Teaching Halle & Marantz (1993)

I can’t remember a phone number for even a second, and when I’m introduced to people, I lose the beginning of their names by the time they reach the end (even for one syllable names, it seems). So any recounting of the origins of Halle & Marantz will necessarily involve rational reconstruction of what must have been going on in the early 1990’s. That being said, attention to the text reveals the many forces that led to the structure and content of the paper, and thus to the structure of canonical Distributed Morphology. Here I want to concentrate on the relationship between the goals of the paper and the various technical pieces of early DM — why there’s morphological merger, vocabulary insertion, impoverishment, fission and fusion.

It should be clear from the number of pages H&M devote to Georgian and, in particular, to Potawatomi, that a main thrust of the paper is a response to Steve Anderson’s A-Morphous Morphology. Following the lead of Robert Beard’s insights into “Separationist” morphology, we wanted to show that item and arrangement morphology could have its realizationism (separation of the syntactic and semantic features of morphemes from their phonological realization) and eat it, too. So, as we stated more directly in “Key Features of Distributed Morphology,” the aim was a marriage of Robert Beard on separation and Shelly Lieber on syntactic word formation — Late Insertion, and Syntax All The Way Down.

However, one shouldn’t forget that I wrote a very long dissertation turned book in the early 1980’s that concerned the relationship between word formation and syntax. The work is titled, “On the Nature of Grammatical Relations,” because it is, in a sense, a paean to Relational Grammar. It re-considers some of the bread and butter issues in RG (causative clause union, applicatives (advancement to 2), ascensions) within a more standard generative framework, with a particular emphasis on the connection between word formation and syntax. So, morphological merger between a higher causative head and the embedded verb might both be the cause of word formation (a verb with a causative suffix) and the structure reduction associated with causative clause union (or “restructuring”). Within this framework, morphological merger is distinct from traditional affix-hopping and from head raising, which don’t by themselves cause structure reduction. Baker’s subsequent work on “Incorporation” tried — and, in my opinion, failed — to unify head raising with the structure reduction associated with morphological merger. These issues are still quite live — see, e.g., Matushanky’s work on head movement.

These days, it might be useful to review the work from the early 1980’s on word formation and syntax. Richard Sproat’s papers are exemplary here. Those of us thinking hard about the issues explicitly connected the “bracketing paradoxes” of inflection (affix hopping creates a local relation between a head and inflection when the syntactic and semantic scope of the inflection is phrasal, not head to head) to similar mismatches between morphological and syntactic/semantic scope exemplified by clitics in particular (so “played” = “the Queen of England’s hat”). While it’s possible to think of all these bracketing mismatches as arising post-syntactically from a PF side morphological merger operation, my book explored the possibility that the same word formation operation of merger could feed the syntax, yielding syntactic restructuring in the case of causative constructions, for example. This may or may not be on the right track, but, as Matushansky makes clear, any phase-based Minimalist Program type syntax adopts an approach to cyclicity that would allow PF-directed morphological merger to feed back into the syntax.
It’s quite remarkable that, given my own pre-occupation with morphological merger and its potential interaction with the syntax, H&M write as if the field had coalesced around the conclusion that syntactic word formation was largely the result of head movement (raising) and adjunction. I’ll blog about this later, but pragmatically, adoption of this assumption about word formation allowed for a straightforward comparison between DM and Chomsky’s lexicalist syntactic theory of the time in the last section of the paper. Nevertheless, H&M are assuming that something like morphological merger/affix-hopping/lowering was necessary to create phonological words. So, for the “syntactic structure all the way down” key feature of DM, H&M are promoting head movement and adjunction as well as morphological merger. H&M leave aside any question about whether morphological merger might feed syntax.

For the “late insertion” key feature, H&M propose a particular technology for Vocabulary Insertion. The empirical target here is contextual allomorphy and (local) blocking relations. One could conclude, then, that the core of DM are the mechanisms of syntactic word formation and the mechanisms of PF realization — and the mechanisms proposed in H&M have been the topic of continuous research for the last 25 years.

What, then, about Fission, Fusion and Impoverishment? For these mechanisms, there were two driving forces at play: empirical domains of interest to Morphologists and the particular research of Eulalia Bonet and Rolf Noyer, which we were convinced by. Fission is a particular approach to the appearance of multiple exponence, and was expertly employed by Noyer in his analysis of Semitic verbal agreement. Fusion involves a head-on tackling of apparent portmanteau vocabulary items. To derive our analysis of syncretism, we required the one to one connection of terminal nodes to vocabulary items, and Fusion was in essence a brute force mechanism for covering situations in which arguably multiple terminal nodes feed the insertion of a single vocabulary item. Impoverishment accounts for two types of phenomena. The first is exemplified in Bonet’s work on Catalan clitics: the use of a unmarked Vocabulary Item in a marked environment. I still believe that the Impoverishment analysis is required to separate standard contextual allomorphy, where a marked VI appears in a marked environment (and a more general VI occurs elsewhere) from situations in which a more general VI occurs in a particular environment — the main argument is that the environment for VI and thus contextual allomorphy is local, while Impoverishment can occur at a distance. The other use of Impoverishment is for systematic paradigmatic gaps — where, for example, gender distinctions are lost in the plural, say. Here, the feature designations of VIs are sufficient to generate the forms without Impoverishment, but Impoverishment explicitly states the underlying generalization (e.g., no gender distinctions in the context of plural).

Jochen Trommer and others have shown that, by playing with the mechanisms of Vocabulary Insertion and with the assumptions about syntactic structure, none of these mechanisms are required to cover the empirical domains for which they were exploited in H&M. That they’re not necessary does not entail that they’re not actually part of the grammar — maybe they were the right approach to the phenomena to which they were applied. Personally, I believe the evidence for Impoverishment is strong, but I no longer adopt Fission and Fusion in my own work (although I’ll happily endorse them in the work of others).

To summarize, H&M lays the foundation for the syntactic word-building and late insertion theory of DM by describing the mechanisms of head movement and adjunction and Morphological Merger for word formation and the mechanisms of Vocabulary Insertion for late insertion. There’s way more of interest going on in the paper, which is, in bulk, a response to A-Morphous Morphology and to Chomsky’s then current version of lexicalism for inflectional morphology. What’s unfortunately largely missing is the concerns of “On the Nature of Grammatical Relations” — the precise interaction of word formation and syntax.