Clearly, in some simplistic, "common sense" way, it is true that speech components are "nested":

  • Phonemes comprise syllables,
  • Syllables comprise words,
  • Words comprise phrases,
  • Phrases comprise utterances

But obviously this is a "Linguistics 101" sort of understanding. The situation is probably somewhat more complex (e.g. Stress, rhythm, intonation, melody, and so on).

But aside from the above-issues, is this an adequate and serious "working understanding" or does it miss some key points known best to linguists?

  • 2
    Your 'comprise' statements all appear to be backwards to me.
    – curiousdannii
    Oct 13 '16 at 8:54
  • 2
    Syllables may be problematic here, depending on the definition of "word" (think of clitics) there can be words that are smaller than a syllable. Oct 13 '16 at 11:58

It is unclear what you are asking (not a plug for closing!), but I think I understand the issue you're pointing to. The last paragraph gives the answer if you don't need the buildup.

One might get the impression that linguistic units are in a strict hierarchy where every lowest-level string groups its elements together into some number of next-higher level units, which are grouped into even-higher level units, until you get to the top. Under that view, if a and b are parsed as an L1 and (following) c and d are parsed as an adjacent L1, then ab and cd will always act as a unit and bc never will. One reason why one might get that impression is that such an idealization has been widely assumed in the profession, so that is how we cook up our units. A second reason for getting that impression is that there is a cognitive reason why the assumption should be truish. We need to look at the formal properties of representations, to see what if any germ of truth there is to the idea. I will assume A Standard Theory in identifying those properties, and there can be substantial disagreements on those properties (I try to hint at those details: a complete deconstruction of all formal properties of all linguistic theories is plainly impossible).

Starting with the lowest accessible unit which is the feature, each feature is immediately dominated (IDd) by some mother node. One theory of mothers is that there is a unique mother(type) for each node (coronal has the mother Cplace); the better-supported theory is that each node has a finite set of possible mothers(type) (coronal has the mothers Cplace and Vplace). Representations are not recursive, which is obviously very different from syntax. Any node may have multiple mothers (features spread), and here syntax and phonology diverge massively. It has been claimed that immediately-preterminal mothers may ID multiple terminals (multiple values of continuant for affricates), but the evidence for such representations is quite weak. Otherwise, below the root node, every mother has a single daughter. So segments, with Root at the top, are formally characterized in terms of limited-choice mothers, multiple mothers, and single daughters.

There are two theories of what is next. One is "skeletal slots" and the alternative is "other prosodic structure". Skeletal slots are basically CV elements or else generic X: the evidence for a type distinction is pretty much nil, and generic X theory seems to be The Standard Theory (if you have any skeletal slots). This is quite simple: every Root is IDd by a non-zero number of skeletal units (generally the choice is held to be 1 vs. 2), and each skeletal unit dominates at most one Root (with the same caveat about claims for affricates). In fact the only reason for having skeletal slots is to handle geminate consonants in inconvenient positions, such as onsets. If there are skeletal slots, there are almost just a projection of Root (a 1-to-1 mapping between Root and X), except that there could also be a 1-to-2 mapping between Root and X. The multiple mother relationship that one finds between nodes below the root does not exist above the root.

There is a big fuzzy zone when it comes to prosodic structure. If you assume the syllable, which is the most popular prosodic constituent, there is a question of what things the syllable IDs (and what those things ID). For instance, is Onset a constituent of a syllable, likewise rhyme, nucleus, coda etc.? Where do moras, if they exist, fit in? Again going with "most popular" as The Standard Theory, syllables ID segments, and they ID moras which ID segments. A segment can be IDd by two morae, by a mora in σ1 plus by following σ2, and semi-theoretically by μ of σ1 and by σ1 itself, so that you may encounter drawings of [wu] being represented as u IDd by μ in a syllable and then an arc to the left from u to σ, meaning "in the onset". But you can also find the same thing for [uw], with the arc going to the right of the straight line. This is, however, formally gibberish. So while the relational properties of syllable-level units are probably different from segment-level units, we still don't know what those properties are.

Above the syllable, it's a little simpler by stipulation: a foot is composed of one or two syllables, and only syllables; all syllables are in exactly one foot. If you have P-words, all feet are gathered into a P-word. The relationship of syllables to feet is the same as the relation of feet to P-words, except for the limit of feet to two (or actually three) syllables in the foot, which you don't find for foot-Pwd relations.

Notice now that the only thing that is actually common to all of these phonological layers of representation is that there is a notion of immediate dominance. That is pretty much a consequence of the metatheoretical decision to represent things as a combination of other things. I could say a but about how syntax is a bit different in certain ways, but I think the point should be clear.

Here's the common sense bottom line. In order to be able to handle long strings of information, it has to be divided up into a very small set of units. You can look at maybe 6 coins and see that there are 6 coins, but you can't look at 12 coins and see that there are 12 coins, you have to count them. A genus-species analysis of reality is how humans deal with the universe, and mentally organizing numerous articulatory and acoustic properties into "segments" and organizing strings of segments into "syllables" and eventually into "words" gives us a manageable way of grasping all of that information. I think what you are reacting to is the fact that information needs to be arranged hierarchically, in order to be practically processable. However, language is not a strict [{(<><>)(<>)(<><>)} {(<><><>)(<>)} {(<>)}{(<><>)(<>)(<>)}] structure.

  • Thanks for such a thoughtful and detailed answer. I am sincerely indebted to the kindness of strangers like you who have done so much to contribute to my understanding of linguistic theory. Thanks a million.
    – Teusz
    Oct 15 '16 at 11:16
  • Aside from the example with [wu] [uw], in what ways is language (for our purposes: phonetics/phonology) not strictly hierarchical?
    – Teusz
    Oct 15 '16 at 11:17
  • Also, is phonetics/phonology divide visible in this hierarchy? It seems that there's what's above the syllable (phonology) and what's below it (phonetics). With the prosodic stuff in the middle, at the level of the syllable, being pretty much the interface between the two. Is this accurate?
    – Teusz
    Oct 15 '16 at 11:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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