The House that Halle Built

…and we’re back. A couple of years ago, there was a lively discussion about phonological opacity that was split between Mr. Verb and phonoloblog. Mr. Verb has now posted a new installment — well, sort of. The post is mostly just a pointer to this MIT News piece on Morris Halle, but Mr. Verb references the earlier discussion (calling it the “Opacity Wars” in a later comment) and explicitly invites some reaction. This is mine.

(Before moving on, you may want to (re-)acquaint yourself with the earlier discussion, all the links to which can be found here. Mr. Verb says “start here and work back”, but that’s hardly helpful; the only link in that post is a totally useless one to phonoloblog‘s main page.)

First, let me get this out of the way. I have nothing but respect and admiration for Morris Halle and his accomplishments. The (sub)field that I call home would probably not exist — certainly not in its current form, warts and all — without Halle, his work, and the work of his many students (all of them, not just and not excluding those who remain loyal to the rules-and-their-ordering program). Halle and his work have deservedly been fêted many times and in many ways: a Festschrift in 1973, another in 1984 (recently republished, with an “and” curiously added to the title), a recent conference, … — I could go on, but I’ll stop there.

My first face-to-face encounter with Halle was 15 years ago, at the “Is the best good enough?” syntax conference at MIT in 1995 (glorified proceedings here). After my talk, Halle walked up to tell me he’d liked it. I thanked him, and then he said: “Tell Alan [Prince] that this stuff [OT] may work for syntax, but it won’t work for phonology.” (After 15 years, I can’t claim that this is a direct quote, but that was definitely the essence of it. I remember thinking at the time that I could interpret this comment either as a deterrent or as a challenge; clearly, I chose the latter.)

Now I’m no mind reader, but it’s a good guess that the problem that Halle saw with OT for phonology is that it doesn’t involve extrinsic ordering of phonological operations, which is what Bromberger & Halle (1989, LI) famously regarded as the essential difference between phonology and syntax. But let’s be clear about one thing: Halle has made lots of significant contributions to phonological theory, and extrinsic ordering ain’t one of them — by his own account of the matter. As Bromberger & Halle point out (beginning on p. 65 of the original and on p. 165 of this reprint):

Extrinsically ordered rules […] were employed in a synchronic account of the phonology of a language by the great Sanskrit grammarian Pāṇini over twenty-five hundred years ago. […] Attempts to utilize extrinsically ordered rules in the description of synchronic […] phenomena date back to the 1930s. One of the earliest is Bloomfield’s (1939) paper “Menomini Morphophonemics.”
[ Get a reprint of that paper here. — EB ]

Bromberger & Halle then go on to point out that Bloomfield’s paper was largely ignored and that Chomsky appears to have re-discovered the value of ordered rules in synchronic descriptions in his 1951 Master’s thesis, The Morphophonemics of Modern Hebrew. In any event, extrinsic ordering of phonological operations is not Halle’s legacy; it’s Pāṇini’s (and, depending on how you look at it, also Bloomfield’s and/or Chomsky’s to some extent).

ANYWAY, I think that Halle and others are essentially right in their vehement defense of rule ordering: there are some interactions among phonological generalizations that have yet to receive a more satisfying account. Contrary to popular belief, however, this set of interactions is only a proper subset of those that fit the definition of “opacity”, as I recently laid out in this paper (currently under revision for publication). The defenders of rule-based serialism should either counter what I say in that paper or concede that rule-based serialism fails at least as much as it succeeds in this area once thought to be the framework’s lynchpin.

It’s of course uncontroversially true that early work in OT had aimed to account for interactions among phonological generalizations by means other than extrinsic ordering (and with objects other than rules). In the MIT News piece, Donca Steriade is quoted as saying that “it’s become obvious that’s not possible.” It appears that at least these three commenters on Mr. Verb’s post see this quote as (further) evidence that “the tide is turning” back to the good ol’ days of ragin’, full-on rule-based serialism and that OT has been shot to shit. They wish — proponents of OT are just as essentially right in their vehement defense of constraint ranking: here, too, there are some interactions among phonological generalizations that have yet to receive a more satisfying account.

Note that I wrote “[…] a more satisfying account” in both cases above, not “[…] a more satisfying explanation” — I think the latter term is inappropriate here, because I believe that there is a deeper insight that underlies both what we have heretofore best accounted for with ordered rules and what we have heretofore best accounted for with ranked constraints. That deeper insight is that the applicability or application of one phonological generalization in some cases depends on the application or applicability of another. Ordering among rules gets you some of these interactional dependencies (e.g. counterbleeding) and ranking among constraints gets you others (e.g. disjunctive blocking) — and of course there is some overlap (e.g. counterfeeding-on-focus) and some places where additional mechanisms are necessary in both cases (e.g. nonderived environment blocking). Some phonologists appear to believe that the answer is a framework with some (principled?) combination of rules and constraints, or serial ranked-constraint evaluation, or some such concoction. But I believe — pretty much as a matter of faith at this point, I fully admit — that there’s some yet-to-be-discovered mechanism that will explain all of the interactional dependencies in some unitary fashion (and, one hopes, other stuff as well).

I think I’ll spend my time thinking about that great big framework in the sky, and caring less about who thinks who pwned what and who openly admitted what failure when. Sure, it’s nice to get our competitive juices flowing, but in the end, we’re only fighting amongst ourselves — and in case you hadn’t noticed, there are people out there trying to take down all of theoretical phonology. A fellow phonological theorist recently wrote to me and said, “we [phonologists] need to stick together.” In the context of our discussion (which I won’t go into here), I disagreed — but I agree with the sentiment here. If we don’t at least have our facts straight amongst ourselves, we can’t hope to get beyond our own petty arguments to mount a worthy and unified defense of the House that Halle Built.

4 thoughts on “The House that Halle Built

  1. Kyle Gorman

    Great post, Eric. A couple comments:

    – any discussion of “Donca vs. Morris” should probably not be just about “declarative vs. emergent”, but also the role of substance in phonology. Donca’s position is well-known, and Morris’ has evolved a bit (a second-hand personal communication of recent provence from Morris says that the articulatory and acoustic content of features are the “non-negiotable” part, which is a bit different than Jakobson, Fant, and Halle 1952, no?)
    – I would strongly suggest Eric’s paper on the opacity/extrinsic ordering dissociation. Though, my own suspicion (once again, just a suspicion) is that the old definition just isn’t valid as an object of study. I of course, have a duty to propose my own…
    – I’m confused about the relationship between extrinsic ordering and disjunctive ordering, the former of which I associate with serial phonology, and the latter with allomorphy in inflection. Does Panini have both? Are disjunctive orderings extrinsic orderings, even if they’re derived[sic?] by the Paninian principle of more specific rules bleeding less specific ones (Halle and Marantz 1993 seem to think so: see p. 119ff.)? Isn’t this kind of thing orthogonal to the complaint from NGP and OT that extrinsic ordering is implausible? Is there an “argument from personal incredulity”?
    – I think those who “trying to take down all of theoretical phonology” crosscuts serialists/optimalogists, not to mention many such people are affiliated with neither school, even casually. So yeah, we of the symbolic cognition school should be working together. Maybe opacity (under any definition?) is the key to dismantling these approaches.

  2. Kyle Gorman

    Just had a chat w/ Chandan Narayan about my #2 above. Apparently Panini does have a sutra about extrinsic ordering of the non-disjunctive type. He also mentioned the “asiddha” principle: “The grandfather does not know what the grandson does”. Note that this particular principle has been under consideration for a while now, most recently with Colin Wilson and Matthew Wolf discussing “counterbleeding from the past”.

    Note that Kiparsky (in the first of many Morris Halle festschriften, none the less) claimed that Paninian disjunctive ordering was superior to generic extrinsic (but non-disjunctive) ordering, though they obtain the same extensionally equivalent languages, because in a disjunctive block, there is no psychologically implausible intermediate step in which, say verbs “hit” and “keep” have past tenses, but “love” doesn’t yet. That point really cross-cuts “declarative vs. emergent”: see Bonet et al. 2007 (in Lingua).

    (Thanks, Chandan and Ashok Aklujkar)

  3. Eric Bakovic Post author

    Thanks for the comments, Kyle. I hope it’s clear from my paper that I basically agree with your suspicion — the only sense in which I think opacity can be a valid object of study is in the original hard-to-learn sense of Kiparsky, but I also have serious doubts that it’s even true that (all) opaque interactions are hard to learn.

    Small correction: it’s counterfeeding from the past. For those who may be wondering what this is, I refer you to Wilson’s paper and Wolf’s handout.

    Finally, this paper of mine addresses so-called “elsewhere effects” (in phonology) in detail. I’m currently working on a book-length version of it, which I will announce here when I’m done.

  4. Ricardo

    I strongly agree with Kyle: the problem of opacity, as addressed in much of the optimality-theoretic literature, is not well-defined. As I observed some time ago in my dissertation (pp. 65ff), having an intensional definition of the concept of opacity is by no means the same as knowing its extension. Deciding whether or not a particular phenomenon is opaque typically raises a host of questions about the theory of representations, the constraint set, the lexicon, the morphosyntax-phonology and phonology-phonetics interfaces, and the relative scope of synchronic and diachronic explanation. Unless we simultaneously engage with all these questions while we talk about "opacity", we will make no progress: for a passage of purple prose expressing the same views, see this paper (p. 37).

Leave a Reply

Your email address will not be published.