Deriving the Complementarity Effect: Relativized Minimality in Breton agreement

of 27
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Graphic Art

Published:

Views: 11 | Pages: 27

Extension: PDF | Download: 0

Share
Description
Breton w-agreement is characterized by the Complementarity Effect, which allows pro-dropped but not lexical DPs to control w-agreement. We contrast verbal and prepositional systems: a lexical DP co-occurs with the root form of a preposition, but with
Transcript
    1 Deriving the Complementarity Effect: Relativized Minimality in Breton agreement *    Mélanie Jouitteau Université de Nantes/Naoned melaniejouitteau@yahoo.fr &  Milan Rezac University of the Basque Country (UPV-EHU) milan_rezac@yahoo.ca January 14, 2005 Abstract : Breton φ -agreement is characterized by the Complementarity Effect  , which allows  pro -dropped but not lexical DPs to control φ -agreement. We contrast verbal and prepositional systems: a lexical DP co-occurs with the root form of a preposition, but with a 3rd.sg. (  frozen agreement  ) form of a verb. We argue that frozen agreement arises through φ -relativized locality: the Breton v P independently shows nominal properties, and thus intervenes for agreement. The φ -probe of T Agrees with the v P for 3rd.sg. rather than the v P-internal subject. In the prepositional system on the other hand, lexical DPs occur with bare stems and φ -inflection spells out affixed pronouns. The mechanics predict that in verbal constructions where the subject srcinates outside the v P, it is local enough to control the agreement of T, which correctly yields  Have  under a prepositional analysis as the sole verb immune to the Complementarity Principle. Finally, we propose a typology of complementarity effects in agreement depending on the interaction of intervention (frozen agreement) and syntactic incorporation past the intervener. Keywords : agreement; locality; Breton 1. The Breton Complementarity Effect and locality Breton φ -feature agreement is set apart from more familiar systems by the C OMPLEMENTARITY E FFECT . To a first approximation, φ -agreement morphology spells out only the features of phonologically null DPs. In (1)a, the person and number features of a  pro -dropped subject are coded on the verb in T; in (1)b, where the subject is an overt DP, the verb bears 3.sg. φ -features regardless of the features of the subject. 1  This pattern holds regardless of the position of the DP subject, which may be post-verbal or in a pre-verbal position, as illustrated in (1)c. The A/  Ā  status of the preverbal subject has no impact on the agreement facts. 2  (1) a. Gant o mamm e karf  ent  /* -e    pro  [3.PL] bezañ. with their mother ® would.love- 3.PL  /* 3.SG  be-INF 'They would like to be with their mother.' b. Gant o mamm e karf  e  /* -ent  Azenor ha Iona bezañ. with their mother ®  would.love- 3.SG/*3.PL  Azenor and Iona be-INF *  We would like to thank the organizers and audience of the Fourth Celtic Linguistics Conference at Selwyn College, in particular suggestions by Stephen Hewitt, James McCloskey, and Ian Roberts; and to David Adger for discussion of a pre-presentation version. Thank you to Dirk Bury for his welcome. Earlier versions of this paper were presented at the Université de Nantes/Naoned, University of Toronto, Université de Paris VIII, and CIL 17 in Prague, and we would also like to thank the audiences there, especially Hamida Demirdache, Ora Matushansky, Alain Rouveret, and Alain Kihm. Finally, we are grateful to two anonymous reviewers for comments which have been both helpful and encouraging in preparing the final version. All errors remain ours. 1  We use the following abbreviations: 1/2/3 for person, SG./PL. for number, MASC./FEM. for gender, INF for infinitive, PROG for the progressive particle o , NEG for both of the Breton discontinuous negation markers ne  … ket  , and ® for the Breton preverbal particles a  and e  ( rannig verb  'small part of verb'). 2  On the A/  Ā   status of preverbal subjects in positive sentences, see Stump (1984, 1989), Borsley and Stephens (1989), Schafer (1995), Rezac (2004b), Jouitteau (forthcoming b, 2003, 2005).    2 'Azenor and Iona would like to be with their mother.' c. Azenor ha Iona 1  a garf  e  /* ent   t  1  bezañ gant o mamm. Azenor and Iona ® would.love- 3.SG  /* 3.PL  be-INF with their mother 'Azenor and Iona would like to be with their mother.' We describe the Complementarity Effect as an alternation of rich and invariant agreement in terms of the relationship between φ -feature targets and φ -feature controllers. This alternation follows (2). (2) Complementarity Effect: (i) R ICH AGREEMENT : the φ -features of a phonologically null DP are coded by φ -agreement morphology. (ii) I NVARIANT AGREEMENT : the φ -features of a phonologically overt DP are not coded by φ -agreement morphology on the target. The target assumes either FROZEN AGREEMENT  (a 3.sg invariant form) or the bare stem. The alternation characterized as the Complementarity Effect also shows up with a preposition and its object, which is the other context in Breton where rich morphology exists. (3) a. gant/*gant i  [o mamm] with/*with.3.SG.FEM. their mother 'with their mother' b. gant i  /*gant  pro [3.SG.FEM.] with.3.SG.FEM./*with 'with her' We observe that the invariant form of the preposition in (3)a is the bare stem rather than frozen agreement as defined in (2). There are significant asymmetries between the verbal and prepositional systems, discussed in detail in section 4 after we lay out the core of our analysis, which are indicative of precisely this difference: a verb in the context of an overt DP controller is 3.sg. (frozen agreement as can be seen in (1)b), while a preposition in the same context is the bare stem. Both frozen agreement and the bare stem instantiate invariant agreement, which shows up when the potentially agreeing argument of a verb or a preposition is a full DP. There is exactly one exception to the Complementarity Effect: 3  the verb  Have  uniformly codes the φ -features of all DP subjects, be they overt or null. (4) Bremañ (, Azenor ha Iona) neus- ont  (Azenor ha Iona) un ti Now (, Azenor ha Iona) have-3.PL. (Azenor ha Iona) a house 'Azenor and Iona/they have a house now.' The Complementarity Effect is of considerable importance in the setting of the minimalist program, where featural agreement plays a central role in syntactic dependencies. It has also 3  We set aside resumptivity, where an overt DP is linked to a null  pro  which itself controls agreement; this yields only apparent overt controllers of rich agreement. Clause-external DP subjects require such resumption by a  pro , which then triggers rich agreement. Subjects that precede the sentential negation always have an Ā  -reading (as opposed to other preverbal subjects which may be in an A-position), and also always trigger rich agreement. For Schafer (1995), Ā  -movement past negation violates Relativized Minimality and the dependency thus requires linking of the DP subject to a resumptive rather than a trace or copy (cf. Rizzi 1990, Cinque 1990, Postal 1998; cf. Ouhalla 1993). For Jouitteau (2005), Neg is a C head and subject movement past it triggers the that  -trace effect. Both analyses predict the rich agreement as resumptive  pro  found with pre-negation subject.    3 always been something of a syntactic anomaly, because it looks as though a central syntactic process, φ -agreement, refers to phonological overtness. Two lines of analysis have been pursued for the Complementarity Effect in the Celtic languages, both carefully developed for Breton by Stump (1984): the Incorporation Analysis and the Agreement Analysis. On the I NCORPORATION A NALYSIS , rich agreement morphology is the spell-out of a pronoun itself, not of φ -agreement with it. Only pronouns stand in the relevant clitic-like relation to T and P to be spelled-out as their affixes; phonologically independent DPs naturally do not. The Incorporation Analysis is pursued for Breton by Anderson (1982), Stump (1984) who rejects it in favor of an Agreement Analysis, and for similar facts in Irish by Pranka (1983), Doron (1988), Ackema and Neeleman (2003) and in Scottish Gaelic by Adger (2000). These treatments differ in whether the amalgamation of T/P and the pronoun happens syntactically as (Stump 1984), as in the usual treatment of syntactic clitic dependencies (see Sportiche 1997 for an overview), or in a post-syntactic, prosody-sensitive component (Pranka 1983, Doron 1988, Adger 2000, Ackema and Neeleman 2003). The Incorporation Hypothesis has the obvious advantage of being rather minimal: very little needs to be said beyond the cross-linguistically familiar fact that weak pronouns but not full DPs can be affixal. The impossibility of agreement morphology tracking an overt DP follows from the Theta Criterion, which disallows a pronoun to co-occur clause-internally with a DP bearing the same theta-role in a language that lacks clitic doubling. 4  However, the Incorporation Analysis has no obvious way of handling the  Have  exception. Furthermore, we will show below that while a post-syntactic version of it seems correct for preposition-object agreement, it cannot work for T-subject agreement because there is in fact 3.sg. φ -agreement (frozen agreement) in the latter context. The alternative is the A GREEMENT A NALYSIS , which posits that Breton uses more or less the same φ -agreement mechanism as that found in languages like English and French. In the framework of Chomsky (2000), T has a φ -probe which is valued by the interpretable φ -set of the DP subject, with the same locality properties as φ -Agree elsewhere. The Agreement Analysis has a certain theoretical appeal in that it does not make Breton a language without φ -probe on T. However, the Agreement Analysis has no explanatory answer for the complementary distribution of φ -agreement and overt DPs which the Incorporation Analysis handles so well. As applied to Breton by Stump (1984, 1989) and to Irish by McCloskey and Hale (1984), the complementarity is handled by a stipulation that φ -agreement in these languages cannot take place with a phonologically overt DP; updating the terminology to Agree, we will call this the * φ -PHON  CONSTRAINT : (5) The * φ -PHON constraint: φ -Agree is limited to phonologically empty goals. Like the Incorporation Analysis, the Agreement Analysis has so far had no good answer to the  Have  exception, and it does not account for the fact that in verbs it is the 3 rd .sg. frozen agreement form rather than the bare stem which appears with overt subjects. In this paper, we derive the Complementarity Effect in borrowing what we see as the good elements of both the Incorporation and Agreement Analyses: the role of the distinction between affixal pronouns and DPs, and the universal presence of a φ -probe on T. However, we propose to derive the Complementarity Effect in the verbal system in a new way: it is a LOCALITY  effect that occurs when there is an INTERVENER  with interpretable φ -features between T and the subject. We propose that the 3.sg. frozen agreement of T as in (1)c in fact is φ -agreement, which takes place with a nominal clausal functional projection containing the subject. Idealizing slightly for the moment by restricting ourselves to transitives and the base position of the 4  The Theta Criterion is not the only principle potentially responsible for complementarity. Anderson's (1982) version of the Incorporation Analysis appeals to the binding theory; cf. Hendrick (1988: 97ff.). Classical accounts of the complementarity of clitics and DPs in non-doubling languages like French rely on Case.    4 subject, frozen agreement is agreement with the v P itself, which in Breton unlike in English or French has interpretable 3.sg. φ -features. Since the DP subject is contained within the v P, its φ -features cannot be reached by T's φ -probe across the intervening φ -features of v P, as in Figure 1. The 3.sg. agreement with the v P constitutes frozen agreement. By contrast, weak pronominal subjects undergo independent syntactic incorporation (cliticization) into T 0 , as in a syntactic version of the Incorporation Analysis. By this, they escape the v P and become the closest goal for the φ -probe of T, as shown in Figure 2. Figure 1: Agreement with an overt DP subject TP 3 T' 3 T 0 [ φ  -]   ↔   v P[ φ =3.SG] 6 …DP subject… Figure 2: Agreement with a clitic/affixal pronoun subject TP 3 T' rp  T 0   v P[ φ =3.SG] 2 6 T 0 [ φ  -] ↔  pronoun 0i  … t  i … This approach to the Complementarity Effect is novel in three ways. First, it claims that there is actual φ -agreement in frozen agreement contexts. We will demonstrate that this is true only for the T-subject system; the P-object system shows all evidence of instantiating a post-syntactic (prosodic) version of the Incorporation Analysis. Second, the controller of the 3.sg. frozen agreement is a nominal functional projection in the complement of T which contains the subject. This implies that the Breton extended VP has a nominal character which is absent in languages like English and French, whose subjects uniformly agree with T. We present evidence for this from the behavior of Breton VPs for Case assignment and Case requirements in section 2. The interpretable φ -features of this nominal clausal projection have the same "default" characters as those of nominalizations, gerunds, clauses, and small clauses in English. Third, the core of our analysis, the distinction between rich and frozen agreement, follows directly from feature-relativized locality applied to φ -features (Chomsky 1995, 2000, cf. Rizzi 1990). Building on the nominal character of the Breton extended VP, we show in section 3 how the logic of locality predicts frozen agreement, and how syntactic incorporation of weak/clitic pronoun subjects evades it. Section 4 then demonstrates that this accounts for φ -agreement asymmetries between the T-subject and P-object agreement system, which provide crucial evidence for the presence of a φ -probe on T. Our analysis predicts that if there were a Breton construction where the nominal functional projection to which we attribute frozen agreement did not intervene between T and the subject, φ -agreement should uniformly be with the subject. In section 5, we show that this is the case for    5 the verb  Have . Following prepositional analyses like Freeze (1992) and Kayne (1993), we show that  Have  in Breton is based on the preposition eus  'from' used as a clausal head, which has the subject of  Have  in its specifier. Since Breton PPs are in no sense nominal and lack interpretable φ -features, the PP projection of [ P   eus ] does not present a goal for the φ -probe of T, which can Agree with the subject in [Spec, PP]: (6) T [ φ  -]  [ PP  subject [ P'   eus P  [ v P  be …]]] It is a significant result of our approach that we predict that precisely  Have  should be an exception for the Complementarity Effect, because of its different clausal architecture. At the heart of our story is the idea that the Complementarity Effect in the verbal system arises from feature-relativized locality, due to the intervention of a nominal projection between the φ -probe of T and the subject. This hypothesis derives the Complementarity Effect from the logic of locality and the independent φ -feature specification of functional categories, without parametrizing or modifying the core Agree relation itself. We discuss the cross-linguistic implications in section 6. 2. The nominal character of the Breton extended VP In this section, we will support the proposal that the Breton extended VP has interpretable φ -features from its nominal behavior. Like expletives, nominalizations, and (small) clauses, the φ -specification is [3.sg.]. 5  The hypothesis that verbal functional projections are nominal receives support from independent data in Breton. Jouitteau (forthcoming a, 2005) discusses the nominal properties of v  and notes that the way the Breton v P licenses its internal argument is systematically parallel with the Breton DP. First, whenever case is morphologically realized on an object, it is uniformly genitive. The data in (7) below illustrates that objects proclitics show the same morphology as possessive pronouns. (7) a. o  distruj   Noun head and its pronoun theme their destruction 'their destruction' b. o  debriñ Infinitival verb and its pronoun theme their eat-INF 'to eat them' c. Alies o  debr Yann. Tensed verb and its pronoun theme often their eat-3.SG Yann 'Yann eats them often.' As in the DP system, direct case is not available for post-verbal object pronouns. Post-verbal free-standing pronoun objects are illicit: (8) a. *distruj int Noun head and its pronoun theme destruction they 'their destruction' b. *debriñ int Infinitival verb and its pronoun theme eat-INF they 'to eat them' 5  'Interpretable' in this sense simply means valued in the lexicon; the same lexically specified but semantically inert φ  -features must be posited for example for the plural of trousers  or the neuter gender of Czech dev č  e  'girl' (Rezac 2004a: 28). By contrast, the φ  -probe of the target of Agree like T has an unvalued φ  -set (Chomsky 2000).
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x