Untitled

Proper answers to polarity questions with aspectual adverbs share the presuppositions of the question (cf. [1], for an introduction to DRT [3], [8]), as in (1a, b). If the presuppositions of the question are not accepted, the reply cannot enforce contradicting presuppositions (1b), but it must use another, stronger form of negation. i.e. denial (1d). In (1a, b) still and no longer share a presupposition, not shared by already in (1c), and denied by not yet in (1d).

If (2 a. b) is assumed to constitute coherent discourse, the information which is presupposed by the first clause cannot be asserted felicitously in the subsequent clause in a simple past tense, eventhough it is obviously assumed to be true.

The incohesion of the discourse in (2) may be repaired by inserting an polarity reversal to shift the reference time in (3 a, b) . This forces disjoint temporal reference of the final clause, hence it no longer constitutes the presupposition of the first clause.

Assuming an overall constraint of coherence of information, adding information by presupposition accommodation must hence be distinguished from asserting information. Asserting dynamic information entailed by the preceding clause creates incoherence. Asserting static information that is inconsistent with content of preceding clause cannot be repaired by shifting the context in accommodating its dynamic presuppositions.

The DRT semantics of aspectual adverbs, abbreviated in (4), reflects the interaction between the reference time, the polarity transitions initiating and ending events, and their anaphoric dependency created by the core temporal adverbs SINCE and UNTIL.

Using the classical notion of logical consequence in DRT as defined in (5), it is straightforward to account for the fact that dynamic temporal presuppositions cannot be asserted at the reference time of the subsequent state, eventhough they are entailed,

Let K, K’ be pure (...) DRSs. K’ is a logical consequence of K ( K |= K’) iff the following condition holds:

Suppose M is a model and f is a function from U_K Θ Fr(K) Θ Fr(K’) into U_M, s.t. M |=_f K. Then there is a function g Ι _UK’ f such that M |=_g K’.

Asserting dynamic information about a polarity transition (e,g, fall asleep, wake up) always shifts the context, as consistency must be maintained. However, when the presupposed information is adjusted to the static perfect tense, it can be asserted later, for it will not affect the context anymore, as in (6).

This supports the view that states described by perfect tense information must be semantically distinguished from states described by simple past tense information, The first, but not the latter, can always coherently be asserted, even when already entailed. This is accounted for by supplementing the DRS construction rules and the classical notion of logical consequence with an inference rule ‘PERF introduction’ which characterizes in which contexts information introduced by simple past tense clauses may be converted to past perfect tense (cf. [4], [5]).In this way validity of temporal reasoning is characterized without any appeal to hairy notions such as a ‘normal’ course of events or ‘normal possible world’ or to common sense about what the world is like or how causal connections arise, as in default logics (cf. [2]) .

[2] Lascarides, A. and N. Asher Temporal Interpretation, Discourse Relations and Commonsense Entailment, Linguistics and Philosophy 16.5 (1993) 437--493

[4] ter Meulen, A. Representing Time in Natural Language. The dynamic interpretation of tense and aspect. MIT Press, Cambridge, (1995),

[5] ter Meulen, A. Chronoscopes: the dynamic representation of facts and events, in J. Higginbotham et al. (eds.) Speaking about events. Oxford U.P., (2000).

[6] Smessaert, H. and A. ter Meulen, Dynamic reasoning with aspectual adverbs. Ms.

[7] van Benthem, J. and A. ter Meulen (eds.), Handbook of Logic and Language. Elsevier Science, Amsterdam, & MIT Press, Cambridge, (1997).

[8] van Eijck, J. and H. Kamp. Representing discourse in context. in [7] 179-237.

Agreement checking at PF

Peter Ackema

1. Chomsky (2001) suggests that agreement is established downward across arbitrarily long distances, as long as no potential target intervenes between the ‘probe’ (with unintepretable phi-features) and the ‘goal’ (with identical, but interpretable, phi-features). The classical spec-head configuration for agreement checking only results if the probe also has an ‘EPP feature’ that triggers movement. On this view, it is surprising that some properties of agreement, such as its morphological incarnation, depend on whether or not the goal moves to spec-probe (Rizzi & Guasti 1999, Carstens 2002). We argue that the difference between the two situations is not one between syntactic checking with or without movement, but one between syntactic checking and postsyntactic checking at PF. In order for agreement to obtain, the agreeing elements must be in a local relationship. However, locality need not be defined syntactically, but can be defined prosodically as well. We show that the properties of some agreement relations are quite unexpected if conditioned by syntax, but follow naturally if conditioned by prosodic phrasing. The particular case we discuss is Germanic complementizer agreement. From our analysis an account of the complementizer-trace phenomenon follows as well, including an explanation of the circumstances under which the structure is condoned.

2. Complementizer agreement is checked at PF, and is conditioned by the prosodic phrasing at this level. In particular, C and the agreeing DP must be in the same phonological phrase (f). If agreement checking consists of identifying the uninterpretable features on one element with the interpretable features on another (Brody 1997), this hypothesis can be expressed as in (1), where braces indicate f-boundaries and F1, F2, and F3 are phi-features:

(1) {[C (F1 F2 F3)] [D (F1 F2 F3)]} ® {[C (F1_i F2_j F3_k)] [D (F1_i F2_j F3_k)]}

Selkirk (1984, 1986) and others have shown that prosodic phrasing is essentially determined by alignment principles: either left edges of XPs are aligned with left edges of fs, or right edges of XPs are aligned with right edges of fs. For the languages we consider the latter holds: prosodic phrases are closed at the right edge of syntactic phrases:

3. From (1) and (2) four predictions follow. (I) C-agreement does not occur when the subject is left-adjacent to C rather than right-adjacent. If the subject precedes C, its right XP-bracket triggers a f-boundary between it and C (as opposed to when the subject follows C). If the subject is to the left of C there can of course still be agreement with C, but this must be syntactically conditioned agreement; PF C-agreement should be impossible. The prediction is borne out by those East-Netherlandic dialects that have different paradigms for the two types of agreement (Van Haeringen 1958, Zwart 1997):

In (3) the –e typical of the C-agreement paradigm is used when the subject is right-adjacent to C, but regular V-agreement in –t shows up when the subject is left-adjacent to C. (II) C-agreement should not occur if a phrase intervenes between C and a subject to the right of C. The right XP-bracket of such a phrase triggers f-closure, so the subject and C do not end up in the same f, see (4). A comparison of (3) with (5) makes clear that this prediction is correct (cf. Haeberli 1999). (This intervention effect is hard to capture in syntactic accounts of C-agreement: the intervening element in (5) is not a potential goal for agreement with C).

(5) Volgens miej loop-t/*lop-e op den wδrmsten dag van ‘t joar ook wiej noar ‘t park

according-to me walk-pl/walk-pl on the hottest day of the year also we to the park

(III) C-agreement must always be subject agreement. Only elements in A-positions are targets for phi-feature agreement. In the languages in question, the only element that can be right-adjacent to C while being in an A-position is the subject. (IV) Tense is uninterpretable in a complementizer, (in contrast to tense in a verb), hence any tense on C should be checked. But given prediction III (PF C-agreement checking can only involve subjects), this is impossible, since subjects are tenseless. It follows that C-agreement checking can never involve tense, so that languages in which agreement and tense morphology are fused cannot show such agreement. Only languages with distinct morphology for agreement and tense allow it (Hoekstra & Smits 1998).

4. Let us now turn to complementizer-trace phenomena. Suppose that two links in a movement chain cannot be in positions in which the same agreement can be checked, a condition that applies at PF. Suppose furthermore that for extraction out of CP to be possible the moving element must move through spec-CP and agree with C there (in order for its features to be visible on CP and hence accessible for an outside antecedent; ‘phase impenetrability’). Then it is impossible to move an element from a position right-adjacent to C (a position in which agreement with C can be checked) through spec-CP (another such position) to a position outside CP – a fair description of the complementizer-trace phenomenon. This account makes three predictions. (I) If the trace is not right-adjacent to C, it is not in a potential PF checking position (see.section 3) and hence the C-trace filter should be voided. This ameliorating effect of intervening phrases can indeed be observed (Culicover 1993):

(6) Who_i did you say t_i that, for all intents and purposes, t_i is the mayor of this city?

Similarly, languages like Italian, in which subjects can occur on the right edge of VP, will not show the complementizer-trace effect (Rizzi 1982). (II) In languages with complementizer agreement, C is already specified for the subject’s phi-features, hence the subject need not move through spec-CP in order to make these features visible on CP. It can be extracted directly from its base position. Thus, C-agreement should void the C-trace effect as well. This too has been observed, for example in Frisian (De Haan 1997):

(III) The above line of argumentation does not apply only to subjects, but to all elements extracted from a position right-adjacent to C. (In contrast to complementizer agreement, the kind of abstract agreement required to make the moving element’s features visible at the CP level is not restricted to elements in A positions – certainly not: spec-CP is not an A-position. Hence, a position right-adjacent to C counts as a potential checking position for such agreement, for any element. That such agreement cannot actually be checked when the element is in situ is because the relevant features must be accessible in syntax for extraction to be possible: PF agreement will not do.) Thus, the complementizer-trace phenomenon should extend to non-subject elements extracted from a positon right-adjacent to C. This is borne out by extraction from clauses with locative inversion in English (Bresnan 1994) and from embedded V2 clauses in Yiddish (Branigan 1998):

(8) In which cities_i do you think t_i (*that) t_i are found the best examples of this cuisine

Mood, Negation and Pronominal Clitics: Evidence from nine Balkan languages

Olga Tomić

There is a general agreement in the literature that Neg is in the IP periphery, lower than CP. I will argue that, in the Balkan languages, there are actually two Neg sites – one to the left of the modality projection (ModP) and another to the right of the mood projection (MoodP). While the former site is operative in indicative clauses, the latter one is operative in subjunctives and imperatives.

First of all, I will give evidence that MoodPs, where subjunctive markers are derived, are distinct from modality phrases ModPs, where the future-tense-marking modal clitics are placed. Then, I will show that verbs with imperative morphology and bare subjunctive constructions share interpretative properties: they both point to “non-veridical” events and can be used with imperative force. This fact argues for checking the imperative mood in the same node in which the subjunctive marker is derived.

The major part of the talk will be devoted to work in progress on the position of pronominal clitics in negative imperatives (with imperative morphology), which occur in all Balkan language except in Modern Greek. Although in the surface structures of the Balkan languages the negation operator appears to the left of the imperative verb, while in subjunctive constructions it appears to the right of the subjunctive marker, the shared interpretative and selectional properties of imperatives and subjunctive argue for analogy in the placement of Neg relative to Mood. Since a prohibition is interpreted as an order not to do something and not as a lack of order to do something, the scope of Mood should be wider than that of Neg, i. e. Neg should be to the left of Mood in both subjunctives and imperatives.

The position of the clitics relative to the negated imperative verb differs from langauge to language. I will discuss various possibilities for “reconciling” these differences.

Measure Phrases and the XP-condition on movement

Maarten Klein

KU Nijmegen, Universitδt zu Kφln

One of Ross’s major contibutions to constraining the power of transformations has been the Left Branch Condition (LBC): no NP which is the leftmost constituent of a larger NP can be reordered out of this NP by a transformational rule. In a number of publications, Corver has tried to show that a number of violations of this constraint can be found, at least in Dutch. In Corver (1999), for example, we find the following sentences that seem to violate Ross’s LBC:

However, in one of his footnotes to his 1999 article, Corver suggests that extraction of the measure constituent from the NP hoeveel uur rust is only possible in case of verbs with a ‘hebben’-like meaning (to have): achterstand hebben, rust nemen. This makes the extraction analysis highly unattractive for several reasons, one of which is that these combinations may be replaced by verbs (achterliggen, rusten), with the measure constituent having the same function (Jan heeft twee uur gerust). My conclusion will be that measure constituents like twee uur in Jan heeft twee uur rust genomen can be either part of the NP twee uur rust or an independent adverbial measure constituent. Preposing is only possible in the case of independent constituents.

Subjects in early Italian-Dutch bilingualism

Manuela Pinto

Thematic Arity Operations and Case: Evidence from Greek

Dimitra Papangeli, UiL-OTS

Romanian Genitive Constructions and the Structure of the Romanian DP

Alexandra Cornilescu, University of Bucharest

Negative Concord and the availability of negative projections

Hedde Zeijlstra, University of Amsterdam

Languages vary both with respect to the syntactic status of the negative marker (adverb, syntactic head, affix) and the way that multiple negative expressions are interpreted (Double Negation vs. Negative Concord). On the basis of Dutch diachronic and dialectological and typological research I will show that there is a unidirectional relation between these two phenomena: languages that have a negative marker that corresponds to a syntactic head position (most Romance and Slavic languages) all exhibit Negative Concord, whereas languages that only have an (overt) negative marker can be Double Negation languages (Dutch, German, Swedish), but can also be Negative Concord languages (Bavarian, Yiddish, Quebecois).

In my analysis I will argue that Negative Concord crucially relies on the availability of a NegP in the language. If a language contains a functional category there is Negative Concord and vice versa. This means that Negative Concord should be thought of as a multiple agree relation between elements carrying uninterpretable [Neg] feature. Double Negation is the result of the absence of NegP in a language in which negative elements obligatorily carry interpretable [Neg] features.

I will show that this analysis explains the difference between n-words and Negative Polarity Items and Negative Quantifiers, and that this analysis can account for the difference between Strict Negative Concord languages and Non-strict Negative Concord languages (which don't allow an n-word in preverbal position followed by a negative marker).

Two types of multiple nominative constructions in Japanese

Reiko Vermeulen, University of Amsterdam

A clause in Japanese may contain more than one nominative phrase. In addition to marking the subject, the nominative marker ga can mark a possessor of the subject (1) or an adjunct (2). In these constructions, the order of ga-phrases is fixed. The following examples hence display the only possible orders. Furthermore, a clause-initial ga-phrase is obligatorily focused (Kuno 1973).

The standard view is that all ga-phrases are NPs or DPs and a tensed head licenses them uniformly in multiple specifier or adjoined positions in one particular projection such as TP, IP, or VP (e.g. Fukui 1986, Heycock 1993, Hiraiwa 2001, Saito 1982, Takahashi 1994, Tateishi 1991, Ura 1996). In other words, a distinction between the two constructions is rarely made. In this paper, I argue that the standard licensing mechanism alone cannot account for several significant differences between them, some of which have so far been unnoticed, and propose an alternative analysis. The acceptability of examples with these constructions is sometimes attributed to the choice of the predicate. I maintain however that their grammaticality is explained in syntactic terms. More specifically, the main claim is that a possessive ga-phrase is an NP, while an adjunct ga-phrase is a PP. The particle ga always functions as a case marker when attached to an NP, but an interpretational rule also treats it as a focus marker in clause-initial position.

A possessive ga-phrase is an NP and is licensed by the clause to its right by predication mediated by a null operator, as in the case of English tough constructions (exemplified in (3) for (1)). A null operator binds a pro, which has the effect that an NP-internal q-role (possessor role) is promoted and assigned to the immediately preceding possessive ga-phrase, which is base-generated in an adjoined position to TP. This operation is potentially recursive. Since a derived predicate cannot contain its subject, a possessive ga-phrase must always precede the NP which contains pro associated with it, resulting in a fixed order among ga-phrases. Ga attached to the possessor functions as a case marker, but the interpretational rule also treats it as a focus marker, since it is licensed in clause-initial position.

By contrast, an adjunct ga-phrase is a PP. This conclusion is based on the observations that a postposition can be spelled out before ga attached to an adjunct and that an adjunct ga-phrase cannot host a floating quantifier. An NP may, but a PP may not, host a floating quantifier in Japanese (Miyagawa 1989). Ga attached to an adjunct cannot be a case-marker, because PPs do not require case. Consequently, it can only function as a focus marker and as such it must appear in clause-initial position in accordance with the interpretational rule, accounting for the rigid word order. (2) has the following structure.

The proposed approach accounts for the following contrasting properties of the two constructions, which are difficult to capture with the standard licensing mechanism alone. (i) A possessive ga-phrase has a possessive relation with the following ga-phrase, but an adjunct ga-phrase does not. (ii) An indefinitely large number of possessive ga-phrases, but only one adjunct ga-phrase, is permitted in a clause. An extra adjunct ga-phrase renders ga on the lower adjunct uninterpretatble, since it cannot function as a case marker and it is not in clause-initial position, where it can be interpreted as a focus marker. (iii) A possessive ga-phrase can host a floating quantifier, but an adjunct ga-phrase cannot. (iv) Pro associated with a possessive ga-phrase, but not with an adjunct ga-phrase, can be overtly realised. PP-pro does not exist in Japanese (Longobardi 1984). (v) A possessive ga-phrase has a subject-predicate relation with the clause to its right, but an adjunct ga-phrase does not. (vi) In an embedded clause, where a clause-initial ga-phrase which is obligatorily focused in a matrix context need not be focused for independent reasons (Kuroda 1987), a non-focused interpretation is available only for a possessive ga-phrase and not for an adjunct ga-phrase. Ga on a possessor can function as a case marker, while ga on an adjunct can only be interpreted as a focus marker.

First Conjunct Agreement and Double Agreement in Dutch dialects: interaction between syntax and morphology

Marjo van Koppen, Leiden University

In this talk I discuss how syntax and morphology interact. I assume, following minimalist guidelines (Chomsky 1995, 2000) that syntax is an economical and purely derivational system. Furthermore, I follow the insights of Distributed Morphology (Halle & Marantz 1993) and take it that syntax operates with feature bundles solely. These bundles of features get spelled out by Morphology, a subsystem that operates at the PF-branch. Morphology in contrast with Syntax does not have as its goal to go for the most minimal option, rather in spelling out feature bundles it will opt for spelling out the least minimal, i.e. the most specified item.

I discuss two empirical phenomena and show that their analyses largely overlap. The empirical phenomena I am dealing with are agreement with the first conjunct of a coordinated subject in Frisian and Tegelen Dutch on the one hand (cf. example (1a-b)) and so-called ‘double agreement’ (cf. a.o. Zwart 1993) in the dialect of Hellendoorn on the other (cf. example (3)).

(1) a. … dat-st [do en Marie] yn Rome west ha.

… that-_2sg[you_2sg and Marie_3sg]_2plin Rome been have_pl

‘…that you and Marie have been in Rome this weekend.’

[Frisian]

b. … de-s [doow en ich] τs treff-e.

… that-_2sg [you_2sg and I_1sg]_1pl each.other_1pl meet-_pl

‘… that you and I could meet.’

[Tegelen Dutch]

These examples show that in Frisian and Tegelen Dutch (a dialect spoken in the province of Limburg), the complementizer shows agreement with the first conjunct of the coordinated subject rather than with the subject as a whole. Interestingly, the complementizer does show FCA, whereas the finite verb does not. This is demonstrated by the ungrammaticality of the example in (2).

(2) * Doow en ich tref-s τs.

[you_2sg and I_1sg]_1plmeet-_2sg each.other_1pl

I assume complementizer agreement to be a reflex of phi-features on C°. I argue that FCA can obtain as C° agrees with both the coordinated subject as a whole, i.e. with the features of the maximal projection and with the first conjunct of the subject, as they are both equally close. Morphology chooses to spell out the agreement relation that results in the most specific agreement morphology. I show that this is the relation between C° and the first conjunct. Furthermore, I argue that the difference between the complementizer and the verb with respect to the ability to have FCA boils down to the fact that the target for agreement moves out of the c-command domain of T° (the controller for agreement on the finite verb), whereas it doesn’t move out of the c-command domain of C°.

The second empirical domain I touch upon is double agreement. Consider the example in (3).

(3) a. …darr-e wiej den besten bint!

…that-agr we the best are

‘…that we are the best!’

b. Wiej bin-t den besten!

we are-_agr the best

‘We are the best!’

c. Binn-e wiej den besten?

are-_agr we the best

‘Are we the best?’

[Hellendoorn Dutch]

This example shows that in Hellendoorn Dutch, the agreement associated with C°, i.e. the agreement on the complementizer and the verb in inversion, has a different ending than the verb in non-inversion contexts. I show that this pattern can be analysed by assuming that the internal structure of the plural pronoun is visible to syntax, much in the spirit of Harley & Ritter 2002). For this phenomenon I also argue that the agreement on the complementizer and the verb in inversion results from an intricate interaction between syntax and morphology. Furthermore, I show that the difference between the agreement on the finite verb in SVO-context and that on the complementizer boils down to the fact that the target for agreement moves out of the c-command domain of T°, whereas it stays put in the agreement domain of C°.

Abstracts

“Black” and “White” languages: parameter clusters in the languages of the world

Leon Stassen

Temporal cohesion and the presuppositions of aspectual adverbs

Alice ter Meulen