I'm developing a library that takes words broken down into phonemes and counts the number of syllables in them.

The phonemes provided by the CMU Pronouncing Dictionary fall into eight types:

  • Fricative
  • Affricate
  • Vowel
  • Semivowel
  • Stop
  • Aspirate
  • Liquid
  • Nasal

My question is as the title: which of these types of phoneme 'create' syllables?

For example, broken breaks down into B R OW K AH N with B and K as stops separating the two syllables. Which other types of phoneme act like this?

  • I don't think this is that universally definable solely by those categories: For example, liquids can, but don't necessarily do form syllables (compare /ˈeɪ.bl̩/ (the liquid /l̩/ is syllabic) vs / /ɔːlˈðəʊ// (the liquid /l/ is non-syllabic), and categories like vowels don't always form syllables e.g. if they are parts of diphthongs. And then again, syllable formation behaves completely differently across languages; the phonemes that an average English syllable consist of are nothing compared to something like Georgian syllables. Dec 15, 2016 at 21:31
  • @lemontree thanks for your reply, it's obvious that I'm out of my depth! if one assumes English syllables, are there any rules that could be used to catch the majority of cases? Do some types act differently depending on the neighbouring phonemes? Or is there a different subset of linguistics I should look at to process syllables?
    – Luke Moll
    Dec 15, 2016 at 21:56
  • I think this might be basically the same question as linguistics.stackexchange.com/questions/2529/… ? Dec 15, 2016 at 22:04

2 Answers 2


Assuming you are using the CMU dictionary then each phoneme of type "vowel" indicates that there is a syllable. It won't tell you where the break is in a sequence of consonants, but it will (quickly) give you the count.


I have a theory about that -- it is not generally accepted. Consonants as well as vowels can occur either stressed or unstressed (this is allowed in SPE features). Phonemic stress tells you where phonemic syllables are. I notice that you don't include stress among your "types of phonemes", but of course stress may be phonemic.

A syllable consists of a longest sequence of stressed nonsyllabics followed by a syllabic, followed by a longest sequence of unstressed nonsyllabics. So if you know the stress and syllabicity of all segments, you can count the syllables and tell where each syllable begins and ends.

Marking stress with a tic preceding the stressed segment, your example is /'B 'R 'OW 'K AH N/. The stress of consonants works something like other features; for instance here there is the possibility of regressive assimilation, where the stressed /'K/ assimilates in stress to the following unstressed vowel. Then we get phonetic ['b'r'owkahn]. Unstressed (i.e. not syllable onset) [k] is subject in some dialects to lenition: ['b'r'owxahn], or, marking syllable boundaries with a dot and omitting predictable stress, we go from /.B R OW . K AH N./ to [.br'owx.ahn.].

A more intuitive term for the stressed/unstressed distinction might be fortis/lenis.

  • "A syllable consists of a longest sequence of stressed nonsyllabics followed by a syllabic, followed by a longest sequence of unstressed nonsyllabics." — could you expand on what is a syllabic and a nonsyllabic? The CMU dictionary I linked in my post does give the stress syllabicity of all the segments, so that should help.
    – Luke Moll
    Dec 15, 2016 at 22:21
  • 2
    With reservations, a syllabic is a vowel, and a nonsyllabic is a consonant. But there are some difficulties with the traditional vowel/consonant difference, and several phonological theories try to deal with it. The feature distinction [+/- syllabic] is an idea of C.-J. Bailey that is discussed and accepted by Chomsky and Halle in SPE. A syllabic sound occurs as the nucleus of a syllable, and there are as many syllables in a pronunciation as there are [+syllabic] segments. When people refer to CV syllable canons, they usually mean the C and V as - and + syllabic.
    – Greg Lee
    Dec 15, 2016 at 23:17
  • so the cases when counting the vowel phonemes does not work are when the word contains vowel phonemes that are - syllablic?
    – Luke Moll
    Dec 15, 2016 at 23:24
  • Right. Like the nonsyllabic glide parts of diphthongs (which may or may not be called "vowels"). Liquids r, l can also sometimes be counted as vowels. But the argument in SPE turns on the formulation of a phonological rule, not just a terminological awkwardness.
    – Greg Lee
    Dec 15, 2016 at 23:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.