Friday, 25 March 2011

strong and weak

Veronika Thir wrote
I am wondering when the KIT vowel (or the FOOT vowel) is actually a
strong vowel and when it is a weak vowel. I think it is weak when occurring in an unstressed syllable […] Thus, it should be strong when occurring in syllables that receive either primary or secondary stress. My questions is: is it weak when occurring in a monosyllabic unstressed structure word like "in" within a sentence?
What about content words which […] do not receive stress for [some] reason?

The general rules about strong (= stressable) and weak vowels in English are that
1. in a stressed syllable you can only have a strong vowel;
2. in an unstressed syllable you can have any vowel.

It would be nice if vowels were always weak in unstressed syllables. But clearly that is not the case in English (unlike, say, Russian), as shown in famous pairs such as modest ˈmɒdɪst (last vowel weak) but gymnast ˈdʒɪmnæst (last vowel strong); informant ɪnˈfɔːmənt but torment (n.) ˈtɔːment; Thomas ˈtɒməs but commerce ˈkɒmɜːs, etc.

Most noticeably, the vowel in the less prominent part of a compound is strong despite being unstressed, e.g. bedsheet ˈbedʃiːt, tentpeg ˈtentpeɡ, kettledrum ˈketl̩drʌm. But notice also non-compounds such as colleague ˈkɒliːɡ, phoneme ˈfəʊniːm, hypotenuse haɪˈpɒtənjuːz.

Some analysts (particularly Americans) argue in the other direction, claiming that the presence of a strong vowel is sufficient evidence that the syllable in question is stressed. In the British tradition we regard them as unstressed.

(There are a few exceptional compounds in which the secondary element DOES weaken, notably some in -man and -land, as milkman ˈmɪlkmən, Finland ˈfɪnlənd. Compare, though, snowman ˈsnəʊmæn, Nagaland ˈnɑːɡəlænd with no weakening.)

The weakening process converts what would otherwise be a strong vowel into a weak vowel. This is most obviously seen in function words with distinct weak forms. (All examples from BrE.)
at strong æt, weak ət
them strong ðem, weak ðəm
from strong frɒm, weak frəm
us strong ʌs, weak əs
are strong ɑː, weak ə
for strong fɔː, weak
NB: there are some function words that do not weaken in this way, e.g. on, always ɒn.

When ordinary lexical words bear no sentence stress, there is no weakening of their vowels. So start, for example, always has ɑː; stop always has ɒ; best always has e; and worst always has ɜː, no matter how non-prominent they may be made in an utterance. Weakening has nothing to do with sentence stress (accentuation), only with word stress.

As Veronica rightly points out, the KIT vowel seems to present something of a problem, since ɪ can be either strong or weak. In bridge it is obviously strong; but in the ending -ed, as in waited ˈweɪtɪd, it is obviously weak, competing as it does with the ed sometimes used in formal singing style.

So with the FOOT vowel. I would regard the ʊ as strong in the lexical word full fʊl, but as weak in the ending -ful, e.g. beautiful. The suffix vowel usually weakens further to ə anyway, though the darkness of the l, combined with syllabic consonant formation, and now its vocalization, may make this difficult to determine. The penultimate ʊ in executive and ambulance is weak, and similarly is alternatively pronounced ə.

Words like in and it indeed present a puzzle. Do these words have a weak form, which happens to sound identical to the strong form? Or are they like on, lacking a weak form?

If I assert that the first vowel in finishing ˈfɪnɪʃɪŋ is strong but the other two weak, can I prove it? Does it matter?

It seems to me that there are two further kinds of evidence that can shed light on the issue of strong and weak ɪ.

One is to look at accents of English that have lost the distinction between ɪ and ə in weak syllables. Australian English is one such: for rabbit, where I say ɪ, Australians say ə, making it rhyme with abbot. So if I want to know whether the last vowel of armistice, my ˈɑːmɪstɪs, is strong or weak, all I have to do is look the word up in my Australian Macquarie Dictionary. There I find this word transcribed with -stəs, proving that the last vowel is weak for them. So presumably the vowel in my -stɪs is weak too (despite being derived from the second part of a Latin compound).

In Australian English it and in, strong forms ɪt and ɪn, DO weaken to ət and ən in contexts where we would expect a weak form.

The other kind of evidence comes from phonological processes that are sensitive to the presence of weak vowels. Take AmE t-voicing, or “flapping” as it is often inaccurately called. Within a morpheme it operates only where the following vowel is weak. So we get the output (= ɾ, more or less) in atom (last vowel weak) but not in latex (last vowel strong). In words such as emphatic we DO get t-voicing, which tells us that the last vowel must be weak ɪ, not strong.

31 comments:

  1. What about the middle /ɪ/ in "caterpillar"? According to Gimson, it is weak (Practical Course of English Pronunciation), but I've heard it pronounced with some sort of secondary prominence (after the main stress) on that same /ɪ/. Is this because "caterpillar" can be considered as a compound?

    ReplyDelete
  2. Nice case for the Australian Proof. (Anyone?)

    I first thought it was clearly strong, for the compound reason you mention. (I once tried to explain to a German NS from Romanis that it's pronounced similar to a made-up German Kettepille. He nodded and continued to say kätäɾˈpiläɾ.)

    But I can easily imagine JWL say kætəplə in connected speech; wouldn't sound any wrong. So maybe the answer is the word exists in two versions.

    ReplyDelete
  3. Regarding 'on'
    I've got the impression that
    I'm far from the only one who's happy to say
    A Cat On a Hot Tin Roof
    as /ə ˈkӕt n ə hɒt tɪn`ruːf/

    ReplyDelete
  4. According to Macquarie, AustE has ˈkætəpɪlə. So it is perhaps indeed perceived as a quasi-compound.

    ReplyDelete
  5. Some analysts (particularly Americans) argue in the other direction, claiming that the presence of a strong vowel is sufficient evidence that the syllable in question is stressed. In the British tradition we regard them as unstressed.

    I'm not sure it's necessarily only a matter of naming. It is possible that informant and torment have identical prosody, being only distinguished by the quality of the vowel, in some accents, and the latter has a longer and/or louder and/or higher-pitched and/or whatever in other accents.

    ReplyDelete
  6. As Veronica rightly points out, the KIT vowel seems to present something of a problem, since ɪ can be either strong or weak.

    So why not speak of the KIT vowel and the happY vowel?

    ReplyDelete
  7. Some analysts (particularly Americans) argue in the other direction, claiming that the presence of a strong vowel is sufficient evidence that the syllable in question is stressed.

    Is it really just a question of differing analytical traditions? AmE preserves many instances of secondary stress that have been generally lost in BrE, for example in words such as "secretary".

    ReplyDelete
  8. So why not speak of the KIT vowel and the happY vowel?

    For some of us the rabbIt vowel is distinct from both the happY vowel and the commA vowel. For non-rhotics, a minimal triple would be "taxis-taxes-taxers" -- can't think of a dialect-independent minimal triple.

    ReplyDelete
  9. can't think of a dialect-independent minimal triple

    Rosie's - roses - Rosa's. Thanks, Wikipedia.

    ReplyDelete
  10. Although NID3 and its descendant m-w.com recognize milkman with a weak vowel as a secondary pronunciation, AHD4 and RHD2 do not, and a strong vowel is my own usage.

    VP: I don't think AmE has secondary stress any more than BrE does. Words in -ary are /-eɪri ~ ɛri/, with (in my view) a strong but unstressed vowel. Compare infantry, which had a very similar suffix -ery, but is now trisyllabic in both varieties.

    ReplyDelete
  11. To me, AmE, even of the East Coast varieties, seems to have a strong tendency for secondary accent, not just because of less reduction. Secondary sounds like a compound sec-'n-dairy.

    I like the idea that gentlemen is gentlem[ə]n in the basic form and gentlem[e]n in the vocative case. (The singular would be gentlem[ə]n and sir.)

    ReplyDelete
  12. vp

    I would describe the second vowel in rabbit as a strong but unstressed KIT vowel — distinct, as you say, from both the happY vowel and the commA vowel.
    25 March 2011 15:22

    ReplyDelete
  13. As Veronica rightly points out, the KIT vowel seems to present something of a problem, since ɪ can be either strong or weak. In bridge it is obviously strong; but in the ending -ed, as in waited ˈweɪtɪd, it is obviously weak, competing as it does with the ed sometimes used in formal singing style.

    Funny you should mention strong and weak ɪ. I was just thinking the other day how pointless the NED (OED1) symbol ĭ was - a 'reduced' form of the ideal vowel i (IPA ɪ), the reduced form being pronounced ɪ! In theory, weak ɪ corresponds to ĭ, ĭ, or ė in NED transcription. Sadly, the editors don't seem to have been aware of the subtleties mentioned in this blog-entry: finishing is fi·niʃiŋ, rabbit is ræ·bit, and armistice is ā·ɹmistis.

    ReplyDelete
  14. It's funny that gymnast ˈdʒɪmnæst and milkman ˈmɪlkmən are given as examples; my pronunciations are exactly opposite: ˈʤɪmnəst and ˈmɪlkmæn. My pronunciations for Finland and snowman match, though.

    Incidentally, my AmE stressed from is frʌm.

    ReplyDelete
  15. I would describe the second vowel in rabbit as a strong but unstressed KIT vowel — distinct, as you say, from both the happY vowel and the commA vowel

    That would be fair enough -- with the proviso that some accents merge unstressed KIT with commA.

    If we're going down that road, we may as well describe happY as an unstressed FLEECE vowel (with the proviso that some accents, such as traditional RP, merge unstressed FLEECE with unstressed KIT). One could account for potential minimal pairs such as "Andy's - Andes" by postulating secondary stress in the latter.

    ReplyDelete
  16. @vp: that's not dialect-independent, because there are many of us who retain the schwa/weak /ɪ/ distinction but have the former in "roses" (and "taxes" and "waited").

    I also don't think it makes much sense to talk about the second vowel in "rabbit" as weak, at least in accents like mine. The rhythm of that word seems to me to be the same as that of "Nissan", which has the same two vowel phonemes in the opposite order. (I've probably now demonstrated that I'm neither Australian nor American.) Similarly, I feel the first /ɪ/ in "armistice" is weak but the second one is strong.

    In general, is there any hope of a rule which tells us when an /ɪ/ is weak, just based on the stress pattern, the other phonemes and morpheme boundaries? If not it always seems like a somewhat messy point.

    ReplyDelete
  17. @JHJ:

    there are many of us who retain the schwa/weak /ɪ/ distinction but have the former in "roses" (and "taxes" and "waited")

    Oh, really? I did not know that. What words would have unstressed /ɪ/ in your variety? And what kind of accent would you describe yourself as having?

    ReplyDelete
  18. vp

    with the proviso that some accents merge unstressed KIT with commA.


    Most accents merge NORTH with FORCE, but John still gave us the two lexical sets.

    ReplyDelete
  19. @vp: pretty much everywhere (modernish) RP has it, I think, except in -es and -ed plurals and verb forms. (Certainly in every other example mentioned in JW's blog today and the comments. The other group of exceptions would be words where I have a strong DRESS vowel before a consonant cluster, like one word in this sentence.)

    Northern English, "near RP". TRAP=BATH, but not FOOT=STRUT. Also not NORTH=FORCE.

    ReplyDelete
  20. vp,
    Rosie's - roses - Rosa's. No thanks, Wikipedia. Not for JHJ's reason, but because I have /ɪ/ in "Rosie's" and "roses".

    «… account for potential minimal pairs such as "Andy's - Andes" by postulating secondary stress in the latter.»

    I did that last time round, and I think I explained even then that I would have to treat the latter type as two phonotagms, which is how I account for secondary stress, among all the other things. I am delighted to say I can now reveal that I think both "Andy's" and "Andes" have two syllables.

    ReplyDelete
  21. So is there a strong ɪ in autism? I say [ˈɑtɪzm̩] even though I have t-voicing elsewhere, like in nautical. Maybe this has to do with morpheme boundaries?

    Pretty sure I've heard other Americans say autism with the voiced version, however. Actually both ways sound a little weird to me.

    ReplyDelete
  22. Yes, I think you've identified an interesting bit of optionality there. I am aware of both possibilities even without t-voicing. Again it's a question of the possibility of secondary stress for me, with the manifestation of that being apparent in the possibility of greater length for both vowels and a slower tempo overall. Like Lipman I have always got the impression that AmE has a stronger tendency for secondary accent, and not just because of less reduction, which I think would be getting the cart before the horse-and-cart. What to him sounds like a compound sec-'n-dairy sounds to me like a compound of two phonological words sekn-derj (to be really annoying and treat the i analogously to the n), and I call it two phonotagms with the paraphonotactic stress on the first. I have no idea how this pans out for the Australian Proof, but it would be nice if it correlated with the ɪ~ə alternation for strong~weak.

    ReplyDelete
  23. For a lot of people, including me, this is can actually be distinctive in moonlit ˈmuːnˌlɪt ~ moonlet ˈmuːnlɪt, and not including me, but I think in some AmE, for ˈperˌmit (permission) ~ ˈpermit (the fish). Again with cases like these Australian Rules give ɪ~ə for strong~weak. I don't see how the British tradition can regard them all as unstressed.

    ReplyDelete
  24. To take a pair involving a more familiar word, I'd love to see someone carry out perception tests on starlit vs. starlet (choosing a speaker who uses ɪ rather than ə in the latter). But I'm not really sure how, if they are consistently perceived as different (which is by no means certain), we would then determine whether the difference is attributable to [±stress] or [±strong], or to the effect of different kinds of morphological boundary.

    ReplyDelete
  25. Oh John, you don't believe me! But I did try to be careful and change "this is actually distinctive" to "this can actually be distinctive". I fumbled it, but for me this would only have been a belt-and-braces job, because I think the concept of distinctiveness already covers cases where they are notconsistently perceived as different. To establish different lexical identities for them we only need to demonstrate that one set of allomorphs contains a form with -lət and the other doesn't. The rest is ising on the cake. In realizational terms this -lət alone makes the first set distinctive with respect to the first, but not vice versa.

    As for perception tests on a speaker who uses ɪ rather than ə in the allomorphs in question, speaking as one myself I would say the degree of distinctiveness would depend on the prominence of the secondary stress (in what after all are obviously compounds) and the difference in tempo, and even for me it's only a potential distinctiveness: I doubt if it's consistent in rapid speech, and these pairs are not exactly commonplace. I couldn’t even guarantee to perceive it in artificial instances such as snippets of my own speech, but if people like me did get significantly significant results, you seem to have been implying all along that you would say the difference is attributable to [±strong], in conformity with the British tradition of not attributing these things to stress. And the only way you could actually determine which of the alternatives you mention it's attributable to is the way you set your phonology up.

    ReplyDelete
  26. Sorry:

    In realizational terms this -lət alone makes the first set distinctive with respect to the second, but not vice versa.

    ReplyDelete
  27. For me the distinction between Andy's and Andes is in tenseness: although I have happY-tensing, my happY vowel is not as tense as my unstressed FLEECE (which is not as tense as my stressed FLEECE, though obviously there are no minimal pairs there). The distinction is just barely perceptible to me, unlike Rosa's/roses, which is nonexistent: both are the same lax vowel, whereas Rosie's is happY-tensed.

    For me, the second vowel of starlit is strong, as if it were star-lit.

    ReplyDelete
  28. @mallamb:
    Rosie's - roses - Rosa's. No thanks, Wikipedia. Not for JHJ's reason, but because I have /ɪ/ in "Rosie's" and "roses".

    Interesting -- do you have /i/ in "Rosie"? Or in any non-word-final environments?

    ReplyDelete
  29. For me the distinction between Andy's and Andes is in tenseness: although I have happY-tensing, my happY vowel is not as tense as my unstressed FLEECE (which is not as tense as my stressed FLEECE, though obviously there are no minimal pairs there)

    For me it's length: the second syllable in "Andes" is about twice as long as that in "Andy's". As far as I can tell both are identical in quality with FLEECE.

    ReplyDelete
  30. vp,
    «do you have /i/ in "Rosie"? Or in any non-word-final environments?»

    If that's John's all-purpose unadorned i I will disregard the solidi, as that is some sort of shape-shifting polyphone/allophone/allomorphophone/archiphoneme/morphophoneme! So I'll say that not only do I have [ɪ] in "Rosie", which may or may not be a realization of /i/, depending on your analysis, but having it in "rosy", I do also have it in "rosier, rosiest", and other non-word-final environments. Offhand I'd say I have it wherever John has the mercurial i in LPD.

    ReplyDelete