What would 'recursion' mean in phonology?

That’s the title of Bob Ladd‘s talk at the Recursion in Human Languages conference, organized by Dan Everett and scheduled to take place in Illinois in late April 2007. (See the full program here; Bob’s talk, scheduled last on Saturday afternoon, is one of only two talks that seem to specifically address phonology.)

I’m glad that a phonetician/phonologist of Bob’s stature and respectability was invited to this conference. I also like the title of his talk (although this may just be the topic of the talk, not the actual title, according to the call for papers — but I hope Bob keeps it).

I’d like to throw this question out to phonologists (those who read phonoloblog, anyway). What do you think ‘recursion’ would mean in phonology? Maybe if we get some good discussion going, we can forward it on to Bob (or invite him to join in on the conversation). Leave your comment below, or if it’s something extensive, consider writing a separate post. (If you’re not already a phonoloblog author, all you have to do is ask me.)

6 thoughts on “What would 'recursion' mean in phonology?

  1. Jessica Barlow

    Someone once asked me this, and my first thought was of language games, like the ‘aga’ game (so, phonology would be [fɑgɑnɑgɑlɑgɑʤɑgɑ]), and each vowel within that could be replaced by yet another ‘aga’, and so on.

    I suppose reduplication could be considered another example.

  2. Kie Zuraw

    Recursion of prosodic structure comes up in lots of analyses. It’s maybe not the best example, because the recursive structure isn’t surface-obvious the way it can be in syntax.

    To pick just two analyses that we discussed in class this week:

    Ito & Mester’s 2003 monograph Japanese Morphophonemics has structures in which a prosodic word can be made up of other prosodic words, as in [[denki][kamisori]] ‘electric razor’ (p. 221), where square brackets enclose p-words. (The reason for calling the whole compound a p-word here is so that compound accent applies at the juncture. I&M propose that the (maximal) p-word is the domain of compound accenting. There are other compounds with different accentuation, and I&M ascribe different prosodic structure to them.)

    Similarly, Ito & Mester’s 2006 paper “The onset of the prosodic word” (people.ucsc.edu/~mester/papers/2006itomesteronspword.pdf) proposes structures in which English function-word proclitics can recursively adjoin to a following p-word: [give] [ya [a [job]]].

  3. Kie Zuraw

    …and Jessica’s comment about reduplication made me think a bit about the limited recursion of reduplication in Palauan.

    In Palauan, you can reduplicate a reduplicated word, as in

    burek ‘dye’, bi-brurek ‘yellow’ (the first r is an infix), be-bi-brurek ‘yellowish’

    me-suud ‘shred’, me-su-suud ‘keep shredding’, me-se-su-suud ‘easy to shred’.

    (examples from Josephs 1990 New Palauan-English Dictionary)

    I don’t think that the Ce/i- and CV-/CeC-/CeCe-/CCe- reduplicants are really distinct morphemes, since their sets of meanings are just about the same (the second pattern has a few extra meanings that the first doesn’t seem to), and it’s partly phonologically predictable which one you get. And I think there’s a plausible phonological reason why an already-reduplicated word can only take Ce/i- (if the other reduplication pattern applied to itself, you’d get Cschwa as the outer reduplicant).

    So that suggests that you can freely add a reduplicant to a word, even it’s already reduplicated. BUT, as far as I know you can’t have triple reduplication! Does that mean that when a second reduplicant is added (REDUP + bi-brurek), it has to look inside what it’s attaching to and check that what comes after the first reduplicant is not another reduplicant?!

    A deflationary note, though: It’s possible that the CV-/CeC-/CeCe-/CCe- reduplicants are all fossilized (or Level I, if you like)–hence their bigger set of meanings–and the Ce/i- pattern is the productive (or Level II) reduplicant, and there really is no recursion here, not even one round. That would also explain why there don’t seem to be any Ce/i-Ce/i-X words, except in cases that could be analyzed otherwise. For example, the apparent Ci- in be-bi-brurek can be analyzed as the allomorph of CV-/CeC-/CeCe-/CCe- that occurs when the stem is CC-initial.

  4. Mike Maxwell

    IIRC, SPE had a stress (prosodic) rule that operated recursively at the phrase level. Do modern theories talk about phrasal accent patterns? I don’t know anything about downdrift, but I assume it is not sensitive to syntactic patterns, only to the (immediately preceding) tones, and is therefore finite state.

    There is the exceedingly limited case of center-embedded morphology, where I suppose you could claim something about the stress patterns. The typical example is [missile], [anti-[missile] missile], [anti-[anti-[missile] missile] missile] etc.

    Everything else about phonology seems to be finite state: vowel harmony, for instance, can be handled by iterative rules just fine. In fact, just this issue came up in Johnson’s work (C. Douglas Johnson, Formal Aspects of Phonological Description, Mouton, The Hague, 1972). He showed that phonological rules can be modeled as finite state transducers provided they don’t re-apply to their own output an unlimited number of times (but they can iterate right or left). AFAIK, no convincing example was ever found of a phonological rule that needed to reapply in a way that would violate this condition. (I suppose the same issue comes up in OT, although I’m not sure how the details work out.)

  5. Diane Lesley-Neuman

    I am mulling over aq problem that seems to require one prosodic word to be infixed inside of another. Is this possible?

  6. Pingback: phonoloblog»Blog Archive » Are we flat?

Leave a Reply

Your email address will not be published. Required fields are marked *