Di Sciullo & Williams 1987 -- On the Definition of Word Outline 0. The notion 'word' 1. Listemes 1. Syntactic Objects 2. Unlisted Morphological Objects 3. The Psychological Lexicon 2. Morphological Objects: Rules of Word Formation 1. Rules of Formation 2. Derivation of Argument Structure 3. Conclusion 3. Syntactic Atoms 1. Syntactic Atom vs. Morphological Object 2. Interface between Syntax and Morphology: shared vocabulary 3. Atomicity 4. What Atomicity holds of and why 5. Pseudopassive 6. Atomicity and the Mirror Principle 7. Inflection 8. Bracketing paradoxes 9. Conclusion 4. Morphological Objects 1. Syntactic Words 2. Coanalysis 3. The Phonological Word 4. Conclusion 0. Word 3 different conceptions of 'word' morphological object syntactic atom listeme Morphological objects: the central task of Morphology is to determine the laws of form that determine membership in this set. A syntactic atom is the smallest unit analyzable by the syntax. Syntactic atom =/= morphological object Listeme == whatever has to be listed. There are listed phrases (V-part) and unlisted words (-ness). Listedness is a very uninteresting property. Productivity and listedness are not grammatical concepts. The main message of the work: Lexicalist hypothesis (the atomicity of words) is not a principle of grammar but a consequence of the conception that grammar contains two modules, with different atoms and different rules of formation. 1. Listemes There is a mistaken view that words are listed and phrases are not == but it's mistaken. Many sizes of things have to be listed Listed syntactic objects == idioms. They are syntactic objects They are listed because of a failure to have some expected property (usually semantic) idioms are always syntactic units (though they may have open positions) (take __ to task, push __ too far) all wet, in the dark about, the cat has got X's tongue 1.2.2 unlisted morphological objects There is a view that syntax is productive and morphology is (mostly) unproductive; thus that new morphological objects are listed, but new syntactic ones are not; thus that "blocking" occurs only between morphological objects. When talking about productivity we have to ask over what domain? words that are formed by a productive rule (over any domain) don't need to be listed: -ion, applying to latinate roots, is extremely productive, perhaps fully productive. If it's fully productive, -ion words don't need to be listed. blocking (p10) There is a section on blocking here, which anybody interested in blocking should read carefully. The central claim is that blocking is not restricted to morphology, but occurs across levels. Some highlights: 'gloriousness' blocks 'gloriosity' (!) [Repeating the error of associating blocking with synonymy] 'hotter' blocks 'more hot' 'their' blocks 'them's' 'yesterday' blocks 'the day after today' (!) (Attributed to Aronoff) (!) paradigms: some paradigms contain syntactic entities: the Latin passive present perfect active amo amavi passive amor amatus est 1.2.3 Why words are more often listed than phrases There is a hierarchy of size: morpheme > word > compound > phrase > sentence listed all most many some about 4-5. The morphemes have to be listed, else you have no way of knowing them. The rest follows naturally. 1.3 the psychological lexicon This is a very interesting section, and it's very short. An experiment by Bradley (1980) showed that the frequency of the *cluster* governs retrieval time in a word-recognition experiment, where a cluster is the set of words related by a set of affixes. When the affixes are of the word-boundary kind (#ness, #ment, #er) the frequency of the whole set governs retrieval time; but when the affix is of the morpheme-boundary kind (+ity, ...), the frequency of the word itself, and not its family, governs retrieval time. An experiment by Stanners et al. (1979) found that stem+inflection primes just as well as stem, while stem+derivation primes somewhat more weakly. DS&W attempt to wave this aside. Read it yourself. Words (up to and including compounds) and phrases (after that) have different rules of formation. WORDS ARE RIGHT-HEADED (phrases are not) [what about fuckup, killjoy, tire-bouchon, sacacorchos, karniyarIk, imambayIldI, sinekkaydI ... ?] Syntactic atomaticity: what is formally a word (ch.2) is syntactically opaque (ch.3). The ice cream vendor eats it. Is Aronoff right in claiming that words are derived from words? (I.e. that WFRs apply only to words?) DS&W say no. Look at pluralism pluralist (neither can be derived from the other) church-goer sad-seeming Words are *not* derived from words. Words are derived from morphemes. If syntax is a theory about potential objects (though some may be listed) morphology is a theory of potential objects in exactly the same sense. 2. Morphological Objects: the rules of Word Formation Morphology is a coherent system distinct from syntax (following Williams ... and Selkirk 1982). Words have heads. Suffixes are the heads of their words. Suffixes belong to lexical categories. Affixes have argument structures. There are "external" arguments. There is no principled distinction between inflection and derivation. In syntax, you can tell the head of a phrase. It has the lowest bar level. But in words, everything is of bar level 0. Right-hand Head Rule: the head of a word is its rightmost member. bartend apple pie jet black parts supplier part suppliers Suffixes belong to categories N, V, A. Inflectional affixes are outside the derivational affixes because they mediate between words and syntax. That's just because they are in head position. There are parts in here (pp 26-27) where it gets pretty thick, and I am not going to try to summarize it, beyond noting that it looks like Williams was having a conversation with his former self. 2.2 Derivation of argument structures This is total voodoo. Read it if you will. At your peril. 2.2.1 Compounds I guess Nico will tell us about this. I refuse to even talk about the stuff in this section. 3. Syntactic Atoms (Syntactic atomicity of words) The syntactic atoms are the primes of syntax. syntactic atom =/= morphological object. Syntax and morphology are about different phenomena with different vocabulary; they interact via a *small* shared vocabulary: the lexical categories, features like 'tensed', argument structure. Rules of syntax do not have access to the parts of words directly, but only to the "topmost" properties of a word: the features and argument structure of the topmost word. 3.4 Atomicity holds of everything from compounds down: S > NP/VP > compounds > affixed words > stems > roots Why? Compounds are Head-final (a property of words) (!) Head-final things are syntactically opaque: these are the "words". One way to insure that syntax doesn't see into words: Order Lexical insertion after all the syntax (at S-structure). vs bracket erasure (Pesetsky) ==> "sentence forms" (p54). I'll have to explain this. [I skip a bit here] 3.6 Baker: Mirror Principle Affix order mirrors order of syntactic operations (implicit suggestion: syntax messes with morphology) DS&W: Baker's rules are not syntactic; they are morphological rules that manipulate argument structure. 3.6.2 How lexical rules differ from syntactic rules A lexical rule applies to a *word*, and can change only that word's argument structure. 3.6.3 Noun incorporation Baker: syntax moves N into V DS&W: compounding is a morphological process; incorporated N places a qualifier on an argument of the V (if I don't get around to explaining what tbis is about, don't shoot me) 3.7 Inflection There is no principled difference between derivation and inflection. (A matter of interpretation rather than form) Anderson: inflection is what is relevant to the syntax DS&W: -ion is relevant to the syntax. Affixes more relevant to the syntax occur outside of affixes less relevant to the syntax. This is just because the syntax only sees the outer layer. 3.8 Bracketing paradoxes I mention this only because if anybody is interested in bracketing paradoxes they should read this. It's not very long or deep. 4. NonMorphological Objects This chapter is about things that live in the space (very narrow, acc to DS&W) between morphology and syntax. A syntactic atom is anything that can be inserted under an X[0] (that's a superscript). Such a syntactic atom is not always the same as a morphological object. essui-glace 'windshield wiper' (cf also the ones I mentioned above: tire-bouchon, sacacorchos) these are syntactic words and also syntactic atoms (i.e. they can be inserted under X[0]. But what about the right-hand head rule? It doesn't look like 'glace' is the head of "windshield-wiper". That's because there is a NON-MORPHOLOGICAL WORD-CREATING RULE: N -> VP This (Rule 6) is a nonmorphological word-creating rule of the periphery of the grammar. The section on coanalysis is interesting, very sketchy, and basically suggests that the anlysis of a construction on one level (e.g. the syntactic) might not line up exactly with the analysis on another level (e.g. the morphological). For a more thoroughly worked out version of this idea, read Sadock's Autolexical Syntax.