnreɪʃn̩] indicates a newly tolerated pattern of ς-σ-ς in that variant of English (ς = semisyllable, σ = nucleus).

    We then speak of constraint rules regarding valid shape of semisyllables, how many semisyllables are tolerated in a syllable, valid positions (whether after/before the moraic vowel or both), etc. However the concept of "syllable" and semisyllable by and large remains on the phonological level.

    When mallamb speaks of a phonetic level, I presume he is hinting at the concept of non-moraic vowels in these semisyllables. An example would be the reconstructed Indo-European word for 'earth', *dʰǵʰōm, which necessarily implies a non-moraic schwa between the two initial stops on the phonetic level. Suggested analysis: It's a one-syllable word with a semisyllable onset [d(ə)-] preceding the mora in [go:m].

    The traditional notion of syllabicity, without this helpful concept of semisyllabicity, is heavily challenged by extremes such as Kartvelian or Salish languages. Semisyllables also add to the theory of "sonority hierarchy".

  17. Thanks for sharing. Very informative.


Note: only a member of this blog may post a comment.