Romance languages and linguistic theory 2000

Section 3 shows that the generic readings of proarb in examples like lb or 1c can be derived from the interaction of the indexical 2a or definite readings of pro 2b with the type of qua

ROMANCE LANGUAGES AND LINGUISTIC THEORY 2000

AMSTERDAM STUDIES IN THE THEORY AND

HISTORY OF LINGUISTIC SCIENCE

General Editor

E F KONRAD KOERNER (University of Cologne)

Series IV – CURRENT ISSUES IN LINGUISTIC THEORY

Advisory Editorial Board

Raimo Anttila (Los Angeles); Lyle Campbell (Christchurch, N.Z.) Sheila Embleton (Toronto); John E Joseph (Edinburgh)

Manfred Krifka (Berlin); Hans-Heinrich Lieb (Berlin)

E Wyn Roberts (Vancouver, B.C.); Hans-Jürgen Sasse (Köln)

Volume 232

Claire Beyssade, Reineke Bok-Bennema, Frank Drijkoningen

and Paola Monachesi (eds)

Romance Languages and Linguistic Theory 2000

Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December

JOHN BENJAMINS PUBLISHING COMPANY

AMSTERDAM/PHILADELPHIA

ROMANCE LANGUAGES AND

SELECTED PAPERS FROM 'GOING ROMANCE'

The paper used in this publication meets the minimum requirements of AmericanNational Standard for Information Sciences — Permanence of Paper for PrintedLibrary Materials, ANSI Z39.48-1984.

8TM

Library of Congress Cataloging-in-Publication Data

Romance Languages and Linguistic Theory 2000 : Selected papers from 'Going Romance' 2000,Utrecht, 30 November – 2 December / Edited by Claire Beyssade [et al.]

p cm (Amsterdam studies in the theory and history of linguistic science Series IV, Currentissues in linguistic theory, ISSN 0304-0763 ; v 232)

Includes bibliographical references and index

1 Romance languages Congrsses I Beyssayde, Claire II Series

I N T R O D U C T I O N

In the nineties, Going Romance has emerged as the major European annual

discussion forum for theoretically relevant research on R o m a n c e languages

Going Romance is an international initiative of the Dutch university

community involved in research on R o m a n c e languages; it is organized by and held at the various universities of the country In 1998 the steering committee judged that the quality of the papers presented w a s so high that this should be reflected in proceedings volumes

The current book is the second proceedings volume T h e articles form a

selection of the papers that have been presented at the occasion of Going

Romance 2000 (XIV) - which was held at Utrecht University on N o v e m b e r 30

through December 2 The three day program included a workshop on Topic and Focus Next to the articles discussing pragmatic factors and their interaction with phonology, syntax and semantics, the volume contains articles dealing with a variety of topics that bear on the specifics of one or more

R o m a n c e languages

The editors would like to thank everyone w h o contributed to the success of

Going Romance XIV Next to the editors, the organization committee consisted

of Sergio B a a u w (Utrecht, U i L - O T S ) , Denis Delfitto (Utrecht, U i L - O T S ) , Jenny Doetjes (Utrecht, U i L - O T S ) , Aafke Hulk (Amsterdam, HIL), Brigitte Kampers-Manhe (Groningen, C L C G ) , Johan Rooryck (Leiden, H I L ) , Jan Schroten (Utrecht, U i L - O T S ) , Henriëtte de Swart (Utrecht, U i L - O T S ) and Els Verheugd (Amsterdam, HIL) T h e selection committee for the more than fifty abstracts consisted of editors and organizers and w a s assisted b y the following independent advisors: Elisabeth Delais-Roussarie (Orléans), D o n k a Farkas ( U C S B ) , Haike Jacobs (Nijmegen, C L S ) , K n u d Lambrecht (Texas), Jean-Marie Marandin (Paris VII), Jean-Yves Pollock (Amiens), Georges Rebuschi (Paris III), Annie Rialland (Paris III), and Roberto Zamparelli (Bergamo)

The Organizers and the editors gratefully acknowledge financial support from

the Royal Netherlands Academy of Arts and Sciences ( K N A W ) and the Netherlands Organisation for Scientific Research ( N W O ) , from the host of Going R o m a n c e 2000, the Utrecht institute of Linguistics-OTS (UiL-OTS), the Holland Institute of Generative Linguistics (HIL), and the PICS project

Finally, w e wish to thank Taka Hara, Erica van Lente and Hiske van der Meulen for their invaluable help in the final phases of preparing the manuscript

Claire Beyssade

Reineke Bok-Bennema

Frank Drijkoningen

Paola Monachesi

Gloria Cocchi and Cecilia Poletto

Rhematic focus at the left periphery: the case of Romanian 77

Alexandra Cornilescu

Multiple focus in European Portuguese: apparent optionality

and subject positions 93

Jộo Costa

External subjects in two varieties of Portuguese: evidence for

a non-unified analysis 109

Jộo Costa and Charlotte Galves

Extreme Non-Specificity in Romanian 127

Donka Farkas

Resultatives: Small Clauses or complex V P s ? 153

Raffaella Folli

C O N T E N T S Viii

Topic, focus and secondary predication: the French Presentational

Relative Construction 171

Knud Lambrecht

Intonative structure of focalization in French and Greek 213

David Le Gac and Hi-Yon Yoo

Splitting up subject clitic-verb inversion 233

Nicola Munaro

Edging Quantifiers: on QP-Fronting in Western R o m a n c e 253

Josep Quer

Partitive constructions and antisymmetry 271

Petra Sleeman and Ellen-Petra Kester

Stress-Focus correspondence in Italian 287

ARBITRARY PRONOUNS ARE NOT THAT INDEFINITE

LUIS ALONSO-OVALLE

University of Massachusetts at Amherst

0 Introduction

Although defining structural constraints o n coindexing proved fruitful, its

early eighties extended the use of indexing to capture the 'arbitrariness' of examples like (la) (Chomsky and Lasnik 1977, C h o m s k y 1980), (1b), (1c) (Casielles 1993) or (1d) (Suñer 1983)

"In that restaurant, you (generic) work like a slave"

"In that restaurant, they work like slaves"

" S o m e b o d y is knocking at the door"

T h e semantic import of this type of indexing is not less unclear Chierchia

reflection of second order predication I will cast s o m e doubts about the

* I a m very grateful to Reineke Bok-Bennema, Francesco D'Introno, D o n k a Farkas, L y n Frazier, Kyle Johnson, Paula Menéndez-Benito, Barbara Partee, María Luisa Rivero, audiences

in Barbara Partee's Fall 99 U M a s s seminar and two anonymous reviewers Thanks to Irene

H e i m and Kai von Fintel for being so generous with their teaching and materials I a m truly indebted to Barbara Partee for more advice, help and encouragement than I could acknowledge Usual disclaimers apply

2 LUIS ALONSO-OVALLE

In section 1 I review an analysis that assumes the existence of proarb and takes it to be an indefinite pronoun (Condoravdi 1989, K i m 1991) I argue against it in section 2 I then show that there is no need to appeal to proarb to derive the 'arbitrary' flavor of (1b-d) Section 3 shows that the generic readings

of proarb in examples like (lb) or (1c) can be derived from the interaction of the indexical (2a) or definite readings of pro (2b) with the type of quantification frequently associated with generic sentences

(2) a E n ese restaurante, pro trabajaste c o m o un esclavo

In that restaurant pro work:2sPAST like a slave

"In that restaurant, you (addressee) worked like a slave"

b E n ese restaurante, pro trabajaron c o m o esclavos

In that restaurant pro work:3SPAST like slaves

"In that restaurant, they worked like a slaves"

Section 4 shows that the existential reading of proarb in examples like (1d) can

be traced back to their asserting the existence of an eventuality of the type assumed to have agents without making any claim whatsoever about its agent

1 Proa r b as a heimian indefinite

Both Condoravdi (1989) and K i m (1991) assume the existence of proarb and independently propose that it is a heimian indefinite I will dub this predominant analysis of proarb 'the Heimian Indefinite Analysis' (HIA).2

Heimian indefinites contribute to L F a variable that is bound b y other operators in the sentence (Heim 1982) The variable is bound by the implicit generic operator 'Gn' in (3b) and existentially b y default in (3d) This explains

w h y the quantificational force of the indefinite varies from (3a) to (3c)

(3) a A Spaniard likes coffee ~ Most Spaniards like coffee

b LF: Gnx [ χ is a Spaniard ] [ χ likes coffee ]3

c A Spaniard has just arrived ~ There is a Spaniard that has just arrived

d LF: [ χ is a Spaniard ] [ χ arrived ]

The H I A proposes that proarb also contributes a variable to LF I take (4) to be a fair explicit rendition of the H I A M u c h as in the system in H e i m (1982) for regular indefinites, only the index on proarb gets interpreted T h e 'arb' -index is

2 HIA refers to the analysis of proarb as a heimian indefinite, not to Heim's analysis of indefinites Cinque (1988:fh29) attributes the H I A to Luigi Rizzi

ARBITRARY PRONOUNS ARE NOT THAT INDEFINITE 3

interpreted as a variable (4i) (4ii) assures that the denotation of the variable is

human Since proarb is taken to be an indefinite, (4iii) assures that the 'arb'

index does not represent a current discourse referent.4

(4) Let 'arb' be a distinguished index, restricted to null elements

(i) [[proarb]]gF = g(arb)

(ii) g(arb) is defined iff g(arb) is h u m a n

(iii) arb D o m (F)

T h e H I A is quite a natural analysis for proarb, given that this element

apparently displays the well-attested quantificational variability effects of

indefinites illustrated in (3) The Spanish examples in (5) show that proarb has

existential force in episodic sentences and quasi-universal force in generic

ones Similar contrasts are shown in Condoravdi (1989) for M o d e r n Greek and

in Chierchia (1995) for Italian The H I A explains the existential (5a) and

generic (5b) readings of proarb right away: according to (4i) proarb contributes a

variable to LF A familiar default existential quantifier binds this variable in

the L F in (5b) and a quasi-universal implicit quantifier 'Gn' in (5d)

(5) a Proarb llaman por teléfono

proarb call-3pl by phone

" S o m e b o d y is calling on the phone"

b LF: [ χ is h u m a n ] [ χ is calling on the phone ]

c E n América, normalmente cuando proarb llaman por

in America usually w h e n proarb call-3PL by

teléfono, proarb disparan a la gente

phone, proarb shoot-3PL to the people

"In America, most people shoot around w h e n they m a k e phone

calls"

d Gns , x [ x makes a phone call in s & s is in America ] [ χ shoots

around in s ]

2 Empirical problems for the HIA

T h e H I A is an appealing analysis in that it unifies the existential and generic

readings of proarb under the single assumption that it contributes a variable to

LF Nevertheless, I will argue against it on the basis that both the generic and

existential readings of proarb differ from those of regular indefinites in several

respects

4 Notation: "φ, variable assignment g and file F, [[Φ]] gF = d e f the extension of φ w.r.t F See

4 LUIS ALONSO-OVALLE

T o start with, in view of (3), if the H I A were right, w e would expect proarb

to have existential readings in the episodic counterparts of generic sentences This is not always so T h e Italian examples in (6) show that second person pro5

has quasi-universal force in generic sentences, but it is a regular indexical in episodic ones:

(6) a In questo ristorante proarb mangi bene (e proarb paghi poco)

In this restaurant proarb eat:2s well and proarb pay:2s few

'In this restaurant you (generic) eat well and don't pay a lot"

b In questo ristorante pro hai mangiato bene (e hai

In this restaurant pro A u x eat:PASTPART well (and A U X

introduces from getting bound b y the adverbial quantifier

(7) a (En las fiestas de ese departamento) raras veces u n estudiante

In the parties of that department few times a student

bebe vino

drink:2s wine

"Tn the parties of that department, few students drink wine"

b (En las fiestas de ese departamento) raras veces proarb bebes vino

In the parties of that department few times proarb drink:2s wine 'Tn the parties of that department, people rarely drink wine" (Not: "In the parties of that department few people drink wine".)

E v e n the very existential reading of proarb differs from the existential readings of ordinary indefinites For one, the former is restricted to non-derived subjects (Jaeggli 1986, Cinque 1988):

ARBITRARY PRONOUNS ARE NOT THAT INDEFINITE 5

(8) a Sp: Pro están siendo golpeados

pro be:3PL beingbeaten

"They are being beaten"

[Not: "somebody is being beaten"]

b It: Pro sono venuti a vedere

pro be:3PL came:PASTPART to see

"They have c o m e to see"

[Not: "somebody has c o m e to see".]

" S o m e b o d y is (/they are) not knocking at the door"

Ø$ / * $ Ø

" S o m e b o d y is (/they are) knocking at each door"

V3/*3V

" M a y b e somebody is (/they are) knocking at the door"

03 / *30

crosslinguistically difficult to get in the absence of an explicit restrictor-like expression (Jónsson 1992, Casielles 1993), quite often a pseudo-locative in the left periphery T h e 'arbitrary' reading is m u c h harder to get in (10) than in (1c),

if not impossible at all

" Y o u (addressee) usually work like a slave"

6 LUIS ALONSO-OVALLE

to the previous one T h e prediction, however, as (11b) illustrates, is not borne out.6

" S o m e b o d y is (/They are) knocking at the door H e is (/They are) asking for help"

In conclusion, the H I A , as it stands, cannot capture the w a y s in which the

supplemented with a series of specific ad hoc principles That m o v e would

which constitutes, to m y view, its main interest

In what follows, I s h o w that there is no need to postulate the existence of

Revisiting the semantics of regular pro and appealing to the interaction of pro and the semantics of the sentences in which it occurs can just do the job This

m o v e still retains the aim to generality that characterizes the H I A : I reduce the 'arbitrariness' to the interaction of the independently needed definite or indexical readings of pro with the independently needed semantics of generic sentences

3 The Indexical-Definite vs Generic Type

I start b y distinguishing two types of arbitrary readings: the generic type,

"In the parties of that department, people rarely drink wine"

"In the parties of that department, people rarely drink wine"

6 A s one a n o n y m o u s reviewer points out, most generalizations about proarb, s e e m to carry over

to unstressed English 'arbitrary' they and you I will be cautious and avoid hasty

extrapolations English unstressed 'arbitrary' pronouns merit further attention b y themselves

Just to point one difference, English 'arbitrary' they disallows the anaphoric pattern

exemplified in (11) (Barbara Partee, p.c.)

7 For a discussion of the alleged arbitrary readings of 1 st person plural pro, see Alonso-Ovalle

ARBITRARY PRONOUNS ARE N O T THAT INDEFINITE 7

(14) Proarb mataron a Juan

Proarb kill:3PLPAST to Juan

" S o m e b o d y killed John"

T h e generic reading is associated with generic sentences w h o s e episodic counterparts license either indexical, in the case of pro2 (15), or regular definite readings, in the case of pro3 (16)

(15) (En las fiestas de ese departamento raras veces pro bebiste vino

In the parties of that department few times pro drink:2sPAST wine

"In the parties of that department, you (addressee) rarely drank wine" (16) (En las fiestas de ese departamento) raras veces pro bebieron vino

In the parties of that department few times pro drink:3PLPAST wine

"In the parties of that department, they rarely drank wine"

T h e existential type is associated with episodic sentences and, as (8) illustrates,

is restricted to non-derived subjects In section 4, I will s h o w that there is no need to appeal to proarb to derive the existential readings of examples like (14) Next, I derive the generic readings illustrated in (12) and (13) from the regular indexical and definite readings of pro illustrated in (15) and (16)

Deriving the readings of examples like (12) and (13) from those of examples like (15) and (16) requires non-trivial ontological commitments Alonso-Ovalle (2000) shows that, instead of worlds, our model must contain

parts of worlds or situations I assume that for each part or situation s there is a

unique maximal situation that s is part of: its world (Kratzer 1989)

This m o v e alone allows the following assumptions First, the situation of

utterance s0 is just a distinguished situation Second, w e can use a partial function that retrieves from a situation s the individuals, if any, that are part

of s B y stipulation, (s0) = {cs, ca}.8 Third, L F s are interpreted with respect to

a reference situation (s/r), an index (s) and a variable assignment g.9 O u r Truth definition says that an L F φ is uttered truthfully b y cs to ca in S0 if and only if its denotation yields the True w h e n evaluated with respect to a supersituation of

s0

These assumptions allow us, in turn, to devise the lexical entries for pro2

and pro3 in (17) Unlike its overt counterpart tú, which always denotes ca, pro2 picks up an individual in s/r (provided she is not the speaker of s/r, if there is one) Pro3 picks up a plurality, provided that it belongs to s but not to s0

8 Notation: cs =d e f the speaker in s0, ca =d e f the hearer in s0

W h e r e F ranges over the set of functions f: (De) → De

c [[pro31]]s/r's,g = g (1), g is only defined if g(l) is a plurality and

g ( l ) e ( ( s ) - (so))

B y using the lexical entries in (17) w e can derive the indexical and definite readings of pro2 and pro3 in episodic sentences, exemplified in (18a) and (18b), respectively

(19) PASTe p i s o d i c Tú/Pro2/Pro3 drink

Then add the following lexical entry to our inventory:

(20) [[PAST_episodic φ ]]s/r,s = [[φ]]s0,s<s0

T h e reader can then verify that our Truth definition plus the semantics for tú

correctly predicts an utterance of (18a) to be true in s0 if and only if ca drank in

s, and s < s0 W e correctly predict an utterance of (18a) with pro2 to be

truth-functionally equivalent to one containing tú (18b) is to be true in s0 if and only

if a plurality of people which is in s<So and does not contain cs nor ca, drinks in

s

W e can also correctly derive the generic readings associated with generic sentences containing pro2 and pro3:

ARBITRARY PRONOUNS ARE NOT THAT INDEFINITE 9

(21) a E n ese restaurante, proarb trabajas c o m o u n esclavo

In that restaurant proarb work:2s like a slave

"In that restaurant, you (generic) w o r k like a slave"

b E n ese restaurante, proarb trabajan c o m o esclavos

In that restaurant proarb work:3PL like slaves

"In that restaurant, they (generic) w o r k like slaves"

For that purpose, it is assumed (i) that a proposition is a set of situations

(Kratzer 1989) and (ii) that adverbial quantification amounts to relating t w o

propositions (von Fintel 1995) Propositions in the nuclear scope are

interpreted with respect to situations of the type defined b y the restrictor T h e

schematic L F in (22) corresponds to the examples in (21)

(22) G ns [in that restaurant (s) ] [pro2/pro 3 w o r k like a slave (s)]

I take an utterance of a generic L F to be true if and only if most situations s that

are minimal prototypical situations in the restrictor are extended to situations s'

that belong to the set of situations of the type in the restrictor and in the nuclear

T h e reader can n o w verify that an utterance of (21a) is predicted to b e true if

and only if most prototypical situations of this restaurant s can be extended to

situations s' that m a k e true the proposition that a person in s works like a

slave.13 A variant of (21a) containing tú instead of pro2 is true if and only if

11 Minimal situations raise serious ontological concerns, but I have n o better alternative to offer

at this point

12 Notation: m i n ({s:(j)(s)}) =d e f the set of minimal φ-situations, s/r in the nuclear scope is

restricted to b e a situation of the type defined b y the restrictor (notational preference: s Î

[[(φ]]/r =d e f any s such that [[(p]](s) = 1) It remains to define what counts as a prototypical

situation w.r.t to a situation s, where w e have n o description whatsoever of s

13 T h e sense of extendability that I use here is Kratzer's mereological relation between

situations (Kratzer 1989) It is important to remark that this relation is not equivalent to

temporal extension Temporal extension entails mereological 'extension', but not the other w a y

10 LUIS ALONSO-OVALLE

most prototypical situations of that restaurant s can be extended to situations s' that m a k e true the proposition that ca works like a slave Finally, (21b) is true iff most prototypical situations of that restaurant s are such that a plurality that

is part of the participants in s (which does not include cs nor ca) works as a slave in s T h e predictions are correct

Note that, given this semantics, even if you cannot fix the s/r parameter, you

can understand the proposition expressed b y a sentence containing tú, unlike one containing pro2 N o w , let someone find a note saying either (24a) or (24b):

(24) a Aquí tú no pintas nada

Here you not are considered at all

"Here, you are not considered"

b Aquí pro no pintas nada

Here pro not are considered at all

"Here, you are not considered"

Both (24a) and (24b) are ambiguous between a generic and episodic reading

Given the denotation of tú, under any reading, the person that finds the note

saying (24a) will take the proposition expressed b y the note to be one about the addressee of the note However, given the denotation of pro2, if she find (24b) she will only feel insulted under the episodic LF

Note also that our semantics correctly predicts that (25 a) is a contradiction, while (25b) is a contingency

(25) a Aquí tú pintas m u c h o y (tú) no pintas nada

Here you are considered a lot and you not considered at all

"Here, they respect you and they don't"

b Aquí pro pintas m u c h o y pro no pintas nada

Here pro are considered a lot and pro not considered at all

"Here, they respect you and they don't"

In sum, the previous discussion shows that the generic readings of pro can

be derived from the interaction of its independently needed indexical or

accounting for exceptions A s an anonymous reviewer points out, In that restaurant, someone

is usually working like a slave entails (21a), but (21a) does not entail it T h e truth conditions

offered above capture the entailments provided w e understand that temporal extension asymmetrically implies mereological extension and that not every situation in the restaurant is

a prototypical one If (i) is true, it is true that m o s t situations of this restaurant s can be temporally extended to a situation s' that m a k e s true that a person in s works like a slave This

m e a n s that, in fact, most prototypical situations s can be extended to situations s' that m a k e

definite readings with the independently needed semantics for generic

sentences There is no need to appeal to proarb to derive the 'arbitrariness' of

examples like those under (21) Next, I s h o w that neither is it needed to

account for the existential readings of examples like (14) or (1d)

4 Underspecified Eventualities

Consider the truth-conditions of (1d), repeated here as (26)

(26) a Proarb llaman a la puerta

proarb knock:3PL to the door

" S o m e b o d y is knocking at the door"

b L F : $ e [knock-at-the-door (e)]

T h e schematic L F in (26b) reflects the fact that this particular instance of

'proarb' m a k e s n o semantic contribution to (26a) It is clear that (26a) is true if

and only if there is an eventuality that can b e truthfully described as an

eventuality of knocking at the door (running at so)

There are different w a y s to arrive compositionally at these truth-conditions

Dropping the external argument is one Kratzer (1996) presents a most

attractive rendition of the widespread assumption that external arguments are

independent of their verbs According to her, external arguments are introduced

b y a functional head called 'VoicePhrase' (VoiceP) that is interpreted as a

dyadic predicate, satisfied b y pairs of individuals and events if the individual

fulfills the description of the predicate with respect to the event Basically, V P s

express properties of events VoicePs can further specify these properties b y

overtly expressing thematic relations.14 VoiceP contributes the underlined

portion of the schematic L F in (27b) T h e portion specifies the agent of the

indefinite description of an eventuality that (27a) conveys

(27) a Los bomberos llaman a la puerta

T h e firemen knock:3PL to the door

"The firemen are knocking at the door"

b L F : B e [Agent (e, the firemen) & knock-at-the-door (e)]

It is possible to define an operation 'Drop Voice' ( D V ) that optionally deletes

VoiceP w h e n headed b y 'agent'.15 If D V were defined only w h e n headed b y

14 In order to be sure that the V P and VoiceP express a property of the same event, Kratzer

resorts to a variety of a regular predicate conjunction principle ('Event Identification')

15 For all I k n o w , it would also be possible for an agent headed VoiceP not to have been

12 LUIS ALONSO-OVALLE

'agent', then, as expected, neither subjects of passives nor those of ergative verbs would license 'arbitrary' readings T h e examples under (8), repeated below as (28), show that this is actually the case with the existential uses of

'prOa r b'

(28) a Sp: Pro están siendo golpeados

Pro be:3PL being beaten

"They are being beaten"

[Not: "somebody is being beaten"]

b It: Pro sono venuti a vedere

Pro be:3PL came:PASTPART to

see-"They have c o m e to see"

[Not: "somebody has c o m e to see".]

Interestingly enough, w h e n w e can determine that an agent headed Phrase is present, 'arbitrariness' is blocked In (29) the agent-oriented adverb forces the presence of such a phrase and, as expected, the 'arbitrary' reading is blocked

Pro knock:3P1 to the door reluctantly

"They (definite) are reluctantly knocking at the door"

The same blocking effect obtains with other types of rationale/volitional complements and agent-oriented free adjuncts:

(30) a Pro llaman a la puerta para P R O sorprenderme

Pro knock:3PL at the door to P R O surprise m e

"They (definite) are knocking at the door to give m e a surprise"

b Pro están arreglando el teléfono para P R O cobrar el seguro Pro be:3PL repairing the phone to P R O collect the insurance

"They (definite) are repairing the phone to collect the insurance"

c Pro entusiasmados por tal tarea, pro arreglaron el teléfono

pro delighted by such task, pro repair:3PLPAST the phone

"Delighted with such a task, they (definite) repaired the phone"

examples like (26) Still, (26) can be paraphrased b y saying that someone is knocking at the door H o w come? If all sentences like (28) denote indefinite

A R B I T R A R Y P R O N O U N S A R E N O T T H A T INDEFINITE 13

descriptions of eventualities, then w e can capture their indefinite-like flavor as

an entailment of their truth-conditions Since it can be s h o w n that the

entailment survives different illocutionary forces, it is just the kind of semantic

entailment one wants to specify as a meaning postulate:

(31) M e a n i n g Postulate: Ve[activity' (e) ® 3x[agent'(x,e)]]

Then, provided that knockings at the door are taken to be activities, it follows

that they have agents

(32) Ve[knock-at-the-door'(e) ® activity' (e)]

$ e[knock-at-the-door'(e)] ® $x[agent'(x,e)]]

If there is s o m e knocking going on, w e are entitled to conclude that there is

s o m e agent responsible for it Activities, after all, happen to have agents

5 Conclusions

T h e semantic import of the 'arb'-indexing is unclear T h e H I A assumes its

existence and unifies the existential and generic readings of the so-called

'arbitrary' examples b y assuming that proarb is a heimian indefinite U n d e r the

H I A , however, there is no explicit connection between the definite and

indexical readings of pro and proarb·

T h e m a i n issue at stake is that the H I A predicts a total interpretive

parallelism between proarb and regular indefinites In section 2 I have s h o w n

that the parallelism does not hold This casts doubts on the validity of the H I A

In sections 3 and 4 I have s h o w n that there is no need to appeal to proarb to

derive neither the generic nor the existential readings of 'arbitrary' examples

In section 3 I have s h o w n that the generic readings of the 'arbitrary' examples

can be derived from the interaction of the regular definite/indexical readings of

pro with generic quantification Section 4 shows that there is no need to appeal

to proarb to capture the existential readings of the 'arbitrary' examples T h e y

can just be seen as inferences m a d e o n the basis of indefinite descriptions of

activities that m a k e no claim about their agents

Capturing the so-called 'arbitrary' interpretations w a s the m a i n motivation

for positing an 'arb'-indexing I have s h o w n that such interpretations can be

captured without resorting to it T o the extent that this enterprise is successful,

the 'arb'-indexing is not semantically motivated It remains to be seen whether

it still has a place in the grammar

14 LUIS ALONSO-OVALLE

References

Alonso-Ovalle, Luis, 2000 "Is the 'arbitrary interpretation' a semantic

epiphenomenon?" UMOP 23, ed by K Kusumoto & E Villalta, 155-183

Amherst, M A : G L S A

Casielles, Eugenia, 1993 "Aspect and arbitrary interpretation" Functional

Projections UMass occasional papers in linguistics 17, 49-61

Cinque, Guglielmo, 1988 " O n si constructions and the theory of arb"

Linguistic Inquiry 19, 521-582

Condoravdi, Cleo, 1989 "Indefinite and generic pronouns" Proceedings of the

Eight West Coast Conference on Formal Linguistics, ed by E Jane Fee &

K Hunt, 71-85 Stanford:CSLL

Chierchia, Gennaro 1984 Topics in the syntax and semantics of infinitives and

gerunds Amherst, M A : G L S A

Chierchia, Gennaro, 1995 "The variability of impersonal subjects."

Quantification in natural languages, ed by E m m o n Bach et al., 107-143

Dordrecht: Kluwer Academic Press

Chomsky, N o a m , 1980 " O n Binding" Linguistic Inquiry 11, 1-46

Chomsky, N o a m & Lasnik, H 1977 "Filters and control" Linguistic Inquiry

Heim, Irene 1998 "Anaphora and semantic interpretation: a reinterpretation of

Reinhart's approach" MITWPL 25, ed by U Sauerland and O Perçus,

Kim, Boome, 1991 "An indefinite analysis of proarb" MITWPL 14, 147-163

Kratzer, Angelika, 1989 "An investigation of the lumps of thought"

Linguistics and Philosophy 12, 607-653

Kratzer, Angelika, 1996 "Severing the external argument from its verb"

Phrase structure and the lexicon, ed by J Rooryck et al London: Kluwer

Academic Publishers

Suñer, Margarita, 1983 "Pro arb" Linguistic Inquiry 14, 188-191

A s is well k n o w n , wh-structures differ across languages in several respects

I will concentrate on 3 aspects where variation is observed: (i) w h - m o v e m e n t , (ii) inversion and (iii) complementizers Confirming Greenberg's insight that variation is co-variation, m y approach aims at giving not only a unified account

of variation in wh-questions across languages but also at unifying mirror

effects in wh-questions vs wh-exclamatives, such as: (i) the possibility of in

situ wh-questions as opposed to exclamatives; (ii) obligatory or optional

inversion in questions vs lack of obligatory inversion in exclamatives; (iii) distribution of complementizers in both constructions

I will address the following main questions: (i) W h y is overt w h - m o v e m e n t obligatory, forbidden or apparently optional in s o m e languages but not in

others? (ii) Are in situ whquestions free variants of questions with fronted w h

-phrases or does their interpretation differ? (iii) W h y is inversion in wh-contexts obligatory, forbidden or apparently optional? (iv) W h y does the comple­mentizer appear in wh-structures in s o m e languages but not in others? (v) W h y

is there absence vs presence of root-embedded asymmetries? (vi) H o w do these p h e n o m e n a correlate? A full answer clearly involves a large w o r k w h o s e detailed presentation is precluded because of space limitations I will just outline the main guidelines of this program of research I will systematically

1 This work has been presented at different stages of its development at Elements Minimaux de

la Variation Linguistique, 1997, 2000 Paris; Going Romance 1999, Leiden; I G G 2000, R o m e ; Going Romance 2000, Utrecht and I G G 2001, Trieste I thank the audiences and organizers of those meetings for their comments and m y Faculty, through its Directive Committee, for financial support Special thanks go to Reineke Bok-Bennema and Carmen Sorin

1997, Rizzi 1997, K a y n e & Pollock 1998, Pollock et alii 1999, a.o.) M y approach will be of the second type

A s proposed in previous works I will assume the following structure for the so-called C P field2:

(i)XP[ EvaluativeP [Evaluative' [AssertiveP [Assertive' [XP [ w h P [Wh' [FocusP LFocus' [XP[IP

(ii) X P is a landing site for dislocated elements (see section 3.4 and fn 21); (iii) W h P is an operator projection where wh-phrases m o v e to; its head has two features3: wh and V-features; w h - m o v e m e n t is triggered b y the need of wh-

feature checking; Verb-Inflection (henceforth V-I) raising is triggered b y the need of V-feature checking; in embedded contexts lexical selection allows the V-feature of the W h - h e a d to be legitimated;

(iv) Wh-phrases have a complex internal structure4: there is a scale of referentiality in wh-phrases, due to the existence of semantic features that

restrict their domain of reference -from the barest wh-phrase que, [que [e]N'-r],

2 The notion of discourse-oriented language has been seen in different ways since at least Li &

T h o m p s o n (1976) In previous works I claimed the projections I assume to exist in the C P field are C o m m o n Ground and U(niverse) of D(iscourse) related (Heim 1982, Calabrese 1985, a.o.) EvaluativeP and AssertiveP are Ground; Topic/Focus involve rather U D A language can be discourse-oriented regarding either Ground or U D or both

3 In an earlier version wh-movement was triggered b y the need for a N+ w h feature checking Reviewer Β observes: "According to most theories, wh-movement is triggered by the need to check the feature + Q ; N-feature checking, being an L-related feature, is not involved in w h -checking This is mainly due to the fact that other categories besides Ν are subject to w h -

movement, such as adverbs and PPs." Let us then assume that whatever 'wh' means it has to be

checked b y a wh-element and this is part of + Q checking Here, however, the + Q feature is associated not only with + w h , but also with a V-element - a natural assumption if w e consider that operators are V-related as proposed by Rouveret (1996)

4 For a syntactic proposal see A m b a r (1988), A m b a r & Veloso (1999) and DiSciullo (1999,

WH-QUESTIONS AND WH-EXCLAMATIVES 17

where [r]=[restricted], to wh-phrases of the form [que N]-> que livro "what

book", with an intermediate degree of referentiality given b y wh-phrases that

are m o r e restricted than que, but less restricted than [que N ] : [que [e]N'+r], [+r]

= + h u m a n > quem "who" In languages of the French and E P type w h

-structures are sensitive to this scale of referentiality; only wh-phrases

sufficiently restricted can check the wh-feature; the barest wh-phrase que can

do it iff it is in a spec-head relation with V-I - this is w h y que cannot enter in

situ wh-questions;

(iv) AssertiveP is a projection located above W h P - it projects whenever assertive properties are involved in the construction5 W e will see that AssertiveP provides a unifying link in the treatment of wh-structures that in a sense have a 'factive' interpretation, namely wh-in-situ, fronted wh-questions without inversion, i.e questions that lack a full blown interrogative reading, embedded wh-questions and wh-exclamatives Only semantically restricted wh-phrases (que Ν and que e+r) can check Assertive; que being [-r] cannot; this is w h y que cannot enter root and embedded wh-questions without

inversion or wh-exclamatives Languages differ in their capacity of moving wh-phrases or remnant IP to that projection, due to the ¿definite status of their wh-phrases (fronted non-inverted wh-questions) or to the properties of their inflection systems (wh-in-situ), respectively E P , B P , French and T e t u m can do

it, Hungarian not However, Hungarian allows checking of Assertive b y merge

of a complementizer (hogy) This hypothesis is supported b y other related

phenomena: a) Hungarian 'interrogatives of confirmation' and exclamatives

with complementizer hogy; b) lack of wh-in situ and wh-Subject-V order in

Hungarian vs their availability in E P , B P , French and Tetum;

(v) EvaluativeP6 is a projection located above AssertiveP, which codifies the speaker's evaluations (usually expressed b y adjectival elements) Only semantically restricted wh-phrases to which an evaluative (adjectival) element

is implicitly or explicitly associated can check the evaluative feature; que

cannot In whexclamatives both EvaluativeP and AssertiveP project: w h exclamatives differ then from wh-questions without inversion in that in exclamatives the wh-element m o v e s higher in the structure to check the [¿evaluative] feature, after having checked the assertive feature (that codifies the determined/factive status of exclamatives - cf G r i m s h a w 1977);

-(vi) I will assume that all features are strong and that m o v e m e n t is always visible, m u c h in terms of K a y n e (1998) and K a y n e & Pollock (1998)

5 Searle (1979) defines assertive as the illocutionary act "where w e tell our hearers (truly or falsely) h o w things are" Another possible label for this projection would be ReferentialP

18 MANUELA A M B A R

A s w e will see the analysis suggests a possible unifying link a m o n g the structures and the languages described: languages where AssertiveP is prominent will display a tendency for moving constituents to that projection,

i.e for having wh-in-situ, wh-structures without inversion and complementizer

forms, which lexicalize the head of AssertiveP (BP and Tetum); languages

where AssertiveP is not prominent will exhibit rather absence of wh-in-situ,

obligatory inversion (Hungarian); mixed languages will have c o m m o n behaviors to both types of languages (French and E P )

Let us then take the 3 aspects w e will concentrate on: wh- movem ent , inversion and complementizer - and consider each one in turn, applied first to wh-questions and then to wh-exclamatives

2 Wh-movement - Wh-questions

2.1 Typology

In root whquestions 4 types of languages can be considered, w.r.t w h movement:

-(i) languages where wh-phrases obligatorily occur in situ, i.e languages that

(apparently) lack overt wh-movement, like Chinese:

(1) a Hufei mai-le shenme (ne)?

Hufei buy-Perf what P R T

" W h a t did Hufei buy?"

b * S h e n m e Hufei mai-le

c * S h e n m e mai-le Hufei

(ii) languages where wh-phrases cannot occur in situ, like Hungarian7:

(2) a *János megvett mit?

John part-bought 3/Sg what Acc

" W h a t did John buy?"

b Mit vett m e g János?

what Acc bought 3/Sg part John

" W h a t did John buy?"

7 In Kiss (1993:99) an in situ multiple wh-question (highly marked) is attested I will not deal

WH-QUESTIONS AND WH-EXCLAMATIVES 19

(iii) languages that allow for wh-in-situ just in case one wh-phrase has been

moved, as English:

(3) a W h o bought what?

b * John bought what?

c W h a t did John buy?

(iv) mixed languages, i.e languages that allow both constructions (e.g French,

E P , B P , Tetum) S o m e (micro)variation is however at stake: B P and Tetum allow overt wh-movement, but, contrary to E P , wh-in-situ are m o r e frequent (Lopes Rossi 1996 for B P ) :

(4) O João foi onde? EP

John went where

(5) Jean a vu qui? French

2.2 Wh-in-situ - different strategies

There is abundant literature on wh-in-situ questions since at least the 70's; one could class it in two main views: (i) the wh-element is in its original position (Baker 1970); (ii) the wh-phrase undergoes L F m o v e m e n t (Chomsky

1977, H u a n g 1982, etc)8

Languages of the French and E P type can choose one of two strategies to construct a wh-question: (i) by fronting the wh-phrase or (ii) b y leaving it in its base position Theoretically this optionality poses a problem for the minimalist program - the system disallows optional operations9 Empirically the question

is to understand h o w those two options interpretatively differ

8 For Pesetsky (1987) D-linked phrases do not undergo L F movement; non-D linked phrases do

wh-9 Recent analyses tried to overcome this problem: Boscovic's (2000) L F insertion of C°, Cheng

& Rooryck's (2000) proposal of an underspecified intonational morpheme, Simpson's (1999)

20 MANUELA AMBAR

2.3 Wh-in-situ - the claim

For the so-called wh-in-situ I assume that: (i) wh-phrases always raise10 to the same projection ( W h P ) ; (ii) licensing applies uniformly in fronted and in

wh-in-situ questions; (ii) the in situ effect is the result of ulterior remnant IP

m o v e m e n t to AssertiveP; (iii) elements that appear at the right of the wh-phrase have been previously topicalized (or focalized)

(8) and (9) illustrate h o w the system proceeds, without and with topicalization, respectively:

(8) a O Pedro encontrou q u e m ?

Peter met w h o

b (i) X P [AssertiveP [Assertive' [XP [whP q u e m i [wh' [FOCUSP t i [FOCUS' [XP

[IP o Pedro encontrou ti ] ] ] ] ] ] ] ] ] (ii) X P [

AssertiveP

[O Pedro encontrou ti ] k [Assertive' [XP [whP quenii

[wh' [FocusP t i [FOCUS' [ X P [IP tk ] ] ] ] ] ] ] ] ]

(9) a O Pedro ofereceu o quê à A n a ?

Peter offered what to A n a

b (i) X P [AssertiveP [Assertive' [ X P [whP O quêi [wh' [FocusP t i [FOCUS' [τορΡ à

A n a m [Tοp' [IP o Pedro ofereceu ti tm ] ] ] ] ] ] ] ] ] ]

(ii) X P [AssertiveP [ O P e d r o Ofereceu ti tm ] k [ Assertive' [ X P [WhP O

quêi [wh' [FOCUSP ti [FOCUS' [TOPP à A n a m [Top' [IP tk ] ] ] ] ] ] ] ] ] ] The wh-phrase m o v e s to spec,WhP to check the wh-feature on its head At this point an option would be to m o v e V-I to check the V-feature, but then w e would obtain a fronted wh-question In wh-in-situ another operation is available: in the numeration IP, more precisely its head, has a [+assertive] feature Thus, AssertiveP projects and the remnant IP m o v e s to its spec to check [+assertive] H o w is the V-feature checked? I will assume that the projection of a higher domain creates a configuration of the lexical selection type and allows licensing of the V-feature on the head of the lower domain, just as in embedded questions11 In 1 (iv) w e stated that only wh-phrases

Earlier versions of this proposal were presented at Elements minimaux de la variation

linguistique, Paris, 1997-2000

11 Reviewer Β notes: " E P and French differ from Chinese in that embedded wh-in-situ is impossible in the former though not in the latter" and wonders h o w this fact is derived in m y analysis Portuguese does allow embedded wh-in-situ, selected by s o m e verbs and contrasts with French in this respect (cf Cheng & Rooryck 2000, fn 16) If AssertiveP projects in

WH-QUESTIONS AND WH-EXCLAMATIVES 21

sufficiently restricted could check the whfeature; thus w e predict that in w h

-in-situ whenever the wh-phrase m o v e d to spec, W h P is que (the [-r] wh-phrase) the sentence will be ill-formed; this is a good prediction since que cannot occur

in wh-in-situ questions as in (10)12:

(10) a A Mariaviu *quê /o quê / q u e m / q u e livro?

M a r y saw *what / the what / w h o / what book

b Marie a vu *que / quoi / qui / quel livre?

M a r y saw *what / what / w h o / what book

In order to be able to check the wh-feature, que needs to be in a spec-head

relation with I; this would be the case of a fronted wh-question, but not of a wh-in-situ, where V-I does not raise to W h P , rather it remains inside IP to ensure checking of the assertive feature w h e n remnant IP is moved T h e

prediction that que cannot occur in wh-in-situ questions is derived in a unified

way

In this view then the difference between a fronted wh-question and a called wh-in-situ resides in the presence of one operation more in the latter: remnant IP movement, optionally with previous topicalization This analysis makes a prediction concerning acquisition: wh-in-situ have more steps than fronted wh-questions, therefore the former should be acquired later T h e

so-that allows it in root has to be attributed to other factors, plausibly to properties of its inflection system, namely to its ±pro-drop status - Ambar, Lois & Obenauer (1986), developing Kayne's (1982) ideas, claimed that V-I cannot end up in an embedded C o m p in a non-null subject language I would like to pursue this line of reasoning A second question concerns the licensing mechanism of the V-feature on W h P He/she notes: "in Rizzi's theory selection by V licenses the embedded [+Q] This assumption is reasonable, given that [+Q] is a semantic feature selected by particular verbs" Recall I a m assuming that both the w h - and the V-features are subparts of + Q Furthermore, in embedded and in wh-in-situ the higher domains are verbal

in nature - the matrix verb in the former and Assertive in the latter (what licenses Assertive is the head of IP, V-I, plausibly the Event part of the Verb) Although I cannot pursue this question here, m y idea is that Assertive selects and is licensed by an Event related element - V-I or V are candidates Thus, assuming that Infl inherits Event specifications from the Verb,

in a language having Infl (the IP head) but no V-I raising, wh-in-situ will be excluded, e.g English In languages lacking Infl morphology the Verb takes over the role of Infl, e.g Tetum, Chinese This is a strong hypothesis that will be confirmed or infirmed b y future w o r k on other languages

12 O quê and quoi in (10) are the tonic forms of E P o que and French que, respectively, a fact

probably related with the structure w e are proposing Research in the prosody field is required

to confirm this In its tonic form quê, que can enter wh-in-situ if it is preceded by a preposition

- for our purposes here it is enough to assume that the preposition s o m e h o w changes the

In other terms, in (12) the speaker knows that John bought something, he wants

to k n o w what This is w h y for some speakers there is a clear contrast between

(11b) vs (12b); (12b) is not a felicitous answer to (12a)14:

(11) a Q u e comprou o Jộo?

what bought John

b N a d a

nothing

(12) a O Jỗo comprou o quê?

John bought what

b ??Nada

(Ambar et alii 1998, Chang 1997, apud Cheng & Rooryck 2000) Let us say that these structures are associated with a stronger presupposition, given b y C o m m o n Ground This strong presupposition is translated in our system b y checking of AssertiveP, just as in other echo-like structures, e.g fronted wh-questions without inversion, as w e will see Further evidence for (8)-(9) is provided by the prosody of wh-in-situ, in two ways: (i) the intonation

of wh-in-situ differs from visibly fronted wh-questions, a fact that in m y

analysis m a y be interpreted as being the result of remnant IP m o v e m e n t to

13 I thank M a r y Kato and Carla Soares for having informed m e that research in progress o n acquisition of wh-questions in B P and E P , respectively, confirms that the child acquires wh-in-situ questions later than fronted wh-questions

14 For s o m e speakers this is a sharp contrast (ok/*), while for others it is weaker Thus it is natural that the informants consulted by reviewer Β don't get the subtle opposition in (11)-(12)

T h e same reviewer points out: "if the judgments are indeed as described then the question arises of h o w to describe the differences between, say, E P and English." I a m not assuming that

in E P wh-in-situ are only possible in echo questions as in English This is w h y I used the

expressions 'echo-flavor' and 'stronger presupposition' to m e a n a certain degree of echo, say,

an intermediate degree of echo, not to be confused with echo For the same reason I hypothesized that full echo wh-in-situ in English (also available in E P ) have a different

derivation, where remnant IP m o v e m e n t to Assertive does not apply M y idea is that these

WH-QUESTIONS AND WH-EXCLAMATIVES 23

AssertiveP - the prosodic relation between these structures and others involving

AssertiveP being necessarily established15; (ii) topicalization in (9) also

provides empirical evidence for the analysis proposed: only apparently the

canonic Portuguese word order given b y the declarative clause (13) is

reproduced in the wh-in-situ (14):

(13) O Pedro ofereceu o livro à Ana

Peter offered the book to A n a

(14) O Pedro ofereceu o quê à A n a ?

In (13) the phonic segment o livr'à Ana has no pause between livro and à Ana,

whereas in (14) there is a pause between o quê and à Ana (*o qu 'à Ana), a fact

that is obviously related with the tonic form that the in situ wh-phrase assumes

Neither in an analysis assuming that the wh-phrase has stayed in situ in (14),

nor in the different versions of L F m o v e m e n t is this contrast accounted for In

our analysis the contrast follows straightforwardly: the order in (14) is not the

basic one à Ana is topicalized and the whphrase is in the position where w h

-phrases are generally licensed

Another empirical fact supporting m y claim is provided b y the following

contrasts:

(15) Ofereceu-lhe o quê, o Pedro / ?*ele / *alguém?

offered h i m what, Peter / he / someone

There are restrictions on topicalization of pronouns and indefinites in

Portuguese (cf K a y n e & Pollock 1998 and A m b a r & Pollock 1998) S o the

ungrammaticality in (15) is expected16

Finally, the analysis makes the desired predictions concerning clitic

placement Wh-constructions trigger proclisis in Portuguese (cf 16) as largely

discussed in the literature (Duarte 1983, Rouveret 1989, a.o.); however w h e n

the wh-phrase is post-verbal enclisis is obligatory - (17):

(16) a Q u e livro lhe ofereceste?

what book him offered

15 This will probably be the case of yes/no questions Prosodic research is needed in order to

confirm this hypothesis

16 Reviewer Β notes the informants consulted "disagree with the judgment" in (15), w.r.t ele

Actually, for s o m e speakers judgments o n topicalization of pronouns are not clear in other

24 MANUELA A M B A R

b * Q u e livro ofereceste-lhe?

(17) a * 0 Jộo lhe ofereceu o quê?

John him offered what

b O Jỗo ofereceu-lhe o quê?

I just want to observe that whatever analysis of clitics w e assume, these facts are expected in this analysis due to remnant IΡ m o v e m e n t - clitics behave as they do in root declarative sentences

Assuming this proposal for wh-in-situ, the following question is: what are the parameters responsible for the other types of languages described above? Regarding Hungarian, one possibility would be to consider that AssertiveP is not projected There is however empirical evidence to assume that it does project In Hungarian, wh-questions lacking a full-blown interrogative reading

exhibit the complementizer hogy: interrogatives of confirmation In our system they involve AssertiveP (cf section 3) Suppose hogy checks assertive b y

merge T h e question turns out to be w h y the remnant IP strategy is unavailable

to derive wh-in-situ in Hungarian T h e hypothesis I would like to explore is

that the IP head is too 'heavy', in a sense to be m a d e clearer in future work, and blocks IP m o v e m e n t ; contrarily to Chinese17, inflection in Hungarian carries different values (agreement, Tense, definite/indefinite object agreement) Finally, consider B P and Tetum Again the availability of wh-in- situ correlates with the status of Inflection: their Inflections are less 'heavy' than in Hungarian In Tetum there is no Inflection morphology (as in Chinese), and B P is loosing different inflection specifications, e.g agreement T h e diachronic work b y Lopes Rossi (1996) has s h o w n that the use of wh-in-situ in

B P has increased in the X X century, in a w a y with no parallel in E P In our system this fact is related with another change in the B P system: the loss of inflection specifications Note that this relation is not established in other accounts of wh-in-situ O n c e Inflection is the locus of different values, the

17 This would also explain w h y Iraquí wh-phrases couldn't be in situ in finite tensed domains,

but can in infinitives (Simpson 1999) - inflection is heavier in finỵtes than in infinitives T h e relevant notion seems then to be Tense Reviewer Β observes " O n the basis of morphology I see no reason to consider inflection in Hungarian heavier than in E P Moreover there are probably quite a number of languages without wh-in-situ and with light inflection - English is

a case at point Wouldn't the parameter here proposed predict these languages to have

wh-in-situ?" A s mentioned in the text, heaviness is here used "in a sense to be m a d e clearer in future

work" Nevertheless I don't say that heaviness is the only parameter involved; it is a necessary though not sufficient condition: a language can have a light Infl and be unable to have remnant

IP m o v e m e n t to Assertive due to other factors See fn 11 for one hypothesis concerning

W e will then return to wh-in-situ to understand w h y in situ wh-exclamatives

to Spec-IP (Dobrovie-Sorin 1987, 1994, Bonet 1990, Contreras 1991, Vallduví

1992, Hulk 1993, D e W i n d 1995, Drijkoningen 1997, Zubizarreta 1998, Barbosa 2000, a.o.)

The short character of this paper precludes a deep c o m m e n t to all aspects involved in one or other perspective In the absence of an algorithm to evaluate economy, I go on assuming some traditional points, namely the I-to-C analysis18 Crucially, m y main goal is to extend the study of inversion to a larger wh-empirical domain and to reach a system of articulated answers to the questions formulated in the introduction

3.2 Typology

Considering inversion in wh-questions 3 types of languages seem to exist: (i) languages where inversion is always obligatory, irrespectively of the form

of the wh-phrase, e.g Hungarian:

18 With Rizzi (1995) I assume the C P system serves as "the interface between a propositional content (expressed b y IP) and the superordinate structure (a higher clause or, possibly, the

26 MANUELA AMBAR

(18) a Mit vett m e g János?

what Acc bought 3/S g part John

"What did John buy?"

b *Mit János vett m e g / m e g vett?

(19) a Melyik könyvet olvasta el János?

which book Acc read 3/Sg part John

"Which book did John read?"

b *Melyik könyvet János olvasta el / elolvasta?

(ii) languages where inversion is impossible, e.g Tetum (maka=licensing

w h o enf part you met

" W h o did you meet?"

BP (in the unmarked case; que is not available anymore in B P - o que, [the +r]

form, replaces it)

(22) *?Que livro / *?o que / *que comprou você?

what book/ the what/ what bought you

(23) Q u e livro / o que / *que você comprou?

which book / the what / what you bought

(iii) languages where inversion coexists with absence of inversion in given contexts - EP, French:

(24) a Q u e livro / q u e m / q u e viu o Pedro?

b Quel livre / qui / qu' a vu Pierre?

what book / w h o / what bought Peter

(25) a Que livro / ?quem / *que o Pedro viu?

b Quel livre / ?qui / *que Pierre a vu?

which book/ w h o / what Peter bought

3.3 Presence of inversion - 'pure' wh-questions

The system proposed uniformly derives fronted wh-phrases too, with and without inversion Consider first the structures with adjacency between the

27

fronted wh-phrase a n d V-I: the wh-phrase m o v e s to S p e c , W h P for the s a m e

reason it does in wh-in-situ - to check the wh-feature o f W h P T h e difference

resides in the w a y the V-feature is licensed: in wh-in-situ through AssertiveP,

w h i c h codifies the property that interpretatively distinguishes wh-in-situ f r o m fronted wh-phrases - the stronger presupposition; in 'pure' fronted w h -

questions AssertiveP is not projected; therefore checking o f the V-feature triggers V - I raising to the W h P head, as illustrated in (26b) b e l o w :

(26) a Q u e comprou o Pedro / ele?

what bought Peter /he

b (XP) [whP quei [wh' comprouv [Fp [F' tv [ IP o Pedro / ele [I' tv ti ] ] ] ] ]]

N o w , que (the barest wh-phrase [-r]) is in a spec-head relation with V - I ; therefore it b e c o m e s specified a n d able to check the wh-feature; the restriction

o n que observed in wh-in-situ disappears A s illustrated in (26b) the derivation

stops in W h P ; w e get then w h a t I will call a 'pure' wh-question1 9 T h e system uniformly applies to Hungarian C o m p a r e (26b) a b o v e with (27b) b e l o w :

(27) a Mit vett m e g János?

b (XP) [whP miti [wh' vettv [FP ti [F' tv [IP m e g tv János ti ] ] ] ] ] Both in Hungarian and in Portuguese a dislocated element can precede the w h - phrase:

(28) a János mit vett m e g ?

John what bought part

b O João que comprou?

the John what bought

(29) a. [XP János j [x' [whP mit i [wh' vett k [FOCUSP t i [FOCUS' t k [IP m e g t k t j t i

]]]]]]]

b. [XP o João j [x' [whP quei [wh' comprou k [FOCUSP t i [FOCUS' t k [IP t J t k t i

]]]]]]]

19 F(ocus)P proves to be necessary to account for inversion in embedded domains in languages

of the French type I cannot discuss this aspect here For the derivation of sentences with 3 place predicates and auxiliaries see A m b a r & Pollock (1998), where an account of the different

28 MANUELA A M B A R

3.4 Absence of inversion - wh-questions ('echo flavor') and wh-exclamatives

Consider paradigm (24)-(25) In previous accounts of these contrasts I

claimed that only wh-phrases of the form [que N ] {que livro) could enter root

non-inverted wh-questions; bare wh-phrases [que e+r] (quem, onde) could

dispense with inversion in embedded contexts, but not in root ones; the barest form [que e-r] required inversion in all cases A finer analysis of the data

reveals that the echo vs non-echo-flavor interpretation of each sentence is crucial for a deeper understanding of these facts: lack of inversion in root wh- questions originates echo-like structures i.e wh-questions lacking a full-blown

interrogative reading This is w h y in the absence of inversion there is a special

intonation contour o n the sentence and o n the wh-phrase in particular, translated b y the italic in (25) above Notice, however, although most speakers

accept the non-adjacency illustrated in (25), they prefer the sentences where the wh-phrase has a phonetically realized Ν and completely exclude the barest w h - phrase from this context It seems then that the scale of grammaticality regarding absence of inversion reflects the scale of referentiality in the content

of the wh-phrase (cf l-(iv)): the m o r e the wh-phrase is restricted, the m o r e the non-inverted wh-question is grammatical20

Note that the analyses that derive obligatory adjacency between que and V-I from the clitic status of que do not account for this scale of grammaticality nor for the fact that que does not

cliticize in wh-in-situ Reviewer Β observes: "The author provides a scale of referentiality that

singles out wh-phrases of the following type: (i) que Ν (ii) que [e]+r [=quem] (iii) que [e]-r

( ) note that the feature -r cuts the hierarchy in two, grouping together (i) and (ii) and leaving

out que H o w e v e r in root clauses there is a clear divide between (i) and (ii), the former

dispenses with inversion, but the latter doesn't, at least according to the informants consulted and to A m b a r (1988)" A s for the judgments, in 1988 I noted that most speakers had observed

that que Ν phrases dispensed with inversion just in the presence of an intriguing special intonation contour The 'echo-flavor' ('flavor' meaning a w e a k degree of echo) had not been

discovered at that time (cf m 14) A finer analysis of the data reveals it does really exist, as claimed b y about one hundred speakers ( m y students included) A n y w a y I do maintain the three w a y distinction (as confirmed b y the reviewer's typology in (i)-(iii)) Actually, non-inverted structures are better with (i) than with (ii) and impossible with (iii) Note that (i) are more restricted than (ii): (i) have all the features of the noun itself while (ii) have just one

feature covering a wider range (+ h u m a n in quern) Thus, this system allows accounting for this

scale of grammaticality, a desirable result in m y opinion Note that in this view it is conceivable that languages m a y vary, depending o n their D P systems, i.e a given language

m a y require wh-phrases of the form (i) for non-inverted structures while others m a y use both (i) and (ii) (cf Pollock et alii 1998 and A m b a r et alii 1998) Reviewer Β also suggests that "a

syntactic treatment of these phrases would be more fruitful" and is curious about the status of o

WH-QUESTIONS AND WH-EXCLAMATIVES 29

I will assume that this echo-flavor is to be interpreted as the result of further wh-movement to AssertiveP combined with raising of the subject to X P2 1: (30) [AssertiveP q u e liVrOi [Assertive' [xP O JộOj [x' [whP ti [wh' C O m p r O Uv [FP [F' tv

[IP tj [ I ' tv] ] ] ] ] ] ] ] ] ]

Again only sufficiently restricted wh-phrases can check [+assertive] T h e

ungrammaticality in (25) follows: que cannot check assertive because it is not

sufficiently restricted22

W e understand n o w w h y no lexical material can intervene between the

wh-phrase and the verb in pure interrogatives, as in (31) below, contrarily to echo­

like wh-questions (32): only in the latter is there an available landing site: (31) *Que na tua opiniäo o Jộo encontrou?

what (in your opinion) the John found

(32) a Que livro / Quern na tua opiniäo o Jộo encontrou?

which book / w h o in your opinion the John met

b Onde na tua opiniäo o Jộo terá ido?

where in your opinion the John has gone

21 In previous works I assumed X P w a s a topic position; I cannot discuss its properties here; for our purposes it is enough to assume it exists Reviewer A points out: "This analysis seems plausible as the m o v e m e n t of the wh-phrase accounts for the echo-flavour, but the m o v e m e n t

of the subject is not motivated Something ought to be said about that." If this position is a topic, m o v e m e n t of the subject is triggered by the need of topic-feature checking, its optional nature (the equivalent of (32) with inversion is possible) and the possibility of having elements other than the subject being derived H o w e v e r if it is a topic w e expect indefinite expressions

to be excluded {Que livro alguém leu? which book someone read) S o m e speakers accept,

others don't This is w h y for the time being I don't give a label to this projection I profit to thank this reviewer for his/her careful corrections on the text

22 Reviewer Β notes: "one of the consequences of the theory proposed is that it assigns the same status to non-inverted questions with a bare wh-phrase [in (25)] and (12a), a wh-in-situ question However, these do not have the same status N o t according to m y sources and not according to the paper itself." I agree with the reviewer in that those structures do not have the same status In the architecture of m y system I had this in mind O n e of m y goals w a s to assign two different derivations to these two different constructions Recall that in non-inverted w h -questions the element that has an assertive feature and that is m o v e d to check Assertive is the wh-phrase, whereas in wh-in-situ it is the remnant IP This is expected to have consequences

on interpretation: in non-inverted wh-questions the speaker 'knows more' o n the content of the wh-phrase, in wh-in-situ he 'knows more' on the content of remnant IP Moreover the different status regarding grammaticality also follows: IP is a better candidate to have an assertive feature than bare wh-phrases, which are weakly restricted, as w e saw So I do not agree with

30 MANUELA A M B A R

Thus, wh-questions without inversion result from a conspiracy between two

operations: w h - m o v e m e n t to AssertiveP and m o v e m e n t of the subject to X P

O n c e the latter is not obligatory (2627) vs (28ab), w e predict that w h

-questions with inversion, but with an echo-like interpretation, m a y exist The

prediction is born out: if an echo-like intonation (and interpretation) is assigned

to the sentence, the judgments in paradigm (31)-(32) are maintained in its equivalent with inversion

Let us turn n o w to Hungarian23 W h y is adjacency between the wh-phrase and V-I always obligatory? T w o hypotheses have to be considered: (i) AssertiveP does not project; (ii) AssertiveP does project If w e choose (i) w e immediately account for lack of wh-questions without inversion and of wh-in- situ However, Hungarian has three constructions providing s o m e evidence for choice (ii): interrogatives of confirmation, exclamatives and embedded questions; all them share the 'factive' reading AssertiveP codifies and the

presence of the complementizer hogy I will assume that hogy checks

[+assertive] b y merge in the three structures W e still have to ask w h y

Hungarian chooses merge to check [+assertive] instead of moving wh-phrases

to spec, AssertiveP I will claim that wh-phrases do not m o v e to this position in Hungarian because they are not definite enough to check [+assertive] Evidence for this claim is provided by overt object agreement in Hungarian: object wh-phrases trigger indefinite agreement A problem arises however: d-

linked wh-phrases of the type melyik Ν "which N " trigger definite agreement;

w e should then predict, contrary to facts, that wh-questions without inversion are possible in this context Nevertheless I will not abandon the claim Other factors m a y conspire for the obligatory adjacency between the wh-phrase and the inflected verb in wh-questions triggering definite agreement in Hungarian

If w e look at the other groups of languages considered, w e find a coherent picture: languages that prefer, or simply allow, wh-questions without inversion (correlating with wh-in-situ) are languages whose determiner systems tend to

be m o r e definite than the Hungarian one: either they do not need visible definite/indefinite agreement24, or their wh-phrases are arguably [+defmite] Unfortunately w e cannot here present all the argumentation because of space limitations

24 Let us assume that the more the overt morphology is available, the less the system is rich e.g non-null subject languages need an overt subject, because the inflection system is poor; Hungarian has a rich morphology regarding definiteness (overt definite/indefinite agreement on the verb), its D P s will not be definite b y default; T e t u m morphology regarding definiteness is null, its D P s (wh-phrases) will be definite b y default In other terms absence of grammaticalization entails other means of supplying the relevant information, b y hypothesis through discourse For other languages exhibiting definiteness in the absence of definite

-WH-QUESTIONS AND WH-EXCLAMATIVES 31

(e.g E P , French) , or the D P system is [+defmite (specific)] by default as in the following Tetum and B P sentences:

(33) a Labarik sosa livru

boy bought/buy book

b Ha'u sosa careta

I buy/bought car

(34) a Criança sofre

child suffers

b V o c e compra carro b o m , n'e

you buy car nice n'e

labarik "boy" in (33) is interpreted as [+defmite] (specific); w e should wonder

w h y this is so The answer is far beyond the scope of this article; a working hypothesis would be to assume that in languages of this type in the absence of any definite exponent, referentiality is assigned to the D P b y discourse, probably by Ground, in a w a y not available in other languages, say, Hungarian (fh 26) Let us assume that both in Tetum and B P wh-phrases are sufficiently definite to check assertive whereas in Hungarian they are not A s for w h - phrases triggering definite agreement on the verb, one possibility is to assume that raising of the subject to X P intervening between the wh-phrase and V-I blocks the overt agreement relation necessary for the wh-phrase to be definite enough to check assertive26 In E P and French wh-phrases can be definite, therefore they can check assertive

At this point w e have seen w h y inversion is obligatory in root wh-questions

in Hungarian, w h y inversion coexists with absence of inversion in E P and French and h o w these facts correlate with the so called wh-in-situ strategy in the languages observed W e should n o w wonder w h y Tetum and B P (in the unmarked case) only have non-inverted wh-questions - the opposite case of Hungarian Curiously, their Inflection systems are poor, also contrarily to Hungarian Although I cannot pursue this question here I would like to suggest that for independent reasons there is no V-I raising to W h P in these

25 In A m b a r & Veloso (1999) it is claimed that o in o que and -/ in quel are the definite article

26 Note that this condition holding on checking of assertive m a y prove to be unnecessary for other contexts Further research o n this topic will s h o w if this is the case or suggest other