boeckx 2000 quirky agr

Upload: ka6ka

Post on 05-Apr-2018

227 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/31/2019 Boeckx 2000 Quirky Agr

    1/27

    Quirky Agreement*

    Cedric Boeckx

    Abstract. Verb agreement with a nominative element is severely constrained inIcelandic when a Quirky ``subject'' is present: only partial (number, not person)agreement obtains. This paper tries to account for this restriction in terms ofBonet's 1994 Person-Case-Constraint, which blocks `object' person agreementwhen agreement with a dative element takes place. I put forward the idea thatagreement obtains with Quirky subjects, but fails to show up morphologicallyon the verb for non-syntactic reasons. The analysis is extended to otherlanguages, which allows me to address the issue of the nature of Quirky Case,and of inherent Case more generally.

    1. Introduction

    The phenomenon of Quirky Case in Icelandic has received a great deal of

    attention since Zaenen, Maling & Thrainsson's 1985 detailed study of it.

    Regardless of whether Case is assigned (GB theory) or checked (Minim-

    alist Program), Quirky Case sits square on conceptions of Case Theory in

    the Principles-and-Parameters framework. A quick perusal of the liter-

    ature reveals various stages in the investigation of Quirky Case. There hasfirst been the question of how Quirky Case (usually, dative) is assigned,

    and why Quirky Case-marked elements surface in subject position despite

    their having no nominative Case features usually associated with `sub-

    jecthood.'1 Next, there has been the fact that Quirky elements, which

    behave like subjects, do not induce agreement on the verb, an otherwise

    prototypical feature of subjects. More recently, the fact that elements

    bearing nominative Case behave like objects2 when a Quirky element is

    present, and yet trigger agreement on the verb has been discussed. Finally,

    the fact that this agreement is partial and optional, and even barred insome cases, has received some attention.

    Studia Linguistica 54(3) 2000, pp. 354380. # The Editorial Board of Studia Linguistica 2000.Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK, and350 Main Street, Malden, MA 02148, USA

    * I am most indebted to Noam Chomsky for his comments, patience, and constantencouragement. Thanks are also due to Z eljko Boskovic, Howard Lasnik, Juan Uriagereka,Halldo r A rmann Sigursson, Christer Platzack, Anders Holmberg, Adolfo Ausn, ananonymous Studia Linguistica reviewer, and the audience of the 14th Comparative GermanicSyntax Conference (Lund, January 1999) for important comments and suggestions. Apreliminary report of this research appears in Boeckx 1998b.

    1 See Harley (1995b: chap. 1) for a useful and critical discussion of phenomena associated

    with subjecthood.2 The term `object behavior' draws on the results of such tests as those applied by Harley1995a,b to determine whether the nominative element hestar in (i) is Quirky or not, and ifnot, where it checks its features.

    i. Konungi hafa veri gefnir hestarkingdat have been given horsesnom`A king has been given horses'

  • 7/31/2019 Boeckx 2000 Quirky Agr

    2/27

    I will not dwell upon why the phenomena associated with Quirky Case

    are important for the Principles-and-Parameters approach to language.

    Nor will I attempt to review the vast amount of analyses put forward to

    deal with the facts just mentioned. Rather, I will focus on one particular

    quirk: the agreement pattern, especially that found with raising predicates

    (seem-constructions) in Icelandic.

    The present paper is organized as follows: section 2 provides the

    necessary background for the analysis to come; section 3 develops a

    new view on how agreement functions in Icelandic; section 4 applies the

    proposal to raising predicates; section 5 extends it to other languages, and

    tries to uncover the parameter underlying the availability of Quirky Case,

    and in so doing gain some insight into the nature of inherent Case.

    Section 6 is a summary.

    2. Some Quirky facts

    This section provides some necessary background on Icelandic Quirky

    subjects. I give a simple overview of the constructions in question,

    highlighting the properties of Quirky elements that make them ``subjects.''

    Quirky subjects behave like ordinary nominative subjects (and not like

    `topicalized' objects) with respect to numerous syntactic phenomena,

    including such familiar subjecthood-tests as subject-oriented reflexiviza-tion, subject-verb inversion, ECM, raising, subject control, and con-

    junction reduction (see Zaenen, Maling & Thrainsson 1985, henceforth

    ZMT, for fuller discussion).3 Compare the behavior of `structural' and

    `Quirky' subjects in (1) and (2), respectively. (Examples from Sigursson

    1992:5f.)

    (1) a. Hu ni sa myndina snai Reflexivization

    shenom saw picture selfH s4

    `She saw her own picture'

    Quirky Agreement 355

    # The Editorial Board of Studia Linguistica 2000.

    The conclusion Harley arrives at is that nominative Case is checked in AGRo, theprototypical position of `objects'. As Harley shows (see also Jo nsson 1996), there aregood reasons to believe the nominative element is an object of some sort: unlike `subjects', itdoes not check the EPP (the Quirky element does), it fails to pass the subjecthood-testsdevised by Zaenen, Maling & Thrainsson 1985 (see our section 2), it undergoes Object Shift,behaves like objects with respect to Negative Polarity items, ECM-constructions, passiviza-tion, wh-extraction, and control (see Harley 1995a,b for data and discussion).

    3 It is generally agreed that, though poorly understood, those phenomena are to be

    associated with subjects. Some of them, like ECM, seem robust (see Boskovic

    1997 andMartin 1996, among many others, for relevant discussion).

    4 Compare:

    i. Henni fannst kennarinn sinn/*hennar leiinlegurHerdat found teachernom her[+refl]/[refl] boring`She found her teacher boring'

    Yip, Maling & Jackendoff (1987:224,(5a))

  • 7/31/2019 Boeckx 2000 Quirky Agr

    3/27

    b. Hefur hu n se myndina? ``Inversion''

    Has shenom seen picture

    `Has she seen the picture?'

    c. E g tel [hana hafa se myndina] ECM

    I believe heracc have seen picture

    `A believe her to have seen the picture'

    d. Hu ni virist [ti hafa se myndina] Raising

    Shenom seems have seen picture

    `She seems to have seen the picture'

    e. Hu n vonast til [a PRO sja myndina] Control

    She hopes for to PROnom see picture5

    `She hopes to see the picture'

    f. Hu n horfi og (hu n) sa myndina Conj. reduc.

    Shenom looked and (shenom) saw picture

    `She looked and saw the picture'

    (2) a. Hennii leiist bo kin sniherdat bores book self's

    `She finds her own book boring'

    b. Hefur henni leist bo kin?

    Has herdat bored book

    `Has she found the book boring'

    c. E g tel [henni hafa leist bo kin]

    I believe herdat have bored book

    `I believe she found the book boring'

    d. Hennii virist [ti hafa leist bo kin]

    Herdat seems have bored book

    `She seems to have found the book boring'

    e. Hu n vonast til [a PRO leiast ekki bo kin]

    She hopes for to PROdat bore not book

    `She hopes not to find the book boring'

    f. Hu n var syfju og (henni) leiddist bo kin

    She was sleepy and (herdat) found the book boring

    `She was sleepy and found the book boring'

    With this much background, we can turn to subject-verb-agreement

    patterns in `Quirky' constructions.

    Agreement in Icelandic finite clauses involving nominative subjects

    proceeds as in many other languages: agreement obtains between the verb

    and the subject, usually in the highest spec (ignoring topicalization, post-

    copular subjects, etc.).

    356 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    5 See Sigursson 1991 for convincing arguments that PRO bears Case in Icelandic (here,nominative Case).

  • 7/31/2019 Boeckx 2000 Quirky Agr

    4/27

    (3) Vi kusum stelpuna

    WenomX1Xpl elected1Xpl girlacc`We elected the girl'

    (4) r virast tHH hafa veri tH kosnar t

    TheynomX3Xpl seem3Xpl have been electednomX3Xpl`They seem to have been elected'

    Sigursson (1992:2, (1), (4))

    However, once we turn to Quirky subjects, facts get murky. If there is

    only one argument the Quirky Case-marked subject agreement does

    not obtain between the verb and the subject. The verb takes what has

    come to be called ``default agreement'' (a form homophonous with the

    third person singular neuter).

    (5) Stelpunum var hjalpa

    The girlsdatXplXfem was3Xsg helpedneuterXsgThe girls were helped'

    (6) eim virist tHH hafa veri tH hjalpa t

    ThemdataX3Xpl seems3Xsg have been helpednXsg`They seem to have been helped'

    Sigursson (1992:3, (5b), (8))

    If there is another argument an ``object'', agreement depends on the

    Case of the latter; if the Case of that argument is dative, genitive or

    accusative, no agreement obtains ((7) (9)).6 If the Case is nominative, the

    verb agrees with the nominative NP, not with the Quirky subject (10).

    (7) Mig irar ess

    Meacc repentsdefault thisgen`I repent this' Yip et al. (1987:230, (9e))

    (8) Henni var skila peningunum

    Herdat wasdefault returned moneydat`She was given the money back'

    (9) Mig vantar peninga

    Meacc lacksdefault moneyacc`They seem to have been helped'

    ZMT (1985:454, (29a); 459, (42b))

    (10) Henni voru gefnar bkurnar

    Herdat were3Xpl given booksnomXpl

    `She was given the books' (Sigursson (1992:5, (5f))

    Quirky Agreement 357

    # The Editorial Board of Studia Linguistica 2000.

    6 These examples do not reflect all the possible Case combinations in Icelandic. See Yip,Maling & Jackendoff 1987 for such a survey.

    Note also that the specific Case of the Quirky subject does not matter. For all ends andpurposes, we can equate `Quirky' with `non-nominative'. See below for discussion.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    5/27

    However, this is not the whole story. As the careful studies of Sigursson

    1996, and Taraldsen 1995, have revealed, agreement with nominative

    objects is only partial. If the object is first or second person, only number,

    not person, agreement obtains (11); if it is third person, `full' (i.e. person

    and number) agreement obtains (12).7

    (11) *?Henni leiddust vi

    HerdatX3Xag bored3Xpl wenomX1Xpl`She was bored with us' Taraldsen (1995:309, (7))

    (12) Henni leiddust/*?leiddist eir

    HerdatX3Xsg bored3Xpla3Xsg theynomX3Xpl`She was bored with them' Taraldsen (1995:307, (1))

    Those facts are puzzling, and have given rise to an explosion of possible

    analyses, which I won't attempt to review here. Suffice it to say that

    two main approaches can be envisaged: one is to rely on Agreement

    projections (for example, on split AGRPs like AGRnumber and AGR-

    person) (see Taraldsen 1995 for an approach along these lines), the

    other is to adopt Chomsky's 1995 Attract-F hypothesis and stipulate

    that only certain features can raise to check agreement without triggering

    pied-piping.

    3. The limits of agreement in Icelandic

    In the present paper, I will focus on agreement patterns with raising

    predicates like seem. I believe that these constructions can give us a

    particularly useful and revealing angle to tackle the problem of Icelandic

    finite agreement, for they involve a Quirky Case-marked experiencer

    (the analog of the to-phrase in it/John seems to me . . . )8 and an

    embedded clause containing a subject, and possibly an object. The

    embedded subject can bear either structural or Quirky Case. Thus, such

    constructions offer a full range of possibilities, and can display matrix vs.

    embedded agreement asymmetries, depending on whether the embedded

    subject (features) raise(s), and on which Case it (they) `check(s)'. Put

    differently, raising constructions add the further complexity of having

    two Quirky arguments, which I will try to show is just the level of

    complexity needed to understand what determines agreement in simple

    root contexts.

    358 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    7

    Here I limit attention to agreement with the finite verb, not with past participles, forinstance. Also, note that the term ``full'' agreement may not be adequate, for it might be thatonly number agreement is involved in Quirky Case constructions, and that the third personis copied from default agreement. See Taraldsen (1995:312) for such a view.

    8 As we will see below, the to-phrase is optional in Icelandic (as in English); we will,however, focus on sentences containing it so as to make the Case-array as complete aspossible within one sentence.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    6/27

    3.1. Basic facts

    Consider the agreement patterns in the following examples, exhibiting the

    full range of argument structures.. One Quirky element, agreement on matrix verb with embedded nomi-

    native element:

    (13) Mer ?virist/virast hestarnir hafa veri gefnir konunginum

    Medat seem3XsgaplX horsesnomXpl have been given kingdat`It seems to me that horses have been given to the king'

    (14) Honum eru taldir hafa veri gefnir peningarnir

    Himdat arepl thought have been givennomXmascXpl moneynomXmascXpl`The money is thought to have been given to him'

    (15) Jo ni ?* virist/virast vera *tali/taldir lkaJohndat seemsgapl be believedneuterXsganomXmascXpl like

    hestarnir

    horsesnomXmascXpl`John seems to be believed to like horses'

    (16) Mer *virist/virast eir vera skemmtilegir

    Medat seem3Xsga3Xpl theynom be interesting

    `It seems to me that they are interesting'

    Schu tze (1997:sec.4.1, passim)

    . Absence of agreement on the matrix verb with embedded nominative

    element, presence of an embedded Quirky subject:

    (17) Mer fannst/*fundust henni leiast eir

    Medat seemed3Xsga3Xpl herdat bore theynom`I thought she was bored with them'

    (18) Mer hefur/*hafa alltaf virst honum lka bkur

    Medat has/have often seemed himdat like booksnomXpl`It has often seemed to me that he likes books'

    (19) Mer virist/?*virast Jo ni lka hestarnir

    Medat seemsgapl Johndat like horsesnomXpl`It seems to me that John like horses'

    (20) Mer virist/*?virast straknum lka essir blar

    Medat seem3Xsga3Xpl boydat like these carsnomXpl`It seems to me that the boy likes these cars'

    (21) Mer hefur/?*hafa alltaf virst honum hafa veri seldar/*selt

    Medat has/have often seemed himdat have been soldplasg

    essar bkur a alltof har veri

    these booksnomXpl at far-too high price

    `It has often seemed to me that he has been sold these books at far

    too high a price' Schu tze (1997:sec.4.1, passim)

    Quirky Agreement 359

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    7/27

    . Default agreement:

    (22) Mer o tti Mara vera gafu

    Medat seem3Xsg Marynom be giftednomXfemXsg

    `I thought Mary was smart' Schu tze (1997:sec.4.1, passim)

    . Failure of all combinations:

    (23) Henni *leiddumst/?*leiddust/?*leiddist vi

    Herdat bored1Xpla3Xpladefault wenom`She was bored with us' Sigursson (1996:28, (56))

    At first sight, these patterns seem to resist any unified account, even if one

    restricts attention to a small class of predicates. Sigursson (1996:29) has

    noted that agreement with raising predicates (his `Dat and Nom infinitive'

    verbs, or `D/NcIs') is ``one notch weaker'' than in the monoclausal Dat-

    nom constructions, where (at least) number agreement is the preferred

    strategy (`default' agreement is marginal, and only found with a few

    verbs, see Sigursson (1996:16ff.)). With raising predicates, default

    agreement seems to be the preferred option.

    Schu tze 1997 has noted another generalization. There seems to be a

    general `adjacency' requirement governing agreement in raising contexts:

    a verb can only agree with a nominative `object' if the latter immediately

    follows the matrix verb. As Schu tze remarks (p. 109), this seems to be a

    minimality effect: the closest lower DP is the only one that can be

    attracted into the matrix clause; if that closer element is Quirky Case-

    marked, the matrix verb takes default agreement. In other words, the

    relevant 0-features of the embedded nominative object cannot be attractedacross an intervening Quirky argument (which can only be the one con-

    tained in the embedded clause, the matrix experiencer (the ``to-phrase'')

    always raises, and so never blocks feature-movement,9 see below).10

    Though fairly straightforward, this minimality effect turns out to be

    difficult to implement within the minimalist program.11 It is indeed

    360 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    9 I continue to use the term `movement', though Chomsky 1998 has argued that no actualdisplacement, but mere agreement, is what is involved in the syntax of features. `Feature-Communication' (a term due to David Pesetsky) might be the most appropriate term. SeeBoeckx 1999a for extensive discussion, and ways to tease apart Agree from Attract.

    10 On the invisibility of A-traces, see Chomsky 1998; for a possible explanation, seeLasnik (to appear).

    11 There is a way of capturing the minimality effects in an Attract-F framework, as arguedby Ochi, to appear and Boeckx 1999c, viz. by capitalizing on the Two-Chain hypothesis(Chomsky (1995:265)) according to which overt movement consists in the creation of two

    chains, the Feature-chain and the Pied-piping chain, with only the latter subject to all thelocality conditions on movement. However, this possibility is of no avail in the case underdiscussion since all that moves are features, hence only Feature chains come into play.

    Chomsky 1998, based in part on the present study, distinguishes `feature' from `feature-value', and claims that only the former is relevant to capture Relativized Minimality effects.The present proposal, however, makes it possible to just talk about feature-value, given thatunlike Chomsky, I eventually argue against a structural/inherent Case distinction.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    8/27

    assumed (Chomsky 1995) that potential attractees compete not because

    of their categorial features, but because of some subcategorial feature

    they share (this is the basic insight of Relativized Minimality rephrased in

    Attract-F terms). As such, we would not expect a Quirky DP to intervene

    in the examples under discussion since there is good reason to believe that

    its 0-features are inert for agreement. Given that, it seems that the``minimality effect'' will have to be recaptured in somewhat different

    terms. At the moment, I am aware of no such analysis in the literature,

    which means that we will have to take a fresh start.12

    3.2. The proposal

    I will first lay out some basic properties of raising constructions in

    Icelandic:13

    As in English, the experiencer is optional in Icelandic raising predicates.

    (24) eim hafi virst O lafur vera gafaur

    Themdat has seemed Olafnom be intelligent

    `They regard Olaf as intelligent'

    (25) O lafur hafi virst vera gafaur

    Olafnom has seemed be intelligent

    `Olaf seemed to be intelligent'

    (modeled on Sigursson (1996:29, (57)))

    However, unlike in English, Icelandic experiencers block raising.

    (26) *O lafur hafi virst eim vera gafaur

    Olafnom has seemed themdat be intelligent

    `They regard Olaf as intelligent'

    This blocking effect is not peculiar to raising predicates: Quirky elements

    always block raising of nominative `objects' (or `sobjects' to use Sigurs-

    son's 1996 term) to the highest (`subject') position, irrespective of theagreement pattern.14

    Quirky Agreement 361

    # The Editorial Board of Studia Linguistica 2000.

    12 One possibility, suggested to me by Z eljko Boskovic (personal communication), as anextension of Boskovic (1998), is to say that in the following abstract configuration, theembedded Infl blocks Attraction from the matrix clause.

    i. Experiencer SEEM Quirky to Inf. Nominative| * |

    However, it is unclear to me why the embedded Infl would block attraction given that it hasa different feature specification from the matrix verb, except if we focus on features, notfeature-values (see previous note), and Boeckx 1999b for discussion within a larger context.

    13

    In addition to virast `seem', synast `appear', finnast `find, consider', bykja `find,consider', and heyrast `sound (as if)' are common raising predicates (Sigursson 1996:29). Tothis list, I add ``ECM''-verbs like telja `believe', which involve raising into the matrix clause(see Thrainsson 1979 and much subsequent work), though here the `experiencer' problem isabsent, but the `Quirky-interceptor' remains valid.

    14 I use complex tenses in the examples because in the simplex tense, (iib), correspondingto (28), would be grammatical with henni as a topic.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    9/27

    (27) Strakarnir hafa leist henni

    boysnom has bored3Xpl herdat`She found the boys boring'

    (28) *Strakarnir hafa henni leist

    (modeled on Taraldsen (1995:307))

    Note further that even if the embedded subject is Quirky, the experiencer

    blocks movement.

    (29) Mer virist Jo ni lika hestarnir

    Medat seemsdefault Johndat like horsesnom`It seems to me that John likes horses'

    (30) *Jo ni virist mer lka hestarnir

    In short, raising of a full category is allowed only when the experiencer is

    absent.

    3.3. Insular Scandinavian is not a secluded island

    The state of affairs reported in 3.2. is by no means restricted to Icelandic.

    In interesting studies of experiencers in Spanish, Torrego 1996, 1998

    shows that many Romance languages (in fact, all except French and

    Italian) prohibit raising if there is an overt experiencer.15

    (Illustrationfrom Spanish.)

    (31) Este taxistai parece [ti estar cansado]

    This taxi-driver seems be tired

    `This taxi-driver seems to be tired'

    (32) *Este taxistai nos parece [ti estar cansado]

    This taxi-driver to-us seems be tired

    `This taxi-driver seems to us to be tired'

    Torrego (1996:102, (2))This parallels the Icelandic facts noted above (26). It thus appears that

    English, French and Italian are the exceptions rather than the rule.16

    (33) This driver seems to be to be tired

    362 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    (i) Henni leiddust strakarnirHerdat bored3Xpl boysnom

    `She found the boys boring'(ii) Strakarnir leiddust henni(modeled on Taraldsen (1995:307))

    15 On Spanish experiencers, and possible connections with Icelandic Quirky subjects, seealso Fernandez-Soriano 1999. For some discussion of Germanic, see section 5.2. below.

    16 I will gloss over the differences among the three languages. See Boeckx 1998a, 1999c fordiscussion.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    10/27

    (34) Ce conducteur me semble etre fatigue

    this driver to-me seems be tired

    `This driver seems to be to be tired'

    (35) Questo autista mi sembra essere stanco

    this driver to me seems be tired

    `This driver seems to be to be tired'

    The Romance data highlight the fact that Icelandic may not be the

    marked/`Quirky' case, but rather the specific instantiation of universal

    processes.

    Further facts from Romance reinforce this position. Schu tze, for

    instance, notes (1997:109 fn.17) that the status of (36) becomes ``myster-

    ious'' once the auxiliary is dropped. Neither the agreeing nor the

    nonagreeing form seems good (37).

    (36) Mer hefur/??hafa alltaf virst honum hafa veri seldar/*selt

    Medat has/have often seemed himdat have been sold3Xpla3Xsgessar bkur a alltof har veri

    these booksnomXpl at far-too high a-price

    `It has often seemed to me that he was sold these books at far too

    high a price'

    (37) ???Mer virast/virist honum hafa veri seldar/selt essar

    Medat seemedsgapl himdat have been sold3Xpla3Xsg thesebkar a alltof har veri

    booksnomXpl at far-too high a-price

    `It has often seemed to me that he was sold these books at far too

    high a price'

    This parallels the French data given in (38)(39):

    (38) Marie a semble a Jean etre fatiguee

    Marie has seemed to Jean be tired

    `Marie seemed to Jean to be tired'(39) ??Marie semble a Jean etre fatiguee

    Marie seems to Jean be tired

    `Marie seems to Jean to be tired'

    For as yet unclear reasons, the French sentence is odd when the auxiliary

    is dropped (it is good if the experiencer is cliticized: Marie lui semble etre

    fatiguee `Marie to-him seems be tired', which is impossible in Icelandic).

    I will postpone an explanation for the blocking effect of experiencers

    until section 5, for I believe that a proper understanding of this phenom-

    enon requires some non-trivial departure from previous assumptions.

    Indeed, even if these facts suggest that Icelandic may not be that `quirky'

    after all (at least, not so `isolated' as is often claimed), we still have to

    explain why agreement, when it obtains, is only partial. Besides, we have

    to address the more general question of why Quirky subjects, which

    Quirky Agreement 363

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    11/27

    behave like true subjects in many respects (cf. section 2), do not trigger

    agreement. The puzzle may be derived from George/Kornfilt's 1981

    theory (or, rather, Chomsky's 1998 interpretation of it) that agreement

    and Case are but one phenomenon (one `feature', see also Raposo &

    Uriagereka 1996, and Schu tze 1997), and that agreement on T is only

    possible with nominative.17 But this leaves us with the question of why

    agreement is not `total' when it obtains with some nominative elements

    (`sobjects'). I will show that crosslinguistic considerations again reveal

    that this is by no means a language-specific `inconsistency', but rather

    that it may be part of a universal constraint.

    3.4. Explaining the agreement constraints: Agreement does obtain with

    Quirky subjects

    3.4.1. The Person-Case constraint. The universal constraint I have in

    mind is the one discussed in Bonet 1994 under the name ``Person-Case

    Constraint'' (PCC). The latter basically states that human language

    disallows the presence of first or second person agreement with a direct

    object when there is also dative agreement.

    (40) The Person-Case Constraint (PCC)

    If Dative (agreement) Accusative (agreement) = 3rd

    Bonet (1994:36)

    Here I illustrate the PCC with French data, but Bonet shows that it is

    valid in many, typologically unrelated languages, which leads her to claim

    that the constraint is universal.

    (41) *Jean me lui a recommande

    Jean me him has recommended

    `Jean recommended him to me'

    (compare: Jean le lui a recommandeJean it him has recommended

    `Jean recommended it to him')

    Bonet was not the first one to observe this constraint (see, e.g., Kayne

    1975 and Perlmutter 1971), but the constraint is still poorly understood.

    Here, I will follow Bonet in giving it a morphophonological treatment,

    assuming a Distributed Morphology framework (Halle & Marantz 1993),

    according to which (some/most) morphological operations take place

    after syntax, in Morphology, part of the PF component.18

    This will allow

    364 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    17 Pesetsky & Torrego (to appear) suggest that one might view Nominative as unin-terpretable T-feature on D/NPs.

    18 Ormazabal to appear, and Ormazabal & Romero 1998 argue that the PCC is bestanalyzed as a syntactic constraint. To the extent that the present analysis is correct, it arguesagainst such a view. It is difficult to determine the different empirical predictions the two

  • 7/31/2019 Boeckx 2000 Quirky Agr

    12/27

    the constraint to be ``inoperative'' in (narrow) syntax; its effects will be

    manifest only in overt morphology. Put differently, nothing syntactically

    bars a configuration flouting the PCC (i.e., a configuration in which both

    the dative and the accusative elements fully agree with the verb); if syntax

    yields such an output, Morphology will use repair strategies (on which

    see Bonet 1994) that will mask the PCC-violation, preventing the deriva-

    tion from crashing at PF. For our present purposes, it may mean that

    some agreement process takes place in syntax but fails to map onto

    Morphology, and so doesn't surface at PF because it would flout the

    PCC.

    I will argue that this is indeed the case in Icelandic (and elsewhere, see

    section 5); that is, I will put forward the view that agreement between the

    Quirky element and the verb actually obtains (an instance of AGRs-checking)19 thus making Quirky subjects ``full-fledged subjects'',

    but that this agreement cannot be morphologically realized for independ-

    ent Economy reasons. By contrast, ``AGRo-checking'' can surface on

    the verb iff a nominative element is involved, because there is an intricate

    link between agreement on T and nominative (cf. George-Kornfilt/

    Chomsky's view). The morphological shape of this agreement relation

    is constrained by the PCC; let us now see how and why the latter holds in

    Icelandic.

    3.4.2. Partial agreement. We have already noted that agreement with

    nominative objects is only partial; the verb takes third person agreement,

    only number agreement varies (being available in some, well-defined

    cases). This is similar to the PCC-result, which states that if agreement

    obtains with a dative element (in the case of Icelandic, a Quirky

    element);20 accusative agreement (here referred to as AGRo) must be

    third person. True, there is no `surface' manifestation of dative agreement

    on the verb, but I would like to claim that there is on the dative elementitself.

    Quirky Agreement 365

    # The Editorial Board of Studia Linguistica 2000.

    views make, for the very place of morphology is not made clear in Ormazabal & Romero'swork. Conceptually, it is, I think, desirable to restrict the scope of the PCC to the PF-component, that way, we avoid look-head (why would syntax care about the output ofmorphology?).

    19 I use `AGRs' and `AGRo' as mnemonics for subject and object agreement, respectively.As a matter of fact, the present study could be viewed as an argument of doing away with

    AGR-projections, given that it captures agreement facts in a language (Icelandic), where ithas been argued AGRs and T should be split (see Bobaljik & Jonas 1996, Bobaljik &Thrainsson 1998).

    20 Here I use `dative' as an umbrella term for `Quirky'. As can be seen in such detailedstudies as Yip, Maling & Jackendoff 1987, dative is indeed the prototypical instantiation ofQuirky Case. See Smith 1994, 1996 for an interesting conjecture about this ``DativeSickness'' phenomenon.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    13/27

    Delsing 1996 has expressed the view that Case has lost most if not all of

    its function in modern Icelandic, and can now be seen as a `decorative'

    device. This corroborates Sigursson's (1996:28) intuition that nomina-

    tive is becoming Quirky like all other Cases in Icelandic. The decoration

    Delsing mentions might be reinterpreted as `sufficient' for agreement

    purposes. That is, Economy considerations might dictate the non-overt

    manifestation of agreement on the verb, because the information is

    already visible on the noun (it seems a general fact about human

    languages that the checked feature is realized on only one member of

    the checking relation), a sort of ``Generalized Doubly Filled Comp

    Filter.''21 More subtly, Dative agreement is also manifested in the

    obligatory third person morphology on the verb.

    Our claim is then that agreement takes place with Quirky subjects, but

    the result is almost `null' for the verb's morphological make-up, it only

    forces third person agreement, leaving morphological room for object

    number agreement with objects (if any). If no syntactic clash (Case-

    mismatch) arises, number agreement obtains with the nominative object

    (George/Kornfilt theory).

    3.4.3. Beyond the PCC: Point of View. Bonet develops the PCC in

    Optimality terms, and goes on to show what its effects are, how they

    can be circumvented, etc., but does not actually try to explain why there

    should be such a thing like the PCC in UG. Here, I can only be

    speculative. A large survey on dative Case assignment/checking (such as

    van Belle & van Langendock 1996) reveals that the interpretive `role'

    associated with the dative (understood here as Quirky, cf. fn. 20) is typical

    of [+human/+animate], which is best characterized in terms of [+person],

    rather than gender or number. [+Person] is intrinsic to humans, which

    might lead to a `correspondence rule' of the type: dative D person. AsSigursson 1996 and many others have claimed, person 1 o r 2

    person, third person is the absence of the [+person] feature; so dativedoes not block the latter, it only blocks 1st or 2nd person, i.e. [+person]

    features, which might be said to encode Point-of-View ([+/ speaker],[+/ addressee]).

    Point-of-view has recently received some attention in the generative

    literature, and is now assumed to be represented structurally via some

    projection (or several such projections, see Rizzi 1997), be it FP (Uria-

    gereka 1995a,b), PolP (Culicover 1992), or P (Laka 1990), dominating

    IP. Uriagereka 1995a,b, especially, has tied clitic distribution, Point-of-

    View, and IP. Further, Chomsky 1998 argues that T's EPP-feature shouldbe kept distinct from its Case/0-feature, and, in the case of what he calls

    366 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    21 This may be the equivalent of `Avoid structure', `avoid morphology' ( . . . up to crash)(i.e., avoid superflous symbols, Minimize a (as much as possible)), familiar from theliterature.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    14/27

    defective T (where the EPP is checked by an element distinct from the

    one that checks Case/Agreement), the EPP might be viewed as an

    impoverished [Person]-feature (which he argues, expletives can check).

    I suggest we relate this EPP/[Person]-feature to the dative/Quirky feature,

    both of which encode Point-of-View. This will favor raising of the Dative/

    Quirky element over the nominative `(s)object' for obvious feature-

    matching reasons.

    3.4.4. Explaining Icelandic agreement. Given the PCC, we can now

    explain why person agreement is unavailable for nominative objects.

    The formulation is misleading; it should be something like: AGRs

    features are checked by Quirky subjects (whenever Quirky Case is

    involved), AGRo features by objects. Some universal morphological

    process says: whenever AGR features are checked by nominative,

    they can be phonologically realized (George/Kornfilt's theory).

    This coupled with the PCC says that AGRo features checked by a

    nominative element can surface but, now that some dative element is

    involved, person feature realization is impossible (restricted to third

    person). Only number features can surface, which is what we get in

    Icelandic.

    It is interesting to note that no matter which agreement pattern is

    chosen, sentences involving a Quirky 3

    rd

    person element and a 1

    st

    or 2

    nd

    person nominative `object' are hopelessly bad.

    (42) *Henni leiddumst/leiddust/leiddist vi

    Herdat bored1Xpla3Xpladefault usnom`She was bored with us'

    Sigursson (1996:28)

    (43) *Henni voru syndir/syndar i

    Herdat were shownmascafem younom`You were shown to her'

    (44) *Henni vorum syndir/syndar vi

    Herdat were shownmascafem usnom`We were shown to her'

    Sigursson (1996:32)

    I take this as a confirmation of the view just expressed: such sentences

    yield a ``Point-of-View'' clash. 1st and 2nd person features (borne by a

    nominative element) express point-of-view, which is also associated with

    dative (`Quirky') elements; if a dative nominal prevents raising of a 1st or

    2nd person nominative element to the projection encoding Point-of-View,

    the sentence crashes.

    In those cases, speakers will resort to another configuration allowing

    the 1st or 2nd person bearing element to express point-of-view and

    agreement syncretically (45)(46) (compare (43)(44)).

    Quirky Agreement 367

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    15/27

    (45) i voru syndir/syndar henni

    younom were shownmascafem herdat`You were shown to her'

    (46) Vi vorum syndir/syndar henni

    Sigursson (1996:32)

    3.4.5. A binding extension. Arguably, the analysis allows us to explain

    why there is no nominative reflexive in Icelandic (Lyle 1997:322).22

    (47) *Hennii finnst REFLi veik

    Her finds REFL sick

    `Shei considers herselfi sick

    (48) Icelandic third person reflexivesNom

    Acc sig

    Dat ser

    Gen sin

    True, this might have to do with independent properties of psych-

    predicates and the binding configurations they allow,23 but, I would

    like to claim, it is also the result of the PCC. Reflexives of the sig-type

    have no inherent number features; we saw that nominative checkingis linked with overt realization of 0-features (inasmuch as they areallowed, given the PCC). The nominative reflexive has to check its

    features (cf. Lebeaux 1983 and much subsequent work for the claim

    that reflexives raise at LF). Since it only has features which are blocked by

    the PCC (person features), the Morphological Component will not know

    how to realize number features (the number features are realized on the

    Quirky element, not on the verb, see above), and the sentence will crash at

    PF. Given that, language resorts to repair strategies and uses anaphoric

    pronouns instead of anaphoric reflexives.24

    (49) Henni finnst hu n veik

    Herdat finds shenom sick

    `She finds herself sick'

    368 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    22 I concur with Lyle that this should receive a deeper explanation than an appeal to somemorphological gap (Maling 1984).

    23

    See Tagalog, though.

    i. Gusto ng kanyang sarilii si Amadoilikes A his self T Amado`Himselfi likes Amadoi' Harley (1995b:205, (39))

    24 This is just tentative, the issue of binding in Icelandic is too complex for me to tacklehere.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    16/27

    It thus seems that our theory on Quirky Case can account for why

    agreement is only partial with nominative `sobjects',25 and for why

    agreement seems to be absent with otherwise `well-behaved' subjects:

    Quirky elements. The basic idea is that syntactic agreement with Quirky

    subjects is morphologically manifested by the anti-person-agreement with

    nominative objects.

    4. Agreement in raising predicates

    Having grasped how agreement works in Icelandic, we can now recon-

    sider the data in (13)(23). Two generalizations have to be account for:

    (a) agreement in raising contexts is ``one notch weaker'' than in mono-

    clausal environments (Sigursson 1996:29), and (b) the minimality effectnoted in Schu tze 1997 that agreement between the matrix verb and the

    embedded nominative element is blocked if there is an embedded Quirky

    subject intervening.

    Quirky Agreement 369

    # The Editorial Board of Studia Linguistica 2000.

    25 Partial agreement in Icelandic brings to mind the agreement patterns in Romance se-constructions discussed in Raposo & Uriagereka 1996. The basic pattern is as follows.

    i. Ontem compraram-se demasiadas salsichas no talho SanzotYesterday bought

    3Xpl-SE too-many sausages at butcher S.

    `Yesterday too many sausages were bought at the butchershop'ii. Essas salsichas compraram-se ontem no talho Sanzot

    Those sausages bought3Xpl-SE yesterday at butcher S.iii. Compra-se sempre demasiadas salsichas no talho Sanzot

    Buy3Xsg-SE always too-many sausages at butcher S.`Too many sausages are always bought at the butchershop S.'

    Raposo & Uriagereka (1996:750, (1)(2))

    In (iii), agreement looks like default. Interesting, Raposo & Uriagereka adhere to theGeorge/Kornfilt hypothesis, and show that due to its morphological defectiveness, se bears aspecial Case (Null Case) and cannot check person agreement, as a result, these are free to berealized with nominative Case, which is borne by the object, which raises across se (as in (ii)),

    when the F feature is strong (F point-of-view, cf. Uriagereka 1995a,b). Thus, [+person]forces projection of a category on top of IP, another instance of our Person/Point-of-Viewcriterion.

    It seems that, like Icelandic, Romance objects can check nominative; the only differencebetween Icelandic and Romance is that person features are not frozen by se (which is notdative, and thus fails to activate the PCC), as a result they will be checked high in the tree(for point of view reasons), the object will have to raise. Objects in Icelandic do not raisebecause `high' positions related to [+person] features are frozen. Romance is thus the mirrorimage of Icelandic, if the present analysis is correct.

    Note also that we achieve a parallelism between Null Case and Quirky Case; both casesare special with respect to agreement; interestingly, both are checked by PRO (Martin 1996)in non-finite clauses, where agreement is not involved. Again, a case of perfection in

    language it seems to me.Other cases of partial agreement come to mind, such as that found in classical Arabic (see

    Aoun, Benmamoun & Sportiche 1994 for detailed discussion), where the subject and the verbagree in `full' if the subject has raised; if the subject appears on the right, person agreement isnot morphologically realized, again as expected under our approach (note, incidentally, thatBenmamoun 1997 has argued that partial agreement in classical Arabic was the result of aPF process).

  • 7/31/2019 Boeckx 2000 Quirky Agr

    17/27

    It will be remembered that the minimality effect could not be captured

    in an Attract-F framework given that the 0-features of Quirky elementswere considered inert for agreement, hence should not prevent attraction.

    But now that we have seen that the 0-features of Quirky subjects areactive in the syntax, the minimality effect follows at once. The 0-featuresthat the attractor (the matrix verb) will encounter are those of the

    embedded Quirky subject. Since attraction is ``myopic'' (Chomsky

    1995, Collins 1997), it cannot go beyond and see the features of the

    nominative element. Given that, agreement will surface on the matrix

    verb as default (third person singular).

    (50) Mer fannst/*fundust henni leiast eir

    Medat seemed3Xsga3Xpl herdat be-bored theynomXpl

    `It seemed to me that she was bored with them'(51) Mer hefur/*hafa alltaf virst honum lka bkur

    Medat has/have often seemed himdat like booksnomXpl`It often seemed to me that he liked books'

    This maximally simple view also accounts for why agreement is ``one

    notch weaker'' in raising predicate contexts. In monoclausal environ-

    ments, there is no intervening Quirky element, and attraction can proceed

    unhindered.

    (52) Henni leiddust strakarnirHerdat bored3Xpl boysnomXpl`She was bored with the boys'

    Interestingly, agreement in the embedded clause of raising constructions

    takes place as in monoclausal contexts (as shown on the participle), there

    being no intervening Quirky element.

    (53) Mer virist/*virast Jo ni vera taldir lka

    Medat seems3Xsga3Xpl Johndat be believednomXmascXpl like

    hestarnir

    horsesnomXmascXpl`I perceive John to be believed to like horses'

    5. Agreement constraints elsewhere

    5.1. PCC effects in English: partial agreement in raising contexts26

    Howard Lasnik (personal communication) has pointed out to me that the

    present analysis might account for the following English agreement facts,

    370 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    26 The `partial agreement' data to be discussed are different from the facts studied inKayne 1995. Kayne assumes that (certain varieties of) English has (have) only numberagreement. The present analysis crucially relies on the availability of both number andperson agreement features.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    18/27

    which, to my knowledge, have gone unnoticed in the literature. (For

    further discussion, see Boeckx 1999a,c).

    (54) John seems to Mary to be the best

    (55) John and Bill/The men seem to Mary to be the best27

    (56) There seems to Mary to be a man in the room

    (57) *?There seem to Mary to be men in the room

    (58) There seems to Mary to be men in the room

    Whatever the [person] content of the nominative element, matrix verb

    agreement obtains when the element raises overtly ((54)(55)). However,

    (assuming a covert raising analysis of the ``associate'', a fact to which I

    return below), agreement is blocked when the nominative element raises

    covertly (57) (I will argue that the third person singular form in (56)should be conceived of as default), which suggests that dative agreement

    takes place in English, activating the PCC.28

    The data in (54)(58) can be accounted for as follows. In (54)(55), the

    nominative element raises overtly, it is the true (syntactic) subject of the

    sentence (it can, for instance, bind the experiencer: Johni seems to hisi boss

    to be lazy), it thus checks the agreement (AGRs) features on T, allowing

    them to be realized morphologically (George/Kornfilt's view). When the

    nominative element is the associate of an expletive, the PCC is operative

    (cf. (57)), but why should this be so? More precisely, given a checkingtheory where inflection is present from the start, why should there be an

    asymmetry between overt and covert operations?

    For the PCC to be operative, we have to say that dative agreement

    takes place, but that should block full agreement in all cases (the

    agreement features are already encoded on the verb). However, this is

    not quite correct. The PCC says that:

    (59) If dative agreement 3 accusative (`AGRo') = third person

    When the nominative element moves overtly, it is a true syntactic subject,and the PCC has nothing to say about nominative ``subjects''. When the

    nominative element moves covertly, it is an associate, not a true subject;

    rather, the expletive is, as can be seen from various tests, such as tags,

    where only syntactic subjects are repeated.29

    (60) There was a man in the room, wasn't there/*a man/*he?

    (61) John was in the room, wasn't he?

    Quirky Agreement 371

    # The Editorial Board of Studia Linguistica 2000.

    27 I use the men as an alternative to avoid problems inherent to coordinated structures.28

    Howard Lasnik (personal communication) notes that the intervention effect is strongerif the experiencer is indefinite. See Boeckx 1999a for an explanation of this fact in terms ofChomsky's 1998 deficient intervention constraint.

    29 As for the status of the expletive, I assume with Chomsky 1995 that it is merged directlyin IP, and not raised from a VP-internal small clause (as in the predicate-raising hypothesis,see Moro 1997 and much related work). See Boeckx 1999a for arguments against thepredicate-raising approach.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    19/27

    As is well-known, expletives have no (full) agreement features, so they do

    not play a role in agreement patterns (one might say they are neutral as to

    Point-of-View); the `associate' status of the nominative element, however,

    does.

    Lasnik 1992, 1995a has argued that the associate of the expletive checks

    a special Case (partitive Case, see Belletti 1988 for the origin of this idea,

    see also Rooryck 1997a for additional arguments in favor of Lasnik's

    approach), in a manner similar to objects, that is, in a specific functional

    projection parallel to AGRo. Capitalizing on this, I propose that only in

    those cases, agreement between the matrix verb and the nominative

    element is an instance of AGRo, and can (actually, is) disrupted by

    dative agreement, in accordance with the PCC.30

    We must now explain how dative agreement (agreement with the

    experiencer) takes place in English. Here I will build upon work by

    Kitahara 1997 (see also Ferguson 1994 for a similar line of reasoning

    372 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    30 Mona Anderson (personal communication) points out that the pattern noted in (57) isnot found when the experiencer is a pronoun.

    i. *There seems to her to be men in the roomii. There seems to her to be men in the room

    This fact does not cast doubt on the analysis developed here. As is well-known, pronouns arespecial. In French, for instance, a full NP/DP experiencer seems to block overt raising of thesubject (at least in the present tense, for reasons that are not clear to me), whereas raising isfine when a pronoun is used.

    iii. ??Jean semble a Marie etre heureuxJean seems to Marie to-be happy

    iv. Jean lui semble etre heureuxJean to-her seems to-be happy

    A similar fact is found in Italian, where a full experiencer NP/DP must topicalize to allowraising, while a clitic pronoun is irrelevant for raising purposes.

    v. A Piero, Gianni non sembra fare il suo devereTo Piero Gianni not seems do the his dutyGianni does not seem to Piero to do his duty

    vi. *Gianni non sembra a Piero fare il suo deverevii. Gianni non gli sembra fare il suo devere

    Gianni not to-him seems do the his dutyTorrego (1996:106107)

    Still, the special behavior of pronouns, though suggestive, does not explain why we do notdetect PCC effect in that case in English. Here is what I would like to suggest (see Boeckx1999c: Appendix for extensive discussion).

    Pronouns, unlike full NPs/DPs in English are morphologically marked for Case; they may

    then be said to realize agreement on themselves, much like Quirky elements in Icelandic, andthus leave morphological room for associates to realize their agreement checking relationsfor number (person being excluded by the PCC), but note that English only has number, notperson morphology, hence the apparent `full' agreement with the associate in (ii). Putdifferently, the morphology of pronouns may be said to tease apart number and personagreement, freezing only person, not number, which can then be realized in accordance withthe associate.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    20/27

    regarding the invisibility of some PP nodes). Kitahara (1997:63) notes the

    following. In (62),

    (62) Theyi seem to him [ti to like John]

    John must be interpreted as disjoint in reference from him. Chomsky

    (1995:304) takes such disjoint reference effects as evidence that him c-

    commands into the embedded clause. Given that, Condition C of the

    Binding theory is violated if him takes John as its antecedent. But, if that

    is so, him must also c-command they prior to raising, and should therefore

    block raising (the features of him are closer for attraction), yielding

    something like:

    (63) *Himi seems to ti [they to like John]

    To prevent this, Kitahara suggests that him does not c-command into the

    embedded clause prior to raising. This is made possible via the following

    Larsonian shell (Kitahara 1997:64).

    (64)

    Assuming the standard definition of c-command (Reinhart 1976)

    (roughly, A c-commands B if the first branching node that dominates

    A dominates B), the PP-node dominates him but not the embeddedclause, hence him does not c-command into the latter; raising is thus

    made licit.

    Quirky Agreement 373

    # The Editorial Board of Studia Linguistica 2000.

    That pronouns behave differently from DPs/NPs with respect to agreement and otherphenomena is well-known. Thus, in classical Arabic, agreement is partial with full NPs/DPslocated on the right of the verb, while it is total with pronouns in the same position.

    viii. Naama/*Naamuu l-?awlaad-uslept3XmascXXplasg the-childrennom

    `The children slept'ix. ?al-?awlaad-u naamuu/*naamax. Naamuu/*naama hum

    slept they`They slept'

    xi. Hum naamu/*naamaAoun, Benmamoun & Sportiche (1994:197, (5); 203, (21))

    TP

    T vmax

    v Vmax

    PP VH

    to him seem TP

    . . . they . . .

  • 7/31/2019 Boeckx 2000 Quirky Agr

    21/27

    As regards the disjoint reference effect, Kitahara assumes that Binding

    applies at LF;31 by then, him must c-command John. To achieve this,

    Kitahara suggests that him raises and adjoins to P to check its Case; it is

    only after this movement that him can c-command into the embedded

    clause, which forces the disjoint reference effect (to avoid a Condition C

    violation). As for raising, since him checks its features covertly, it does not

    bleed any overt operations.

    I agree in essence with Kitahara's account, but will suggest a revision

    that will turn out to have wide-ranging empirical consequences.32 Fujita

    (1996:157) proposes that prepositional objects check their Case in a Spec-

    head relation (AGRprep for him, or a specifier position adjoined to PP,

    as in Kitahara 1997, following Chomsky 1995), and that this might

    explain the well-known reanalysis facts of English.

    (65) Johni was talked about ti(66) Whoi did you talk to ti

    Fujita argues that in the case of reanalysis, P and V form a complex

    predicate and that AGRprep AGRo; put differently, the prepositionalobject checks its Case with the [V-P] head, not with P alone.

    This is what I would like to suggest for `seem-to' constructions: to

    incorporates into (`reanalyzes with') V (seem), equating AGRprep and

    AGRo; this mechanism does not affect Kitahara's account: reanalysis

    takes place at LF, the c-command relations he discusses are maintained;

    but the revision allows us to claim that seem checks the Case of the

    experiencer, in other words, agrees with a `dative' element, which

    activates the PCC, as desired.

    Not only does this simple mechanism account for intricate data from

    English, it also allows us to explain the blocking effect of experiencers for

    overt raising in languages like Icelandic or Spanish (cf. (26), (32)) if we

    assume that reanalysis takes place in the overt component in those

    languages, which makes the experiencer close for attraction, barring

    overt raising of the embedded nominative. In English, reanalysis takes

    place covertly, so raising is allowed.33

    One question that arises given this line of reasoning is why there is no

    blocking effect with clitics in French (cf. the discussion below example

    (39)). Torrego 1996, 1998 has shown that those languages where overt

    raising is impossible when an experiencer is present obligatorily double

    the clitic with the equivalent of a to-phrase, Italian and French being the

    374 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    31 As far as I can see, nothing in Kitahara's analysis necessitates that assumption; Binding

    Theory could be said to apply everywhere, the relevant relation being formed at LF in theexample under discussion.

    32 See Boeckx 1999c for a full-fledged account of `conflicting c-command requirements' ofthis kind. The account in Boeckx 1999c is slightly different from the one presented here,which will do for present purposes given the intricacies one would have to go into.

    33 See Boeckx 1998a for extension of that approach to other constructions and otherlanguages.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    22/27

    exceptions to the rule. A plausible explanation is that clitics are base-

    generated as heads on, say, T, and therefore do not block overt raising

    from below (see Boeckx 1998a for extensive discussion).

    5.2. The nature of Quirky Case

    The overt/covert reanalysis asymmetry (which falls within the range of

    generally assumed parameters) might potentially account for why certain

    languages have Quirky experiencers, and others don't. The question is

    very broad, and cannot be fully explored here, but I will suggest a possible

    line of reasoning, focusing on German and Icelandic, which differ

    minimally on the surface, though only one language (Icelandic) has

    Quirky Case, as originally observed by Zaenen, Maling & Thrainsson

    1985.

    (67) Ihm wurde geholfen

    Himdat was3Xsg helped

    `He was helped'

    (68) *Ihm hofft, PRO geholfen zu werden

    Himdat hopes PROdat helped to become

    `He hopes to be helped'

    (69) *Er kam und (ihm) wurde geholfen

    He came and (himdat) became helped`He came and (he) was helped'

    These are but a few subjecthood tests (control, conjunction reduction)

    which German dative elements fail to pass (compare Icelandic `Quirky'

    subjects in section 2). Given that, ZMT put forward the now widely

    accepted view that fronted dative elements in German are topicalized

    constituents, not subjects.

    What I would like to suggest is that the Icelandic/German contrast

    results from an overt/covert asymmetry. The reanalysis process (whichmight underlie every instance of dative agreement) takes place covertly in

    German, but overtly in Icelandic. One piece of evidence in favor of this

    hypothesis comes from raising predicates; German experiencers do not

    block raising, which suggests that reanalysis does not take place prior to

    Spell-Out in the language.

    (70) Peteri scheint mir, ti zufrieden zu sein

    Peternom seems medat happy to be

    `Peter seems to me to be happy'

    Sentences like (71) are well-formed in German:

    (71) Mir scheint Peteri, ti zufrieden zu sein

    medat seems Peternom happy to be

    `Peter seems to me to be happy'

    Quirky Agreement 375

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    23/27

    But this is no evidence for `experiencer-blocking'. The Verb-second

    nature of German obscures cases of Topicalization; but once we are

    aware of that fact (it is difficult to see how German would allow both (70)

    and (71) otherwise, except if one allows for optional renalaysis, which

    should be discarded on minimalist grounds), (71) becomes identical to the

    Italian facts mentioned in note 31 (repeated here under 72)).

    (72) A Piero Gianni non sembra fare il suo devere

    To Piero Gianni not seems do the his duty

    `Gianni doesn't seem to Piero to do his duty'

    The covert reanalysis option in German immediately explains why dative

    elements fail all subjecthood tests like control, binding, etc. These

    phenomena are associated with full category movement, not with feature

    movement (cf. Lasnik 1995b,c), hence their incompatibility with covert

    processes.

    5.3. The nature of inherent Case

    The present study has argued that agreement takes place with dative

    elements in the syntax, but that this process cannot be morphologically

    realized, roughly because the information is already contained on the

    nominal item. As a result, agreement surfaces as default (third personsingular). This casts doubt on the well-established inherent/structural

    Case distinction. The latter was empirically motivated on agreement facts

    from passives in German like (73)(74).34

    (73) Die Kuchen wurden/*wurde gegessen

    The cakesnomXpl became3Xplasg eaten

    `The cakes were eaten'

    (74) Den Kindern wurde/*wurden geholfen

    The childrendatXpl became3Xsgapl helped

    `The children were helped'

    The PCC does not force us to view inherent Case as special, or `inert' for

    agreement; rather, it leads us to argue in favor of viewing inherent Case as

    equally `active' as structural Case in the syntax. In other words, our

    approach unifies structural and inherent Cases a welcome result from a

    minimalist point of view (see Collins & Thrainsson 1996, and Lasnik

    1995a for further arguments that the inherent/structural distinction is

    spurious and should be done away with).

    376 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

    34 Apart from agreement facts, Chomsky 1986 also mentions the fact that inherent Case isintricately linked to theta-selection, thus differing from structural Case, which entertains amuch freer relation with theta-role assignment. I don't find the argument compelling;English has theta-selection processes similar to German, yet is often said to lack inherentCase (thus allowing prepositional passives, preposition-stranding, etc., see Hornstein &Weinberg 1981 for a classic argument along these lines).

  • 7/31/2019 Boeckx 2000 Quirky Agr

    24/27

    Obviously, the consequences of such a unification remain to be

    investigated, but it seems to me that it might overcome many problems

    which the inherent/structural distinction had encountered over the years.

    6. Conclusion

    I started by investigating some of the properties of raising predicates in

    Icelandic, concentration on agreement patterns between the matrix finite

    verb and the nominative element in the embedded clause.

    By resorting to the Person-Case Constraint studied in Bonet 1994,

    according to which object agreement is limited to third person if dative

    agreement takes place, I have been able to explain why only partial (i.e.

    number) agreement takes place with nominative objects, given the factthat agreement with Quirky subjects take place.

    I was able to maintain (indeed, strengthen) the George & Kornfilt

    theory that agreement and Case are but one feature. At no point have I

    felt the need to resort to Agreement projections a desirable result (cf.

    Chomsky 1995).

    The claim that agreement takes place with experiencer (dative/Quirky)

    elements has proven empirically correct, not only for Icelandic, but also

    for English, Spanish, and other languages. The present theory has been

    shown to have wide-ranging empirical and conceptual consequencesfor (among other things) the nature of the associate of expletives

    (which I have argued should be regarded as an object of some sort), for

    Relativized Minimality effects in an Attract-F framework, for the avail-

    ability of `Quirky' elements, and for the inherent/structural Case-theoretic

    distinction.

    In view of this empirical coverage and conceptual simplicity, the path

    taken here seems very promising.

    References

    Aoun, J., Benmamoun, E. & Sportiche, D. 1994. Agreement, word order, andconjunction in some varieties of Arabic. Linguistic Inquiry 25, 195220.

    Belle, W. van & Langendock, W. van 1996. The dative, vol. 1. Amsterdam:John Benjamins.

    Belletti, A. 1988. The Case of unaccusatives. Linguistic Inquiry 19, 135.Benmamoun, E. 1997. Agreement in Arabic at the PF-interface. Proceedings of

    WCCFL XV, ed. B. Agbayani & S.-W. Tang, 3348, Stanford, CA: CSLI.Bobaljik, J. 1995. Morphosyntax: the syntax of verbal inflection. Ph.D. diss.,

    MIT.Bobaljik, J. & Jonas, D. 1996, Subject positions and the role of TP. Linguistic

    Inquiry 27, 195236.Bobaljik, D. & Thra insson, H. 1998. Two heads aren't always better than one.

    Syntax 1, 3771.Boeckx, C. 1998a. Raising in Romance. Ms, U. of Connecticut. A celebration for

    Noam Chomsky's 70th birthday, http://mitpress.mit.edu/celebration.

    Quirky Agreement 377

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    25/27

    Boeckx, C. 1998b. Agreement constraints in Icelandic and elsewhere. Workingpapers in Scandinavian syntax 62, 135.

    Boeckx, C. 1999a. Remnants of movement. Ms, U. of Connecticut.Boeckx, C. 1999b. Extractions. Ms, U. of Connecticut.

    Boeckx, C. 1999c. Conflicting c-command requirements. Studia Linguistica 53,227250.Bonet, E. 1994. The Person-Case Constraint: a morphological approach. The

    Morphosyntax connection, MIT Working Papers in Linguistics 22, ed. H. Harley& C. Phillips, 3352.

    Boskovic , Z . 1997. The syntax of non-finite complementation. An Economyapproach. Cambridge, MA: MIT Press.

    Boskovic , Z . 1998. LF-movement and the minimalist program. Proceedings ofNELS 28, ed. P. N. Tamanji & K. Kusumoto, 4357. U. of Massachusetts,Amherst: GLSA.

    Chomsky, N. 1986. Knowledge of Language. Its nature, origin, and use. New-York:

    Praeger.Chomsky, N. 1995. The Minimalist Program. Cambridge, MA: MIT Press.Chomsky, N. 1998. Minimalist inquiries: the framework. MIT Occasional Papers

    in Linguistics 15. Cambridge, MA: MITWPL. [To appear, Step by step: essayson minimalism in honor of Howard Lasnik, ed. R. Martin, D. Michaels &J. Uriagereka. Cambridge, MA: MIT Press.]

    Collins, C. 1997. Local Economy. Cambridge, MA: MIT Press.Collins, C. & Thra insson, H. 1996. VP-internal structure and Object Shift in

    Icelandic. Linguistic Inquiry 27, 391444.Culicover, P. 1992. Polarity, inversion, and focus in English. Proceedings of

    ESCOL'91, 4668. Dept. of Linguistics, Ohio State U., Columbus.

    Delsing, L-O. 1996. The reduction of a Case system Old Swedish. Talkdelivered to the Linguistic Institute, U. of Troms.

    Ferguson, K. 1994. Deriving the invisibility of PP-nodes for Command fromAGR-P+P Case checking. Harvard Working Papers in Linguistics 4, ed. S. D.Epstein, H. Thrainsson & S. Kuno, 3036. Harvard U.

    Fernandez-Soriano, O. 1999. Two types of impersonal sentences in Spanish:locative and dative subjects. Syntax 2, 101140.

    Fujita, K. 1996. Double objects, causatives, and derivational economy. LinguisticInquiry 27, 146173.

    George, L. & Kornfilt, J. 1981. Finiteness and boundedness in Turkish. Bindingand Filtering, ed. F. Heny, 105127. Cambridge, MA: MIT Press.

    Halle, M. & Marantz, A. 1993. Distributed Morphology and the pieces ofinflection. The View from Building 20: Essays in Linguistics in Honor of SylvainBromberger, ed. K. Hale & S. J. Keyser, 111176. Cambridge, MA: MIT Press.

    Harley, H. 1995a. Abstracting away from abstract Case. Proceedings of NELS25, ed. J. Beckman, 20721, U. of Massachusetts, Amherst: GLSA.

    Harley, H. 1995b. Subjects, events, and licensing. Ph.D. diss., MIT.Hornstein, N. & Weinberg, A. 1981. Case theory and preposition stranding.

    Linguistic Inquiry 12, 5592.Jo nsson, J. 1996. Clausal architecture and Case in Icelandic. Ph.D. diss., U. of

    Massachusetts, Amherst.Kayne, R.

    1975. French Syntax: the transformational cycle. Cambridge, MA: MITPress.Kayne, R. 1995. Agreement in three English dialects. Studies in comparative

    Germanic Syntax, ed. H. Haider, S. Olsen & S. Vikner, 15965. Dordrecht:Kluwer.

    378 Cedric Boeckx

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    26/27

    Kitahara, H. 1997. Elementary operations and optimal derivations. Cambridge,MA: MIT Press.

    Laka, I. 1990. Negation in syntax. Ph.d. Diss., MIT.Lasnik, H. 1992. Case and expletives: notes toward a parametric account.

    Linguistic Inquiry 23, 381406.Lasnik, H. 1995a. Case and expletives revisited: on Greed and other humanfailings. Linguistic Inquiry 26, 61533. [Reprinted in Lasnik 1999, 7496.]

    Lasnik, H. 1995b. Last Resort. Minimalism and linguistic theory, ed. S. Haraguchi& M. Funaki, 132. Tokyo: Hituzi Syobo. [Reprinted in Lasnik 1999,120150.]

    Lasnik, H. 1995c. Last Resort and Attract F. Papers from the Sixth AnnualMeeting of the Formal Linguistics Society of Midamerica, ed. L. Gabriele,D. Harison & R. Westmoreland, 6281. Indiana University Linguistics Club,Indiana U., Bloomington.

    Lasnik, H. 1999. Minimalist analysis. Oxford: Blackwell.

    Lasnik, H. To appear. Chains of arguments. Working minimalism, ed. S. D.Epstein & N. Hornstein. Cambridge, MA: MIT Press.

    Lebeaux, D. 1983. A distributional difference between reciprocals and reflexives.Linguistic Inquiry 14, 72330.

    Lyle, James. 1997. Oblique subjects and nominative anaphors. Proceedings ofWCCFL XV, ed. B. Agbayani & S.-W. Tang, 32136. Stanford, CA: CSLI.

    Maling, J. 1984. Non-clause-boundedness reflexives in Modern Icelandic.Linguistics and Philosophy 7, 21141.

    Martin, R. 1996. A minimalist theory of PRO and control. Ph.D. diss., U. ofConnecticut, Storrs.

    Moro, A. 1997. The raising of predicates. Cambridge: Cambridge U. Press.

    Ochi, M. To appear. Some consequences of Attract F. Lingua.Ormazabal, J. To appear. Toward a minimalist conception of Case. Step by step:

    Essays on minimalism in honor of Howard Lasnik, ed. R. Martin, D. Michaels &J. Uriagereka. Cambridge, MA: MIT Press.

    Ormazabal, J. & Romero, J. 1998. On the syntactic nature of the me-luiand thePerson-Case constraint. Ms, U. of the Basque Country & MIT.

    Perlmutter, D. 1971. Deep and surface structure constraints in syntax. New-York: Winston, Rinehart and Holt.

    Pesetsky, D. & Torrego, E. To appear. T-to-C movement: causes and con-sequences. Ken Hale: A life in language, ed. M. Kenstowicz. Cambridge, MA:MIT Press.

    Raposo, E. & Uriagereka, J. 1996. Indefinite se. Natural Language and LinguisticTheory 14, 749810.

    Reinhart, T. 1976. The syntactic domain of anaphora. Ph.D. diss., MIT.Rizzi, L. 1997. The fine structure of the left periphery. Elements of grammar, ed.

    L. Haegeman, 281337. Dordrecht: Kluwer.Rooryck, J. 1997. Passive as partitive quantification. Projections and interface

    conditions, ed. A. Di Sciullo, 20137. Oxford: Oxford U. Press.Schu tze, C. 1997. INFL in child and adult language: agreement, Case, and

    licensing. Ph.D. diss., MIT.Sigursson, H. 1991. Icelandic Case-marked PRO and the licensing of lexical

    arguments. Natural Language and Linguistic Theory 9, 32764.Sirursson, H. 1992. The Case of Quirky subjects. Working Papers inScandinavian Syntax 49, 126.

    Sirursson, H. 1996. Icelandic finite verb agreement. Working Papers inScandinavian Syntax 57, 1 46.

    Quirky Agreement 379

    # The Editorial Board of Studia Linguistica 2000.

  • 7/31/2019 Boeckx 2000 Quirky Agr

    27/27

    Smith, H. 1994. ``Dative sickness'' in Germanic. Natural Language and LinguisticTheory 12, 675736.

    Smith, H. 1996. Restrictiveness in Case theory, Cambridge: Cambridge U. Press.Taraldsen, K. 1995. On agreement and nominative objects in Icelandic. Studies

    in Comparative Germanic Syntax, ed. H. Haider, S. Olsen & S. Vikner, 30727.Dordrecht: Kluwer.Thra insson, H. 1979. On complementation in Icelandic. New-York: Garland.Torrego, E. 1996. Experiencers and raising verbs. Current issues in Comparative

    grammar, ed. R. Freidin, 10120. Dordrecht: Kluwer.Torrego, E. 1998. The dependencies of objects. Cambridge, MA: MIT Press.Uriagereka, J. 1995a. Aspects of the syntax of clitic placement in Western

    Romance. Linguistic Inquiry 26, 79123.Uriagereka, J. 1995b. An F-position in Western Romance. Discourse-configura-

    tional languages, ed. K. E . Kiss, 15375. Oxford: Oxford U. Press.Yip, M., Maling, J. & Jackendoff, R. 1987. Case in tiers. Language 63, 21750.

    Zaenen, A., Maling, J. & Thra insson, H. 1985. Case and grammatical functions:the Icelandic Passive. Natural Language and Linguistic Theory, 3, 44183.

    Cedric BoeckxDepartment of Linguistics

    University of Connecticut, U-1145Storrs, CT 06269-1145

    USAemail: [email protected]

    Received January 9, 1999Accepted September 21, 1999

    380 Cedric Boeckx