No Null Nouns Alex Alsina UPF April 15, 2010 Aim of the talk To argue that apparently headless NPs are really headless. Examples: (1) a. M’agrada més aquest llibre antic amb tapes de cuiro. b. M’agrada més aquest antic amb tapes de cuiro. c. M’agrada més aquest amb tapes de cuiro. d. M’agrada més aquest antic. e. M’agrada més aquest. Aim of the talk Structure of a headed NP: DP D′ D NP N′ AP N A aquest llibre antic Alternatives to choose from A. Structure of an NP B. Structure of an NP with an empty head: without a head: DP DP D′ D′ D NP D NP N′ AP AP N A A aquest ∅ antic aquest antic Existing work The majority position is in favor of alternative A: Bosque 1989, Bosque i Gutiérrez-Rexach 2009, Brucart 2002, Brucart i Rigau 2002, Wheeler 1991, among others. At this point I can’t cite any work explicitly assuming alternative B, although it is the natural assumption in analyses within LFG, HPSG, and other frameworks that factor grammatical information into different structures. Structure of the talk Argument 1: positing empty nouns has the effect of making a potentially empirically contentful hypothesis unfalsifiable. Argument 2: two types of determiners: those that may, and those that may not, introduce headless NPs. Argument 3: prenominal adjectives: not possible in headless NPs. Argument 1 The empty noun reduces a hypothesis/claim to unfalsifiability. The claim that every XP is the projection of an X0 (the projection claim) is a strong claim (a falsifiable claim) provided an X0 is always a word—a lexical item with a phonological representation and a syntactic/semantic representation. If we have evidence that some string of words is an XP, we predict that it should include a word of category X in the right position. Argument 1 What counts as a counterexample to this claim? Eg: a phrase that behaves in all respects like a PP but does not include a preposition. Headless NPs are exactly this case. They are counterexamples to the projection claim. Two responses: (a) keep the claim but make it untestable, (b) abandon the claim. The majority position has been (a): we allow empty Ns so that headless NPs are no longer counterexamples to the projection claim. Argument 1 In fact, once we allow empty X0s, the projection claim is no longer a strong claim: it makes no predictions; it just tells us how to represent phrases. What’s the use of having untestable hypotheses (hypotheses that do not lead to testable predictions)? Better to abandon the projection claim: not all phrases have a head. This spares us from introducing a new hypothesis (the empty X0 hypothesis) whose function is to make another hypothesis vacuous. Argument 2: two types of Ds (2) a. Hem insistit en cada situació (compromesa) (que es va presentar). b.*Hem insistit en cada (compromesa) (que es va presentar). (3) a. Hem insistit en aquella situació (compromesa) (que es va presentar). b. Hem insistit en aquella (compromesa) (que es va presentar). Argument 2: two types of Ds Weak determiners: cada, sengles, qualque, cert, mon, ton, son, llur, tot*, etc. Strong determiners: aquest, aquell, quin, quant, cap, un, algun, tant, molt, etc. If we allow empty nouns, we either cannot account for the difference in behavior or we have to restrict the null N in the appropriate position in relation to a strong determiner. The assumption that a strong D can license an empty N is ad hoc. * Tot, as in tota reclamació, tot document, not as in tots els llibres, tota l’estona, etc., where it is a predeterminer. C-structure and f-structure An alternative that assumes syntactic information is factored into a categorial structure (c-structure) and a feature structure (f-structure) as parallel, co- present levels of representation. A string is grammatical if it satisfies the well- formedness conditions on both structures. Let’s start with the c-structure: a syntactic tree of the usual sort, satisfying Lexicalism: every X0 dominates a fully formed word—a unit satisying morphological integrity and consisting at least of a phonological representation—of category X. C-structure The usual inventory of categories: V, N, P, A, D, C, VP, NP, DP, etc. Functional categories ok, provided they comply with Lexicalism. X-bar Theory: X′′ YP X′(′) X′ X(′) ZP Economy of Expression: all c-structure nodes are optional and are only used if needed for semantic reasons or to satisfy well-formedness conditions. (Bresnan 2001) C-structure of the noun phrase Words of category D in Cat.: el/la, cada, sengles, qualque, aquest, aquell, quin, cap, molt, etc. Spec of NP is reserved for SpecA (specificational adjectives): altre, dos, tres, primer, segon, últim, seu, nostre, etc. D always has an NP as its c-structure complement, if it has any complement. Spec of DP is reserved for predet. tot/s and adv. such as fins i tot, exactament, només, almenys, etc.: tots els llibres, només aquelles notes, etc. Example c-structures DP D′ D NP SpecA N′ N PP aquell altre castell de cartes Example c-structures DP DP DP D′ D′ D′ D NP D NP D SpecA N′ aquell aquell altre PP aquell de cartes F-structure However, the principles given so far allow headless noun phrases with weak dets: cada *(proposta) nova cada *(paquet) que arriba cada *(estudiant) de doctorat *cada They are excluded by well-formedness conditions on the f-structure. F-structures: formally, attribute-value matrices that satisfy a uniqueness condition requiring each attribute to have a unique value. F-structure Every word form has a vocabulary entry that specifies its c-structure information and its f- structure information. In forming the f-structure of a phrase: The f-structure of a phrasal category is that of its head; The f-structure of a functional category is that of its complement phrase. Thus, the f-structure of a DP is the unification of the f-structure of the head D and of the N head of the complement NP. Sample vocabulary entries aquell: D SPEC [DEM +] NOMB sg GÈN masc aquella: D SPEC [DEM +] NOMB sg GÈN fem cadira: N PRED ‘cadira’ NOMB sg GÈN fem calaix: N PRED ‘calaix‘ NOMB sg GÈN masc Sample det-noun sequences DP D′ PRED ‘calaix‘ SPEC [DEM +] D NP NOMB sg GÈN masc N′ N aquell calaix Sample det-noun sequences DP D′ PRED ‘cadira‘ SPEC [DEM +] D NP NOMB sg GÈN masc fem N′ N aquell cadira The two types of Ds Another condition that f-structures must satisfy: Completeness: every f-structure with a thematic role must have a PRED feature. (Bresnan 2001) Words of categories N, V, A generally have a PRED feature. Dets split into two groups: some don’t have this feature (the weak dets) and some have an optional [PRED ‘pro’] feature (strong dets). See vocabulary entries of a D of each type: Example Ds of each kind aquell: D SPEC [DEM +] NOMB sg GÈN masc (PRED ‘pro’) cada: D SPEC [DISTR +] NOMB sg això: D SPEC [DEM +] NOMB sg GÈN masc PRED ‘pro’ Nounless DPs of both kinds DP D′ SPEC [DEM +] NOMB sg GÈN masc D PRED ‘pro’ aquell DP D′ SPEC [DISTR +] NOMB sg Ill-formed f- D structure: violates completeness cada The role of null nouns Having empty categories would play no role in explaining the contrast between the two kinds of determiners. We would need some additional principles to tell us when the null noun is possible and when it is not. In this theory, null nouns are excluded by Economy of expression and Lexicalism. Argument 3: prenominal adjectives Prenominal adjectives: must appear before a noun, follow SpecA, if there are any Examples: bon, mer/a, mal/a, presumpte/a, antic/ga (‘former’), trist/a (‘insufficient’). (4) a. un bon amic, *un amic bon, *aquest amic sí que era bon, *era un bon. b. una mera suposició, *una suposició mera, *aquesta suposició és mera, *era una mera. Strings not allowed by X-bar theory given. Explaining prenominal As Two subtypes of A: A [+preN]. Ex: bon, mer, mal, etc. A [-preN]. Ex: dolent, bo, policial, etc. Many As are unspecified: belong to both subtypes. New c-structure rule (Sadler and Arnold 1994): N A N [+preN] Postnominal APs are projections of A [-preN]. Structures with prenominal As DP DP D′ D′ D NP D NP N′ SpecA N′ N N A N A N [+preN] [+preN] un bon amic els dos presumptes culpables Excluding illicit structures Why are *un bon, *la mera illicit DPs? [+preN] adjectives must appear in the configuration [N A[+preN] N ]. By Lexicalism, both categories in this structure must be words (or dominate words) of the same category. Words have phonological representation. N N N A A N A N [+preN [+preN [+preN bon bon bon ∅ What does the null noun do for us? If we have a null noun as a vocabulary item, we cannot explain these facts. A representation such as the following is possible: N A N [+preN] bon ∅ What does the null noun do for us? A hypothetical construct such as a null category should only be posited if there is clear evidence for it, if its presence is crucial for explaining some facts. There is no evidence for the null noun. In addition having a null noun complicates the explanation of certain facts considerably. It complicates the explanation of the behavior of the two classes of determiners. And it makes the explanation of preN adjectives practically impossible. Therefore We are much better off without the null noun. Thank you References Bosque, Ignacio (1989): Las categorías gramaticales. Relaciones y diferencias. Madrid: Síntesis. Bosque, Ignacio; Gutiérrrez-Rexach, Javier (2009): Fundamentos de sintaxis formal. Madrid: Ediciones Akal. Bresnan, Joan (2001): Lexical-functional syntax. Oxford: Blackwell. Brucart, Josep M. (2002): «Els determinants». A: Solà, Joan et al. (dir.), vol. 2: 1435-1516. Brucart, Josep M.; Rigau, Gemma (2002): «La quantificació». A: Solà, Joan et al. (dir.), vol. 2: 1517-1589. Sadler, Louisa and Douglas J. Arnold. 1994. Prenominal adjectives and the phrasal/lexical distinction. Journal of Linguistics 30:187– 226. Wheeler, Max W. (1991): «Dels quantitatius i altres elements especificadors». Els Marges 43: 25-49.