Romanian al and the syntax of case heads

ROMANIAN AL AND THE SYNTAX OF CASE HEADS Ion Giurgea* Abstract: After briefly presenting the distribution of the Romanian genitival agreeing particle al and the most important results of the previous research, I compare three recent analyses of al that are based on the idea that al is essentially a genitive marker and make use of a K (Case) projection: (I) al is a complex of functional heads (K-P+Agr) in the extended projection of the possessee; (II) al is a K head that forms a constituent with the genitive DP; (III) al is an Agr morpheme projected at PF by a genitival K head that forms a constituent with the genitive. I first compare analysis (I) with analyses (II)-(III) and conclude that analysis (I), although it offers a straightforward explanation for agreement, is contradicted by some distributional facts which indicate that al and the genitive form a constituent. Moreover, it needs an important modification in order to account for the fact that al-genitives can appear outside DPs, in predicative position. Analyses (II) and (III), in which al forms a constituent with the genitive, do not have these empirical problems, but require some modifications of the current minimalist assumptions about structural case in order to deal with the alternation between al and prepositional genitives. I then compare analyses (II) and (III) and I conclude that (II) is preferable because it can account for the loss of agreement of al in some varieties of Romanian. Keywords: genitive, possessors, case (case markers, structural case), agreement (possessor agreement) 1. The distribution of al In Romanian, genitives marked by the oblique (i.e. genitive-dative) morphological case and agreeing possessors1 are introduced by the so-called “possessive (or genitival) article al”. This element agrees with the head noun (the ‘possessee’) in gender and number2, having the forms M.SG al, F.SG a, M.PL ai, F.PL ale: (1) a. b. o parte a oraşului a part-F AL.F.SG city-the.OBL aceste rude ale noastre these relatives-F AL-F.PL our.F.PL Al does not appear if (and only if) the genitive/possessive immediately follows the suffixal definite article of the matrix DP: (2) a. b. * prietena (*a) mamei friend-F.the al.FS.G mother-the.OBL prietena bună *(a) mamei friend-F.the good al.F.SG mother-the.OBL The “Iorgu Iordan – Alexandru Rosetti” Institute of Linguistics of the Romanian Academy, . 1 I use this term for the category traditionally labeled “possessive adjective”. The fact that they compete for the realization of structural adnominal case and that they evne occupy the same position as genitive-marked DPs in some languages, including Romanian, indicate that “possessive adjectives” are not adjectives, but rather pronouns, i.e. DPs (see Dobrovie-Sorin and Giurgea 2011). 2 Case agreement is sometimes possible in the DP-initial position, see section 4.4. 70 Ion Giurgea Al-constituents are normally postnominal; they can appear DP-initially, in which case they mark the DP as definite. With an overt possessee, this position is normal only with wh-genitives (see (3)), otherwise it is obsolete, archaic and poetic (see (3)); with pronominal possessors, it is still found in regional varieties: (3) a. b. c. [ale cărei] rude al-F.PL whose.F.SG relatives-F ‘whose relatives / the relatives of which’ ![ai noştri] fraţi (high style/ironic/regional) al-MPL our.M.PL brothers-M ‘our brothers’ ![a lumii] boltă (archaic, poetic: Eminescu, Scrisoarea I) al-F.SG worldthe.OBL vault-F ‘the world’s vault’ Al-phrases can also appear without an expressed possessee, in which case they are usually interpreted as relying on ellipsis of the possessee, as in (4a); as in other DPs without an overt N, the possessee can also be interpreted non-anaphorically, as [+human], in the masculine plural, as in (4b). (4) a. b. Casa Mariei e mai mare decât a Ioanei (N ellipsis) house-F.the Maria-the.OBL is more big than al-F.SG Ioana-the.OBL ‘Maria’s house is bigger than Ioana’s’ Ia adus pe toţi ai lui (non-anaphoric empty N) CL.ACC has brought PE all al-M.PL his ‘He brought all his people/folks’ These DPs are interpreted as definite. Thus, (3) and (4) can be subsumed under the generalization that a DP-initial al in surface order (i.e. not considering any possible covert material) marks the DP as definite. This seems to suggest that al-phrases in (4) occupy the same position as in (3), but we should notice that the elliptical use is fully acceptable with any kind of possessors, as opposed to the prenominal use in (3) – see (3b-c). We will come back to this issue in 4.4 below. 2. Results of previous research 2.1 Al dropping as a PF-phenomenon Several studies have shown that the absence of al in adjacency with the suffixal definite article – see (2) above – is to be analyzed as a PF-phenomenon, as it involves linear adjacency rather than a different structural position of the al-less genitive (Ortmann Romanian al and the syntax of case heads 71 and Popescu 2000a, Dobrovie-Sorin and Giurgea 2005, Cornilescu and Nicolae 2011, Beavers and Teodorescu 2012, Giurgea 2012, 2013a, Giurgea and Dobrovie-Sorin 2013)3. There are several possible formulations of the linear adjacency conditioning of al dropping. Cornilescu (2003) proposes that the head represented by al (a K-P + Agr head) adjoins to D+def when the two are adjacent in linear order, by a head-adjunction operation of the same type as the one found in P + D complexes like fr. du (= de le ‘of the.MSG’), it. col (= con il ‘with the.M.SG’). In the Distributed Morphology framework, al dropping can be analyzed as the contextual insertion of a null allomorph of al. The rule can be formulated as in (5) (see Giurgea 2013a; the condition (ii) is necessary because al dropping only obtains if the preceding article is the article of the matrix DP, with which al agrees, and not the definite article of another DP, see Giurgea 2012, Teodorescu and Beavers 2012)4: (5) al has a null allomorph iff (i) it is (linearly) adjacent with -L (ii) al and -L share -features as a result of agreement 2.2 Al as a genitive marker The generative analyses of al have either privileged its agreement features and its use in DPs with no overt possessee N, as in (4) above – hence the treatment of al as a (matrix) D + pro-N that takes a genitive specifier (Dobrovie-Sorin 1987, 2000, 2002, d’Hulst et al. 1997) – or the fact that it only introduces genitives and possessives, alternating with prepositional genitive markers – hence the idea that it is essentially a genitive marker (Cornilescu 1993, 1994, 2003). The first type of analysis (as D + pro-N) analyzes postnominal al, see (1), (2b), as a relativizer (Dobrovie-Sorin 2000, 2002), but this hypothesis is contradicted by the fact that postnominal al- phrases can be 3 This can be shown, inter alia, by the fact that in coordination of genitives following the definite article, al appears on the second adjunct (in (i), the singular agreement on the verb shows that the (...truncated)