If we look at Halle & Marantz (1993) for the origins of Distributed Morphology, we see a framework designed to show that inflectional morphemes should be pieces, distributed syntactically and realized phonologically after the syntax. In a sense, Halle & Marantz (1993) is an immediate expansion of Chomsky’s Syntactic Structures, as I will Blog about later. In his analysis of the English auxiliary system employing “affix hopping,” Chomsky does syntactic word formation and late insertion for inflectional morphemes, in much the same way as DM (nihil sub sole novum, particularly in Noam’s shadow).
But Harley & Noyer 1999, in their introduction to DM, actually emphasize thinking that went into my 1997 “No Escape from Syntax” paper and my 2000 WCCFL talk that spawned the heavily cited “Words” manuscript. Thus they present DM as essentially anti-“lexicalist.” There’s a concentration of issues surrounding derivational morphemes and roots, rather than inflection, and one sees there the strands of thought that Heidi will pursue in her later work.
Before turning to the “lexicalism” that DM was “anti-“ in the 1990’s, I should clarify two uses of “lexicalist,” only one of which is relevant to this post. On the non-relevant reading, a “lexicalist” endorses the notion that morphemes are “lexical items” in the sense of units that relate or contain both sound and meaning, with “meaning” broadly construed. This is the “morphemes are signs” position that often gets derided in the literature (see e.g. recent work by Blevins). It’s in this sense of “lexicalist” that, e.g., Lieber in Deconstructing Morphology is a lexicalist and Kayne is a lexicalist. These guys are serious, and their work cannot and should not be dismissed by waving at reduplication, zero morphs, patterns of syncretism in paradigms, etc. (see e.g. recent work by Blevins). It’s great for the field that serious linguists pursue this lexicalist hypothesis.
However, the lexicalism of anti-lexicalism in the 1990’s was the MIT-style lexicalism that was being explored in the Lexicon Project, a position associated with, e.g., certain versions of Lexical Morphology and Phonology, a position inspired by Wasow’s work contrasting lexical and syntactic rules, a position driving early versions of Lexical-Functional Grammar, and a position also endorsed by Chomsky at various points (see the discussion at the end of Halle & Marantz 1993). This view said that there was a difference between word formation before syntax (in the lexicon) and word formation that might be post-syntactic, and that the syntax operated on morphologically complex words from the lexicon, rather than on morphemes. There was a spirit around that a nexus of Chomsky’s “Remarks on Nominalization,” Kiparsky’s Lexical Morphology and Phonology, Wasow’s Lexical vs. Syntactic Rules, Lieber’s work on word formation, and Levin’s work on lexical semantics was creating a coherent and compelling picture of a Lexicon for generative grammar. The “No Escape” paper was meant to pop the bubble specifically by questioning the correlation of properties associated with “wordhood” that underlay the apparent consensus on the Lexicon. To perhaps oversimplify the conclusions of that paper, I argued that the “special behavior” that was claimed to distinguish lexical properties from syntactic properties was better understood as the local determination of properties of roots in the context of the first category node merged above them. Phonological wordhood per se was largely irrelevant to the syntactic, semantic, and morphological properties of a language.
By the end of the 1990’s, any notion of a transtheoretical consensus on a “lexicon” had vanished, and arguments for versions of lexicalism became more nuanced – less vulnerable to the Wreck-It Ralph treatment of “No Escape…” There’s less of a notion that a striking set of correlations converge on properties of a word as opposed to a syntactic phrase, and perhaps more of a notion that the lexicon allows for the unruly (the “lawless” for Di Sciullo and Williams) whereas the syntax plays by the rules. I hope to Blog on more recent anti-anti-Lexicalist positions later in the semester (e.g. papers by Kiparsky and by Bruening). As previewed in the Harley & Noyer article, the 2000’s saw an emphasis on the syntactic treatment of derivational morphology and of uncategorized roots of words. This work was and is not specifically or essentially anti-lexicalist – depends on the particulars on one’s theory of the lexicon. The work does however reject the notion that word formation is ever “lawless” – the adoption of syntactic word formation along with the strict locality implications that go along with this adoption in certain theoretical worlds is supposed to reduce the wiggle room for morphological analysis. So, less “anti-lexicalist” and more “pro-decomposition into minimal syntactic units organized hierarchically and subject to the locality constraints visible in syntax” or some such.
Thanks for this! When I starting studying these things, I came at it from the other way, thinking first about word formation, roots, idiosyncrasy/systematicity, etc. Later, when I got around to the early DM stuff, I was surprised at how inflection-heavy it was. I remember wondering where all the roots were!
I wonder what you think about modern incarnations of the “word”. Peter Svenonius has been emphasizing that words are special (within the spanning framework), and Dave Embick has often emphasized the complex head as the “m-word” which constrains certain processes at PF. Both are treating words as special only morphophonologically, as far as I can tell, which is in contrast to, say, Marit Julien’s “words as syntactic habits” perspective. One could imagine these domains constraining allosemy at LF as well. Bruening’s treatment of the semantic idiosyncrasies of adjectival passives seems reminiscent of Julien’s perspective, only on the LF side. Assuming we can distinguish allosemy from idioms, I wonder if there are any “phrase/word” differences in the locality of alloseme conditioning. If the LF branch allows other PF-like operations that form complex heads, like lowering, m-merger, etc., it may be impossible to tell. (I don’t know if anyone has ever suggested this, but Anton Karl Ingason has a really interesting article where he proposes that you can have dissociated morph insertion at LF.)