May 302020

Katie van Baarsen & Terrance Gatchalian


  24 Responses to “The syllabification of VV-sequences in Dàgáárè”

  1. Do Anttila & Bodomo (2009) propose an infix? Could you explain their analysis?

    • Thank you for the question. Anttila & Bodomo (A&B) analyze both kinds of roots in the Epenthetic V data (word-final mid vowels in a/b and word-medial high vowels in c/d) as part of the same pattern on CV-roots — namely, epenthesis, with the variation in vowel quality and position due to constraints on (i) phonologically-derived word-final high vowels and (ii) adjacent mid vowels. In fact, A&B explicitly reject an affixal analysis (specifically of an underspecified V) for these forms because of the word-medial surface position of the vowel in (c/d).

      Formally, the epenthesis is accomplished with FtBin > Dep-IO > *I] > *mid-mid.

      Our reanalysis, in light of the generalizations on tone melody, takes these two be distinct classes of singular marking.

      • Ah, very cool. Thanks for that explanation; I understand the contribution here a little more! So: A&B’s analysis assumed that morphemes must be concatenated. Your poster is pointing out that A&B didn’t consider the distribution and restrictions on tones on adjacent moras. Very neat!

  2. Under ‘singular marking with epenthetic V’, what is the morphological structure of the sg. forms in a-d? For example, what is the structure of [bíé] ‘child-SG’; is it [bí-é] ‘noun-sg’ or ‘child-sg’?

    • Great question! Going with your example [bíé], the structure would be [bí-é] child-SG.

      • I really enjoyed this poster!

        To follow-up: You claim that the singular examples at bottom-right of page 1 fall into two groups: (a-b) have a suffix which changes quality due to the restriction on high vowels word-finally, while (c-d) have an infix. I can see that the tone is distinct, but the vowel pattern seems to be identical looking from the surface, except that in the examples (a-b) we see lowering of the second vocalic element, whereas in the examples (c-d) we see the raising of the first.

      • Whereas, for words like [gbìé] ‘forehead-SG’ and [dìé] ‘room-SG’ the singular is not realized overtly. So we get a structure like dìé-∅ ‘room-SG’.

  3. Hi, interesting poster! I have a couple of questions.

    1) I noticed that in the data that you shared, there aren’t any coda consonants. I was wondering if this is a general property of the language (i.e. if syllable structure is restricted to (C)V syllables). If not, do coda consonants contribute to syllable weight? I was wondering how this might interact with the ALIGN constraint that you propose. For example, for a CVC root, would the final C resyllabify as an onset if followed by a vowel suffix? /CVC-V/ -> [CVC.V] or [CV.CV]?

    2) The forms in a-d regarding epenthesis (under “Singular Marking with Epenthetic V”) are pretty interesting given the two tonal patterns. To clarify, are you proposing that c-d have a suffix whose form is a bare mora, while a-b are unsuffixed? In c-d, is the mora suffix then contributing the meaning of ‘sg’ in these forms? If so, are there different suffixes for ‘sg’ in a-b or does the mora suffix not get realized in these cases? It would be interesting to see if you can get the analysis to work out if you assume that the ‘sg’ suffix is always this bare mora but sometimes it doesn’t get realized in the same way. Have you thought about this? (Feel free to follow up if this isn’t clear!)


    • 1) Coda consonants do appear in Dagaare, but are restricted in roots to {m, n, ng, l, g, d, r}, but only {ng} appears word-finally. Resyllabification of these forms would lead to a violation of our ALIGN constraint – thank you for bringing this up. Anttila & Bodomo (A&B, 2019) assume that /CVC-V/ will always be syllabified as [CV.CV]. They suggest that this might be implemented with an undominated, locally conjoined Onset & *Coda (p 35, fn).

      2) That is exactly our proposal! Number marking in Dagaare has a few different exponents. The basic patterns A&B outline are: (i) polar marking in -rI, as can be seen in the first table, meaning either SG or PL depending on the root, (ii) the other value in these polar-pairs is either -E or (iii) a copy of the root vowel (except when high; then the vowel is a mid vowel with the same back value).

      A&B group the cases in (iii) together, which are the “Singular Marking with Epenthetic V” data here. We haven’t thought about how to get the mora suffix to be unrealized, but we will keep this in mind. My intuition behind this analysis was that since the difference in tonal pattern also accompanied the difference in surface position of the vowel, the analysis of the tonal pattern in -E cases would translate nicely. Of course, there are other differences in these forms as well: for one, the root vowel qualities in a/b are high, while c/d are mid.

      If both classes of forms do contain a mora suffix, this is still compatible with our analysis, but we definitely need say something more about exactly how these forms get their tones and possibly something more about the role of the OCP in Dagaare.

      Thanks for the questions!

  4. You write (on page 2, right-side):

    “In each optimal candidate, the syllable will associate with the tonal
    material in a one-to-one fashion. These forms are monomorphemic
    and monosyllabic, so we expect only a single tone to associate with
    the syllable, capturing the generalization that the tone of a
    phonologically lengthened/diphthongized vowel is consistently a level

    I don’t quite understand the reasoning in the second sentence. Why would we expect monomorphemic forms to have only a single tone? (Does this follow from the Dàgáárè data?) But more importantly, the lengthened vowel forms you provided on page 1 (top-right) are NOT monomorphemic; the only reason to assume the lengthening is phonological rather than morphological (via a floating suffix mora) is tone. If so, then your argument is a little circular. To paraphrase your last sentence, above: “Lengthened vowels have a single tone, which explains that lengthened vowels have a single tone.”

    Thanks so much for your answer in advance!

    • Thank you for the question!

      On the point about monomorphemic forms containing a single tone, there are very few forms in Dàgáárè that contain contours. I did mid-state the process with the “monomorphemic and monosyllabic” comment. Thank you for pointing that out. The biggest gap right now is, as you point towards, that we don’t present a clear account of the underlying tones or of how to derive the tone in morphologically complex forms.

      Our argument here is that lengthening will always produce a monosyllable. In environments which the root vowel lengthens, such as before the /-rI/ number morpheme, the action nominalizer and the imperfective, we see that the tone contour over the VV-sequence is level regardless of whether we get a long-vowel or a diphthong. In other cases, such as the plural in /-E/, we get a VV-sequence that does not have the same restriction on the surface tone contour. The goal was to leverage this asymmetry to clarify whether there was a syllable-boundary in these VV-sequences.

      Lengthening environments restrict VV-sequences in tone and quality, which are derived from the fact that they are tautosyllabic. When the forms do not fall in line with the generalizations drawn from the lengthening environments, specifically in the plural in /-E/ cases (but also in the action nominalization suffix), the morpheme boundary occurs between the VV-sequence. We would definitely like some more independent evidence to support our case. But assuming that the VV-sequences in both these cases are equal, I think we lose our explanatory power over the asymmetry.

      I hope this answers your questions, but please let me know if they don’t!

  5. What about the case of underlying (as opposed to derived) morpheme-internal VV sequences? Or does that not occur? Your abstract’s introduction talks about “derived, underlying, and morphologically concatenated VV-sequences”, which seems to imply that it is possible.

    • There are cases of underlying morpheme-internal VV sequences. For example, [túú] ‘forest-SG’ and [túúrí] ‘forest-PL’ which is underlyingly /tuu/. We also see this with some mass nouns such as [tɪ̰̀ɪ̰́] and [tɪ̰̀ɪ̰̀nɛ̰́ɛ̰́] ‘(types of) medicine’ which is underlyingly /tɪ̰ɪ̰/. I hope this answers your question.

  6. In your abstract you state, ” Both types of VV-sequences permit unlike tones only when a morpheme break intervenes. If the VV-sequence occurs in the same morpheme and is derived, the tone on both Vs is identical. This asymmetry provides evidence for the syllable as the tone bearing unit.”
    Could you expand on the reasoning of this evidence, i.e. why the mora could not be the TBU?

    • The biggest reason for proposing the syllable as the TBU is the complete lack of HL- or LH-tone on VV-sequences in lengthening environments. In these cases, there are two moras which should in principle be allowed to bear distinct tones — in fact, they should bear distinct tones considering the OCP. The only time we do see HL or LH tones on adjacent vowels is when there are additional factors suggesting that they also occupy adjacent syllable nuclei (for example, the presence of sequences distinct from the diphthongs which appear in lengthening environments; Generalization 2). Assuming the mora was the TBU, I think we would lose our ability to explain why this is the case. Thank you for the question, and I hope this answers it!

  7. This is very neat data! I’m still reading through, but have a basic question: what’s the evidence that the roots at the top are underlyingly monomoraic? Are there root alternations for these stems? Or is it simply a static distribution, e.g. there are no roots with a surface shape like CV?

    • Good Question! The evidence for the roots being monomoraic is that we see the monomoraic forms in other instances within the language. For example, with the word for child, we see it in unlengthened environments, such as [bíé] ‘child-SG’ and [bì-fáá] ‘bad child’. Anttila and Bodomo (2009) suggest that the-rI suffix, [bíí-rí] ‘child-PL’, triggers vowel lengthening.
      I do not know of any noun roots that surface as CV, but according to the analysis proposed by Anttila and Bodomo, this is a result of the need to satisfy bimoricty. On the other hand outside of the imperfective forms you can have CV surfacing verbs. For example,
      à sáá- nà ɲɛ́ lá kṹṹ
      D stranger-SG see FOC dead.person
      ‘The stranger saw the dead person’

      I hope this answers your questions!

  8. I also have some questions about your constraints.

    1) I see a lot of markedness and alignment constraints (e.g. constraints on surface forms) but no faithfulness constraints here. Is there a reason for this? For example, the optimal candidate in (1) has a long vowel but the UR is short. So maybe this violates something like Dep-mora.

    2) *I] seems awfully specific! A constraint against word-final high vowels, but only if they are derived… are there any non-derived word-final high vowels in Dagaare? (I don’t see any in your poster, but maybe elsewhere in the grammar.) Is there some way to get the same effect by using a positional markedness constraint (no word-final high vowels) if you also include faithfulness constraints? (Maybe underlying word-final high vowels “escape” the effects of this constraint because they must remain faithful.)

    • 1) We definitely made a lot more implicit assumptions than we should have in the tableau. There’s not particular reason for that oversight, except for (maybe) space. But yes, something like Dep-mora ought to be present in all the tableau.

      2a) There are word-final high vowels in Dagaare, for instance in all /-rI/ suffixed forms.

      2b) *I] is very specific. We imported this constraint from Anttila & Bodomo (2009) as a sort of placeholder to capture their descriptive generalization that instances where we expect an epenthetic high-vowel word finally instead surface as mid vowels. Thank you for the idea about the positional constraints. Another possibility might be a markedness constraint on high vowels generally, with positional faithfulness constraints on earlier/root positions (and relaxed on word-final elements), though again this runs into immediate problems with /-rI/ and /-U/ suffixed forms. We’ll look more into these analyses.

      Thank you for the questions and suggestions!

  9. Like!! Great article post.Really thank you! Really Cool.

  10. I like the valuable information you provide in your articles.

  11. Good one! Interesting article over here. It’s pretty worth enough for me.

  12. I love looking through a post that can make people think. Also, many thanks for permitting me to comment!

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>