USC Graduate Phonology ✳︎ Fall 2019 ✳︎ Smith

The syllable

• Syllable boundaries are marked with “.” in the IPA.
• Every syllable in every language has a nucleus - usually a vowel, but sometimes a consonant.
  • The feature [+syllabic] is commonly used for sounds that can be the nucleus of a syllable (usually but not always vowels).
• Classic syllabification rules, which are pretty universal:
  • Step 1. Build nuclei of syllables.
  • Step 2. Add consonants as onsets (as many as legal).
    • Maximal Onset Principle: VCV is always syllabified as V.CV
  • Step 3. Join unsyllabified consonants as codas.
• Languages differ in the inventory of permitted syllable types, including answers to questions like:
  • What are possible nuclei?
  • Are onsets required?
  • Are codas permitted?
  • Are complex onsets permitted? Complex codas? Both?
  • What are legal onsets?
  • What are legal codas?
  • How are medial consonant clusters syllabified?
    • E.g., VCCV syllabified as VC.CV or V.CCV?
  • Does a coda make a syllable heavy?

Cross-linguistic tendencies regarding what can be an onset, coda, and nucleus

• Codas tend to have fewer contrasts than onsets (Itô 1986, Goldsmith 1990)
• Sonority sequencing principle (Selkirk 1984)
  • Syllables tend to have highest sonority in the nucleus, and lowest sonority at the margins.

• The syllable contact law (Murray & Vennemann 1983)
  • There is a cross-linguistic preference for syllable contacts with falling sonority.
• Onsets independently tend to be less sonorous (e.g. Gnanadesikan 1995)
  • Example: acquisition of English (Gnanadesikan 1995)
    • blue = [bu], sky = [gaɪ], snow = [noʊ], balloon = [bun]
  • Example: Sardinian has a ban on rhotic and glide onsets, but allows rhotic and glide codas. (Smith 2003)
• More sonorous codas are more likely to be counted as heavy (see Zec 2007)
  • Example: In Gonja and Kwak’hala syllables with coda nasals and liquids are heavy, but syllables with coda obstruents are light.

Contrastive syllabification?

• Syllabification is never contrastive.
  • No language has a minimal pair in which the words differ only in syllabification.
  • E.g., [a.ma] ‘orange’ vs. [am.a] ‘lemon’.
  • Instead, syllabification is completely predictable within a language.
  • When it looks contrastive, there’s always an explanation in terms of other differences in underlying forms (e.g., morpheme boundaries in night rate vs. nitrate).
• Consequence: in OT, there are no faithfulness constraints that refer to the syllabification of the input, and syllabification is usually taken not to be in the input at all.

Why do phonologists think syllable structure exists?

• What is the phonetic reality of syllables? Psychological reality?
  • No perfect phonetic correlates.
  • Syllable counting is unreliable.
  • Speakers' application of syllable-based rules sometimes clashes with syllabification judgments (see work by Marie-Hélène Côté for examples).
• In fact, some phonologists (Steriade, Côté, most prominently) have proposed doing away with syllables entirely.
  • The thinking: if you can model all of the data that seem to require syllables without using syllable structure, the syllable-less model is better due to parsimony. (Recall: “Do not posit complexity without necessity.”)
  • In SPE, there were no syllables at all!
• Here’s one theory-internal argument for syllables: we see rule after rule that looks like this:
  • C → … / __ {C, #}
  • The use of braces makes us think we’ve missed something - the two cases must have something in common, especially when they keep occurring together.
  • How might we translate this rule into syllable terms?
    • This requires invoking cross linguistic tendencies in syllabification.
• Syllables also make it easier to describe stress patterns.
  • “Stress the penultimate syllable if it has a coda. Otherwise, stress the antepenultimate syllable”
  • CV.VC.CCVC → CV.VC.CCVC
  • CV.VC.CVC → CV.VC.CVC
  • CV.V.CVC → CV.V.CVC
  • CV.V.VC → CV.V.VC
  • How could we write the rule without syllables?
    • Note: Donca Steriade has worked out a syllable-less OT analysis of stress patterns like this in a series of papers (Steriade 2009 et seq.), so it is possible.
• Syllables also make it easier to state patterns of reduplication and truncation, e.g. “The output of truncatation is heavy syllable”, “Copy the first two syllables.”

Typology of onsets and codas

• Data below are from Blevins 1995, and the typology is originally described in Jakobson 1962.
  • I’m excluding complex codas and onsets for now.

• What hypothetical language types are missing from this typology? (Don’t look below yet!)
• Typology of syllable structure
  • There are languages lacking syllables with initial vowels and/or syllables with final consonants, but there are no languages devoid of syllables with initial consonants or of syllables with final vowels. (Jakobson 1962: 526)
  • If a language has syllables without onsets, it also has syllables with onsets.
  • If a language has syllables with codas, it also has syllables without codas.
  • Every language has at least some CV syllables.
  • If-then statements like the two above are called implicational universals.

Acquisition of codas and onsets

• In acquisition, children (mostly) mirror this typology, using onsets before they use codas.

OT constraints on syllables

• There are two major constraints:
  • Onset: Assign one violation for every syllable lacking an onset.
  • NoCoda: Assign one violation for every syllable with a coda.
  • The mirror-image constraints - NoOnset or Coda - aren’t in CON.
• Let’s calculate the factorial typology.
  • Four constraints
    • (1) Onset (2) NoCoda (3) Max-C (4) Dep-V
    • How many possible rankings?
    • Let’s enumerate them.
  • Let’s try a few rankings…
  • Four inputs:
    • /V/
    • /CV/
    • /CVC/
    • /VC/
  • Four candidates per input:
    • a. /V/
    • b. /CV/
    • c. /CVC/
    • d. /VC/
  • Are any candidates harmonically bounded?
• Let’s test our factorial typology using software (OT-Help)

Complex margins

• “margin” = onset or coda (i.e., anything except a nucleus)
• An onset or coda is said to be “complex” if it contains more than one consonant.

• What’s missing?
• Phonologists have proposed…
  • *ComplexOnset: Assign one violation per complex onset (with >1 consonant)
  • *ComplexCoda: Assign one violation per complex coda (with >1 consonant)
• …or just *Complex: Assign one violation per complex margin (onset or coda)
• Again, there are no constraints such as OnsetsAreComplex or CodasAreComplex.

Recap of important ideas

• The syllable and its structure
• Why have syllables in our model if we can’t really hear them?
• The sonority scale and syllables
• Syllable structure typology
• How to do factorial typology
• Using factorial typology to justify a particular constraint set