What Counts: Focus and Quantification (Linguistic Inquiry Monographs, 36)

What Counts

Linguistic Inquiry Monographs Samuel Jay Keyser, general editor 6. 10. 12. 13. 15. 16. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36.

Some Concepts and Consequences of the Theory of Government and Binding, Noam Chomsky On the Nature of Grammatical Relations, Alec Marantz Logical Form: Its Structure and Derivation, Robert May Barriers, Noam Chomsky Japanese Tone Structure, Janet Pierrehumbert and Mary Beckman Relativized Minimality, Luigi Rizzi Argument Structure, Jane Grimshaw Locality: A Theory and Some of Its Empirical Consequences, Maria Rita Manzini Inde®nites, Molly Diesing Syntax of Scope, Joseph Aoun and Yen-hui Audrey Li Morphology by Itself: Stems and In¯ectional Classes, Mark Arono¨ Thematic Structure in Syntax, Edwin Williams Indices and Identity, Robert Fiengo and Robert May The Antisymmetry of Syntax, Richard S. Kayne Unaccusativity: At the Syntax±Lexical Semantics Interface, Beth Levin and Malka Rappaport Hovav Lexico-Logical Form: A Radically Minimalist Theory, Michael Brody The Architecture of the Language Faculty, Ray Jackendo¨ Local Economy, Chris Collins Surface Structure and Interpretation, Mark Steedman Elementary Operations and Optimal Derivations, Hisatsugu Kitahara The Syntax of Non®nite Complementation: An Economy Approach, Zeljko BosÏkovic Prosody, Focus, and Word Order, Maria Luisa Zubizarreta The Dependencies of Objects, Esther Torrego Economy and Semantic Interpretation, Danny Fox What Counts: Focus and Quanti®cation, Elena Herburger

What Counts

Elena Herburger

Focus and Quanti®cation

The MIT Press Cambridge, Massachusetts London, England

( 2000 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. This book was set in Times New Roman by Asco Typesetters, Hong Kong and was printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Herburger, Elena. What counts : focus and quanti®cation / Elena Herburger. p. cm. Ð (Linguistic inquiry monographs ; 36) Includes bibliographical references and index. ISBN 0-262-08287-X Ð ISBN 0-262-58185-X (pbk.) 1. Focus (Linguistics) 2. Grammar, Comparative and generalÐQuanti®ers. 3. Grammar, Comparative and generalÐDeterminers. 4. Semantics. 5. Pragmatics. I. Title. II. Series. P299.F63 H47 2000 415Ðdc21 00-020404

To my parents, Luise and Helmut Herburger

Contents

Series Foreword Preface

xi

xiii

Chapter 1 Overview and Background

1

1.1

Overview

1

1.2 Sentences as Descriptions of Events 3 1.3 Adverbials as Predicates of Events 4 1.4 Why So Much Decomposition? 8 Chapter 2 Negated and Nonnegated Sentences 11

2.1

Introduction

11

2.2 The Role of Nonfocused Material 12 2.3 Focal Presuppositions Rethought 15 2.4 Structured Davidsonian Decomposition 17 2.5 Backgrounded Focal Entailments and Scope 21

viii

Contents 2.6 Postverbal Spanish n-Words as a Diagnostic 23 2.7 Focus and Negation: Bound, Free, and Wide Readings 29 2.8 What Is the Negation of a Sentence? 34 2.9 De®nite Descriptions, Focus, and Negation 35 2.10 Other Negative Contexts

38

2.11 How the Semantics Relates to the Syntax 40 2.12 A Comparison with Earlier Analyses 47 Appendix: (Non)contrastive Focus and Intonation 50 Chapter 3 Adverbial Quanti®ers

59

3.1

Introduction

59

3.2 Assocation with Focus: An Instance of Structured Davidsonian Decomposition 59 3.3 When There Is More than One Adverb 61 3.4

Second Occurrence Focus

3.5

Embedded Clauses

63

69

3.6 Quanti®cational Variability of Inde®nites 73

Contents

ix 3.7 Focus or VP-Internal versus VPExternal? 80 3.8

Events or Situations?

4.1

Introduction

82

Chapter 4 Only and Even

85

85

4.2 Some Syntactic Facts about Only 86 4.3 Conservativity, Focal Mapping, and Q-Raising 89 4.4 The Existential Force of Only 93 4.5 The Existential Force of Every 97 4.6 Downward Monotonicity: Weak and Strong 100 4.7 Only Quantifying over Events 105 4.8

Only's Partner Even

108

4.9 Syntactic Similarities between Only and Even 108 4.10 Implicatures and Being Noteworthy 110 4.11 Ambiguous Even-Sentences 111 4.12 The Existential Implicature as Pragmatic Inference 114

x

Contents 4.13 Syntactic and Crosslinguistic Support for the NPI Approach 116 4.14 Conventional Implicature versus Literal Meaning 118 4.15 Summary of the Discussion regarding Even 121

Chapter 5 Determiners

123

5.1

Introduction

123

5.2 Focus-A¨ected Noun Phrases: The Basic Phenomenon 124 5.3 Focus-A¨ected Readings and Q-Raising 129 5.4 Unary Weak Determiners and Q-Raising 131 5.5 Unary Weak Determiners Are Not Adjectives 133 5.6 Back to the De®niteness E¨ect: QR versus Q-Raising 135 Notes

141

References Index

161

155

Series Foreword

We are pleased to present the thirty-sixth in the series Linguistic Inquiry Monographs. These monographs present new and original research beyond the scope of the article. We hope they will bene®t our ®eld by bringing to it perspectives that will stimulate further research and insight. Originally published in limited edition, the Linguistic Inquiry Monographs are now more widely available. This change is due to the great interest engendered by the series and by the needs of a growing readership. The editors thank the readers for their support and welcome suggestions about future directions for the series. Samuel Jay Keyser for the Editorial Board

Preface

This monograph grew out of my dissertation (Herburger 1997b). Chapters 1, 2, 3, and 5 stem from that work, but have been substantially revised. Chapter 4 was written following a reviewer's suggestion and is entirely new. The phenomenon discussed in chapter 5 (focus-a¨ected readings of determiners) is also discussed in Herburger 1997a. The description of the data remains the same, but the analysis that I give here di¨ers in several ways. In my attempt to describe focus and how it interacts with quanti®cation I have received help from many sides. My deepest gratitude goes to Barry Schein. His comments and answers to my countless questions have been of tremendous help. While writing my thesis, I also bene®ted from the generous feedback I received from other linguists, especially Kai von Fintel, Irene Heim, Norbert Hornstein, Richard Larson, Renate Musan, Orin Percus, Tim Stowell, Anna Szabolcsi, and Jean-Roger Vergnaud. The suggestions of two reviewers for MIT Press aided and guided me in preparing this monograph, as did many hours of discussions with Paul Pietroski, Paul Portner, and Juan Uriagereka. Regine Eckardt's careful review of my dissertation in Glot International 3 gave me a further sense of what needed more work and clari®cation. I also appreciate the assistance I received from Laura Barker, who helped me with the research on even, and the input of those who participated in my classes on the de®niteness e¨ect (spring 1998) and on focus (fall 1998) at Georgetown University; they gave me an opportunity to try out and rethink some of the ideas discussed in this monograph. Finally, I want to thank Amy Brand for her patience and support and Anne Mark for her meticulous copyediting, which has greatly improved the readability (and in some places the literal content) of what follows.

Chapter 1 Overview and Background

1.1

Overview

It is a basic linguistic fact that the meaning of a sentence depends not only on the meanings of its words, but also on how the words are combined. Oswald killed Kennedy is obviously not the same as Kennedy killed Oswald. The words are identical, but the syntactic structures are di¨erent, and so are the meanings. Although fundamental, syntactic structure is not alone in determining how the meanings of words combine to give the meaning of the sentence that contains them. We know that focus also has a considerable semantic e¨ect. Thus, by shifting the main stress within a sentence, we shift the interpretation as well. Compare, for instance, OSWALD didn't kill Kennedy, where Oswald is stressed, with Oswald didn't KILL Kennedy, where the stress falls on killed. The interpretations of the two examples di¨er. Whereas the ®rst can be understood as denying that the person who killed Kennedy was Oswald, the second denies that the relation that ties Oswald to Kennedy is one of murder. In this monograph I explore how focus a¨ects meaning and speci®cally how focus a¨ects quanti®cational structure; I argue that the latter is rather ubiquitous, and present in cases where we have no overt indication that it is there. One of the main points I will make is this: when a quanti®er has only one syntactic argument at LF, focus reshapes its quanti®cational structure in that the nonfocused material in its scope also provides a restriction on it. I argue that this is true not only of the tacit neo-Davidsonian event quanti®er, but also of overt adverbial quanti®ers, only and even, and certain determiners (weak ones). The latter fact has interesting consequences for the analysis of the de®niteness e¨ect.

2

Chapter 1

The discussion is organized as follows. In chapter 1 I introduce the semantic background, summarizing various arguments that have led to the view that sentences are descriptions of events. In chapter 2 I turn to focus and its e¨ect on quanti®cation, in particular its e¨ects on the tacit quanti®cation over events introduced in chapter 1. I propose that focus at LF reshapes the quanti®cational structure of unary quanti®ers by making the nonfocused material in their scope part of not only their scope but also their restriction (focal mapping). I use the structured Davidsonian decomposition that results from focal mapping to explain certain pragmatic consequences of focusing. A central feature of the proposed analysis is that it maintains a simple, two-valued semantics, one that does not rely on semantic presupposition. In this context I also consider in some detail the interaction between focus and negation. In an appendix to the chapter I look at the di¨erence between contrastive and noncontrastive focus, attributing it not to the existence of di¨erent kinds of focus, but to the pragmatic properties of varying the intonation contours with which focus can be pronounced. Having dealt with covert quanti®cation over events in chapter 2, in chapter 3 I examine overt event quanti®cation involving adverbs, showing that the analysis proposed in chapter 2 carries over straightforwardly to this phenomenon. I explore several predictions the analysis makes with respect to the quanti®cational variability of inde®nites. I also propose an account of so-called second occurrence focus that is consistent with the overall idea that the quanti®cational structure of adverbial quanti®ers is very directly shaped by focus. Gradually moving from covert and overt adverbial quanti®ers to determiners, in chapter 4 I take up the interpretation of only and its close relative even, which resemble adverbs in some respects, but determiners in another. I address various questions regarding their semanticsÐfor example, whether only has existential force, whether only is conservative, and whether even is lexically ambiguous. Extending the analysis of focus and quanti®cation developed in chapters 2 and 3, I propose that only and even are adverbial-like at LF; should they not surface in an adverbial position, they covertly move to such a position by a process of Q-raising. Q-raising also plays an important role in chapter 5, where I investigate how focus a¨ects the interpretation of determiners. I show that weak noun phrases are more susceptible than strong noun phrases to the e¨ects of focus. I propose that this is so because weak determiners can take their scope through Q-raising, whereas strong determiners always have to take

Overview and Background

3

their scope through (an equivalent of ) QR. I argue that the de®niteness e¨ect is intrinsically tied to Q-raising. 1.2

Sentences as Descriptions of Events

The analysis of focus and quanti®cation that I propose in this monograph assumes that sentences are best thought of as descriptions of events (where ``event'' is meant in a generous sense that also includes states). Developing an analysis of focus that relies on event semantics, I try to provide a further argument for this general claim. In what follows I brie¯y summarize some of the arguments in favor of an event-based semantics. Traditionally, we think of a sentence like (1) as a predication where the verb love is a two-place predicate taking Mary and John as its arguments, as shown in (2). (1) John loves Mary. (2) Love( john,mary) Although (2) is simple and gives an intuitively correct interpretation, it has been argued that sentences should be accorded more involved interpretations than assumed in (2). The proposal is that the verb decomposes into a ``subatomic'' semantics (e.g., Parsons 1990). The sentence John loves Mary now describes an event (again, in the loose sense) of loving where John is the lover and Mary is the loved one. (3) be (Love(e) & Lover(e,john) & Loved-one(e,mary)) On this view all verbs translate as one-place predicates of events, regardless of whether they are syntactically unaccusative, intransitive, transitive, or ditransitive. Like verbs, adverbial modi®ers are direct predicates of events (see below). Now that verbs are analyzed as one-place predicates, what are traditionally considered the arguments of the verb (e.g., John and Mary in (2)) are tied to the verb only indirectly through a relation that links an event described by the verb to the participants in that event. In (3) this relation is expressed by ``Lover'' and ``Loved-one,'' respectively. More standardly, the relations are provided by theta-roles, as in (4) (see Parsons 1990). (4) be (Love(e) & Experiencer(e,john) & Theme(e,mary)) Even though (4) makes use of theta-roles, it is worth noting that a full¯edged event semantics is not committed to the existence of a general

4

Chapter 1

de®nition of the individual theta-roles agent, experiencer, theme, and so on. All that is needed is that the event participants can be distinguished relative to each verb (see Parsons 1990, 101¨., Schein 1993, 331¨.). This is just as well, because a general de®nition of the individual theta-roles is notoriously di½cult, perhaps even impossible (see Dowty 1989). Why should we accept this more elaborate way of analyzing John loves Mary when the traditional ``Love( john,mary)'' seems to do the job? In chapters 2 and 3 I show that the semantics of focus and adverbial quanti®cation makes event decomposition attractive because it allows for a simple and empirically detailed analysis of the e¨ects of focus. In what follows here I discuss some of the independent reasons indicating that it makes sense to analyze sentences as descriptions of events. This includes both Davidson's (1967) classical argument and a more recent argument for treating the arguments of the verbs as forming their own conjuncts (``separation''). 1.3

Adverbials as Predicates of Events

Davidson's (1967) original argument for thinking of sentences as descriptions of events stems from the semantics of adverbs. He proposes to treat adverbs like silently, yesterday, with a knife as predicates that apply to an event introduced by the verb. The empirical reason behind this proposal is that it allows us to explain certain entailment patterns that we ®nd with adverbs. Notice ®rst that (4a) entails all of (4b), (4c), and (4d). Moreover, (4b) and (4c) entail (4d). (4) a. b. c. d.

Brutus Brutus Brutus Brutus

(5)

stabbed stabbed stabbed stabbed

Caesar in the back with a knife. Caesar in the back. Caesar with a knife. Caesar.

(4a) (4b)

(4c) (4d)

This entailment pattern, represented in the diamond in (5), is explained if verbs describe events and adverbs are predicated of these events because,

Overview and Background

5

simply put, the longer conjunctions entail the shorter ones. If there was a stabbing where Brutus was the agent, Caesar was the theme, it was a stabbing in the back, and it was done with a knife, then it follows that there was a stabbing where Brutus was the agent, Caesar was the theme, this stabbing was a stabbing in the back; and so on. (6) a. be (Stab(e) & Past(e) & Agent(e,brutus) & Theme(e,caesar) & In-the-back(e) & With-a-knife(e)) ``There was a stabbing whose agent was Brutus, whose theme was Caesar, which was a stabbing in the back, and which was a stabbing with a knife.'' (cf. (4a)) b. be (Stab(e) & Past(e) & Agent(e,brutus) & Theme(e,caesar) & In-the-back(e)) ``There was a stabbing whose agent was Brutus, whose theme was Caesar, and which was a stabbing in the back.'' (cf. (4b)) c. be (Stab(e) & Past(e) & Agent(e,brutus) & Theme(e,caesar) & With-a-knife(e)) ``There was a stabbing whose agent was Brutus, whose theme was Caesar, and which was a stabbing with a knife.'' (cf. (4c)) d. be (Stab(e) & Agent(e,brutus) & Theme(e,caesar)) ``There was a stabbing whose agent was Brutus and whose theme was Caesar.'' (cf. (4d)) Similarly, the analysis accounts for the fact that (4a) entails the conjunction of (4b) and (4c); if there was a stabbing of Caesar by Brutus with a knife in the back, it follows that there was a stabbing of Caesar by Brutus in the back and that there was a stabbing of Caesar by Brutus with a knife. Conversely, it also follows that the conjunction of (4b) and (4c) does not entail (4a). This is so because (4b) and (4c) may refer to two di¨erent events, one where Caesar was stabbed in the back (with an ice pick) and one where he was stabbed with a knife (in the thigh), whereas (4a) requires there to be one single event of stabbing Caesar such that the instrument used was a knife and the place he was stabbed was in the back. A competing analysis of adverbs, the ``polyadic'' analysis, accords each adverb its own argument place in the verb. Compared with the Davidsonian account, this one is less attractive because it is committed to the claim that (for instance) the verb in (4a) (Brutus stabbed Caesar in the back with a knife) is a four-place predicate, whereas the verb in (4d) (Brutus stabbed Caesar) is a di¨erent (but homophonic) two-place predicate.

6

Chapter 1

Multiplying verb meanings, this analysis obscures the logical entailment between (4a) and (4d). Actually, there is one way to treat adverbs as arguments of the verb and still capture the entailment between (4a) and (4d): if we assume that (4d) contains a four-place predicate, just like (4a), and that, unlike the predicate in (4a), the one in (4d) is elliptical. Even though (4d) contains no visible modi®cation for time or place, it would mean ``Brutus stabbed Caesar somewhere, with something.'' Including somewhere and someone in the interpretation of (4d) would get the meaning right in (4d) and would capture the logical relation between (4a) and (4d). Still, it would not work in the general case because, in principle, a great many adverbs can be added (at seven, in front of everyone, on a night in March, etc.). However, this makes it impossible to state how many adverbial argument places a given verb has, and we would not know how to interpret it semantically. A related shortcoming of this improved version of the polyadic analysis is that many adverbials are genuinely optional (Parsons 1990). Stabbings, for instance, can be described as taking place next to someone, but they do not have to (most stabbings do not take place next to anyone). On a view that treats adverbials as arguments of the verbs and existentially quanti®es into the place of the ``missing'' modi®er, this is unexpected; if one stabbing takes place next to a guard, for instance, then every instance of stab has to be analyzed as taking place next to someone. On the other hand, if verbs introduce event arguments and if adverbials are analyzed as predicates of these events, it is not problematic that adverbials can, in principle, be iterated forever. Nor is it problematic that many adverbials are genuinely optional in the sense that when they are left out, no implicit existential quanti®cation can be assumed to take the place of the argument within the adverbial. Thus, analyzing adverbials as predicates of events accounts for the entailment pattern in (5) and captures the fact that adverbial modi®cation can be genuinely optional. A simple polyadic analysis that treats adverbials as arguments of the verb cannot capture the entailment patterns; and although a more elaborate version of this analysis does capture the entailments, it has a hard time accounting for optionality. But the possibilities have not been exhausted yet. A third account of adverbsÐone that treats them neither as arguments nor as predicatesÐ analyzes adverbials as operators that apply to VP meanings and return

Overview and Background

7

VP meanings. Since this account does not treat adverbials as arguments of the verb, it can easily capture their readiness to iterate; in other words, there is no polyadicity problem. Nor is it di½cult to do justice to the fact that they can be genuinely optional. The di½culty with this kind of analysis is rather subtle. Although it captures some of the entailments observed in (4), it does not capture all the entailments it should. This is shown by the following argument from Parsons 1990. Treating (for instance) both with a knife and violently as operators predicts that one modi®er will be able to take scope over the other. (7), for instance, should be scopally ambiguous between (8a) and (8b): in one case with a knife takes wide scope, in the other case violently does. (7) Brutus stabbed Caesar violently with a knife. (8) a. [with z(violently(stabbed(y)))](x) b. [violently(with z(stabbed(y)))](x) (8a) entails that x stabbed y violently, but not that x stabbed y with z. In contrast, (8b) entails that x stabbed y with z, but not that x stabbed y violently. The point is that (7) entails both, namely, that x stabbed y violently and that x stabbed y with z. On the operator analysis, this would follow only if we could point to a scopal ambiguity of (7), as shown in (8a) and (8b). Yet speakers do not perceive (7) as being ambiguous in that way. (7) is not scopally ambiguous, and yet it has both entailments. The two entailments follow directly on Davidson's analysis because each adverbial forms its own conjunct. Leaving out the knife conjunct, it follows that Brutus stabbed Caesar violently; leaving out the violent conjunct, it follows that Brutus stabbed Caesar with a knife. (9) be (Stab(e) & Past(e) & Agent(e,brutus) & Theme(e,caesar) & Violent(e) & With-a-knife(e)) ``There was an event of stabbing whose agent was Brutus, whose theme was Caesar, which was violent, and which was a stabbing with a knife.'' In sum, an analysis that treats adverbials as predicates of events explains all the logical entailments that we ®nd in examples like (4) and (7), and it ®ts naturally with the facts that adverbials can be iterated freely and that many of them are genuinely optional. Competing analyses of adverbialsÐthe polyadic analysis and the operator analysisÐfare less well.1

8

1.4

Chapter 1

Why So Much Decomposition?

All the entailments in (4) would also be explained on a less radical account: sentences are descriptions of events; only the adverbial modi®ers form their own conjuncts; the traditional arguments of the verb are not separated (i.e., the verb remains polyadic). This is in fact the kind of analysis proposed originally by Davidson (1967). On this view, (4a) translates as follows: (10) be (Past(e) & Stab(brutus,caesar) & With-a-knife(e) & In-the-back(e)) In keeping with the principle of least e¨ort, one may wonder why decomposition should be more radical than this, as I have assumed so far and will in fact assume throughout. One reason for full-¯edged decomposition involving the separation of arguments into their own conjuncts is that there are theta-marked arguments that are genuinely optional. These arguments do not have to be lexically realized; moreover, they are not semantically ``implicit'' when they are absent. Thus, in a relevant sense these optional arguments behave like the optional adverbial modi®ers that were discussed above. Parsons (1990, 97¨.) argues that the theme of stab is genuinely optional because we can coherently say Brutus stabbed and missed. There is nothing that he stabbed. Many arguments that are realized as datives are genuinely optional as well. For instance, it is possible to say Mary wrote a note without implying that she wrote the note to someone. Similarly, we can say He said something without entailing that he said it to anyone. If arguments form their own conjuncts and are tied to the verb only through their theta-roles, we can capture this alternation without having to posit that verbs like stab, write, and say are ambiguous, which certainly seems desirable. This is not possible on a polyadic analysis of verbs; on this view there would have to be two verbs say, for instance, say x and say x to y.2 Thus, separation captures the entailment patterns found with genuinely optional arguments. An independent argument for separation comes from sentences like (11) (Schein 1993, chap. 4). (11) Three video games taught every quarterback exactly two new plays. This sentence has various readings. On one interpretation it is construed as saying that each of three video games is such that it taught every

Overview and Background

9

quarterback two new plays, where the plays may vary according to the quarterbacks. On a similar but inverse scope reading it says that for every quarterback it holds true that three possibly di¨erent video games taught him two possibly di¨erent new plays. Here, the video games can vary relative to each quarterback, and the new plays can vary relative to each of the video games. Although possible, these readings are not the ones that come to mind ®rst. The most salient reading of (11) seems to be this: As in the ®rst reading, the quanti®ers take the scope that corresponds to their surface order. Unlike in the ®rst reading, though, here it is left vague how the video games relate to the quarterbacks. The reading distributes new plays with respect to quarterbacks, but is not committed in any way as to which video game taught what to whom; all that is claimed is that three video games altogether were teaching games. This reading can be paraphrased as follows: ``There was a teaching where three video games were the teachers, and that teaching resulted in every quarterback's learning two new plays each.'' As the paraphrase makes clear, what the video games do (namely, teach) and what happens to every quarterback (he is taught two new plays) have to be treated as separate (though related) matters. The point now is that we can keep what the video games do apart from what the quarterbacks experience if the subject (three video games) is separated from the rest into its own little event descriptionÐthere was an event where three video games did some teachingÐand that event results in a related event whereby every quarterback learns two new plays. But if the subject is to describe its own little event, it cannot be a direct argument of the verb. Thus, the third interpretation of (11) o¨ers an independent argument for the separation of arguments.3 The claim then is this. Although Davidson's original arguments only show that sentences are descriptions of events and that adverbs are direct predicates of events, there are reasons to assume even more decomposition. Not only do adverbial modi®ers form their own conjuncts; what are traditionally considered the arguments of the verb should also be interpreted as separate conjuncts. The analysis of focus developed in chapters 2 and 3 will make extensive use of this kind of event semantics. I trust it will also provide a further argument showing that the traditional polyadic analysis of verbs is too simple-minded and that investing in a somewhat more complex Davidsonian decomposition has empirical payo¨s that more than make up for the initial investment.

10

Chapter 1

One ®nal thing worth noting about separation is this. To make sure that the translations get the truth-conditions right, theta-roles have to be seen as being assigned exhaustively (see, e.g., Larson and Segal 1995, 485). (12), for example, is interpreted as in (13). (12) Romeo kissed Juliet. (13) be (Agent(e,romeo) & Kiss(e) & Past(e) & Theme(e,juliet)) We want (13) to capture that Romeo was the one kissing and that Juliet was the one being kissed by him. The concern now is that unless something further is said, (13) does actually not deliver this result; concentrating on Romeo, (13) only claims that there was a kissing and that Romeo was an agent in it. But this could be true if Romeo persuaded someone else to kiss Juliet and he himself did not kiss her! Clearly, this is not what (12) means. It follows that Romeo is not just some agent or other in the kissing event, but in fact the kisser, if theta-roles are assigned exhaustively such that when a theta-role is assigned to a given noun phrase, the individuals denoted by that noun phrase are all its bearers and its only bearers. Now that Romeo is the sole agent in the kissing event, it correctly follows that heÐand not somebody elseÐkissed Juliet.

Chapter 2 Negated and Nonnegated Sentences

2.1

Introduction

Pragmatically, it would make sense to utter (1a) in a stretch of discourse where we are discussing who wrote a poem. It would be less felicitous to use this sentence if what we are discussing is what RosalõÂa wrote; instead, it would be appropriate to use the minimally di¨erent (1b). (1) a. ROSALIA wrote a poem. b. RosalõÂa wrote A POEM. The reason (1a) and (1b) are felicitous in di¨erent discourse environments is that even though they have the same logical entailments, they are intuitively about di¨erent things: (1a) is about writing a poem, (1b) is about what RosalõÂa wrote. In this chapter I propose to capture the e¨ect illustrated in (1) by analyzing focus as imposing a certain amount of structure on the event quanti®cation introduced in chapter 1. I do not deny that focus has pragmatic repercussions. However, I argue that these are derivative on the restructuring e¨ect of focus. I also discuss the interaction between focus and negation. Whereas in simple a½rmative sentences the quanti®cational structure that focus imposes only a¨ects what the sentence is about, in negated sentences the very same e¨ect not only shapes the sentence's ``aboutness,'' but alsoÐdepending on the way the restricted event quanti®er interacts with the scope of the negationÐgives rise to di¨erent logical entailments. I would like to make a preliminary point here to delimit the scope of the discussion in this chapter and to perhaps avoid misunderstandings. I will concentrate on the meaning e¨ects of focus, leaving as a separate and largely undiscussed issue its phonological realization through stress.

12

Chapter 2

Throughout, I use capital letters to mark the semantic domain of focus, rather than the main stress. Even though I will not be concerned here with the phonological realization of focus through stress, it will be helpful to be aware of the following fact in order to evaluate the data: although focus is marked through stress, there is not always a one-to-one correlation between a word's carrying the main stress of the sentence and its being the focus. This is so because in certain cases one stress assignment can in fact signal more than one focus. Thus, in English focus on any given phrase is generally realized as increased stress on its rightmost and most deeply embedded word (see Chomsky 1971, Cinque 1993). However, this has the consequence that stress on shirt in (2), for instance, can be interpreted as giving rise to all the di¨erent foci that are represented in (3a±e) since in all of the relevant phrases stress falls on the rightmost and most deeply embedded word. (2) The police arrested an ex-convict with a red shirt. (3) a. b. c. d.

The police arrested an ex-convict with a red SHIRT. The police arrested an ex-convict with A RED SHIRT. The police arrested AN EX-CONVICT WITH A RED SHIRT. The police ARRESTED AN EX-CONVICT WITH A RED SHIRT. e. THE POLICE ARRESTED AN EX-CONVICT WITH A RED SHIRT.

For more detailed discussion of how focus is phonologically realized and how this relates to syntactic structure, see (among others) Selkirk 1984, 1995, Cinque 1993, and Zubizarreta 1998. With these two preliminary points made, let us now turn to the way focus a¨ects meaning. 2.2

The Role of Nonfocused Material

Among the semantic characterizations of focus that can be found in the literature, the following has been particularly successful. It claims that the primary function of focus is to express a contrast between the element that is denoted by the focus and other elements that are present in the discourse and that, had things been di¨erent, could have taken the place of the focused element. Conversely, the absence of focus is held to signal that the information encoded by the nonfocused material is ``the function under discussion.'' This characterization of focus is prominent in Jackendo¨ 's (1972) discussion; it also underlies to a considerable extent the

Negated and Nonnegated Sentences

13

``alternative semantics'' approach to focus (see, e.g., Rooth 1985, 1992, von Fintel 1994). Jackendo¨ 's analysis seems a good place to start. Roughly, the proposal works like this. If we hold that focus serves to introduce alternatives, we predict that the focused element RosalõÂa in (1a), for example, contrasts with other people salient in the discourse of whom we might have said that they wrote a poem (Emily, William, Federico, etc.). To capture this contrastive e¨ect of focus, the focused element is abstracted over and replaced by a variable that matches it in semantic type, thus creating the equivalent of a ``p-set.'' (4) lx Wrote-a-poem(x)

``p-set'' of (1a)

This lambda-abstraction also provides the mechanism generating the ``focus-semantic'' value for the sentence, which in the alternative semantics account exists alongside the regular semantic value, and which is used to account for various e¨ects of ``association with focus'' (see Rooth 1985, 1992, and chapter 3). Originally, Jackendo¨ considered a somewhat di¨erent claim, namely, the possibility that the type-matching variable that replaces the focus is bound not by a lambda-operator, as in (4), but by an existential quanti®er. (5) bx (Wrote-a-poem(x))

``focal presupposition'' of (1a)

The idea was that (5) would capture the ``focal presupposition'' of (1a), something (4) does not do. Ultimately, though, Jackendo¨ discarded (5) and adopted (4) instead. To see the reasoning behind adopting (4), we can ®rst note that examples like (1a,b) suggest the generalization in (6), where focus splits the sentence into ``presupposition'' and ``assertion'' (Chomsky 1971). (6) The nonfocused part of a sentence expresses an existential presupposition; the focused part expresses the assertion. To appreciate (6), it is worth noting what is not meant by ``focal presuppositions.'' Importantly, Jackendo¨ (1972) distinguishes focal presuppositions from other phenomena referred to as presuppositions. Consider for instance the ``lexical presuppositions'' of the word again, where Mary won again presupposes that Mary has won before. These kinds of presupposition are independent of focus.1 Taking ``presupposition'' in an intuitive, pretheoretic sense, (6) seems to describe the e¨ect of focus in (1) well enough; and it is this generalization that the analysis sketched in (5) aims to account for. There is a problem with (6), however. As Jackendo¨ observes, the generalization in

14

Chapter 2

(6) is not fully general. It is falsi®ed by examples like (7) (Jackendo¨ 1972, 246). (7) NOBODY likes Bill. Contrary to what (6) would lead us to expect, (7) cannot be seen to presuppose that somebody likes Bill, then going on to assert that it was nobody. If it did, (7) would be contradictory; but it is perfectly sensible. In light of this, Jackendo¨ concludes that (6) is too strong and opts for something like (8) instead. (8) The nonfocused part of a sentence expresses the property (function) that is under discussion. Speci®cally, making use of the abstraction over the focused element, Jackendo¨ argues for (9). The variable x here stands in for the focus, and Presupp stands for the nonfocused material of the sentence. 9 8 is a coherent set > > > = < is well defined in the present discourse > . (9) lx Presupps (x) > > is amenable to discussion > > ; : is under discussion The purpose of (9) is to relieve the nonfocused part of any existential commitment. The nonfocused part simply denotes a function. Whether or not this function applies to an individual, resulting in truth, is crucially left open. The aim is to make (9) consistent not only with (1), where focus may be seen as giving rise to an existential focal presupposition that somebody wrote a poem, but also with examples like (7), where it clearly would make no sense to assume that the nonfocused part gives rise to a presupposition that somebody likes Bill. Despite its initial appeal, (8) turns out to have certain empirical limitations. By aiming to be so general that it can encompass both (7) and the examples in (1), (9) o¨ers no account of the di¨erence between these two cases. In addition, (7) may actually not provide an argument in favor of (9), despite Jackendo¨ 's conclusion to the contrary. If we take at face value the claim that the focus is replaced by a type-matching variable, then by existentially binding this variable as in (5), we would actually be able to draw a distinction between (1) and (7);2 since what is focused in (7) is a quanti®er (nobody), a type-matching variable replacing it would be a variable over quanti®ers (rather than a variable over individuals). Consequently, existentially binding this variable will not have the unwelcome

Negated and Nonnegated Sentences

15

consequence of generating the focal presupposition that someone likes Bill; instead, it would generate a focal presupposition that some quanti®er applies to likes (x,b). Since the quanti®er in question might be decreasing (e.g., nobody; few; if any, of his colleagues), this would be harmless insofar as it would not signal any commitment to the existence of someone who likes Bill.3 Given that (7) provides no motivation for (9), Jackendo¨ could have held on to the view that the distribution of ``focal presuppositions'' is captured by existentially binding the type-matching variable that replaces the focus. This view would have had more empirical bite than (9) because, unlike (9), it would capture a di¨erence between (7) and (1). An analysis along these lines is in fact contemplated in Rooth 1994a, as we will see in section 2.10. 2.3

Focal Presuppositions Rethought

Although the presuppositional view is empirically stronger than the ``function-under-discussion'' view stated in (9), it appears that the presuppositional view is not quite strong enough. As a next step in the argument, we can observe that the presuppositional view does not seem to make correct predictions in all cases. The problem is that it suggests that (7) has a harmless focal presupposition because what is focused is a quanti®er. It turns out, however, that it makes a di¨erence what kind of quanti®er is focused (i.e., whether it is decreasing or not), and this is something that the presuppositional view does not capture, which would have to be added by stipulation. Compare (7) with (10). Unlike in (7), in (10) the focus is on a nondecreasing quanti®er. (10) MANY OF HIS COLLEAGUES like Bill. The point is that (10) can be smoothly used in a context where it counts as previously established that Bill is liked. To one committed to the idea that the nonfocused part can give rise to a presupposition, this would suggest that (10) has the focal presupposition that somebody likes Bill. In contrast, this cannot be said of (7); if we use (7) in this kind of context, we create a con¯ict with what was said before. This indicates that the nature of the focused quanti®er matters. In particular, it matters whether it is decreasing as in (7) or nondecreasing as in (10); only when focus is on a decreasing quanti®er like nobody as apposed to a nondecreasing quanti®er

16

Chapter 2

like many of his colleagues can one say that a ``focal presupposition'' is absent. Thus, although the presuppositional analysis fares better than the function-under-discussion analysis in that it successfully distinguishes between (1) and (7), it lumps (7) together with (10), which does not seem quite right. Apart from this very concrete empirical problem and some related ones that will be discussed below, the presuppositional analysis of focus has what can be considered a more general drawbackÐthe very fact that it relies on the notion of presupposition. This might seem harmless enough if ``presupposition'' is taken in an intuitive sense, which is the sense in which it was intended in Herburger 1997b. But a number of nontrivial issues arise as soon as ``presupposition'' is taken in a more technical sense (see Eckardt 1998). One important empirical consideration is that ``presupposition'' may not refer to a distinct, well-de®ned class of things. As Boer and Lycan (1976) show in a series of case studies, the semantic presuppositions that various linguistic phenomena have been said to give rise to turn out to behave rather di¨erently upon close inspection. Concretely, Boer and Lycan show this by investigating the alleged presuppositions of nonrestrictive relative clauses, cleft constructions, factive verbs, implicative verbs, counterfactives, de®nite descriptions, and counterfactuals. Stalnaker (1974) makes what I think is a similar point. A second, more general problem involves the failure of presuppositions. On a Fregean view of presupposition, for the sentence S to have a meaningÐto be true or false, that isÐthe sentence's presuppositions must hold. Thus, regardless of whether S is true or false (in which case the negation sS is true), its presuppositions have to be true. This is captured by saying that both S and its negation sS entail p, where sS entails p because p ``projects'' across the negation. It turns out, however, that the notion of presupposition cannot be phrased in terms of entailment (see Katz 1973, Boer and Lycan 1976).4 The reason is that if both S and sS entail p (S ! p & sS ! p) then, should p fail, we would have to conclude that S & sS is true. This being a contradiction, it follows that a presupposition p could never be false, that presuppositions are tautological. But presuppositions do fail. To maintain the idea that a sentence whose presuppositions fail is neither true nor false, it is necessary to give up a bivalent semantics and introduce a third truth-value alongside ``true'' and ``false,'' namely, ``valueless'' (Boer and Lycan 1976). A sentence S now presupposes p not

Negated and Nonnegated Sentences

17

if S and sS entail p, but if S and sS necessitate p. One sentence necessitates another when there is no model in which the ®rst sentence is true but the second sentence is untrue. If the only truth-values are ``true'' and ``false,'' being untrue coincides with being false and necessitation boils down to entailment. On the other hand, when there is a third value ``valueless,'' a sentence can be untrue either because it is falseÐor because it is valueless. Saying now that a sentence S presupposes p just in case S and sS necessitate p means that there can be no model where S and sS are true but where p is untrueÐfalse or valueless, that is. What happens if p is untrue, either because it is false or because it is valueless? In such a case S and sS cannot both be true, nor can they both be false, since both situations would result in contradiction. The only remaining option when p is untrue is that S and sS are valueless. Thus, saying that only when its presuppositions are true can a sentence be judged true or false comes at the price of giving up a two-valued semantics where the only truth-values are ``true'' and ``false.'' This kind of analysis might seem worth its price if the phenomena that are called presuppositions formed a natural class and the empirically best analysis were clearly one that says that if the presupposition fails, the sentence receives the value ``valueless.'' But, as already mentioned, in trying to sort out the di¨erences among the many alleged presuppositions, Boer and Lycan go to great lengths to show that ``presupposition'' is mainly useful as cover term, since presuppositions are not a uni®ed phenomenon. Emphasizing the heterogeneous nature of the data involved, they show that whereas some alleged presuppositions are cancelable and better analyzed as instances of conversational implicature, others involve logical entailments where the material that gives rise to the entailment is backgrounded, deemphasized, or not part of what the sentence ``says'' or ``asserts.'' With this in mind, let us return to focus. 2.4

Structured Davidsonian Decomposition

Having noted both an empirical problem and a theoretical concern that a presuppositional analysis of focus would face, let us explore a di¨erent analysis of the behavior of the nonfocused part of a sentence, an analysis that does not rely on the (technical) notion of presupposition and that successfully di¨erentiates not only between (1a) and (7), but also between (7) and (10).

18

Chapter 2

(1) a. ROSALIA wrote a poem. (7) NOBODY likes Bill. (10) MANY OF HIS COLLEAGUES like Bill. At the heart of the analysis I want to explore lies the following claim. Holding that both stative and eventive sentences are descriptions of events as argued in chapter 1, we can say that focus imposes a certain amount of structure on this quanti®cation. (11) All the nonfocused material in the scope of the event quanti®er Q also restricts Q. Along with the event quanti®er Q that is restricted by the nonfocused material, there may be other event quanti®ers whose quanti®cational structure is not directly shaped by focus. Which event quanti®er Q is thus restricted is an optional matter. In the case of (1) the structured Davidsonian decomposition in (11) will give us the translations that are shown in (12). (For concreteness, I am assuming here that the quanti®ed noun phrase a poem takes narrow scope.) (1) a. ROSALIA wrote a poem. b. RosalõÂa wrote A POEM. (12) a. [be: C(e) & Write(e) & Past(e) & [a x: Poem(x)] Theme(e,x)] Agent(e,rosalõÂa) & Write(e) & Past(e) & [a x: Poem(x)] Theme(e,x) b. [be: C(e) & Agent(e,rosalõÂa) & Write(e) & Past(e)] [a x: Poem(x)] Theme(e,x) & Agent(e,rosalõÂa) & Write(e) & Past(e) The interpretation of (1a) that is represented in (12a) states that some relevant event of writing a poem in the past was such that it was a past event of writing a poem and its agent was RosalõÂa, which amounts to saying that some relevant past event of poem writing had RosalõÂa as its agent. In contrast, the interpretation of (1b) given in (12b) states that some relevant past event of writing whose agent was RosalõÂa had a poem as its theme (and was a past event of writing whose agent was RosalõÂa). Several comments are in order here. First, (12a) and (12b) could be thought of as enriched expressions of the object language that closely correspond to their metalanguage translations (cf. Schein's (1993) logical forms). Alternatively, (12a) and (12b) can be seen as part of the metalanguage, as the output of semantic derivation. In that case each one represents the right-hand side of a T-sentence, which captures the truth-

Negated and Nonnegated Sentences

19

conditional equivalence between a sentence and its semantic translation: for example, Val(t, ROSALIA wrote a poem, s) if and only if [be: C(e) & Write(e) & Past(e) & [an x: Poem(x)] Theme(e,x)] Agent(e,rosalõÂa) & Write(e) & Past(e) & [an x: Poem(x)] Theme(e,x). This is what I will assume in what follows. Another assumption concerns be. I assume that every sentence has a syntactically represented adverbial quanti®er. Sometimes it is overtly realized, as I will show in chapter 3, but often it is tacit, as in (1a,b). Tacit event quanti®ers are existential except in generic sentences, where they are (quasi)universal. Regarding C(e), I assume following standard views that every quanti®er is restricted by a context predicate C whose value is ®xed by the context of utterance. In the case of the event operator, this kind of context predicate ensures that we are only looking at those events that are relevant in the context in which the sentence is used. By restricting the event operator to elements that are relevant in the discourse, C(e) functions much the same as C(x) functions in the translation of Everybody stood up, where it helps encode that we interpret this sentence relative to a given, salient group of people, perhaps the people in the room, not relative to all people absolutely. In addition to structured Davidsonian decomposition, the analysis has a second component. There is a widely held intuition that the restrictions of quanti®ers ``set the scene,'' that they encode what the rest of the sentence will be about. This does not follow from the semantic axioms themselves. As far as the lexical meaning of a restricted existential quanti®er is concerned, both the restriction and the scope translate as sets (or classes) and are in this sense on equal footing; all that the lexical meaning of the quanti®er requires is that the intersection between the two sets not be empty (see chapter 5). We can encode the aboutness expressed by the restriction of a quanti®er by saying that similar to the selection relation S that links an utterance to the ``right'' sequences for the evaluation of the sentence in question (see Larson and Segal 1995, chap. 6), there is another semantics-pragmatics relation that links the quanti®er restrictions of a sentence to the pragmatic notion of what the utterance of the relevant part of the sentence is about. (13) Quanti®er structure and aboutness For any utterance of a structure with a quanti®er Q with restriction A and scope B, a relation R links A to what the utterance of [Q A] B is about. B, in turn, expresses the relevant assertion.

20

Chapter 2

Aboutness and assertion are pragmatic notions, which presumably can be formalized by studying their relations to the previous discourse. I think that for present purposes this is not necessary and that an intuitive understanding su½ces. Clearly, a sentence can contain various instances of restricted quanti®cation, all of which encode some degree of aboutness according to (13), suggesting that aboutness is a relative notion. It seems reasonable to assume that the more lexical material that is involved in restricting a quanti®er, the more aboutness that quanti®er captures. Since the restrictions of event quanti®ers can be quite largeÐas we will see, they can reach even across clausesÐthey have the potential of capturing a great deal of it. It is as a result of (13) and structured Davidsonian decomposition that (1a) is about an event of writing a poem and that it moreover gives rise to what I would like to call a ``backgrounded focal entailment'' that such an event did indeed take place. I refer to this as a backgrounded entailment in the spirit of Boer and Lycan (1976) because although it is a genuine logical entailment, it does not arise in the sentence's assertion (see also Atlas 1991, 1993). To say that (1a) is about some event of writing a poem and entails that there was such an event is similar but not identical to saying that it presupposes, in the technical sense, that someone wrote a poemÐthe description of the nonfocal part that the presuppositional view gives. An important di¨erence between the presuppositional analysis of (1a) and the present one is that in case there was no event of poem writing (1a) does not have the value ``valueless,'' as it would on the technical version of the presuppositional view; instead, because the semantics remains twovalued, it has the value ``false.'' Under such a circumstance (1a) will be logically as false as it would be if a relevant event of poem writing did not have RosalõÂa as its agent. What is of course di¨erent is the pragmatic feel of the lack of truth. If RosalõÂa was not the writer, then the falsity lies in what the speaker asserted. If no relevant writing of a poem transpired, then the falsity lies in the backgrounded focal entailment. The reason the latter feels pragmatically worse has to do with our understanding of how discourse works. Adapting ideas from Stalnaker (1979) to our purposes, we can say that in a felicitous discourse backgrounded focal entailments are entailed either by what was said earlier or by the Common Ground; or if they are entailed by neither of these, they are such that they can be accommodated without much trouble; if (1a) is used in a context where we know that no poem was written, this discourse felicity is clearly not met.

Negated and Nonnegated Sentences

21

Apart from the truth-value di¨erence, a presuppositional analysis of (1a) and the present analysis di¨er in that on the former the nonfocused part expresses that someone wrote a poem. On the latter it says only that there was some relevant event of writing a poemÐnot that someone wrote a poem. The intuition that (1a) is about someone writing a poem does not follow from (12a) itself. Rather, it follows from our understanding of how poems come to be written; the way the world is, events of writing poems require writers. Clearly, if we conceived the writing process di¨erently, then, leaving language the same, this inference would not follow. Some independent support for characterizing the nonfocused material as being semantically as weak as structured Davidsonian decomposition suggests it is will be o¨ered in section 2.6. So far, then, I have proposed that event quanti®cation is not always unrestricted, as assumed in chapter 1. Rather, an event quanti®er can be restricted by nonfocused material in its scope. It is because focus a¨ects the quanti®cational structure of a sentence that it has an e¨ect on the felicity of the sentence in a given discourse; being a quanti®cational restriction, the nonfocused part marks what the relevant part of the sentence is about. Moreover, in the cases discussed up to now, the restricted event quanti®er gave rise to a backgrounded focal entailment.5 2.5

Backgrounded Focal Entailments and Scope

Now that we have seen how structured Davidsonian decomposition deals with the examples in (1), let us turn to (7), the example that posed a problem for the function-under-discussion view. (7) NOBODY likes Bill. On the present account of (7), the decreasing quanti®er nobody takes scope over the event operator. (14) [nobody x] [be: C(e) & Like(e) & Theme(e,bill)] Experiencer(e,x) & Like(e) & Theme(e,bill) ``Nobody is such that some (relevant) event of liking Bill had him or her as its experiencer.'' Since the nonfocused material likes Bill restricts the event quanti®er, (7) is pragmatically about whether or not Bill is liked. Crucially, (7) does not entail that he is. This follows from the fact that the negative quanti®er NOBODY takes scope over the existential event quanti®er. By the same

22

Chapter 2

token, although what is focused in (10) is a quanti®er, the sentence is correctly predicted not only to be about whether or not Bill is liked but in fact to entail in its background that Bill is liked. Even though the focused quanti®er takes scope over the event, it is a nondecreasing quanti®er and hence does not rob the narrow scope event quanti®er of its existential import. (10) MANY OF HIS COLLEAGUES like Bill. (15) [many x: Of-his-colleagues(x)] [be: C(e) & Like(e) & Theme(e,bill)] Experiencer(e,x) & Like(e) & Theme(e,bill) Nothing in what has been said so far forces a decreasing quanti®er to take wide scope over the event operator. The question now is whether there are any cases where we can see that a decreasing quanti®er in fact takes narrow scope. By the logic of our analysis of (7), we expect that under such a scope constellation the focally restricted event quanti®er will have existential import; since (7) lacks such an entailment precisely because the decreasing quanti®er nobody takes wide scope over the event, when an analogous decreasing quanti®er takes narrow scope, an entailment should become available. One example we might consider in this context is (7) itself. Does it have a reading along the lines of ``There is a liking of Bill and nobody is the liker in it''? It is hard to tell, because the coherent wide scope interpretation of nobody will always be in the foreground, masking this nearly incoherent one. A better example may be this one: (16) Regina sang a song. But she sang it TO NOBODY. In accordance with what the ®rst sentence states, we can say that the second sentence is about an event where Regina sang. The existence of such an event is entailed, and it is asserted that it was not performed for anybody's bene®t. Since the focused dative here is genuinely optional (see chapter 1), we can argue that in (16) the negative quanti®er takes narrow scope with respect to the event operator. As a result, the sentence is correctly predicted to have the backgrounded entailment that Regina sang a song. That there is a di¨erence between (7) and (16) is expected under the present analysis. Since the backgrounded entailments that replace the focal presuppositions derive directly from the existential import of the restricted event quanti®er, their distribution is sensitive to the scope that operator takes. As a result, when the event quanti®er occurs within the scope of a decreasing element, no backgrounded entailment will be avail-

Negated and Nonnegated Sentences

23

able; but when the event quanti®er takes wide scope over a decreasing element, one will be. There is a problem with the example in (16), however. One could argue that in the second sentence the negative quanti®er in fact takes scope over the entire sentence, including the event operator, and that the reason we think that this sentence entails that there was singing has only to do with the content of the ®rst sentence, which explicitly states that a song was sung. There are examples that are immune to this kind of objection. In these examples a decreasing quanti®er unmistakably takes narrow scope and, in line with our prediction, these examples do indeed have backgrounded focal entailments. Particularly clear cases of this can be found in Spanish because in this language the narrow scope of a negative quanti®er can be read o¨ the surface syntax of the sentence. Showing this will require a little detour through the quanti®er system of Spanish. 2.6

Postverbal Spanish n-Words as a Diagnostic

It is generally assumed that when so-called n-words in languages like Spanish and Italian occur preverbally, they do not require a c-commanding negative element, but that they do require such an element when they appear in postverbal position (see, e.g., Laka 1990, Zanuttini 1991, Ladusaw 1992, Uribe-Etxebarria 1994; see also chapter 4). (17) a.

Nadie vino. n-body came ``Nobody came.'' b. No vino nadie. not came n-body c. *Vino nadie. came n-body

Yo nunca habõÂa estado en CoÂrdoba. I n-ever had been to CoÂrdoba ``I had never been to CoÂrdoba.'' b. Yo no habõÂa estado en CoÂrdoba nunca. I not had been to CoÂrdoba n-ever c. *Yo habõÂa estado en CoÂrdaba nunca. I had been to CoÂrdaba n-ever

(18) a.

24

Chapter 2

Descriptively, the standard paradigm illustrated in (17) and (18) suggests that preverbal n-words pattern with negative quanti®ers (nobody, never, etc.) whereas postverbal n-words correspond to negative polarity items (NPIs; anybody, ever, etc.). There has been an extensive debate in the literature about whether n-words should be considered as fundamentally negative quanti®ers (e.g., Zanuttini 1991) or as underlying NPIs (e.g., Laka 1990). The issue is not important here.6 What is important is that the standard paradigm is in fact incomplete: under some conditions it is possible to ®nd postverbal n-words that are not c-commanded by any negative element. As discussed in Zanuttini 1991 and Herburger 1998, this is possible when the n-word is interpreted as a negative quanti®er that takes narrow scope with respect to the event described by the verb. Consider for instance (19). (19) MarõÂa se fue con nada. MarõÂa left with nothing Since nada in (19) can be interpreted as a narrow scope negative quanti®er, when (19) itself is negated, it is ambiguous between a simple negation (where the n-word is interpreted as any) and a double negation (where the n-word corresponds to a negative quanti®er). (20) MarõÂa no se fue con nada. ``MarõÂa didn't leave with anything.'' (simple negation) ``MarõÂa didn't leave with nothing. (i.e., She left with something.)'' (double negation) In the examples discussed so far, the postverbal negative quanti®ers occur in adjuncts. However, these quanti®ers are not limited to such environments. (21) provides an instance where the postverbal negative quanti®er is an indirect object. (21) No se movõÂa ni una brizna de hierba, ni una triste hoja. ``Not a strand of grass moved, not a sad leaf.'' Todo era tan tierno que no tenõÂa bastantes ojos para mirar. ``Everything was so touching that I didn't have eyes enough to see.'' Al ®nal, con los brazos extendidos hacia adelante, ``Then, at the end, with my arms stretched out in front of me,'' dije bajito a nadie que todo era mõÂo. said-I softly to n-body that everything was mine ``I said softly to nobody that everything was mine.''

Negated and Nonnegated Sentences

25

(from ``ParecõÂa de seda'' by MerceÁ Rodoreda, translation and emphasis mine) If an event of saying is not directed toward anyone, we normally do not explicitly say so, but simply leave out the genuinely optional goal speci®cation. As (21) illustrates, however, we do not necessarily have to do this; the last sentence states not only that the narrator of the story said something (namely, that everything was hers), but also that this event of saying was not directed toward anyone. In this context, consider also the contrast in (22), which makes use of the di¨erence between mirar a and ver. (22) El pobre Juan se ha vuelto loco. Se pasa los dõÂas mirando a nada/ *viendo nada. ``Poor Juan has gone insane. He spends his days looking at nothing/ *seeing nothing.'' It is possible to look and to look at nothing (even if a sane person might not spend his days doing that). It is not possible, though, to see and to see nothing.7 This di¨erence between see nothing and look at nothing explains the contrast that is illustrated in (22); a postverbal n-word without a licenser is acceptable when it can be interpreted as taking narrow scope with respect to the event described by the verb (mirar a nada) but not when such a narrow scope interpretation is unavailable (ver nada). Last but not least, consider (23), an example that brings out clearly the narrow scope properties of postverbal negative quanti®ers in Spanish.8 (23) Hemos leõÂdo en alguna parte que la meditacioÂn para los budistas ``zen'' ``no consiste en `no pensar en nada' sino en `pensar en nada' ''. ``We have read somewhere that for Zen Buddhists meditation `does not consist in ``not thinking of anything'' but in ``thinking of nothing.'' ' '' (23) rejects the proposition that Zen Buddhists do not think when they meditate, a proposition expressed by no pensar en nada, where the negation takes wide scope over the event operator. Rather, it asserts that the meditation of Zen Buddhists involves a particular kind of thinking: namely, thinking that is not directed toward anything, described by pensar en nada. Here again, the negative quanti®er clearly takes narrow scope with respect to the event described by the sentence.

26

Chapter 2

The point of exploring the Spanish quanti®er system is that it provides examples where the surface syntax of the sentence reveals directly whether a negative quanti®er takes narrow scope with respect to the event. This is the case when an n-word occurs postverbally and there is no c-commanding negative element. English is much less helpful in this respect because to the extent that speakers accept postverbal negative quanti®ers (many ®nd them rather stilted except with have and remain; see Zanuttini 1991), they freely assign them wide scope. Going back to the main thread of the argument, the Spanish facts allow us to test the prediction that if a negative quanti®er takes narrow scope, the sentence has a backgrounded entailment that the event described by the nonfocused material took place. The prediction is indeed borne out: (24) entails that MarõÂa left, (25) that the person denoted by the subject said that everything was hers, (26) that Juan spends his days looking, and (27) that for Zen Buddhists meditation consists in thinking. (24) MarõÂa se fue CON NADA. backgrounded focal entailment: ``MarõÂa left.'' assertion: ``MarõÂa left with nothing.'' (25) . . dije bajito A NADIE que todo era mõÂo. backgrounded focal entailment: ``I said that everything was mine.'' assertion: ``I said that everything was mine to no one.'' (26) Se pasa los dõÂas mirando A NADA. backgrounded focal entailment: ``He spends his days looking.'' assertion: ``He spends his days looking at nothing.'' (27) . . . la meditacioÂn para los budistas ``zen'' . . . consiste . . . en pensar EN NADA. backgrounded focal entailment: ``For Zen Buddhists meditation consists of thinking.'' assertion: ``Their thinking consists of thinking of nothing.'' The interpretations that result from structured Davidsonian decomposition capture these backgrounded entailments straightforwardly. Unlike in (14) (the interpretation of (7)) in (28) and (29) (the interpretations of (24) and (25)) the decreasing quanti®ers are interpreted inside the scope of the event operator. Consequently, the event operator that is restricted by the nonfocused material has existential import. (28) [be: C(e) & Leave(e) & Past(e) & Agent(e,marõÂa)] [nothing x] With(e,x) & Leave(e) & Past(e) & Agent(e,marõÂa)

Negated and Nonnegated Sentences

27

``Some (relevant) event of leaving by MarõÂa was a leaving with nothing.'' (29) [be: C(e) & Say(e) & Past(e) & Agent(e,y) & Theme(e,that everything was mine)] [nobody x] Goal(e,x) & Say(e) & Past(e) & Agent(e,y) & Theme(e,that everything was mine) ``Some (relevant) event of my saying that everything was mine had nobody as its goal.''9 Earlier I noted that according to my analysis the event quanti®er in the interpretation of ROSALIA wrote a poem literally describes an event where poem writing took place, not an event where someone wrote a poem. I further suggested that the de®cit of the backgrounded entailments that are derived by structured Davidsonian decomposition is made up by our world knowledge; since we know that events of poem writing have an agent, we can infer that if a poem was written, then there also was somebody writing it. I also promised to give some independent evidence for this. It is provided in (25). As structured Davidsonian decomposition predicts, (25) entails that in an event of saying, something was said by the individual denoted by the subject. Importantly, however, our world knowledge need not add any further information concerning other participants in the event; although events of saying require a theme, the way the world is we know that they do not necessarily require a goal. This is a welcome result because if the entailment included that something was said to someone, the sentence would be contradictory; something was said to someone, and it was to no one. The fact that the sentence is not contradictory supports the claim that the weak backgrounded entailments that structured Davidsonian decomposition derives are exactly weak enough; anything stronger would get the meaning wrong. Whereas structured Davidsonian decomposition accounts for (24)±(27), the presuppositional analysis does not. As we saw, on this proposal the nonfocused part gives rise to a focal presupposition, which is derived by existentially binding the type-matching variable that is said to replace the focus. When focus falls on a quanti®er, the focal presupposition that is derived is that some quanti®er applies to the nonfocused part of the sentence. As noted earlier, this kind of analysis fails to distinguish the behavior of focused decreasing quanti®ers from that of nondecreasing ones. It is also unclear that an analysis like that stated in (8) and (9) can capture the di¨erence between the cases where a decreasing quanti®er takes wide scope or narrow scope. Presumably, sentences like (24)±(26)

28

Chapter 2

would be described as having a focal presupposition; thus, for instance, (24) would presuppose that MarõÂa left, and assert that she left with nothing. But the focal presupposition that the presuppositional analysis would derive is merely that some quanti®er applies to MarõÂa left. If the quanti®er is decreasing, it does not follow that MarõÂa left. Thus, an empirical advantage of the present analysis of focus over the presuppositional one is that it can distinguish between focus on a decreasing quanti®er and focus on a nondecreasing quanti®er and can also capture the fact that the scope of a focused decreasing quanti®er matters. A last point I would like to make in this connection is that although I have concentrated on the behavior of focused quanti®ersÐbecause I was presenting my proposal against the background of the Jackendo¨-style analysisÐit in fact does not matter for the distribution of backgrounded entailments that are due to focus whether a quanti®er is focused or not. Regardless of how focus is assigned, as long as a decreasing quanti®er takes scope over the restricted event operator, the sentence is predicted to not license a backgrounded focal entailment. And this is indeed what we ®nd. For instance, when we shift the focus in (7) away from the quanti®er to Bill, the sentence does not entail that there is an event of liking of which it asserts that the one liked is Bill. The felicity of (30) illustrates this. (30) Nobody likes BILL. In fact, nobody likes ANYONE. IT'S A LOVELESS WORLD! Nobody likes BILL lacks a backgrounded focal entailment since it is interpreted along the lines of ``Nobody is such that some relevant event involving him or her and Bill is a liking event.'' Thus, although I have concentrated on how focused quanti®ers a¨ect the existential import of the nonfocused material in a sentence, my account also captures the fact that nonfocused quanti®ers have analogous e¨ects. This is not the case with the presuppositional approach, which is concerned only with the e¨ects that focused quanti®ers have on focal presuppositions, o¨ering no predictions concerning the e¨ects of nonfocused quanti®ers.10 In summary, so far in this chapter I have argued that it is empirically not enough to say that the nonfocused part of a sentence merely expresses what is ``a function under discussion'' at the point in the discourse where the sentence occurs. Nor does it seem entirely satisfactory to say that the nonfocused part gives rise to a ``focal presupposition'' that is formed by existentially binding a type-matching variable that replaces the focus.

Negated and Nonnegated Sentences

29

Instead, I proposed that a better analysis becomes available if the nonfocused material is seen as restricting an event quanti®er. When and how the nonfocused part has existential import and sentences have what I called backgrounded focal entailments was reduced to scope interaction between the restricted event quanti®er and decreasing quanti®ers. I supplemented the proposal with the claim that quanti®er restrictions set the scene by encoding a relative degree of pragmatic aboutness. 2.7

Focus and Negation: Bound, Free, and Wide Readings

I would now like to extend the empirical scope of the analysis, while staying with negative contexts, which are the most interesting for our purposes. The overall claim I make in this and the following sections is that structured Davidsonian decomposition o¨ers a simple account of the interaction between focus and negation. On this account negation remains a univocal propositional operator. The basic fact I will be considering is this. As many linguists have noted, there is more than one way to interpret a sentence like (31) (e.g., Jackendo¨ 1972, Jacobs 1991). (31) Sascha didn't visit MONTMARTRE. On its most salient interpretation, we understand (31) along the lines of (32); here not negates the focus, letting the rest of the sentence escape from its e¨ect. I will refer to this as the ``bound'' reading. (32) ``What Sascha visited wasn't Montmartre.'' [bound reading] Although the bound reading of (31) is the one that seems to come to mind ®rst, the sentence has another reading as well, namely, the ``free'' reading paraphrased in (33). Intuitively, in this reading the e¨ect of the negation is restricted to the verb. (33) ``What Sascha didn't visit was Montmartre.'' [ free reading] The two readings are quite di¨erent. Whereas in the bound reading the nonfocused part entails that Sascha did some sight-seeing, in the free reading the nonfocused part licenses the inference that he skipped some sight. Given this di¨erence in their backgrounded focal entailments and aboutness, it is not surprising that the two readings are felicitous in different pragmatic contexts. Thus, the bound reading appears when we are assuming that Sascha engaged in some sight-seeing and we are wondering

30

Chapter 2

what in particular he saw. The free reading becomes readily available if we imagine uttering (31) in a di¨erent context, one where it counts as established that Sascha is shunning the main tourist sites of Paris and we are wondering which one he managed to avoid on a particular, contextually salient occasion. There is a simple way to force the two di¨erent readings without construing pragmatic context, namely, by adding di¨erent continuations to the sentence. Continuing (31) with but . . . forces the bound reading; continuing it with and not . . . or but not . . . forces the free reading. (34) a. Sascha didn't visit MONTMARTRE, but THE LOUVRE. [bound reading] b. Sascha didn't visit MONTMARTRE, and not THE LOUVRE. (He DID in fact visit the Louvre.) [ free reading] Curiously, intonation also helps tease apart the two readings (Jackendo¨ 1972). A ``fall-rise'' contour, where the sentence melody rises at the end, typically (though not always) signals a bound reading. A ``fall'' contour forces the free reading. Why this should be so is a question I would like to set aside for the time being. I will take it up again in the appendix to this chapter. It is inviting to attribute the di¨erences among bound, free, and wide readings to the scope of the negation (see HajicÏova 1984, Jacobs 1991). A scopal account of the various readings is readily available on the analysis I am proposing; because of the extensive Davidsonian decomposition this account assumes, which includes separation, the various readings reduce to scope interactions where negation can remain a propositional operator regardless of the semantic nature of the element that is e¨ectively negated (visit vs. Montmartre). Beginning with the bound reading, this interpretation results when the negation e¨ectively takes scope over the conjunct containing the focus. (35) [be: C(e) & Visit(e) & Agent(e,sascha) & Past(e)] sTheme (e,montmartre) & Visit(e) & Agent(e,sascha) & Past(e) [bound reading] ``Some (relevant) visit by Sascha yesterday was such that it was not a visit by Sascha of Montmartre.'' In (35) the conjuncts containing the nonfocused material form the restriction of the event operator. Since the negation only takes scope over the conjunct interpreting the focus, the nonfocused material escapes its

Negated and Nonnegated Sentences

31

e¨ect. This way, (35) captures both that the bound reading of (31) is about Sascha doing some visiting yesterday and that the bound reading entails that such an event took place. Incidentally, the backgrounded focal entailment of the bound reading of (31) is the same as that of the nonnegated counterpart of (31), Sascha visited Montmartre yesterday. I will come back to this later. Unlike in the bound reading, where the negation takes scope over the restricted event quanti®er, in the free reading the negation takes scope only over the verbal conjunct, as in (36). As a result of this scope constellation, we have a negative description of an event as a ``not visiting.''11 (36) [be: C(e) & sVisit(e) & Agent(e,sascha) & Past(e)] Theme(e,montmartre) & sVisit(e) & Agent(e,sascha) & Past(e) [ free reading] ``Some (relevant) event of not visiting by Sascha was an event of his not visiting Montmartre.'' Summarizing, (35) and (36) illustrate how the bound and free readings result from the negation's taking di¨erent scope: whereas in the bound reading the negation takes scope over the focus, in the free reading it takes scope over the verb. Since the verb and the focus both form conjuncts in the semantic translation, negation remains a propositional operator in both cases. Support for making the contrast between the bound and free readings follow from scope comes from languages where the overt syntax is scopally more revealing than in English. German is to some extent such a language. As the following minimal pair illustrates, in German the bound reading and the free reading are distinguished by word order in the ``Mittelfeld.'' In particular, if a focused phrase like Montmartre follows the negation, only a bound reading is possible and a but continuation seems necessary; see (37). On the other hand, when the focused noun phrase precedes the negation (i.e., if it is scrambled), the sentence has only the free reading. It is sometimes thought that scrambled XPs in German cannot be focused, but (38) suggests that this is not the case. (37) . . . daû der Sascha gestern nicht MONTMARTRE besichtigt that the Sascha yesterday not Montmartre visited hat, sondern DEN LOUVRE/aund nicht DEN has but the Louvre/aand not the LOUVRE. [only bound reading] Louvre

32

Chapter 2

(38) . . . daû der Sascha gestern MONTMARTRE nicht besichtigt that the Sascha yesterday Montmartre not visited hat, und nicht DEN LOUVRE/asondern DEN has and not the Louvre/abut the LOUVRE. [only free reading] Louvre In relevant respects similar examples can also be found in languages with focus movement such as Hungarian and Basque. For instance, the (a) examples in (39)±(40) are reported to have the bound reading. When the relation between focus and negation is reversed, as in the (b) sentences, the interpretation shifts to a free reading.12 veri Peter. (39) a. Nem MARIAT not MARY(acc) beats Peter ``It isn't Mary who Peter is beating.'' b. MARIAT nem veri Peter. MARY(acc) not beat Peter ``It is Mary who Peter is not beating.'' (40) a. Ez da MIREN etorri. not aux MIREN come ``It wasn't Miren who came.'' b. MIREN ez da etorri. MIREN not aux come ``It was Miren who didn't come.'' Given that the overt syntax of Hungarian and Basque is more revealing than that of English, it is not surprising that the di¨erence between bound and free readings in these languages should be manifested in word order if the di¨erence is indeed scopal, as the present account suggests. Along with the bound and free readings that are observed in the literature (e.g., Jackendo¨ 1972), (31) seems to have a third interpretation. Like the bound reading, this ``structured wide reading'' requires a fall-rise contour. Unlike in the bound reading, however, in the structured wide reading the e¨ect of the negation is not restricted to the focus, but seems to extend across the entire sentence. (41) ``It is not the case that what Sascha visited yesterday was Montmartre.'' The structured wide reading can be found in a scenario such as (42), for example.

Negated and Nonnegated Sentences

33

(42) A: Yesterday, Sascha visited MONTMARTRE. B: No, Sascha didn't visit MONTMARTRE, because he in fact didn't visit ANYTHING! He FELT SICK AND STAYED IN THE HOTEL ALL DAY. The ®rst sentence in (42B) states that it is not the case that some visit of Sascha yesterday was a visit of Montmartre. And, as the second sentence makes clear, this is so because Sascha did no sight-seeing at all. In other words, B claims that A is mistaken because A's utterance entails something in its background that is not right, namely, that Sascha ventured out of the hotel to take in some sights.13 The representation for the structured wide reading is this: (43) s[be: C(e) & Visit(e) & Agent(e,sascha) & Past(e)] Theme(e,montmartre) & Visit(e) & Agent(e,sascha) & Past(e) [structured wide reading] ``It is not the case that any (relevant) visit undertaken by Sascha was a visit of his to Montmartre.'' Focus in (43) does have an e¨ectÐas usual, it structures the event description, as a result of which the sentence is taken to be about Sascha's sight-seeing. Because of the wide scope of not, however, the event quanti®er does not have existential import. This means that not in (43) functions in the same way as the wide scope decreasing quanti®er does in the example Nobody likes Bill; by taking wide scope, it undermines any backgrounded focal entailment. A last descriptive point. A reading similar to the structured wide reading becomes available if we stress the negation itself (or the auxiliary supporting it). As in the structured wide reading, in this reading the negation takes scope over the entire sentence. With the stress falling on the negation itself, however, there is no additional focus e¨ect. I will call this fourth kind of reading the ``unstructured wide reading.'' (44) Sascha DIDN'T visit Montmartre yesterday. (45) s[be: C(e)] Visit(e) & Agent(e,sascha) & Past(e) & Theme(e,montmartre) [unstructured wide reading] ``It is not the case that there is some (relevant) event where Sascha visited Montmartre yesterday.'' The unstructured wide reading can appear in a context like this: (46) A: I bet SASCHA VISITED MONTMARTRE YESTERDAY! THAT'S why there was such a tra½c jam!

34

Chapter 2

B: No, Sascha DIDN'T visit Montmartre yesterday. HE COULDN'T HAVE! HE IS STILL IN MOSCOW.14 Finally, one may wonder how the di¨erent scopes that not takes on the analysis proposed here come about. If LF is scopally disambiguating, then presumably there should be an LF process that distinguishes between negation taking scope over an entire clause (wide reading), negation taking scope over the verb (free reading), and negation taking scope over the focused part of the sentence ( bound reading). What the exact nature of this process would be is something I cannot address here, but its existence would ®t well with the present view of things. 2.8

What Is the Negation of a Sentence?

Having seen four ways in which the sentential negation not can be interpretedÐthe bound, free, structured wide, and unstructured wide readings Ðwe may wonder whether all four have equal status, or whether we should single out one as the paradigmatic ``negation of a sentence.'' If we set aside the unstructured wide reading, which di¨ers from the other readings in requiring stress on the negation, then I think there may indeed be reason to say that one of the remaining three readings is more paradigmatic than the others; often, when we refer to the ``negation of a sentence,'' we seem to have in mind a reading that amounts to the bound reading, where the negation takes scope only over the focus and the nonfocused material escapes its e¨ect, thus giving rise to a backgrounded focal entailment. The bound reading appears to be the default in that the other readings are pragmatically more marked. The structured wide reading seems to be pragmatically quite special in that the nonfocused material is backgrounded, in accordance with the previous discourse, but, contrary to the previous discourse, the nonfocused material does not give rise to a backgrounded focal entailment. The free reading also seems quite marked, I think because it involves a negative description of an event. As we saw earlier (see note 11), unless such descriptions are interpreted relative to a rich context, they run the danger of being vacuously true of all sorts of irrelevant things. The structured wide reading and the bound reading being, each in its own way, pragmatically special, we are left with the bound reading. Here the negation neither gives rise to a negative description nor takes scope over already backgrounded material; instead, it negates only the conjunct(s) containing the focus.

Negated and Nonnegated Sentences

35

As mentioned earlier, the nonnegated version of (31) and the bound reading of the negated version share the same backgrounded focal entailment. Taking the bound reading to be the prototypical negation of a sentence now o¨ers an account of why a sentence and ``its negation'' are traditionally held to have the same presupposition and therefore to be felicitous in parallel discourse environments: although the assertions of a nonnegated sentence and its negated counterpart on the bound reading crucially di¨erÐone is not negated, the other isÐtheir backgrounded focal entailments are the same, and consequently so are many of the demands they impose on the previous discourse to form a felicitous utterance. Thus, on the present analysis a sentence and ``its negation'' seem to share the same ``presupposition'' not because there are focal presuppositions that project across a wide scope negation but because a sentence and ``its negation'' are felicitous in parallel discourse contexts because they have the same backgrounded focal entailments; they do so because by the ``negation of a sentence'' we typically mean its bound reading. 2.9

De®nite Descriptions, Focus, and Negation

With the analysis of focus and negation relying so heavily on the varying scope of not, it is interesting to see whether and how this interacts with the interpretation of de®nite descriptions. According to the Russellian view, singular de®nite descriptions like the king of France are quanti®ers (e.g., Russell 1905, Neale 1990). On a relational view of determiners, we can say that the de®nite determiner the functions like other binary determiners (e.g., most, all ) in that it is equivalent to a relation between the restriction A and the scope B. (47) a. [the A] B is true i¨ [the x: A(x)] B(x) b. [the x: A(x)] B(x) i¨ A J B & jAj ˆ 1 (47a) disquotes a de®nite description. (47b) relates the disquotation to an equivalence, which states that everything that is an element of the restriction B of D is also an element of the scope A. Moreover, there is exactly one such thing. We can now say that what was noted earlier concerning the pragmatic status of quanti®er restrictions applies here as well: the restriction of the encodes what the relevant part of the sentence is about. Much like the quanti®cational analysis of focus that has been proposed here, the quanti®cational analysis of de®nite descriptions maintains that

36

Chapter 2

when a description like the king of France appears in an environment that is not downward entailing, it will have existential import; it will give rise to an entailment that there is a king of France. Conversely, when the description appears in the scope of a negative element, no backgrounded entailment that there is a king of France is available and the sentence can be true, even if pragmatically misleading, simply because there is no king of France. Using what we now know about how negation interacts with focus, we can make certain predictions regarding the behavior of de®nite descriptions in negated sentences (see Jacobs 1991, HajicÏova 1984): in a sentence with a wide reading a de®nite description will not have existential import, since on this reading everything ends up inside the scope of the negation. Similarly, no existential entailment will be available if the de®nite description is part of the focus and the sentence is interpreted as having a bound reading because, again, the quanti®er will end up in the scope of the negation. On the other hand, in a sentence that exhibits a bound reading and where the de®nite description is not (part of ) the focus the description will have existential import, so that if nothing satis®es the description the sentence will be false. The next examples are intended to illustrate that these predictions are correct. Consider (48) and (49). (48) The king of France ISN'T bald. There IS no king of France! (49) s[be: C(e)] [the x: King-of-France(x)] & Bald(e) & Theme(e,x) [unstructured wide reading] (48) represents the paradigmatic Russellian example. It shows how a de®nite description that fails to hold of anything can be interpreted inside the scope of negation, resulting in the value ``true'' for the sentence. Unlike the standard version of this example, however, (48) makes it clear that on a natural reading of this sentence the negation is stressed. As we have seen, this type of focus assignment with stress on the negation causes the sentence to have an unstructured wide reading. As a result of this kind of reading, the de®nite description is interpreted inside the scope of the negation and the sentence does not entail that France has a king. Put di¨erently, if France has no king, as the second sentence asserts, then the ®rst sentence is true. Now consider (50). (50) The king of France isn't BALD, because the king of France isn't ANYTHING. There IS no king of France!

Negated and Nonnegated Sentences

37

Whereas (48) is an instance of an unstructured wide reading, in (50) the ®rst clause has a structured wide reading, along the lines of ``It is not the case that some event involving the king of France is a being bald.'' Since in the ®rst clause of (50) the de®nite description is interpreted inside the scope of the negation, the sentence is correctly predicted not to entail that there is a king of France. The ®rst clause of the ®rst sentence states that it is false that something that can be attributed to the king of France is baldness (structured wide reading). Then the second clause of the ®rst sentence denies that there is any property at all that can be attributed to him. This, in turn, explains the coherence of the second sentence, which explicitly denies the existence of a French king. If there is no king of France, there is nothing that can be predicated of him: he is neither tall, nor short, nor a hemophiliac, nor, in particular, bald.15 The third type of case, where the de®nite description itself is focused and the sentence has a bound reading, is given in (51). What we expect is that here, too, the sentence will not entail that France has a king because the de®nite description ends up in the scope of not. (51) THE KING OF FRANCE isn't bald. There IS no king of France. The person who is bald is THE PRESIDENT OF THE REPUBLIC! (52) [be: C(e) & Bald(e)] s[the x: King-of-France(x)] Theme(e,x) & Bald(e) [bound reading] It is only because the ®rst sentence does not entail the existence of a French king that the second sentence, which explicitly denies the existence of such a king, makes sense and o¨ers an explanation of the ®rst sentence, rather than contradicting it. The third sentence then takes up another aspect of the ®rst sentence, namely, its backgrounded focal entailment that someone other than the French king is bald (see (52)). It asserts that the person who can be so described is actually the president of the republic. Finally, looking at a bound reading where the de®nite description is not part of the focus, we ®nd that, consonant with our predictions, there has to be someone of whom the description is true; otherwise, the sentence is false. (53) Clinton didn't give A MEDAL to the king of France, but A COLLECTION OF EARLY AMERICAN PAINTINGS. As its continuation with but ensures, in (53) the negation takes scope over the focused theme, but not over the rest of the sentence (ˆ bound reading).

38

Chapter 2

Unlike in the other examples we have seen, here the de®nite description ®nds itself outside the scope of the negation. Consonant with what we expect, (53) is indeed false, and it is false because part of what it entails in its background is false: throughout Clinton's life France has been a republic. Summarizing, existential descriptions of events and de®nite descriptions of individuals behave similarly in that their restrictions encode what the relevant part of the sentence is about and have existential import only if they occur outside the scope of a negative element. If they occur inside, the aboutness is preserved but the backgrounded existential entailment disappears and the sentence is true (even if pragmatically marked) if nothing satis®es the restriction. The present account adds predictive power to a quanti®cational analysis of de®nite descriptions in that it shows that whether or not a de®nite description is interpreted in the scope of negation systematically depends on how negation interacts with focus. 2.10

Other Negative Contexts

The negative elements considered so far are decreasing quanti®ers (in particular, nobody, nothing) and the negation not. Our analysis should also explain the distribution of backgrounded focal entailments in negative or downward-entailing contexts that are created by other negative elements. When an existential event quanti®er appears in the scope of, for instance, unlikely, deny, or ifÐwhich also create downward-entailing environments Ðthe sentence should lack a backgrounded focal entailment. Let us consider (54) (an example due to Rooth (1994a)). (54) A: Did anyone win the football pool this week? B: Probably not, because it's unlikely that MARY won it, and she is the only person who ever wins. (54B) does not suggest that the football pool was won. Why? As mentioned earlier, Rooth (1994a) considers the possibility of a presuppositional analysisÐin particular, an analysis where the type-matching variable that on the alternative semantics view replaces the focus is bound existentially rather than with a lambda-operator. Rooth then points out the following problem. If presuppositions project across negative elementsÐ and, following standard views, he assumes they doÐthen (54B) should have a focal presupposition. It should presuppose that someone won the football pool, but it clearly does not.

Negated and Nonnegated Sentences

39

To overcome this problem, Rooth contemplates treating the focused proper name not as a referring expression but as a quanti®er, which ``trivializes'' the focal presuppositions of (54): all that is predicted to be presupposed is that some, possibly decreasing quanti®er applies to what is denoted by the nonfocused part of the sentence. In other words, the presuppositional analysis handles the lack of focal presupposition in (54) the same way it handles the lack of focal presupposition in (7) (NOBODY likes Bill ); an existential commitment to there being some quanti®er that applies to a predicate is not a commitment to there being an individual that satis®es the predicate. At this point a question arises. If we trivialize the presuppositions in (54) by treating Mary as a quanti®er, what stops us from doing the same in a case like (1a)? (1) a. ROSALIA wrote a poem. Someone who subscribes to the presuppositional analysis would presumably not want to give up the claim that (1a) presupposes that a poem was written. Yet it seems that by ``trivializing'' focal presuppositions, we would immediately lose on one side what we have gained on the other: the analysis of (54) would undermine a plausible analysis of (1a). The analysis of (54) resulting from the present account does not have this di½culty. Just as in the structured wide reading of a negated sentence, unlikely takes scope over the entire that-clause, which is internally structured by focus. (55) unlikely [be: C(e) & Win(e) & Past(e) & Theme(e,it)] Experiencer(e,mary) ``It is unlikely that some (relevant) event of winning had as its experiencer Mary.'' Like negation, unlikely creates a negative, or downward-entailing, environment for everything that appears in its scope. Thus, It is unlikely that it will rain entails that it is unlikely that it will rain hard. Given the semantic nature of unlikely, it follows that the event description that appears in its scope in (55) need not pick out any event of winning. It then follows that even though sentence (54) is about winning the football pool, it does not entail that the football pool was actually won. The discussion of (54) is intended to show that the proposed account of when and where backgrounded focal entailments occur is not limited to negation and quanti®ers like nobody, but can be extended to other nega-

40

Chapter 2

tive contexts. (54) provides but one example of this. There are obviously other words that would be interesting to investigate in this connection: predicates like deny and doubt, the conjunction if, and so on. Although it is not possible to discuss all the relevant cases here, the proposed analysis makes a clear prediction. Whenever the restricted event quanti®er appears in a negative (downward-entailing) environment, then no matter what element creates this environmentÐbe it negation itself or some other element such as unlikelyÐthe sentence should lack a backgrounded focal entailment. 2.11

How the Semantics Relates to the Syntax

Now that we have seen how the proposed analysis handles focus in nonnegated and negated sentences, it may be a good moment to consider how the proposed semantic representations relate to syntactic representations. One type of analysis of focus maintains that in a language like English a focused constituent moves covertly to a clause-initial position, for instance, to a particular functional projection, Focus Phrase. This type of approach ®nds empirical support in the classic crossover argument for focus movement (Chomsky 1976).16 On the other hand, there are various reasons to think that focus does not involve LF movement of focused elements of this type (see Rooth 1985). Some of these reasons are syntactic, others semantic. Beginning with the syntactic considerations, one issue is that it is known that focus can a¨ect all kinds of elements, some of whichÐsuch as the adjective red in (56a) and the pre®x ex in (56f )Ðare not generally assumed to undergo syntactic movement. (56) a. b. c. d. e. f.

The police arrested an ex-convict with a RED shirt. The police arrested an ex-convict with a red SHIRT. The police arrested an EX-CONVICT with a red shirt. The police ARRESTED an ex-convict with a red shirt. THE POLICE arrested an ex-convict with a red shirt. The police arrested an EX-convict with a red shirt.

A second argument against covert movement of the focused element(s) is that the semantic e¨ects of focus are notoriously nonlocal and are not subject to syntactic island constraints (see Rooth 1985 and references cited therein). The nonlocality of focus e¨ects can be shown in many

Negated and Nonnegated Sentences

41

ways. For present purposes, we can consider the possible interpretations of (57). (57) [CP1 Mary told me about [ DP the rumor [CP2 that Bill had said [CP3 that SUE was going to India]]]]. This sentence can have three di¨erent readings. On the ®rst reading, the focus on Sue is interpreted within the most deeply embedded clause CP3 , as in (58a). But (57) has two other, more interesting readings: one where the event quanti®er that is a¨ected by focus is the one introduced in CP2 (that is, one clause higher up than the focus), and another where the focus structures the event quanti®er of the matrix clause, CP1 . These two readings are paraphrased with the help of clefts in (58b) and (58c), respectively. (58) a. ``Mary told me about the rumor that Bill had said that it was SUE who was going to India.'' b. ``Mary told me about the rumor that it was SUE who was such that Bill had said that she was going to India.'' c. ``It was SUE who was such that Mary told me about the rumor that Bill had said that she was going to India.'' Taking (58b), for example, we ®nd that the event description that is a¨ected by focus is the one introduced by the verb say in CP2 . Since the focus falls on the subject of the embedded CP3 , it cannot be assumed to undergo syntactic movement to interact with the event description it a¨ects (the event description introduced in CP2 ) because movement out of the subject position of a tensed clause is illicit. More drastically even, in (58c), where the focus a¨ects the event description introduced in the matrix clause, CP1 , the focus would presumably have to be moved out of the tensed clause and then across the complex NP. It might appear that the problem could be overcome by pied-piping the entire island that contains the focus, instead of moving just the focus (see Drubig 1994). However, unless some further movement of the focused element out of the pied-piped island is posited, a pied-piping analysis will not work because it predicts that it makes no semantic di¨erence what is focused inside an island; all sentences with varying foci within the same island should have the same interpretation, contrary to fact. Last but not least, a semantic argument against simply moving the focused element(s) to a clause-initial position is that this kind of movement, in itself, would not have the desired semantic consequences. In

42

Chapter 2

particular, it would not allow for a straightforward mapping into interpretations using structured Davidsonian decomposition. As it seems doubtful that simple movement of the focus to a clauseinitial position will o¨er the right LF structure for semantic interpretation, we are left with the question of how the semantic translations provided here do relate to LF structure. One possibility is that the mapping from a sentence's LF syntax to its truth-conditional equivalent does not entirely mirror the syntactic structure; on this approach the syntax-to-semantics mapping not only is a function of the covert syntactic structure but also is mediated by focus, which overrides syntactic hierarchy. This kind of analysis would be compatible with an in-situ analysis of focus, where focus does not a¨ect the LF syntax (cf. the tripartite mapping analysis of focus proposed in Partee 1991; also see Herburger 1997a). Although this type of analysis works, it invokes a mapping mechanism that has the power to override the normal syntax-semantics relation. This may not be the conceptually most desirable solution. Another possibility is to say that the syntax-semantic mapping is as we think it commonly is, transparent and a direct function of the syntactic structure and the lexical meanings of the words it contains. On such a view, focus a¨ects not the mapping itself but its input, the LF syntax. The idea is that much the way QR (or its minimalist equivalents) can be motivated on interpretive grounds in that it facilitates a semantically transparent interpretation where scope and quanti®cational structure directly depend on syntactic hierarchy, LF may have another process that rearranges syntactic hierarchy in such a way that structured Davidsonian decomposition mirrors the covert syntax. The choice between the previous option and this one is not easy because, at least at present, there is no overt syntactic evidence that would independently support the syntactic approach. However, because the syntactic approach seems more forthright and makes semantic interpretation easier, I will adopt it here. Assuming that focus a¨ects LF syntax, the next question is how it does so. On standard assumptions a quanti®er's internal argument is interpreted as its restriction and its external argument is interpreted as the scope of the quanti®ed phrase (e.g., Larson and Segal 1995). This phrase is dubbed ``QP'' in (59), although nothing much hinges on that name; all that matters for present purposes are the hierarchical relations. (Similarly, ``XP,'' ``WP,'' and ``ZP'' are merely abbreviatory devices.)

Negated and Nonnegated Sentences

43

(59)

Adopting the syntactic route, we can now say that the e¨ects of focus on quanti®cation that we have observed amount to there being a process P whereby the nonfocused material inside the c-command domain of quanti®er Q winds up contributing to Q's restriction in that it forms Q's internal argument. Let us call this process P ``focal mapping.'' (60) Focal mapping The nonfocused material in the c-command domain of Q also provides an internal argument for Q. The only kind of quanti®er I have considered so far is the tacit existential quanti®er over events (be). I am assuming that this quanti®er is already present in the syntax and not merely introduced through the semantic axioms. To make clear that it is phonologically silent, I represent it here as sometime in the syntax. (Analogously, the tacit generic operator would be represented as generally.) Until focal mapping applies, sometime is unary in that it has no internal argument. Focal mapping makes the quanti®er binary in that the nonfocused material in its external argument, which translates as its scope, also appears in its internal argument, which denotes the restriction. Thus, prior to focal mapping the LF representation of (61) looks like (62a); after focal mapping it looks like (62b). (61) Paul sometime ordered SALMON. (62) a. [sometime [Paul ordered SALMON]] b. [[sometime [Paul ordered]] [Paul ordered SALMON]] Focused subjects behave just like focused objects and other phrases relative to the event quanti®er. This suggests that the tacit event quanti®er sometime is interpreted in a position from which it c-commands the subject. As it does not matter for my purposes which is correct, I will remain agnostic about whether an adverbial quanti®er canonically appears between the subject and the verb phrase and then raises to clause-initial position at LF or whether the subject reconstructs to a VP-internal position.

44

Chapter 2

The structure that is the result of focal mapping in (62b) now translates into (63) in a transparent fashion. (63) [be: C(e) & Agent(e,paul) & Order(e) & Past(e)] Theme(e,salmon) & Agent(e,paul) & Order(e) & Past(e) One might doubt the claim that a quanti®er acquires an ``argument'' (the internal one) in the course of the derivation. The same, however, is the case with QR. A quanti®ed noun phrase that undergoes QR e¨ectively obtains part or all of one of its semantic argumentsÐnamely, its scopeÐ in the course of the derivation. Next, to see how focal mapping deals with the nonlocal e¨ects of focus, let us consider the reading of (57) that is illustrated in (58c), repeated here. Recall that in this reading the e¨ect of focus is not limited to the clause it occurs in, but reaches up to the matrix clause. This means that the quanti®er that is a¨ected by focal mapping is the topmost one. The relevant LF structure is thus as shown in (64). (57) [CP1 Mary sometime told me about [ DP the rumor [CP2 that Bill sometime had said [CP3 that SUE sometime was going to India]]]] (58) c. ``It was SUE who was such that Mary told me about the rumor that Bill had said that she was going to India.'' (64) [sometime [Mary told me about the rumor that sometime Bill had said that sometime was going to India] [Mary told me about the rumor that Bill sometime had said that SUE sometime was going to India]] Let us now extend the scope of the analysis a bit further. It seems descriptively accurate to say that every sentence has a focus. Often this focus falls on one constituent of a sentence, but it need not. Thus, focus can fall on an entire sentence, as is quite common with presentational sentences or sentences that are all assertion, often used after questions that pragmatically convey ``What happened?''17 In such instances the restriction of the event operator remains empty (except for the context predicate C, of course). (65) A: ANYTHING NEW? B: RICARDO sometime GOT OFFERED A JOB! The structure that focal mapping provides for (65B), shown in (66), gives rise to the semantic interpretation in (67).

Negated and Nonnegated Sentences

45

(66) [[sometime] [ RICARDO GOT OFFERED A JOB]] (67) [be: C(e)] Get-o¨ered(e) & Past(e) & [a x: Job(x)] Theme(e,x) & Experiencer(e,ricardo) ``Some relevant event was an event of Ricardo's getting o¨ered a job.'' In addition to consisting solely of one big focus, sentences can contain more than one instance of focus. These foci may associate with di¨erent quanti®ers (``multiple focus''), in which case focal mapping applies to di¨erent quanti®ers, or they may all associate with the same quanti®er (``discontinuous focus''). The latter results when focal mapping applies to just one quanti®er, as illustrated in (69). (68) PAUL sometime ordered SALMON, (IVETTE sometime chose TROUT). (69) [sometime [ordered]] [ PAUL ordered SALMON] (70) [be: C(e) & Order(e) & Past(e)] Agent(e,paul) & Theme(e,salmon) & Order(e) & Past(e) ``Some relevant past event of ordering was an ordering whose agent was Paul and whose theme was salmon.'' Although the proposed account o¨ers an analysis of various types of focus assignment, there is also a type that literally falls outside its scope, namely, focus in sentence topics. By sentence topics I mean constructions such as these: (71) Dr. Johnson, I didn't like. (72) As for Dr. Rothkopf, he really seems like a good dentist. There are subtle and perhaps not so subtle di¨erences between these kinds of constructions (see, e.g., Portner and Yabushita 1998 and references cited therein; see also note 5). One thing they have in common, though, is that when these topicalized phrases contain focus, its e¨ect is limited to the topicalization itselfÐthe focus does not structure the quanti®cation of the event operator of the clause. This would follow if sentence topics are necessarily structurally higher than the position that the quanti®er over events is interpreted in, which seems reasonable. In chapter 3 we will see that scrambled XPs in German behave similarly to sentence topics in the sense that they are interpreted outside the scope of the event quanti®er that they have scrambled across.18

46

Chapter 2

As soon as focal mapping is regarded as part of LF syntax, the question arises how it interacts with other LF processes, in particular with QR. We know that the scope quanti®ers take relative to the restricted event operator is crucial in predicting the right entailments; when the event quanti®er appears in the scope of a ``negative'' element, which may be a quanti®er, the nonfocused part restricting the event operator only encodes aboutness; it does not give rise to a backgrounded focal entailment. The wide scope negative quanti®er can be focused, as in (7), or nonfocused, as in (30). In either case the sentence will lack a backgrounded focal entailmentÐthat Bill is liked in (7), and that there is some liking in (30). (7) NOBODY likes Bill. (14) [nobody x] [be: C(e) & Like(e) & Theme(e,bill)] Experiencer(e,x) & Like(e) & Theme(e,bill) ``Nobody is such that some (relevant) event of liking Bill had him or her as its experiencer.'' (30) Nobody likes BILL. (73) [nobody x] [be: C(e) & Like(e) & Experiencer(e,x)] Theme(e,bill) & Like(e) & Experiencer(e,x) ``Nobody is such that some relevant event of his or her liking is an event of his or her liking Bill.'' The fact that in both (14) and (73) the negative quanti®er binds both into the restriction of the event operator and into its scope suggests that QR takes place prior to focal mapping. If focal mapping applied ®rst, the interpretation of (30) would contain two instances of nobody, one in the restriction of be and one in its scope; regardless of the scope they took afterward, the desired interpretation would not be available. One way of saying that focal mapping applies after QR is to say that whereas QR is a process of LF, focal mapping is a process of LF 0 (see May 1985). Finally, I want to note that the present analysis remains neutral about whether focal mapping is a primitive syntactic process of LF (LF 0 ) or whether it can be reduced to a syntactic process that is also manifested overtly. It has generally been deemed theoretically desirable to reduce semantically motivated movement such as QR to an overt syntactic phenomenon. Thus, at various stages of its history QR was likened to overt movement. Whereas Government-Binding accounts likened it to overt  wh-movement (A-movement), with which it shares various properties (cf. crossover phenomena), some minimalist approaches exploit simi-

Negated and Nonnegated Sentences

47

larities to Case movement (A-movement), which is like QR in being clause bound (e.g., Hornstein 1995). The fact that QR is neither exactly like wh-movement nor exactly like Case movement and yet attempts have been made to reduce it to one or the other illustrates, I think, how important syntacticians have considered the notion that covert movement is the same as overt movement, just invisible. Even if it is of considerable theoretical import whether there exists an overt counterpart for focal mapping, this is something that I am unable to establish here. Thus, just as QR can be argued for on semantic grounds, so the argument for focal mapping is at this point a semantic one.19 Summarizing, the overall picture that emerges looks like this. In languages like English focus on a syntactic phrase has a phonological re¯ex in that focus on any given phrase is generally realized as stress on its rightmost, most deeply embedded word. Focus is not a pragmatic or information structure phenomenon; instead, it has a direct e¨ect on the semantic interpretation of the sentence. In particular, focus rearranges LF structure through focal mapping after quanti®er scope is assigned. The rearranged structure then maps into translations involving structured Davidsonian decomposition. The event quanti®er that can be a¨ected by focus can be either the one that directly binds into the conjunct containing the focus (local cases) or any ``higher'' event quanti®er (nonlocal cases). A sentence has a backgrounded focal entailment exactly when the restricted event operator is interpreted in a nonnegative (non-downward-entailing) context. Negative contexts are created by wide scope decreasing quanti®ers (e.g., nobody), negation, and words like unlikely. Even in a case where a restricted event quanti®er appears in a negative context there is still a pragmatic asymmetry between the focused and the nonfocused material. As it contributes to the quanti®er's restriction, the nonfocused material expresses what the relevant part of the sentence is about; the focused part, on the other hand, contributes only to the assertion. 2.12

A Comparison with Earlier Analyses

No analysis exists in a vacuum. In a number of ways this analysis is indebted to previous ones. This is what I will try to show next, in necessarily brief fashion. The interpretation that structured Davidsonian decomposition accords to (1a), for instance, is e¨ectively equivalent to saying that some poem writing was done by RosalõÂa. One way of thinking about this is that the

48

Chapter 2

focus marks the main semantic predication in the sentence (in a particular sense of ``main predicate'') and the nonfocused part forms an event quanti®er that combines with the focused conjunct(s) much as an inde®nite quanti®er combines with a verb phrase on the traditional view. There are actually analyses according to which the focus is literally the main predicate of the sentence, whereas the nonfocused part forms the argument to which that predicate applies. In spirit, at least, structured Davidsonian decomposition can be viewed as similar to such directly ``predicational'' analyses of focus. Von Stechow (1991) observes that a directly predicational view of focus goes back at least to the nineteenth-century work of Hermann Paul, who termed the focus the ``psychological predicate'' and who contrasted it with the nonfocused part, which he thought of as the ``psychological subject.'' More recently a predicational approach has been pursued by Ogihara (1987) and LoÈbner (1990). On their accounts the denotation of the focused element is type-shifted into a function from individuals to truth-values (the type of the verb phrase in the normal case) and the nonfocused part is type-shifted into a de®nite description, where de®nite descriptions are thought of in a Fregean way, as referential expressions (type e). Thus, ROSALIA wrote a poem is interpreted as in (74). (74) lx [x ˆ rosalõÂa] (ix Wrote-a-poem(x)) Analyzing the nonfocused part as denoting a de®nite description is intended to capture the semantic and pragmatic e¨ects of focus by assimilating them to those of de®nite descriptions. Although this analysis works very well when the focused element is a referential expression, as in (74), it works less well when the focused element is a quanti®er. One possibility is to say that since a quanti®er takes the rest of a sentence as its argument, focus has no e¨ect when it falls on a quanti®er. NOBODY likes Bill would then be interpreted as if there were no focus on NOBODY, with the quanti®er simply applying to its scope (see LoÈbner 1990, von Stechow 1991). (75) [nobody x] (Likes-Bill(x)) If so, however, it is no longer the case that the nonfocused part denotes a de®nite description. This not only cuts into the generality of the analysis; it also does not predict the existence of a backgrounded focal entailment in examples where focus falls on a nondecreasing quanti®er (see (10)) and in examples where focus falls on a decreasing quanti®er but that quanti-

Negated and Nonnegated Sentences

49

®er takes narrow scope (see (25)±(28)). As a result, this version of the predicational analysis encounters empirical limitations similar to those of what was called earlier the presuppositional analysis. On a di¨erent version of the predicational analysis, one developed by Ogihara (1987), it is proposed that a focused quanti®er should be assimilated to a focused individual-denoting expression (e.g., a proper name) so that when focus falls on a quanti®er, the quanti®er is ®rst type-shifted into an individual-denoting expression. The individual-denoting expression is then type-shifted to a function from individuals to truth-values. Thus, a quanti®er that is focused goes from type hhe,t,i,ti to type hei, and then to type he,ti. This analysis predicts that only those quanti®ers that can be typeshifted into individuals can also be focused. Ogihara (1987) argues that this is essentially true as far as the Japanese ``obligatory focusing'' phenomenon is concerned. However, the prediction does not seem to extend to English. Although quanti®ers that are introduced by the determiners some, no(body), many, and most cannot generally be type-shifted to individuals, they can clearly be focused. We saw this earlier in the case of no(body) and many.20 (7) NOBODY likes Bill. (10) MANY OF HIS COLLEAGUES like Bill. The same holds true of other quanti®ers that cannot be type-shifted into individuals, like most and some in (76). (76) In order to decide how a corporation is run you have to own MOST/SOME SHARES. Thus, although in spirit structured Davidsonian decomposition resembles the predicational analyses of focus, the means it uses to derive the result are quite di¨erent. Various arguments have been given to show that it thereby o¨ers an empirically more ®ne-grained analysis of the distribution of backgrounded focal entailments. Whereas in the roles it assigns to the focused part and the nonfocused part structured Davidsonian decomposition resembles predicational approaches to focus, in its reliance on event semantics it resembles Larson and Lefebvre's (1991) analysis of focus in Haitian Creole. Essentially, structured Davidsonian decomposition di¨ers from Larson and Lefebvre's analysis only in that they hold that it is the focus (rather than the nonfocused part) that provides the restriction of the event operator, and,

50

Chapter 2

conversely, that it is the nonfocused part (rather than the focus) that provides the second argument of the event operator. On this view (1a) (ROSALIA wrote a poem) comes out not as in (77a), which is equivalent to (77b), but as its mirror image in (78). (77) a. [be: C(e) & Write(e) & Past(e) & [bx: Poem(x)] Theme(e,x)] Agent(e,rosalõÂa) & Write(e) & Past(e) & [bx: Poem(x)] Theme(e,x) b. [be: C(e) & Write(e) & Past(e) & [bx: Poem(x)] Theme(e,x)] Agent(e,rosalõÂa) (78) [be: C(e) & Agent(e,rosalõÂa)] Write(e) & Past(e) & [bx: Poem(x)] Theme(e,x) (Larson and Lefebvre 1991) Whereas (77) loosely parallels the predicational analysis, the analysis in (78) recalls the ``structured meaning'' approach, where the focus is interpreted as an argument of a function formed by the nonfocused part (see von Stechow 1991; also see Jackendo¨ 's (1972) analysis, summarized in section 2.2). Distributing the labor of focus as in (78), rather than as in (77), is supported by the fact that it ®ts in with a widely held analysis of clefts. This analysis maintains that in a (pseudo)cleft like (79) the clefted (and focused) element forms a semantic argument of a function denoted by the nonfocused part (e.g., Williams 1983, Partee 1986). (79) Who John saw was MARY. At the same time, saying that the focus forms the restriction of the event operator, rather than contributing to the matrix, makes it necessary to stipulate the pragmatic e¨ects of focus.21 It is also not clear how this kind of analysis would deal with the numerous facts that I have attributed to a scope interaction between the restricted event quanti®er and other scopetaking elements (not, decreasing quanti®ers, unlikely). Appendix: (Non)contrastive Focus and Intonation A1 The Fall-Rise Intonation of Bound Readings Examining the interaction between focus and negation reveals a striking relation between the type of intonation contour a sentence is pronounced with and the kind of reading it receives. When a bound reading is intended, the utterance involves a fall-rise contour (marked by ``@ ''), whereas when a free reading is intended, it typically involves a fall contour (marked by `` =''); see Jackendo¨ 1972.

Negated and Nonnegated Sentences

51

(80) a. Sascha didn't visit @MONTMARTRE yesterday. [ fall-rise contour] ``It wasn't MONTMARTRE that Sascha visited yesterday (but something else).'' [bound reading] = ONTMARTRE yesterday. [ fall contour] b. Sascha didn't visit M ``It was Montmartre that Sascha didn't visit yesterday.'' [ free reading] Two questions arise at this point. First, why do we ®nd this pattern, and, second, what does it mean for the analysis of focus? In principle, the relation between intonation and type of reading could be taken to suggest that there are actually two kinds of focus: a ``fall'' focus, which induces a free reading, and a ``fall-rise'' focus, which induces a bound reading. If so, this would create a conceptual problem for the present account of the contrast between (80a) and (80b), which, after all, reduces the meaning di¨erence to the scope of the negation, letting focus itself function the same way in both readings. It turns out, however, that there is no one-to-one correlation between intonation contour and type of reading. Although a bounded reading always seems to require a fall-rise intonation contour, and a fall contour is typical of a free reading, there are free readings that have a fall-rise contour. Put another way, the fall-rise intonation is not limited to bound readings, but also appears in nonnegated sentences. For example, (81) can be pronounced with a fall-rise intonation even though there is no negation with which the focus could interact in a bound reading. (81) Sascha visited @MONTMARTRE. An independent explanation is needed for this pattern. An important observation that I think helps explain this matter is o¨ered by Carlson (1984). He points out that a fall-rise contour creates the sense that the sentence has a continuation. The fall contour, on the other hand, creates a sense of informational completeness (also see Pierrehumbert and Hirschberg 1990). Speci®cally, Carlson says that ``[a] sentence with non®nal intonation cannot constitute an end point of a well-formed dialogue game'' (1984, 554). The relation between intonation and type of reading that is illustrated in (80) now becomes less of a mystery. A bound reading has an assertion one part of which is negated. Unless we have reason to think that we are dealing with a negatively described event, which is rare, we are left wondering whatÐif not the focused elementÐcould explain why the sentence has a backgrounded focal entailment. Put di¨erently, the negation in the

52

Chapter 2

assertion generally creates an explicit informational gapÐhence the signaling that the sentence has a continuation. All things being equal, in the free reading no continuation is required. Since the negation forms part of the quanti®er restriction, the assertion is not negated, and there is no intrinsic informational gap that leaves the hearer wondering; since the sentence is informationally complete in itself, a fall-rise intonation is not necessary. In this chapter we also saw that the structured wide reading parallels the bound reading in requiring fall-rise intonation; see (82) (ˆ (42)). (82) A: Yesterday, Sascha visited MONTMARTRE. B: No, Sascha didn't visit @MONTMARTRE, because he in fact = NYTHING! He FELT SICK AND STAYED IN didn't visit A THE HOTEL ALL DAY. This also ®ts with the observation that a fall-rise contour signals that there is going to be a continuation. Intuitively, the ®rst sentence of B's answer is meant to say that the issue of Sascha's doing some visiting yesterday may well be what's in the background (it restricts the event), but it did not take place, contrary to what A's comment entails in its background. However, for it to be clear that the ®rst clause in B's reply is meant to deny all of A's utterance on the grounds that its backgrounded focal entailment does not hold, a speci®c continuation is necessary, namely, . . . he didn't visit ANYTHING. It is the need for this continuation that explains the fall-rise contour on the ®rst clause, the one having the structured wide reading. A2 Contrast: An E¨ect of the Fall-Rise Contour To say that sentences with a ``free'' focus (including nonnegated sentences and negated sentences with a free reading) do not require a fall-rise contour does not mean that they cannot ever be pronounced that way, only that they need not be. After all, we saw that (81), which involves a free focus, can easily be pronounced with a fall-rise contour. As with the bound and the structured wide readings, a free focus with a fall-rise contour suggests that the sentence has a continuation. There seem to be two such types of continuation for sentences with a free reading. One is what Carlson (1984) calls the ``at least'' interpretation. (83) A: What did Sascha see? B: Well, he visited @MONTMARTRE. (And maybe some other places too.)

Negated and Nonnegated Sentences

53

Because a fall-rise contour in an utterance with a free focus can suggest that there are other events that di¨er from the one described only in the focused element, it has the e¨ect of making the focus nonexhaustive: it suggests that the value of the focus might not be the only thing that would make the rest of the sentence true (for (83), there may in fact be others: the Louvre, Notre Dame, etc.). If so, exhaustiveness is not a characteristic property of focus itself. A free focus with a fall contour is understood as being exhaustive only because the fall contour signals that the sentence is informationally complete; and it is informationally complete if it obeys the Gricean maxim of quantity, along the lines of ``Say as much as you = ONTMARTRE, the fall concan.'' So if a speaker says Sascha visited M tour signals that this is all there is to be said concerning his past visits, which in turn implicates that he visited no site other than Montmartre. Conversely, saying Sascha visited @MONTMARTRE leaves room for the possibility that he visited other Paris tourist sites as well. To summarize: One e¨ect a fall-rise contour has on a sentence with a free reading is that of signaling an ``at least'' interpretation. This overrides the conversational implicature that focus is exhaustive. Besides signaling an ``at least'' interpretation, the fall-rise contour on a free focus can signal a continuation of the type and not . . . or but not . . . (see Carlson 1984). (84) Sascha visited @MONTMARTRE (and/but not THE LOUVRE). The result is a contrastive focus. Consider also C's answer in (85), which is uttered with a fall-rise contour, and which has a distinctive contrastive ¯avor. It can be understood as something like ``It was Karen that she loaned her bike to; she did not loan it to Peter.'' This contrastive ¯avor is absent in (85B), which is uttered with a regular fall intonation. Consequently, (85B) is taken as a simple answer to A's question. Here the fall contour implicates that the focus is exhaustive. (85) A: Who did Hilde loan her bike to? B: She loaned it =TO KAREN. C: She loaned it @TO KAREN (but not TO PETER). If this characterization of contrastive focus is on the right track, then, contra Jackendo¨ (1972) and contra Rooth's (1985, 1992) alternative semantics approach, contrastiveness is not a property of focus. Contrastive focus is really just a subcase of a more general phenomenon, focus itself. It has the same basic interpretive properties as regular, noncontrastive focus, forming the main assertion of the sentence by contributing to the

54

Chapter 2

scope of the event quanti®er that is restricted by the nonfocused material. On top of that, the fall-rise contour adds that there will be a continuation that makes the sentence informationally complete (e.g., by excluding likely alternatives that are implicit in the preceding discourse). Similarly, exhaustiveness is also not an intrinsic part of the meaning of focus, but is really due to a fall contour signaling that the sentence is informationally complete. When a fall contour on a free focus is changed to fall-rise, a resulting ``at least'' interpretation undermines the exhaustivity of focus. Finally, it seems worth noting that if we set aside the ``at least'' e¨ect of the fall-rise contour, then this type of contour always suggests that there is a ``polarity-reversing'' continuation. If the sentence is negated, having a bound reading, the fall-rise contour signals a positive continuation, along the lines of ``not X, but Y''; and if the sentence is a½rmative, a fall-rise contour signals ``X, but not Y.'' Fall-rise contour thus seems like a tonal way of saying but. A3 More Complex Examples Some interesting, rather complex examples showing how a fall-rise intonation can also occur in nonnegated sentences are provided by Ladd (1980). Let us see how the analysis of the pragmatic e¨ect of di¨erent intonation contours works in these cases. (86) A: Did you feed the animals? B: I fed @THE CAT. (87) A: You have a VW, don't you? B: I've got @AN OPEL. (88) A: Do you want a glass of water? B: I'll have @A BEER. Ladd gives a rather complex characterization of the fall-rise intonation. He says it marks ``something like focus within a given set.'' It picks something out of a set of possibilities and focuses on it, but it speci®cally notes the connection of the set of possibilities to the context. Furthermore, Ladd's discussion suggests that the discourse antecedent that the focus element is related to tends to appear in the same (or a similar) focus frame. In addition, the focus is related to its so-called antecedent through some kind of subset relation. Evidence for the relevance of a subset relation is said to come from examples like (86). Here the focused element the cat denotes a subset of

Negated and Nonnegated Sentences

55

the noun phrase the animals, to which it is said to be anaphorically related. Even if the subset relation is not entirely obvious in (87) and (88), Ladd argues that it nonetheless exists. He suggests that the focus in (87) and (88) is interpreted to be anaphoric to what amounts to a discourse antecedent, which in these cases is derived through some kind of accommodation on the part of the hearer, namely, ``a German/small car'' and ``something to drink.'' Finally, it is the anaphoric connection of the fallrise contour that is held responsible for the polite softening that distinguishes B's reply in (88) from the same reply uttered with a simple fall intonation, which would be felt to be a rather abrupt rejection of an o¨er. This account recognizes the existence of fall-rise intonation in nonnegated sentences, but it is not obvious that it also explains why the fall-rise intonation is obligatorily found with bound readings in negated sentences. Ladd o¨ers the following suggestion, which does not seem e¨ective. Using John doesn't drink @BECAUSE HE'S UNHAPPY as an example, he states, . . . because he's unhappy is focused on as one reason out of a set of possible reasons; the combination of the negative, the focus on one reason, and the reference to a set of other reasons causes us to interpret the negative as associated with the focused reason, and we infer that John drinks, but not because he's unhappy. (Ladd 1980, 161)

The way in which the because-clause ends up becoming a subset would seem to allow for virtually any focus to function as a subset; hence, we should always expect a fall-rise focus, which is not the case. Furthermore, it is not clear why a subset understanding of the because-clause should force the negation to associate with the focus. Yet another issue needs to be addressed, relating to the analysis of fallrise intonation in nonnegated sentences. Even though a subset relation is present in (86), and arguably in (87)±(88), such a relation is not what characterizes a fall-rise contour in the general case. For example, (89) and (90) would presumably satisfy all of Ladd's requirements for a fall-rise melody: the focus is related through a subset relation to its antecedent; moreover, focus and antecedent share the same focus frame. Still, a fallrise tone is not needed hereÐindeed, it sounds odd. (89) A: What would you like to drink? = BEER/a@A BEER, please. B: I'll have A (90) A: You drive a small car, don't you? = N OPEL/a@AN OPEL. B: Yes, I drive A

56

Chapter 2

It seems then that Ladd's (1980) account of fall-rise intonation does not clearly explain why bound readings typically occur with this kind of intonation contour, nor does it explain the oddness of a fall-rise intonation in examples like (89) and (90). The account of fall-rise intonation given earlier can be argued to carry over to Ladd's examples in the following way. There is an important difference between (86)±(88) and (89)±(90). In (89)±(90) B's response does not con¯ict with A's question; in fact, B agrees with A. By contrast, in (86)±(88) B does not agree with A. Rather, preserving the backgrounded focal entailment of A's claim, B asserts that it is true not of what A suggests, but of something else. This other element may be a subset of the element focused in A's utterance, as in (86), but it need not be, as is in fact the case in (87) and (88). The point now is that whereas (89) and (90) require no continuation, (86)±(88) are intended to be continued along the lines of . . . but I DIDN'T feed the animals, . . . but I DON'T have a VW, . . . but I DON'T want any water, respectively. It is this need for a continuation that explains the fall-rise contour in (86)±(88). The polite softening that Ladd observes with a fall-rise intonation in these cases can thus be attributed to having left room for a continuation that answers the question, without having had to say ``no'' explicitly. When the continuations are actually pronounced, B's replies sound considerably less polite. (91) A: Do you want a glass of water? B: I'll have BEER, but I don't want any water. Also, if (86)±(88) are pronounced using the fall contour, no continuation is implicit, and the result is a bit rude. This is so because the answer has the pragmatic impact of saying ``no'' without acknowledging that the question that was asked was not answered directly, thus violating the turn-taking compliance that governs polite discourse (Barry Schein, personal communication). The claim that fall-rise intonation signals the need for a continuation also helps explain the di¨erent intonation patterns that appear in (92) (which is also due to Ladd). (92) A: Harry's the biggest fool in the state of New York. B: In @ITHACA, maybe. C: In =THE WHOLE WORLD, maybe. B's answer shows skepticism concerning A's claim. The fall-rise intonation suggests an implicit continuation along the lines of . . . but not in the

Negated and Nonnegated Sentences

57

state of New York. Unlike B's answer, C's answer indicates that C agrees with A's claimÐin fact, agrees with it emphatically. A must be right because something even stronger, which entails A's claim, holds true: if Harry is the biggest fool in the whole world, he necessarily is also the biggest fool in the state of New York. Since C very much agrees with A, nothing further needs to be said, so there is no need for a fall-rise contour. Let us next consider the subtle contrast between (93) and (94) (also due to Ladd). (93) A: I suppose it was pretty rough meeting all the linguists, wasn't itÐthere were probably a few who wouldn't talk to you. B: Hah! =MOST OF THEM wouldn't talk to me! (94) A: It wasn't so bad meeting all those linguists, was it. You didn't think anyone would talk to you, but it seemed like Beveldown and Wandervogel were being pretty friendly. B: Well, @MOST OF THEM wouldn't talk to me. Notice that in (93) B agrees with A's assessment; after all, not only were there a few who would not talk to B, most of the linguists actually ignored B. Since there is no disagreement in (93), there is no need for a continuation, and the fall contour is su½cient. By contrast, in (94) A and B do not agree. Speaker A suggests that B should not have worried that no one would talk to him, because, so she claims, some people actually did. But B replies that he was justi®ed in his worry because in fact most people would not talk to him. The continuation that the fall-rise contour conveys here can be understood along the lines of ``but it wasn't the case that many people would talk to me, as you seem to be suggesting.'' Interestingly, what is being disagreed with is not necessarily the literal meaning of the original sentence. As Ladd's example in (95) shows, the disagreement signaled by a fall-rise intonation can in fact be a disagreement with the sentence's conversational implicature (see Horn 1989). When we assert that someone was able to solve a problem, we implicate, by the Gricean maxim of relevance, that he solved it. (95) a. He was @ABLE to solve the problem (but he didn't solve it). b. She was @CLEVER ENOUGH to ®gure out the solution (but she didn't do it). In summary, I have adopted the view that the meaning e¨ect of di¨erent intonational contours is this: whereas the fall-rise contour suggests that the sentence should be understood as having a continuation, the fall

58

Chapter 2

contour conveys a sense of informational completeness. In a case where the assertion contains a negationÐin a bound reading of a negated sentenceÐa fall-rise intonation is obligatory because the sentence, as it stands, is not informationally complete. Similarly, the fall-rise contour is necessary in structured wide readings to make clear where the disagreement lies. But the fall-rise contour is not limited to contexts of negation. It can also appear in cases where the focus is free (in free readings of negated sentences, and in nonnegated sentences). There are two reasons why a continuation may be needed in such a case. One is to hint that there might be a whole list of elements that would also make the event description true, thereby undermining the exhaustivity of focus, which was explained in Gricean terms. The other is to explicitly rule out elements present in the discourse that one could have expected to take the place of the focused element. It is this kind of continuation that causes certain foci to be understood as contrastive. On this view, both exhaustivity and contrastiveness are not so much intrinsic properties of focus as e¨ects that result from the pragmatics of intonation contours.

Chapter 3 Adverbial Quanti®ers

3.1

Introduction

In the preceding chapter I showed that the e¨ect that focus has on the entailments and discourse requirements of nonnegated and negated sentences can be explained rather simply if focus is seen as imposing a certain amount of structure on tacit event quanti®cation. In this chapter I extend the empirical scope of structured Davidsonian decomposition to sentences where the quanti®er over events is not tacit, but is realized overtly as usually, always, rarely, never, sometimes, and the like. After showing how the analysis accounts for simple cases, I turn to more complex ones. In these cases there is more than one adverb in the clause; they include instances of second occurrence focus, which have received considerable attention in the recent literature. I also consider the restricting e¨ects of if-clauses and preverbal when-clauses. After showing how the quanti®cational structure of adverbs is a¨ected by focus, I discuss the quanti®cational variability e¨ect; I argue that its locus varies with the assignment of focus. Finally, I compare the analysis proposed here with two previous accounts of the quanti®cational structure of adverbs, the Mapping Hypothesis (Diesing 1992) and a situation-based account (von Fintel 1994). 3.2 Association with Focus: An Instance of Structured Davidsonian Decomposition One striking property of adverbial quanti®ers, by which they are said to di¨er from determiners, is that the syntax seems to remain noncommittal with regard to their quanti®cational structure. Instead, the task is left to the focus assignment. This fact was pointed out by Rooth (1985, 1995a)

60

Chapter 3

and has caught the eye of many semanticists since (see, e.g., de Swart 1991, Krifka 1992, von Fintel 1994). An important exception is Diesing (1992), who, as we will see, argues that the quanti®cational structure of adverbs primarily depends on syntactic factors. To see how strong the e¨ect of focus on adverbial quanti®ers is, we can compare the interpretation of (1a) with that of (1b). The two interpretations not only do not divide the sentence in the same way; they in fact result in di¨erent truth-conditions. (1) a. MEMBERS OF MINORITIES rarely trust the police. b. Members of minorities rarely trust THE POLICE. (1a) can be paraphrased as ``Rarely when someone trusts the police is that person a member of a minority.'' (1b), in contrast, means ``Rarely when a member of a minority trusts someone does he or she trust the police''. The same kind of contrast also distinguishes (2a) from (2b). (2) a. Louise always said hi TO ALBERT. b. LOUISE always said hi to Albert. (2a) is true if all of Louise's greetings were directed toward Albert. (2b), however, requires that all greetings that were directed toward Albert were initiated by Louise. If it so happens that Louise greeted someone other than Albert, (2b) can still be true (as long as he was not greeted by anyone else), but (2a) is automatically false. The simple generalization that emerges from these examples of ``association with focus'' is that the nonfocused part of a sentence restricts the adverb. Since adverbial quanti®ers are nothing but phonologically realized quanti®ers over events, structured Davidsonian decomposition as introduced in chapter 2 captures association with focus of adverbs in a straightforward way. (3) Structured Davidsonian decomposition All the nonfocused material inside the scope of the event quanti®er Q also restricts Q. The only di¨erence between the examples considered so far and the ones considered here is that here the quanti®er is not tacit, but phonologically realized (always, usually, seldom, often, never, etc.). In every other respect the two sets of cases are identical, as the interpretations in (4) indicate. (4) a. [all e: C(e) & Say-hi(e) & Past(e) & Agent(e,louise)] To(e,albert) & Say-hi(e) & Past(e) & Agent(e,louise)

Adverbial Quanti®ers

61

``All (relevant) events of Louise's saying hi were directed to Albert.'' b. [all e: C(e) & Say-hi(e) & Past & To(e,albert)] Agent(e,louise) & Say-hi(e) & Past & To(e,albert)] ``All (relevant) events of saying hi to Albert had Louise as their agent.'' The proposed analysis predicts that because of their di¨erent foci (2a) and (2b) will be appropriate under di¨erent circumstances. Whereas (2a) can be felicitously uttered in a discourse where what is at issue is Louise's greeting, (2b) is appropriate when the topic of conversation is Albert's being greeted. As (4a,b) are entirely analogous to the examples of focus discussed in chapter 2, it seems fair to say that structured Davidsonian decomposition as presented there provides a particular account of a wellknown fact about overt adverbial quanti®ers and extends it to cases where there is no overt quanti®er, but where on present assumptions there is in fact an unpronounced sometime that translates as be. To underline how similar these adverbs are to determiners when we assume quanti®cation over events, I render the adverbs as ``all'' (always), ``most'' (usually), ``few'' (seldom), ``many'' (often), ``no'' (never), and so on, in the interpretation. If we want to further spell out the meaning of these quanti®ers, we can say that [most e: F(e)] G(e) is true exactly when it is true that more than half of the things that are F are also things that are G. Similarly, [some e: F(e)] G(e) amounts to saying that there is a thing that is both F and G.1 3.3

When There Is More than One Adverb

Along with the kind of reading illustrated in (1) and (2), where the adverb associates with focus, there is a second, less salient kind. Here the overt adverb is not restricted by the nonfocused part but is instead interpreted within the nonfocused part. As a result, it contributes to the backgrounded focal entailment of the sentence. Consider (5), for example. (5) FRIEDRICH usually comes late. (5) can mean that usually if someone comes late, it is Friedrich (here usually ``associates'' with FRIEDRICH ). I will call this the bound reading, in analogy with the bound reading of negation. (5) can also mean that the one who usually comes late is Friedrich, a reading that arises when we are discussing who has the habit of showing up late. I will call

62

Chapter 3

this the free reading. As in (4), in the ®rst reading of (5) the nonfocused part restricts the overt adverbial quanti®er (usually). (6) [most e: C(e) & Come-late(e)] Theme(e,friedrich) & Come-late(e) ``Most (relevant) events of arriving late are events of arriving late that have Friedrich as their theme.'' In contrast, in the free reading it is not usually that is restricted by nonfocused material. Rather, the adverb seems to be embedded within the restriction of a higher event quanti®cation, one of the tacit kind, and it is the latter description that is structured by the focus. (7) [be: C(e) [most e 0 : C(e 0 )] Come-late(e 0 ) & Contain(e 0 ,e)] Theme(e,friedrich) & [most e 0 : C(e 0 )] Come-late(e 0 ) & Contain(e 0 ,e) ``Some relevant event e such that most relevant events e 0 are late comings contained in e is such that its theme is Friedrich (and most relevant events are late comings contained in it).'' This interpretation makes sense if we conceive of the event quanti®ed over by be as a serial event, one containing a number of other events, namely, the frequent late comings. Of this event it is asserted that its theme is Friedrich. Like the bound and the free readings of negation, these two readings are syntactically disambiguated in German, and also in Hungarian and Basque. For instance, the German (8a) has only the reading where Friedrich associates with the overt adverb. (8b), on the other hand, has only the other reading. By the logic of the earlier argument concerning negation, we can take this to suggest that the di¨erence between the bound and the free readings is due to scope. Unlike in the bound reading, in the free reading the focused element takes scope over the relevant adverbial quanti®er. (8) a. weil meistens DER FRIEDRICH zu spaÈt kommt because usually the Friedrich too late comes b. weil DER FRIEDRICH meistens zu spaÈt kommt In light of the proposed analysis of the free reading of (5), it would be interesting to ®nd an example where one overt adverbial quanti®er co-occurs with another. This would provide an instance where what we have said corresponds to a tacit be in (5) would actually be realized overtly; one adverb would be una¨ected by focus (analogous to usually in (7)) whereas the other one would be structured by focus (analogous to be in (7)). (9) is an example of this sort.

Adverbial Quanti®ers

63

(9) Rarely did BRIAN always interrupt. Usually, JACK would do that (that is, always interrupt). The ®rst sentence in (9) contains two overt adverbial quanti®ers, rarely and always. The focus on Brian associates with rarely. Rarely, in turn, includes in its restriction always, which itself is not structured by focus but whose restriction is entirely left to the pragmatics. The resulting interpretation says that ``Few events such that for all relevant events there was an interruption that formed part of that event had as their agent Brian''Ð few instances of always interrupting involved Brian. As the free reading of (5) and this interpretation of (9) show, all that the present account requires is that a focus that occurs in the c-command domains of various event quanti®ers directly a¨ects the quanti®cational structure of one of them. Importantly, the converse does not hold. Not all event quanti®ers, be they tacit or overt, need to have their quanti®cational structure shaped directly by a focus in their c-command domain. 3.4

Second Occurrence Focus

With this in mind we can now observe that sometimes an adverb seems to associate with something that is not a focus, at least not at ®rst sight. These are cases of so-called second occurrence focus. The phenomenon is illustrated in (10), which is an example that is prominently discussed, among other authors, by Partee (1991). (10) A: Eva only gave xerox copies to THE GRADUATE STUDENTS. B: (No,) PETR only gave xerox copies to the graduate students.2 The interest of this example lies in the fact that even though in B's reply the graduate students is not focused, curiously enough, only is interpreted as if it were associating with the graduate studentsÐjust as in (10A). What should we make of this kind of example? VallduvõÂ (1990) and Dryer (1994) maintain that similar examples pose a problem for the view that the quanti®cation structure of adverbs is directly shaped by focus. The particular kind of account that these authors take to be disproved by such examples is the alternative semantics account proposed by Rooth (1985, 1992). In an alternative semantics account, association with focus takes place in two stages. The sentence is interpreted as if no focus were present, receiving its ``ordinary semantic value.'' At the same time it is given

64

Chapter 3

a second semantic valueÐnamely, a ``focus-semantic'' oneÐformed by abstracting over the focus (see chapter 2). The e¨ects of focus on an adverb like always or only are now captured by using the focus-semantic value to restrict the context variable C of the adverb. As a result of this, the nonfocused part de facto restricts the adverb, even if the process is indirect.3 The problem is thatÐcontrary to what the alternative semantics analysis predictsÐin (10B) only does not seem to be associating with focus; after all, it associates with the graduate students, which is seemingly nonfocused. In addressing this issue, Rooth (1995a) disputes that there is no focus on elements like the graduate students in (10B) and that hence there is no association with focus between only and the graduate students.4 He claims, instead, that there is a phonologically realized ``second occurrence focus'' on the noun phrase. It is not marked the way a regular focus is, becauseÐunlike a regular focusÐit does not carry a pitch accent. But he cites experimental data showing that it does have a longer vowel and a greater waveform amplitude than a regular unstressed word. The graduate students in (10B) being focused after all, it can be argued to associate with only as a regular instance of association with focus. A question that immediately arises is why the second occurrence focus is not realized in the regular way with a high pitch accent. Clearly, it is possible to realize two pitch accents within a clause. (11) JOHN invited MARY, and SUE BILL. Examples like (12B), which presumably involve a second occurrence focus on Bill, raise a further issue (Dryer 1994, Rooth 1994b). (12) A: John only introduced BILL to Mary. B: He also only introduced Bill to SUE. The problem here is that only in (12B) is understood to be associating with Bill (as in (12A)). But it is not immediately obvious why the ``real'' focus SUE does not associate with only as well. This association would have the consequence that only quanti®es over introductions all of which are said to be introductions of Bill to Sue; instead, though, only seems to associate with Bill, leaving Sue to associate with also. Rooth (1994b) suggests that in (12) the problem can be solved by moving SUE out of the reach of only. However, as Rooth also points out, the same strategy would be necessary to account for (13), where it seems less plausible.

Adverbial Quanti®ers

65

(13) A: We only recovered [the diary entries [that MARILYN made about John]]. B: We also only recovered [the diary entries [that Marilyn made about BOBBY]]. In order to make sure that BOBBY does not associate with only, we would have to move it, too, out of the reach of only. Yet such movement would blatantly violate a syntactic island (a complex NP island). It is of course possible to bite the bullet and say that because of second occurrence foci, focus moves after all, even if this movement violates syntactic islands. This is what Rooth (1994b) in fact seems to advocate. Ironically, doing so would undermine one important argument for the alternative semantics approach, namely, the claim that precisely because it is an in-situ theory, it has no problem with nonlocal focus relations and need not have focus move across syntactic islands (see chapter 2). Another way of tackling the problem is explored by von Fintel (1994) (for related ideas, see VallduvõÂ 1990, Dryer 1994). This line of reasoning maintains that the e¨ects of focus on adverbs are rather indirect and pragmatically mediated. In some cases an adverb's quanti®cational structure happens to be molded by focus, but, crucially, that is not necessary and other factors may play a role as well. Von Fintel follows Rooth's analysis over wide stretches. Where his analysis seems to di¨er from Rooth's is in the claim that the relation between the focus-semantic value and the context variable C is even more indirect than in the standard alternative semantics account. He proposes that the focus-semantic value of a sentence, which denotes what is a function under discussion, is anaphoric to a ``discourse topic'' in the preceding discourse. At the same time the context variable C of the adverb is also an anaphor looking for an antecedent. It may turn out that this antecedent is in fact the discourse topic, which also serves as the antecedent for the focus-semantic value, in which case we get the illusion, it is claimed, that the focus-semantic value directly helps restrict the adverb. To illustrate, the association of MARY with always in (14) is accounted for as follows. (14) MARY always took John to the movies. Assuming that always takes sentential scope, the focus-semantic value of the sentence is lx Took-john-to-the movies(x). It is anaphoric to a discourse topic (e.g., the question Who took John to the movies?). This

66

Chapter 3

discourse topic, in turn, also happens to provide an antecedent for the context variable of always. (15)

Discourse topic Who took John to the movies? Always C

lx Took-john-to-the movies(x) ``focus anaphor''

The crucial point now is that it is not necessary for the context variable C of the adverb to connect with the discourse topic; that is just one option. It can also take other items as its antecedent. The idea is that this is what happens with only in (10B). (10) A: Eva only gave xerox copies to THE GRADUATE STUDENTS. B: (No,) PETR only gave xerox copies to the graduate students. The adverb only appears to be associating with the graduate students, not because the graduate students is in any way focused, but because the context variable of only in (10B) is anaphorically linked to the context variable of only in (10A). Assuming that only in (10A) takes scope over the VP, which is an option, it is restricted by givings of xerox copies. This restriction can now serve as the antecedent of the context variable of only in (10B). Consequently, this occurrence of only also quanti®es over givings of xerox copies, and not over giving xerox copies to graduate students, as desired. Although this conspiratorial analysis of association with focus is attractive because it seems ¯exible enough to account for examples like (10B), von Fintel (1994) shows that it ultimately faces a di½culty similar to the one facing the proposal considered by Rooth: it requires movement of focus out of islands. For there to be an anaphoric relation between the context variables of two adverbs, it is essential that the two adverbs contain the same nonfocused material in their scope. This, however, requires movement in examples like (12). (12) A: John only introduced BILL to Mary. B: We also only introduced Bill to SUE. Unlike in Rooth's account, where SUE had to be moved out of the reach of only in (12B), now Mary has to be moved out of the reach of only in (12A); if it is not, it does not follow that only in (12B) quanti®es only over

Adverbial Quanti®ers

67

introductions, which is what is needed. Instead, it will quantify over introductions to Mary, which would mean that introductions of Bill to Mary would be to Sue, which is not right. It would not be problematic to move Mary in (12A). The problem is that the same argument can be made for (13); to make sure that only in (13B) quanti®es over recovering diary entries, rather than recovering diary entries made about John, John would have to be moved out of the scope of only in (13A). This time, however, the movement would violate an island. (13) A: We only recovered [the diary entries [that MARILYN made about John]]. B: We also only recovered [the diary entries [that Marilyn made about BOBBY]]. Thus, although the conspiratorial analysis of association with focus initially seems to have exactly the right kind of ¯exibility to account for examples like (10B), a closer look reveals that it requires that a nonfocused element in a previous sentence sometimes move out of a syntactic islandÐa serious problem. What can we say then about second occurrence foci? I think they have two striking properties that we should take into account. First, they are phonologically realized, but barely so (recall Rooth's observation that they have longer vowels than regular unstressed words and a greater waveform amplitude). Second, their distribution is very limited in that they only seem possible when an utterance mimics the preceding one very closely. This is emphasized by Krifka (1995), for instance, who o¨ers the following paradigm: (16) Mary only supports AFRICAN-AMERICAN job candidates. (17) a.

So what? Even JOHN only supports African-American job candidates. b. ?So what? Even JOHN only supports AFRICANAMERICAN job candidates. c. ??So what? Even JOHN only supports Black job candidates. d. So what? Even JOHN only supports BLACK job candidates.

The contrast between (17a) and (17c) shows that a second occurrence focus is possible only when the second sentence mimics the ®rst, word by word.5 Furthermore, a second occurrence focus cannot be realized as a regular focus with a pitch accent; compare (17a) with (17b). (Contra

68

Chapter 3

Rooth, Krifka argues that there is no phonological marking on AfricanAmerican at all; but in light of Rooth's experimental observations, I do not think this is correct. Thus, I am assuming that African-American in (17b) is pronounced with the particular reduced marking of second occurrence focus.)6 Given structured Davidsonian decomposition and the two characteristics of second occurrence focus just notedÐits phonological weakness and its being restricted to mimicking contextsÐwe can now analyze the interpretation of the second occurrence focus in (10B) in the following manner. (10) A: Eva only gave xerox copies to THE GRADUATE STUDENTS. B: (No,) PETR only gave xerox copies to the graduate students. Both in example (10A) and in (10B) only quanti®es over the same thing, namely, the handing out of xerox copies. I assume there is indeed a focus on the graduate students in (10B), but because (10B) mimics (10A), the second occurrence focus is barely realized; it seems that the internal structure of the mimicked chunk (only gave xerox copies to the graduate students) is invisible to the rules of stress assignment and the whole forms only one cycle. The mimicking relation between the ®rst and second sentences also explains why the quanti®cation structure remains parallel, that is, why only in (10B) associates with the graduate students analogous to the manner in which only associates with THE GRADUATE STUDENTS in (10A) even though only the latter noun phrase is clearly stressed, whereas the former is barely marked, as it is a second occurrence focus. The event description that is structured by the focus is a tacit one that embeds an overt one in its restriction. So far everything is the same as in (8). What is di¨erent is that in (18) (the interpretation of (10B)) the embedded event quanti®er only is itself structured by the second occurrence focus on the graduate students. (18) [be: C(e) [all e 0 : C(e 0 ) & Give(e 0 ) & Past(e 0 ) & Theme(e 0 ,xerox copies)] Goal(e 0 ,the graduate students) & Contain(e 0 ,e) & Give(e 0 ) & Past(e 0 ) & Theme(e 0 ,xerox copies)] Agent(e,petr) & [all e 0 : C(e 0 ) & Give(e 0 ) . . . (18) states that some event such that all relevant instances of giving xerox copies were to graduate students and were contained in the event had as its agent Petr.

Adverbial Quanti®ers

69

To sum up, the proposed analysis does not require that every unary quanti®er that has a focus in its c-command domain associate with it, but only that focus structure some unary quanti®er in whose c-command domain it occurs at LF (see p. 43, note 2). Because second occurrence foci are viewed as the result of an utterance mimicking a previous one, the analysis o¨ers an explanation of why a second occurrence focus is phonologically reduced and why its association properties are preserved. Given this, it is now not so surprising that only in (10B) associates with the graduate students instead of something else. The same kind of account extends to the examples in (12) and (13). Only in (13B), for instance, associates with Marilyn because (13B) mimics (13A). As a result, it inherits the focus on Marilyn, which is therefore only realized very weakly. But the mimicking relation also forces the quanti®cation structure to remain parallel. Consequently, the weakly focused Marilyn must associate with only. No focus movement out of the island is needed. Krifka (1995) proposes a similar analysis, arguing that second occurrence foci are a so far unrecognized kind of anaphor. I think the intuition behind ``anaphor'' and ``mimicking'' is essentially the same. However, his analysis di¨ers substantially from mine in the following way: whereas he assumes that second occurrence foci have no phonological re¯ex, I have argued here that the weak phonological marking of second occurrence foci not only exists (see Rooth 1995a), but in fact is intimately tied to the mimicking.7 3.5

Embedded Clauses

Up to this point we have only considered examples where the focus and the adverb it associates with are clausemates. What happens when embedded clauses (e.g., when-clauses) enter the picture? One could think that when-clauses necessarily serve to restrict adverbial quanti®ers and that in this respect they behave like if-clauses (see below). As Rooth (1985) shows, this is actually not the case. Postverbal whenclauses that are focused do not restrict the adverb, but enter its matrix instead. Comparing (19a) and (19b), for instance, we ®nd that when the when-clause is focused, as in (19a), we are quantifying over events of Mary's reading, of which it is claimed that most of them take place when she is on the subway. On the other hand, when the matrix predicate is focused, as in (19b), it is the when-clause that restricts the adverb, and we

70

Chapter 3

are quantifying over Mary's being in the subway. Whereas (19a) tells us something about Mary's reading habits, (19b) tells us about what Mary does on the subway. (19) a. Mary usually reads WHEN SHE IS ON THE SUBWAY. b. Mary usually READS when she is on the subway. (20) a. [most e: C(e) & Read(e) & Agent(e,mary)] Read(e) & Agent(e,mary) & When(be 0 ,e) & Theme(e 0 ,she) & On-the-subway(e 0 ) ``Most (relevant) events where Mary reads are such that Mary reads when she is on the subway.'' b. [most e: C(e) & Theme(e,mary) & When(e 0 ) & On-the-subway(e 0 )] Agent(e,mary) & When(be 0 ,e) & Theme(e 0 ,she) & On-the-subway(e 0 ) & Read(e) ``Most events involving Mary when she is on the subway are events where Mary reads when she is on the subway.'' (I am glossing over the semantics of when, as they are not essential to the point. For relevant discussion, see for example Johnston 1994.) Cases where only a subpart of a postverbal when-clause is focused behave analogously; in (21), for instance, all the nonfocused material, including that of the when-clause, ends up restricting the adverb. (21) Mary usually reads when she is SITTING in the subway. ``Most (relevant) events where Mary reads when she is in the subway are such that she is sitting.'' Thus, embedded postverbal clauses introduced by when behave just like the that-clauses we encountered in chapter 2: when they are focused, or contain a focus, the conjunct(s) interpreting the focus form part of the quanti®er's scope, but not part of its restriction. Unlike postverbal when-clauses, preverbal when-clauses can never contribute to the scope of the adverb (Rooth 1985). Regardless of where we assign the focus in the when-clause in the following example, it cannot enter into the matrix of usually: (22) When Mary is on the subway, she usually reads. This suggests that preverbal when-clauses are outside the scope of an adverbial quanti®er in the main clause. Moreover, it seems they are interpreted de facto as if they were in the restriction. This is a property they share with if-clauses, except that if-clauses are said to be interpreted

Adverbial Quanti®ers

71

as restricting an adverbial quanti®er regardless of whether they surface preverbally or postverbally. In both (23a) and (23b), for instance, usually quanti®es over events of snowing, but cannot quantify over events of my staying inside. (23) a. If it snows I usually stay inside. b. I usually stay inside if it snows. The interpretation of these sentences can be stated like this: (24) a. ``Most events where every one-to-one related event is a snowing are such that I stay inside.'' b. [most e: C(e) & [Ee 0 : C(e 0 ) & One-to-one(e,e 0 )] Snowing(e 0 )] Stay-inside(e) & Theme(e,I ) To make sure that usually e¨ectively counts snowings, which is what the truth-conditions require, we have to assume that there is a relation that relates the events that are quanti®ed over in the if-clause to those described in the main clause in a one-to-one fashion. The interesting question now is what happens when focus appears in an if-clause or a preverbal when-clause. Since preverbal clauses cannot contribute to the scope of an adverb in the main clause but seem to e¨ectively enter the restriction of the adverb, we expect that the e¨ects of focus will be found only within the description in the restriction; the focus will shape the quanti®cational structure of that description, but will not enter into the scope of the overt adverbial quanti®er. Consider (25). (25) If THE IRS calls, Donald usually worries. Here, the direct e¨ect of focus is indeed limited to the embedded clause. Given structured Davidsonian decomposition, we can render its e¨ect as in (26). (26) ``Most events where every related event of calling has the IRS as its theme are such that they (the ®rst events) result in Donald's getting worried.'' Focus here a¨ects only the embedded clause, which restricts the adverb usually; the event description in the embedded clause is structured in the regular way. (27) [most e: C(e) & [Ee 0 : C(e 0 ) & One-to-one(e,e 0 ) & Call(e 0 )] Agent(e 0 ,irs) & Call(e 0 )] Worry(e) & Experiencer(e,donald)

72

Chapter 3

(27) says that most events such that all related callings come from the IRS are events that involve Donald's worrying. Even more natural than the focus assignment in (25) is one where there is also a focus inside the matrix clause. (28) If THE IRS calls, Donald usually WORRIES. This alters the quanti®cation structure only insofar as the nonfocused material in the scope of usually (Donald ) joins the restriction. (29) [most e: C(e) & [Ee 0 : C(e 0 ) & One-to-one(e,e 0 ) Call(e 0 )] Agent(e 0 ,irs) & Call(e 0 ) & Experiencer(e,donald)] Worry(e) & Experiencer(e,donald) ``Most relevant events where all one-to-one related calls are calls from the IRS that are experienced by Donald are such that they are worryings experienced by Donald.'' Summarizing, we have seen that postverbal when-clauses behave straightforwardly: their nonfocused material joins the restriction of an event quanti®er, whereas their focused material is interpreted as part of that quanti®er's scope. In contrast, preverbal when-clauses, which are not in the scope of the adverb, are interpreted as e¨ectively restricting the adverb. In case they contain focused material, it structures the event description denoted by the when-clause within the restriction of the quanti®er of the main clause. Regardless of the position they surface in, be it preverbal or postverbal, if-clauses seem to pattern with preverbal when-clauses. Although already quite complex, the picture that I have drawn concerning focus and when-clauses may still not be su½ciently accurate. Johnston (1994) argues that there are some instances where a when-clause is postverbal and contains a focus, but whereÐcontrary to what we might expectÐthe nonfocused main clause contributes not to the adverb's restriction but to its scope. According to Johnston, this is the case when the predicate in the main clause is atelic, as in (30). (30) Marty is always in the shower when he SHAVES. The observation is that (30) can only mean that all events where Marty shaves are events where he is in the shower; it cannot mean that on every occasion where Marty is in the shower, there is an event where he shaves. (30) minimally contrasts with (31), which has the second reading. (31) When he SHAVES Marty is always in the shower.

Adverbial Quanti®ers

73

In line with VallduvõÂ (1990) and Dryer (1994), Johnston argues that instances of second occurrence focus show that the quanti®cational structure of adverbs is not directly sensitive to focus. He takes (30) to point in the same direction. We saw earlier that the lack of association with focus in cases of second occurrence focus is only apparent. In contrast, Johnston's observation may indeed indicate that besides focus, aspect can directly a¨ect quanti®cational structure. Apart from noting this possibility, I will not pursue it further here as the scope of my study is limited to the e¨ects of focus.8 With the e¨ects of focus in restricting clauses in mind, I would now like to turn to a di¨erent but ultimately related phenomenon exhibited by adverbial quanti®ers that has received a great deal of attention in the literature, namely, the quanti®cational variability e¨ect. 3.6

Quanti®cational Variability of Inde®nites

So far we have seen how the quanti®cational structure of adverbs like always, rarely, and usually interacts with the way focus is assigned and to what extent this directly follows from structured Davidsonian decomposition. One phenomenon that depends on the quanti®cational structure of adverbs is the quanti®cational variability e¨ect (QVE) exhibited by inde®nite noun phrases. Given this, we now expect that there will be a direct relation between focus assignment and the QVE. After summarizing what the QVE looks like and how it was ®rst explained, I will show that this relation indeed exists. As is well known, Lewis (1975) notes that in a sentence like (32) the interpretation of the inde®nite noun phrase riders is linked to the interpretation of the adverb seldom. (32) Riders on the Thirteenth Avenue line seldom ®nd seats. This sentence can be true when during the few peak hours most of the riders in these hours do not get seats. It is irrelevant if during most other hours (namely, the o¨-peak hours, when there are few passengers) all of the passengers do in fact get seats. In other words, (32) is similar in meaning to Few riders on the Thirteenth Avenue line ever ®nd seats. Lewis takes the dependence of riders on seldom to show that seldom here is quantifying not over times, but unselectively over both times and riders. To this end, he proposes that inde®nites like riders are devoid of any quanti®cational force and that instead they introduce a free variable

74

Chapter 3

that can be unselectively bound by adverbs like seldom. (32) is now interpreted as in (33). (33) Few hx,yi where x is a rider on the Thirteenth Avenue line and y a stretch of time are such that x ®nds a seat at y In an extension of the Lewis-style analysis, Heim (1982) views adverbs as having a tripartite structure with a restrictive clause and a nuclear scope. Following Lewis, on Heim's account inde®nites are not quanti®cational but only introduce a free variable that needs to be bound somehow. All the inde®nites that are interpreted in the restrictive clause of the adverb are unselectively bound by it and hence show quanti®cational variability. As for inde®nites that only appear in the nuclear scope, they receive their quanti®cational force through existential closure; consequently, they do not show quanti®cational variability. We know now that which part of a sentence goes in the restriction of an adverb and which part marks its matrix depends crucially on focus. This also applies to Lewis's example given in (32); as (34) shows, by changing the focus assignment, we change the intuitive meaning of the sentence and, in particular, the locus of quanti®cational variability. (34) a. RIDERS ON THE THIRTEENTH AVENUE BUS seldom ®nd seats (unlike RIDERS ON THE LA SUBWAY). b. Riders on the Thirteenth Avenue bus seldom FIND seats (they often GRAB them from somebody). c. Riders on the Thirteenth Avenue bus seldom FIND SEATS (they usually HAVE TO STAND). (34a) is roughly equivalent to saying that few of those people who ®nd seats are riders on the Thirteenth Avenue bus, because, perhaps, the Thirteenth Avenue bus is notoriously overcrowded. In contrast, (34b) asserts that few matchups between a Thirteenth Avenue bus rider and a seat come about in a civilized wayÐby the seat's being found. (Instead, most riders on that bus are really aggressive and simply grab a seat from someone.) Finally, (34c), which has the ``unmarked'' focus on the verb phrase, has the reading that Lewis observes, where few riders on the Thirteenth Avenue bus are such that they ®nd seats. Here the interpretation of riders seems to depend directly on seldom; that is, riders shows quanti®cational variability. Another set of examples that illustrate how the quanti®cational variability of inde®nites depends on focus is given in (35).

Adverbial Quanti®ers

75

(35) a. A CLAUSTROPHOBIC usually avoids an elevator. b. A claustrophobic usually avoids AN ELEVATOR. c. A claustrophobic usually AVOIDS an elevator. Varying with the assignment of focus, in (35a) the adverb quanti®es over instances of avoiding elevators, in (35b) it seems to count claustrophobics, and in (35c) it seems to count pairs of claustrophobics and elevators. The Lewis-style analysis gives interpretations of the following sort: (36) a. Most hx,yi where x is a time, y is an elevator, and y is being avoided are such that the one avoiding at x is a claustrophobic b. Most hx,yi where x is a time, y is a claustrophobic, and y avoids something are such that what is being avoided at x is an elevator c. Most hx,y,zi where x is a time, y is a claustrophobic, and z is an elevator are such that y avoids z at x As this shows, the QVE is tied to the quanti®cational structure of the adverb, and this structure in turn depends on focus. On the unselective binding account, adverbs bind variables over both times and individuals; the e¨ects of focus on the QVE can be handled by mapping the nonfocused inde®nites into the restrictive clause and the focused inde®nite into the matrix. For a relevant proposal see Partee 1991; also see Rooth 1995b for discussion of the intricacies involved in combining an unselective binding account with an alternative semantics treatment of focus.9 I will not pursue this here; instead, I adopt a di¨erent view. All along we have been treating quanti®cational adverbs as selective quanti®ers over events, not as unselective quanti®ers over n-tuples. On this view, inde®nites are genuine quanti®ers and do not introduce free variables (also see chapter 5). It is not di½cult to capture the QVE on these assumptions; it follows naturally from the fact that when an inde®nite quanti®er is interpreted in the restriction of an event quanti®er, the assignments to the inde®nite's variable vary with the assignments to the event variable of the adverb (e.g., de Swart 1991, von Fintel 1994). Considering the examples in (35) again, along with their interpretations in (37), we see that since only the nonfocused material enters the restriction of the adverb, only a nonfocused inde®nite (not a focused one) shows quanti®cational variability. This follows straightforwardly from the interpretations provided by structured Davidsonian decomposition.

76

Chapter 3

(35) a. A CLAUSTROPHOBIC usually avoids an elevator. b. A claustrophobic usually avoids AN ELEVATOR. c. A claustrophobic usually AVOIDS an elevator. (37) a. [most e: C(e) & Avoid(e) & [an x: Elevator(x)] Theme(e,x)] [a y: Claustrophobic(y)] Experiencer(e,y) & Avoid(e) & [an x: Elevator(x)] Theme(e,x) ``Most (relevant) events of avoiding an elevator are events of a claustrophobic's avoiding the elevator.'' b. [most e: C(e) & [an x: Claustrophobic(x)] Experiencer(e,x) & Avoid(e)] [a y: Elevator(y)] Theme(e,y) & [an x: Claustrophobic(x)] Experiencer(e,x) & Avoid(e) ``Most (relevant) events of avoidance that a claustrophobic experiences are events of avoidance experienced by the claustrophobic that involve an elevator.'' As (37a) shows, in (35a) the interpretation of an elevator is tied to the interpretation of the event quanti®er usually because the inde®nite appears in the restriction of the event quanti®er such that in most cases of an elevator's being avoided, the one avoiding it is a claustrophobic. To the extent that in each case of avoiding an elevator a di¨erent elevator is involved, the interpretation of an elevator will vary directly with the assignments to e and hence will show quanti®cational variability. On this account it is not problematic that the nonfocused inde®nite an elevator is interpreted twice, once in the restriction and once in the matrix. This is so because theta-roles are assigned exhaustively (see chapter 1). As a result, the same elevator is picked out in both instances; there is no problem of ``requanti®cation'' here (see Rooth 1995b, von Fintel 1994; also see note 7). To highlight the uniqueness of the theta-roles, I use the in the paraphrases of the interpretations. Analogous to (35a), in (35b) the events quanti®ed over are those of a claustrophobic's avoiding (something); see (37b). As these examples illustrate, analyzing adverbial quanti®ers as selective quanti®ers over events, rather than as unselective quanti®ers over n-tuples, poses no problem for the analysis of the QVE. On the contrary, keeping the traditional analysis of inde®nites as quanti®ers, their quanti®cational variability follows directly from their being interpreted in the restriction of the event quanti®er, as a result of which the assignments to the variable of the inde®nite vary relative to the assignments to the variable of the

Adverbial Quanti®ers

77

event quanti®er. Whether or not an inde®nite ends up in the restriction of an event quanti®er depends on whether it is nonfocused or focused. If it is nonfocused, it goes into the restriction; if it is focused, it is only interpreted in the second argument of the event quanti®er. One well-known challenge for the unselective binding analysis of adverbs is posed by asymmetric quanti®cation in if-clauses, which gives rise to the proportion problem. As if-clauses are the paradigmatic restrictive clauses of adverbs, any inde®nite that appears in such a clause should be unselectively bound by the adverb. Examples like (38), however, are known to show that this is not the case. (38) If a farmer buys a donkey, he is usually poor. Here usually does not seem to count farmer-donkey pairs, but only farmers, or farmers buying something. Intuitively, (38) can be true if the only rich farmer buys the majority of donkeys and all the poor farmers, of whom there are many, buy just one donkey each. What forces the quanti®cation of usually to be asymmetric in this case? Or, put di¨erently, why does a farmer show quanti®cational variability, but not a donkey? One possible answer (mentioned by Heim (1990), who attributes it to BaÈuerle and Egli (1985)) is that the presence of the pronoun is responsible. Notice, for instance, the contrast between (38) and (39). (39) If a farmer buys a donkey, it is usually in good hands. Whereas (38) says something pertaining to farmers (or farmers buying something), (39) quanti®es over donkeys (or donkeys being bought). The di¨erence between the two is that in (38) the pronoun he is anaphoric to a farmer, whereas in (39) the pronoun it is anaphoric to a farmer. Given these data, there seems to be a correlation between an inde®nite's showing quanti®cational variability and its serving as the antecedent for a donkey anaphor. It may come as little surprise at this point that there is another di¨erence between (38) and (39), namely, a di¨erence in focus (Kadmon 1987, Heim 1990, Krifka 1992). Whereas in (38) a donkey is focused (or part of the focus) and a farmer is not, in (39) the opposite is true: a farmer is focused and a donkey is not. Compare (40) and (41) in this respect. (40) a. If a farmer buys A DONKEY, he is usually poor. b. If a farmer BUYS A DONKEY, he is usually poor. (41) If A FARMER buys a donkey, it is usually in good hands.

78

Chapter 3

The following paradigm indicates that (for whatever reason) pronouns prefer taking nonfocused noun phrases as their antecedents, rather than focused ones: (42) a. If a farmer buys A DONKEY, he is usually poor. b. aIf A FARMER buys a donkey, he is usually poor. c. aIf a farmer buys A DONKEY, it is usually in good hands. d. If A FARMER buys a donkey, it is usually in good hands. Yet, as Heim (1990) shows, when the focus on a noun phrase is legitimized through preceding discourse, a pronoun can easily be anaphoric to a focused noun phrase. (43) Drummers mostly live in crowded dormitories. But if a drummer lives in an APARTMENT COMPLEX, it is usually half empty. Crucially, in such a case the adverb's restriction contains the nonfocused inde®niteÐand not the inde®nite that serves as antecedent for the pronoun. (43), for instance, would be false if 200 out of 299 drummers who live in apartment complexes live in apartment complexes that are fully occupied (Heim 1990, 152). Similarly, it is argued that in (44) it is the nonfocused a donkey that shows quanti®cational variability, rather than the antecedent of the pronoun, A FARMER. (44) Donkeys that belong to peddlers generally are in miserable shape, whereas those that belong to farmers mostly have a comfortable life. The reason is that if A FARMER owns a donkey, he is usually rich (and uses tractors and other modern equipment for the hard work on his farm). This con®rms, then, that it is really not so much a noun phrase's functioning as an antecedent to a pronoun that forces the noun phrase to be interpreted in the restriction of the adverb, as its being nonfocused. There is an interesting relation between antecedence and the lack of focus, but, whatever the reason for the relation is, it can be overridden. On the analysis of adverbial quanti®ers developed here the asymmetric readings of if-clauses are expected. The asymmetric quanti®cation of the adverb is predicted by the way focus structures quanti®cation over events and by the restrictive power of if-clauses, which we have already seen in (25), which is interpreted as in (27).

Adverbial Quanti®ers

79

(25) If THE IRS calls, Donald usually worries. (27) [most e: C(e) & [Ee 0 : C(e 0 ) & One-to-one(e,e 0 ) & Call(e 0 )] Agent(e 0 ,irs) & Call(e 0 )] Worry(e) & Experiencer(e,donald) Since the events quanti®ed over by usually relate one to one to those quanti®ed over by if (Ee 0 ), it is as if the adverb in an example like (25) directly quanti®ed over the events quanti®ed over in the if-clause. It follows that an inde®nite in the nonfocused part of the if-clause will show quanti®cational variability, but not an inde®nite in the focused part. When (38) is pronounced with focus on the verb phrase (If a farmer BUYS A DONKEY, he is usually poor), it is interpreted as in (45). (45) [most e: C(e) [Ee 0 : C(e 0 ) & One-to-one(e,e 0 ) & [an x: Farmer(x)] Agent(e 0 ,x)] [a y: Donkey(y)] Theme(e 0 ,y) & Buy(e 0 ) & [an x: Farmer(x)] Agent(e 0 ,x)] Poor(e) & Theme(e,he) (45) states that most events where every one-to-one related event involving a farmer was a buying of a donkey were such that the farmer was poor. To summarize: As a number of examples in this section have shown, the QVE on inde®nites initially observed by Lewis, which motivated the unselective binding analysis, depends on the quanti®cational structure of adverbs, which in turn depends on focus. Nonfocused inde®nites show quanti®cational variability because they appear in the restriction of an adverb; focused ones do not, because they appear only in the scope of the quanti®er. Structured Davidsonian decomposition predicts which inde®nites will show quanti®cational variability. Even though the non-local nature of focus e¨ects makes it necessary to interpret the non-focused material, including narrow scope quanti®ers, both in the restriction and the matrix of the relevant adverb, no problem of requanti®cation arises on the present analysis. Since theta-roles are assigned exhaustively, they force the same individuals to be picked out in both instances. We also considered the asymmetric readings of if-clauses and preposed when-clauses. These were argued to occur precisely when the clause contains a focus. Since these clauses seem to necessarily restrict the overt adverb, structured Davidsonian decomposition only structures the event quanti®cation within the embedded clauses if-clause. This, in turn, accounts for the indirect way focus here a¨ects the overt adverbial quanti®er.

80

3.7

Chapter 3

Focus or VP-Internal versus VP-External?

Empirically, the facts discussed in this chapter have shown the great extent to which the quanti®cational structure of adverbial quanti®ers depends on focus. Theoretically, my aim has been to show how structured Davidsonian decomposition accounts for these facts. Not everyone agrees that the quanti®cational structure of adverbs depends primarily on focus assignment. I have already addressed the skepticism voiced by VallduvõÂ (1990) and Dryer (1994) concerning second occurrence focus. I have also discussed some of Johnston's (1994) observations regarding e¨ects of aspect. Another view that di¨ers very markedly from a focus-based analysis is the one developed by Diesing (1992). Diesing (1992) assumes a Lewis-Heim style analysis where adverbial quanti®ers are unselective quanti®ers that can bind free variables introduced by nonquanti®cational inde®nites. Adverbs are interpreted as having a tripartite structure consisting of the adverb, the restrictive clause, and the nuclear scope. Inde®nites that appear in the restrictive clause are unselectively bound by the adverb. In contrast, inde®nites that appear in the nuclear scope are bound by a default existential quanti®er. Building on this view of adverbial quanti®cation, Diesing proposes the Mapping Hypothesis. It states that what ends up in the ``restrictive clause'' of a tripartite structure is not the nonfocused material, but the material that is VP-external at LF. Similarly, the ``nuclear scope'' of the adverb does not contain the focused and nonfocused material, but the material that is VP-internal at LF. This type of analysis and a focus-based analysis make di¨erent empirical predictions. Let us consider some of the di¨erences. (For related discussion, also see BuÈring 1994.) A substantial part of the argument for the Mapping Hypothesis comes from the behavior of bare plural noun phrases. Diesing notes that German bare plurals that are scrambled cannot be interpreted as contributing to the scope of the adverb that they have scrambled across, but seem to form part of its restriction. She takes this to mean that scrambled elements are VP-external and must therefore be interpreted in the restrictive clause of the adverbial quanti®er. Support for this claim comes from the type of contrast found in (46a) and (46b). (46) a. weil because b. weil because

Bananen selten bananas rarely selten Bananen rarely bananas

erhaÈltlich available erhaÈltlich available

waren were waren were

Adverbial Quanti®ers

81

(46b) can mean that on few relevant occasions were bananas available. (46a) cannot have that meaning. It can, however, mean something very close to ``There were few bananas available.'' This reading arises if the scrambled Bananen is interpreted in the restrictive clause and the verb phrase erhaÈltlich waren is mapped into the nuclear scope. The subjects of individual-level predicates like be altruistic and know Spanish, which scramble in German, are also said to be VP-external material. On the other hand, VP-internal material consists of everything that surfaces within the verb phrase, and, optionally, the subjects of stagelevel predicates like be available and arrive, which can be interpreted inside or outside the verb phrase, resulting in weak and strong readings, respectively (see chapter 5). The present analysis agrees with the predictions made by the Mapping Hypothesis with regard to scrambled elements if we assume that they cannot enter the scope of the adverb they have scrambled across. This follows if adverbs do not c-command the scrambled elements at LF (see p. 43). In other respects, however, the two approaches make di¨erent empirical predictions. Those of the focus-based view seem to be more accurate. The Mapping Hypothesis claims that by being VP-internal an element will automatically contribute to the adverb's (nuclear) scope. This does not seem true. Although focus often falls exactly on the verb phrase, we have seen ample evidence that it does not always do so. Crucially, when a verb phrase contains nonfocused material, this material contributes not to the adverb's scope but to its restriction. We can see this by comparing (47a) with (47b) and (47c). (47) a. David rarely READS THE NEWSPAPER. b. David rarely reads THE NEWSPAPER. c. David rarely READS the newspaper. Whereas (47a) can indeed be argued to mean that few events involving David are newspaper-reading events, as the Mapping Hypothesis predicts, (47b) and (47c) are not interpreted in that way, contra the hypothesis. In (47b) reads, though VP-internal, contributes to the restriction of rarely (``Few events of David's reading are events of him reading the newspaper''). The same is true of the newspaper in (47c) (``Few events that involve David and the newspaper are reading events''). Similarly, (46b) can indeed have a reading where the entire verb phrase contributes to the scope but not to the restriction. But it has that reading only if the entire verb phrase is focused. When only part of it is focused, the nonfocused part

82

Chapter 3

must form part of the adverb's restriction. Thus, weil selten BANANEN erhaÈltlich waren can only mean that rarely when something was available was it bananas that were available.10 3.8

Events or Situations?

Finally, there is a treatment of adverbial quanti®ers that is in many ways similar to the event-based analysis given here and that may merely seem like a di¨erent way of doing the same thing: namely, a situation-based approach. An analysis of this sort is worked out by von Fintel (1994). Adopting an alternative semantics treatment of focus, von Fintel (1994) o¨ers an account of adverbial quanti®ers with which the present account has much in common: both accounts view focus as a major factor in determining quanti®cational structure, and both treat adverbs as selective quanti®ers and inde®nites as quanti®cational.11 But whereas on the present account adverbs quantify over events, on the situation-based account they quantify over situations. These two views of adverbs di¨er. Events are individuals in the ontology (albeit special ones). They are individuated by their participants, which, as we have seen, are exhaustively picked out by theta-related noun phrases. Situations are not individuals, but parts of worlds. Being parts of worlds, they are ordered with respect to each other in a part-whole relation. As a result of the part-whole nature of situations, a sentence like Martin fed his two cats at seven denotes not just one situation, but a whole number of situations; apart from the minimal situation that contains just Martin and the two cats he feeds at seven o'clock, it picks out a situation where everyone in the village, not just Martin, feeds their cats at seven. On the other hand, according to the event-based account, Martin fed his two cats at seven describes just that: an event of feeding that took place at seven where Martin was the sole feeder and his two cats were the only ones being fed. If other people in the village where Martin lives also happened to feed their cats at seven, that was a di¨erent event, one not picked out by Martin fed his two cats at seven. The onionlike structure of situations has an important consequence for the analysis of quanti®cation. It must somehow be ensured that the only situations counted are the minimal ones; otherwise, too many will enter the restriction of an adverb. The general schema for adverbial quanti®cation therefore takes the form shown in (48) (see Berman 1987, Heim 1990).

Adverbial Quanti®ers

83

(48) [Q A] B i¨ Q-many minimal situations where A is true are extendable (or part of ) a situation where B is true. A situation s is minimal if it contains no other situation that is also in the set of situations that support the proposition. It is this minimality requirement that ensures that a sentence like Martin always feeds his two cats at seven WHEN IT RAINS will come out true as long as all of Martin's cat feedings at seven take place when it rainsÐregardless of whether the other people in the village are such that they, too, only feed their cats at seven when it happens to rain. The following argument, which is due to Barry Schein (personal communication), suggests that although the minimality requirement works in the case just discussed, it does not work in the general case. The example crucially involves the interpretation of quanti®ed noun phrases such as two or more women and most guests in the scope of an adverbial quanti®er. Because the situations that are quanti®ed over have to be minimal, the situation-based account of adverbs cannot account for these cases. The event-based approach does not run into this problem. Imagine a situation where a caterer barbecues twelve chickens for a party. The host thinks this is too many chickens and says the following to the caterer: (49) Never, even if most people eat, WILL THEY EAT MORE THAN TEN CHICKENS ALTOGETHER. There is a lot of OTHER food! As it turns out, the caterer is right and the host is wrong. All twenty-three guests arrive very hungry, none are vegetarians, and they eat half a chicken each, eleven and a half chickens altogether. On the situation-based account, never has to count minimal situations, for the reasons given earlier. The minimal situations that never quanti®es over in (49) are those where most of the guests eat and that furthermore do not contain any other situation where most of the guests eat. This, however, means that never has to quantify over situations where exactly twelve people are eating. This is so because twelve is more than half of twenty-three and thus counts as most, and because a group of twelve eaters contains no other group of eaters that would also count as mostÐa group of eleven would not do. Hence, a minimal situation where most guests eat is a situation where exactly twelve guests eat. But in a situation where twelve guests eat, it is in fact true that more than ten chickens do not get eaten; since each guest eats only half a chicken, only six chickens

84

Chapter 3

are consumed. Consequently, the situation-based analysis predicts that the host is right. As a matter of fact, though, we con®dently judge that the host is mistaken and that the caterer is right. When a determiner like most is interpreted within the restriction of an adverb as in (49), the situationbased account gives incorrect truth-conditions in that it forces us to limit the domain of quanti®cation to the minimal situations where most people eat, which are not all the cases where most people eat. There are other, nonminimal situations that also matter for the truth-conditions of the sentence. On an analysis where adverbs count events rather than situations, we are not forced to restrict ourselves to cases where only twelve guests are eating when considering a circumstance where most guests are eating. Consequently, the problem described above is avoided. As (50) shows, (49) is interpreted as meaning that no eating is such that if most people are eaters in that eating, it will be the case that they eat more than ten chickens. (50) [no e: C(e) & [most x: People(x)] [Ee 0 : C(e 0 ) & One-to-one(e,e 0 ) & Eat(e 0 ) & Agent(e 0 ,x)]] Eat(e) & Agent(e,they) & [more-than-ten y: Chickens(y)] Theme(e,y) (50) is equivalent to saying that no event where most people eat in a oneto-one related event is such that the people eat more than ten chickens. There is, however, such an eating: the very dinner where all twenty-three guests eat and where they do eat more than ten chickens altogether, namely, eleven and a half. Nothing in the analysis prevents this event from being taken into account, so it correctly follows that the caterer has the right intuition. Thus, although the view that adverbs quantify over events and the view that they quantify over situations seem similar, these two views do not have the same empirical coverage. The event-based analysis accounts for the interpretation of quanti®ers like most guests in the scope of adverbial quanti®ers. In contrast, the situation-based analysis runs into di½culty here, becauseÐgiven the part-whole structure of situationsÐit imposes a minimality requirement on situations that are quanti®ed over, which turns out to be too restrictive.

Chapter 4 Only and Even

4.1

Introduction

A discussion of the e¨ects of focus on quanti®ers should address not only adverbial quanti®ers (chapter 3) and determiners (chapter 5), but also only and even, which in many ways resemble adverbs but in some ways recall determiners. As I will show, just like adverbial quanti®ers and weak determiners, only and even are quanti®ers whose quanti®cation structure is highly sensitive to focus. The contrast between (1a) and (1b) illustrates the focus sensitivity of only. (1) a. When asked who he considered an important in¯uence, Picasso only mentioned GOYA. b. When asked who he considered an important in¯uence, Picasso only MENTIONED Goya. The minimal di¨erence in focus assignment between the two sentences creates a di¨erence in truth-conditions. If (1a) is true, then Picasso claimed to have been in¯uenced by Goya and no one else. According to (1b), however, Picasso may very well have mentioned other painters and in fact elaborated on their greatness and the inspiration they provided him. With respect to Goya, however, Picasso did no more than mention him, suggesting perhaps that he did not consider him all that brilliant. This chapter is split into two main parts, the ®rst dealing with only and the second with even. The analysis I propose relies heavily on the structuring e¨ect of focus already observed in chapters 1 and 2. A new component of the analysis that I introduce here and that interacts in important ways with focal mapping is Q-raising, whereby a quanti®er takes scope by moving on its own in a very local fashion. (Q-raising plays an important role not only in connection with only and even, but also in

86

Chapter 4

connection with determiners; see chapter 5.) I also investigate several aspects of the lexical meaning of only and even. Only a small part of the considerable literature on only and even can be discussed here. My view of only owes much to the debate between Horn (1996) and Atlas (1991, 1993) and to the discussion by Bonomi and Casalegno (1993). My discussion of even focuses particularly on the recent debate concerning its ambiguity (see Karttunen and Peters 1979, Rooth 1985, Wilkinson 1996, Rullmann 1997). 4.2

Some Syntactic Facts about Only

Even though we will be mainly interested in the semantic aspects of only, it will be important to take into account certain syntactic facts. A rather curious one is that it is not obvious what syntactic category only belongs to. Given that only can occur in typical adverbial position somewhere between the subject and the verb phrase, as in (2), it might seem that only is an adverb. (2) a. b. c. d. e.

Albert only grows VEGETABLES. Albert only likes trees with EDIBLE fruit. They only disagree on where to plant THE CHERRY TREE. Albert only refuses to plant CORN. I only WATER the geraniums.

Clearly, though, only is not limited to adverbial positions. As (3) shows, it can be attached to a noun phrase. (3) a. b. c. d. e.

[Only BIRDS] have wings. Albert grows [only VEGETABLES]. [Only shrubs that love ACIDIC soil] grow well in this garden. [Only THE TOMATOES] did well this year. [Only A FEW camellia shrubs] survive winters in this city.

Especially in (3a±c) only might look like a determiner on a par with some, few, all, and the like. Yet, as (3d,e) show, only can also attach to noun phrases that are overtly quanti®ed. If we try to be parsimonious, this suggests that only is not a determiner but, whatever it is, it attaches to a noun phrase that is already quanti®ed (either overtly as in (3d,e) or tacitly as in (3a±c)) (see von Fintel 1997). The examples in (2) and (3) clearly do not exhaust the syntactic possibilities of only. It can also attach to entire sentences, as in (4), to prepositional phrases, as in (5), and to adverbial phrases, as in (6).1

Only and Even

87

(4) a. [Only WHEN IT RAINED FOR THREE WEEKS] did the shiitake ®nally grow. b. Nancy will give up on a plant [only if it is CLEARLY dead]. (5) a. [Only after MAY 1st] can you safely plant anything in this area. b. We went [only to ALE'S house]. (6) Use this sauce [only SPARINGLY], it's very hot. Finally, it is worth noting that only can also be used as an adjective, as in (7). (7) The only rival Shakespeare had was Marlowe. Only here seems similar in meaning to sole and single and virtually synonymous with German einzig and Spanish uÂnico. Syntactically, it patterns with adjectival instances of one and few (e.g., the few remaining problems, his one and only trip to Madagascar).2 Setting aside this adjectival use, it seems that the quanti®cational only attaches to almost anything, so much so that it would make sense to classify it syntactically as an ``admanythings'' if such a category existed. Only is not alone in this. Even behaves similarly, as we will see. In all quanti®cational uses of only that we have seen so far, only counts individuals of some sort. There is also a quanti®cational use of only where it does not count individuals, but instead is scalar. Consider (8), for instance. It does not mean that Juliet was nothing other than drugged; rather, it means that she was nothing worse than drugged (i.e., she was not dead). (8) Juliet was only DRUGGED. Although the various instances of quanti®cational only look rather different, in one respect they all are the same (see Jackendo¨ 1972, von Fintel 1997). (9) Only has a focused element in its c-command domain. When only is in adverbial position, as in (2), the focus can be on any element in its c-command domain, be it a verb, noun phrase, an adjective, or something else. In the cases where only is attached to a noun phrase, prepositional phrase, or adjective phrase, it requires that there be a focused element within that phrase (cf. (3)±(7)). Even if the generalization in (9) holds of all the examples introduced so far, it seems there are others that contradict itÐnamely, instances of second occurrence focus. Thus, (10) (from Partee 1991) appears to run

88

Chapter 4

counter to (9) because in (10B) only semantically associates with the graduate students, even though this noun phrase does not carry any visible focus. (10) A: Eva only gave copies TO THE GRADUATE STUDENTS. B: No, PETR only gave copies to the graduate students. Fortunately, (10) is not as problematic as it looks. As argued in chapter 3, examples of second occurrence focus can be given an independent explanation, one that is consistent with (9). In a nutshell, only indeed associates with a focused element in its domainÐin (10B), the graduate students. However, unlike other instances of focus, this focus is phonetically very reduced because it appears in a mimicking context, where a previous utterance is repeated. This means that what is exceptional about instances of second occurrence focus is merely the way focus is phonetically realizedÐthey do not disobey (9). Another interesting generalization about the syntax of only concerns its scopal behavior. The central observation (known as Taglicht's observation; see von Stechow 1991) is that when only attaches to a noun phrase, its scopal properties are di¨erent than they are when it is attached to a verb phrase. For instance, (11a) is reported to be ambiguous, but (11b) and (11c) are not; each of the latter two has only one (distinct) reading. (11) a. The man at the nursery told us to water [only THE AZALEAS]. b. The man at the nursery told us to only water THE AZALEAS. c. The man at the nursery only told us to water THE AZALEAS. (11a) is reported to have two possible readings. On its ®rst reading it says that azaleas are the only kind of plant the man at the nursery told us to water; he didn't say anything about the rose bushes or the rhododendrons, maybe because he assumed we already knew we needed to water those.3 On its second reading (11a) says that the man explicitly told us to water nothing else but the azaleas, saying something like ``Be careful to water the azaleas only!'' In contrast with (11a), neither (11b) nor (11c) is ambiguous; (11b) has only the second reading, and (11c) only the ®rst. The scopal properties of only that are illustrated in (11) seem to be derivative. An only attached to a noun phrase seems to piggyback on the possible scopes of the noun phrase. If in (11a) the noun phrase takes matrix scope, we get the ®rst reading; if it takes embedded scope, we get the second one (see von Stechow 1991, Rooth 1985). By the same token, when only appears in adverbial position, its scope is ®xed, because quanti®ers in adverbial position seem to be scope rigid. Thus, in (12) the scope

Only and Even

89

of the adverb always relative to the negation directly re¯ects the overt word order. (12a) denies that Brian always interrupts; (12b) asserts that he always refrains from interrupting.4 (12) a. Brian does not always interrupt. b. Brian always does not interrupt. Summarizing the syntactic peculiarities of only, we have seen that the quanti®cational only is an ``admanythings'' in that it attaches not only to verbal phrases, but in fact to all sorts of constituents. Whatever the scopal possibilities of the constituent it attaches to, only seems to inherit them. We furthermore noted that only requires that there be a focus in its c-command domain. The instances of second occurrence focus, which seem to run counter to this generalization, can in fact be argued not to be counterexamples after all. 4.3

Conservativity, Focal Mapping, and Q-Raising

One important question that arises in connection with only's syntactic behavior is how the surface syntax of only-sentences relates to their quanti®cational structure.5 Put di¨erently, how does only obtain its semantic arguments, its restriction, and its scope? This turns out to be an important question that goes beyond the analysis of only because it relates directly to the classi®cation of what constitutes a possible natural language quanti®er. Common wisdom has it that only is an oddball. Unlike determiners (some, every, most, etc.), which are held to be conservative, only is said not to be conservative. A quanti®er is conservative if it obeys (13). (13) Conservativity [D A] B i¨ [D A] A X B That the determiner all, for instance, is conservative can be seen from the equivalence in (14). Here the right-hand side seems just a clumsy way of saying the same thing that is said on the left-hand side. (14) All cats have whiskers i¨ all cats are cats that have whiskers. Examples like (15a,b), where all is replaced by only, are claimed to show that only is not conservative. (15) a. Only cats have whiskers i¨ only cats are cats that have whiskers. b. Only cats shed fur i¨ only cats are cats that shed fur. Unlike the equivalence stated in (14), those stated in (15) do not hold. If it turns out that some animals that are not cats also have whiskers (e.g., sea

90

Chapter 4

lions), then the left-hand side of (15a) will be false, but the right-hand side will still be true; what else but a cat can be a cat and have whiskers! Similarly for (15b). Whether or not a quanti®er is conservative is considered an important question in Generalized Quanti®er Theory, as one of its main concerns is to distinguish among the possible generalized quanti®ers those that occur in natural language. Setting aside adverbial quanti®ers for the moment, were it not for only (and some instances of few and many discussed in chapter 5), Conservativity would seem to work very well. Noting that only is not conservative in the way just illustrated, some authors have said that it is ``right-conservative'' or ``neoconservative'' instead (de Mey 1991, Horn 1996). (16) Neoconservativity [D A] B $ [D A X B] B This is supported by the validity of (17). (17) Only cats have whiskers i¨ only whiskered cats have whiskers. It is worth noting that the argument that only is not conservative but neoconservative is based on the following implicit assumption: (18) Interface assumption about only What counts as A in the de®nitions in (13) and (16) are the noun phrases that only attaches to in the overt syntax (its c-command domain), and what counts as B are the c-command domains of those noun phrases. It is due to (18) that we assume that in (15a), for instance, the bare plural noun phrase cats corresponds to A and have whiskers corresponds to B. It is easy to see what motivates (18); it parallels the way quanti®ed noun phrases are interpreted in general, where, as we saw in chapter 3, the internal argument of D denotes its restriction, and the external argument its scope. (19)

Only and Even

91

Because (18) so strongly recalls (19), it is clearly the null hypothesis. But that does not necessarily mean that it is right. (18) runs into a problem as soon as we extend the data from cases where the noun phrase contains only one element (which, because of (9), is necessarily the focus) to cases with more complex noun phrases, where focus can be assigned to a subpart of the noun phrase. (20) [Only cats with LONG tails] shed fur. According to (18), in this example the restriction A should be denoted by cats with long tails and the scope B by shed fur. Given Neoconservativity, we would then predict that (20) is equivalent to (21), where (21) is interpreted to mean that no creatures that are not long-tailed, fur-shedding cats shed fur. (21) Only long-tailed, fur-shedding cats shed fur. But (20) and (21) are not equivalent. (20) rules out that cats with short tails shed fur, but it does not say anything about dogs, for instance. It is beside the point whether they, too, shed fur. On the other hand, for (21) to be true it has to be the case that no creature other than a long-tailed cat sheds fur; any dog getting rid of winter fur falsi®es this. For (21), dogs are not irrelevant. Does this mean that only is sometimes neoconservative, and sometimes not? This seems hardly an attractive conclusion to draw. It seems more reasonable to give up the interface assumption in (18). Giving up (18) is not a big sacri®ce because already in the discussion of free focus and adverbial quanti®ers we saw that the surface syntax of a sentence need not reveal its quanti®cational structureÐfocus plays an important role as well. When only appears in adverbial position, focal mapping applies directly, giving a sentence like Albert only grows VEGETABLES a structure like (22). (22) [[Only [Albert grows]] Albert grows VEGETABLES] Matters are a bit more involved when only does not surface in adverbial position, but forms a constituent with the DP or PP to which it is adjoined. If we assume that in such cases only moves on its own to a local adverbial position via a process of Q-raising, then it will be syntactically unary. This, in turn, allows focal mapping to apply. The various steps are illustrated in (23)±(24) for Only CATS have whiskers, and in (25)±(26) for Only cats with LONG tails shed fur. (23) [[Only CATS] [have whiskers]] [surface syntax]

92

Chapter 4

(24) a. [Only [CATS have whiskers]] [Q-raising] b. [[Only [have whiskers]] [CATS have whiskers]] [ focal mapping] (25) [[Only cats with LONG tails] [shed fur]] [surface syntax] (26) a. [Only [cats with LONG tails shed fur]] [Q-raising] b. [[Only [cats with long tails shed fur]] [cats with LONG tails shed fur]] [ focal mapping] We can think of Q-raising as movement of an adjoined head only to the speci®er of an appropriate local functional projection, or we can think of it as adjunction to some local XP (e.g., IP in the above examples). The choice does not matter for our purposes. Either possibility will result in movement that is licit in the sense of Chomsky (1995); whether Q-raising involves adjoining the head only to an XP or moving it into the speci®er position of a functional projection, the movement results in a uniform chain because neither in its base position nor in its landing site does the head project. Assuming that only is lexically similar to a universal quanti®er (see below), the structures in (22), (24b), and (26b) map into interpretations that can be paraphrased as in (27), which capture the truth-conditions of the corresponding sentences. (27) a. All that Albert plants are (things that he plants that are) vegetables. b. All that have whiskers are cats (that have whiskers). c. All cats with long tails that shed fur are cats with long tails (that shed fur). Given the structures that are derived by focal mapping and Q-raising, the mapping from the syntax to the semantics is transparent or, put di¨erently, it is conservative. What the covert syntax suggests should be the restriction of the quanti®erÐnamely, the nonfocused part that is c-commanded by onlyÐis indeed the semantic restriction. And what the (covert) syntax suggests should be the semantic scopeÐnamely, the focus together with the nonfocused partÐis indeed the scope.6 Similar to the way that focal mapping predicts that unary only is conservative, it also predicts that the unary adverbial quanti®ers that were discussed in chapter 3 are conservative (see de Swart 1991). Since Conservativity is a good candidate for a universal of natural language quanti®ers, this is a welcome result. An analysis similar to the present analysis of only will be given in chapter 5 concerning the e¨ects of focus on weak

Only and Even

93

determiners, showing that they, too, are conservative in certain cases where they have been said not to be. 4.4

The Existential Force of Only

Regarding the interpretation of the word only itself, there has been a great deal of debate about whether only is presuppositional or whether it has genuine existential force. Everyone agrees that (28) entails (29). (28) Only GOD can make a tree. (29) No one other than God can make a tree. There is disagreement, however, about whether (28) also entails (30) in the same way, or whether it merely presupposes it. (30) There is someone that can make a tree. Horn (1996) argues that (30) is a presupposition; Atlas (1991, 1993) claims that it is a logical entailment, a kind of entailment that is very similar if not identical to a backgrounded focal entailment. Let us consider various arguments for and against these claims. One of Horn's reasons for adopting a presuppositional or asymmetricalist approach, as he calls it, derives from the contrast illustrated in (31). (31) a. Only KIM can pass the test, and it's possible even SHE can't. b. aOnly KIM can pass the test, {and/but} it's possible that SOMEONE ELSE can. The felicity of (31a) is said to show that it is possible to suspend the conclusion that Kim can pass the test. In contrast, it is not possible to suspend the conclusion that someone else cannot; (31b) sounds nonsensical. Horn and other asymmetricalists have taken this to show that unlike ``no one but Kim can pass the test,'' which is a genuine entailment of (31a), ``Kim can pass the test'' is not a logical entailmentÐit is something weaker. Exactly what it is has not been easy to establish, for it has variously been analyzed as a (semantic, pragmatic) presupposition, a conventional implicature (Rooth 1985), and a conversational implicatureÐand, most radical of all, as not playing any part in the interpretation (Geach 1962) (see Horn 1996 for more details). Horn (1996) proposes that ``Kim can pass the test'' is actually an inference. Assuming that only is the neoconservative converse of all (see above), he proposes that only, like all, carries an existential presupposition (``someone can pass the test''). If this pre-

94

Chapter 4

supposition is satis®ed, then from the exclusionary part of only's meaning it can be inferred that Kim can pass the test. Part of the di½culty of saying that only carries an existential presupposition is that it would be a rather unusual presupposition because it crucially has to be suspendable; unlike what happens with a regular presupposition, its not being true does not lead the sentence to lack a truthvalue (see chapter 2). Also, apart from showing in detail just how tricky it is to ®gure out what the weaker-than-logical-entailment notion would be, Atlas (1991, 1993) points out that there are numerous examples that are parallel to (31), but that are clearly incoherent. (32) is one of them. Regarding (32b), Atlas (1993, 314) observes, ``The idea that simultaneously I love only you could be true while I love you is false just seems crazy . . .'' (32) a. aOnly HILLARY could reform health care, and even she couldn't. b. aI love only YOU, but (maybe) I don't love you either. These examples seem enough to falsify the asymmetricalists' view, which would predict them to be acceptable. Let us then assume that a sentence like Only GOD can make a tree does indeed logically entail that someone can make a tree. But at this point we may ask, what about the suspension examples? How can they be explained on a ``symmetricalist'' account? This is something that Atlas (1991, 1993) does not seem to address, but that is interesting to consider if one wants to adopt the view, as I do here, that only has genuine existential force. I have the following thoughts to o¨er on the matter.7 The suspension that we observed in (31a) essentially comes about when Only KIM can pass the test is taken to suggest ``the test is very di½cult.'' Put di¨erently, the suspension of the conclusion that someone, and therefore Kim, can pass the test is pragmatically sanctioned because it serves argumentation toward a higher goal, namely, that of saying that the test is hard. By contrast, it is not easy to imagine what someone could mean by I love only YOU other than what she is literally saying; it is di½cult to suspend the conclusion that the speaker who uses these words plainly means that she loves the person so addressed (and nobody else). But, contrived though it may be, if the speaker wants to convey that she is incapable of loving, of loving even the very lovable person she is addressing, then the suspension seems possible again. Consider in this context the following example cited by Horn (1996): All the world is queer save thee and me, and sometimes I

Only and Even

95

think thee is a little queer (attributed to an unidenti®ed Quaker speaking to his wife). Here suspension is possible because the Quaker has the higher goal of arguing that everyone (except possibly he himself ) is strange. Next, the fact that suspension is possible under certain pragmatic conditions (argumentation toward a higher goal) does not mean that what is suspended is a presupposition and not a bona ®de logical entailment. Thus, the sentences in (33) sound acceptable with a certain tone of voice. (33) a. You'll see, Maris eats like a bird! She'll eat A COUPLE OF TINY LITTLE CANAPESÐand maybe not even THAT! b. Don't be fooled! He is no math genius! He is an AVERAGE accountant, and, who knows, maybe not even THAT! It would be unconvincing to say that Maris will eat a couple of tiny little canapes does, as such, not logically entail that she will eat one. And yet, all of (33a) together is clearly consistent with her not eating anything at all! Similarly, if it is true that the person mentioned in (33b) is an average accountant, then it is entailed, given what we think accountants know how to do, that he knows how to calculate compound interest. But the continuation of (33b) very intentionally leaves open whether even these lowered expectations are met. In both examples the ®rst clause describes a situation that is ranked low on a scaleÐthe scale of eating amounts in the ®rst example, and the scale of being able to do calculations in the second. The continuation then corrects what was said in the ®rst clause: even if what was asserted there was very low on a scale, it was not low enough to make the more general point. Both examples could contain an only, just, or merely, but they obviously do not have to. Even without only, just, or merely there is suspension, and, crucially, what is suspended here is clearly something that the sentence entails. I see no reason not to assume that what maybe not even does in the examples in (33)Ðnamely, somehow conversationally suspend certain entailments of the ®rst clauseÐis also what goes on in Horn's original example. If so, it seems fair to assume that Only KIM can pass the test, as such, does indeed entail that Kim can pass the test. The examples in (34) (taken from Horn 1996) illustrate another asymmetry that has been used to argue in favor of the asymmetricalist view and that any symmetricalist should address. (34) a. If only HILLARY trusts Bill, all is well. b. I just discovered that only HILLARY trusts Bill. c. It's too bad that only HILLARY trusts Bill.

96

Chapter 4

The asymmetricalist view claims that what is hypothetical in (34a) is not whether someone trusts BillÐthat is taken to be presupposedÐbut that nobody but Hillary does. Similarly, in (34b) it is presupposed to be old news that someone trusts Bill; what was discovered was merely that no one other than Hillary does. And that is also what is deemed unfortunate in (34c). These facts are explained by saying that ``someone trusts Bill'' is a presupposition of Only HILLARY trusts Bill and that, assuming presupposition projection, it projects past the scope of if, discover, too bad. But the intuition that the presuppositional view aims to account for is not all there is to (34). One consequence of the Monica Lewinsky scandal of 1998 is that it is now easy to imagine a situation where it cannot be assumed that Bill is trustworthy on matters of marital ®delity. Even now, when judged against such a background, the sentences in (34) do not automatically result in a lack of truth-value, contrary to what we would expect if ``someone trusts Bill'' were indeed a presupposition of Only HILLARY trusts Bill. Even if no one trusts Bill, (34a), for instance, can be trueÐand it probably is, if we think of Hillary's great intelligence and how vital her support seems for her husband's career. Thus, Bill's being trusted need not necessarily be construed as being outside the scope of if, discovered, too bad. This is unexpected on the asymmetricalist view, but consistent with the symmetricalist one. Since ``someone trusts Bill'' is not a presupposition of Only HILLARY trusts Bill but an entailment, albeit a backgrounded one, it does not ``project'' when the sentence is embedded in the scope of if, discovered, and too bad. These observations do not mean that it is not generally the case that the scope of elements like if and discovered appears to be limited to the focus so that the nonfocused part seems to have wider scope. A possible symmetricalist explanation along lines suggested by Boer and Lycan (1976) and Atlas (1993) is that what is by pragmatic default taken to be conditional, discovered, too bad, and so on, is what is ``said'' or ``asserted'' by uttering the clause in the scope of these elements, not necessarily all the things that the clause entails. As we saw in detail in chapter 2, on the present analysis (as well as Atlas's) the focus does not contribute to what the sentence is about (ˆ quanti®er restriction), but since the focus contributes only to the quanti®er scope, it forms only part of what the sentence asserts. Consequently, it is not surprising that sentences like (34a±c) should at ®rst glance be interpreted the way the asymmetricalists claim they areÐwith the nonfocused part being outside the scope of if, discovered, and too bad.8

Only and Even

97

The suspension facts in (31) and the alleged projection facts in (34) are not the only kind of data that Horn (1996) uses to argue in favor of an asymmetricalist analysis. He also cites the fact that only can license negative polarity items and that only triggers negative preposing. After considering some quanti®er properties of only and other universal quanti®ers, we will see that these facts are actually compatible with the view argued for here, whereby only has existential force.9 We have now reached the following stage. Only receives its restriction and scope through focal mapping (and Q-raising should only not surface in an adverbial position). One result of this rearrangement of the covert syntax is a conservative interpretation of only. We also want to say that a sentence of the form Only GOD can make a tree entails but does not assert that someone can make a tree (or that a tree can be made). This follows if only is a quanti®er that has existential force; it requires that there be something that satis®es its restriction, which is expressed by the nonfocused part. 4.5

The Existential Force of Every

Is only unique in being a universal quanti®er with existential import? Perhaps not. Even if it is often held that only's counterpart all (along with other strong quanti®ers) is presuppositional (Horn 1996, Diesing 1992), this Strawsonian view is not written in stone. An argument can be made that universal quanti®ers have existential force. It is part of their literal meaningÐnot a presupposition or implicatureÐthat their restriction is true of something. On this Aristotelian view (see Moravcsik 1991), every, for instance, is not interpreted as in (35), which represents the standard logic meaning of E (which can be supplemented with a presupposition requirement); instead, it is interpreted as in (36). (35) Val (hY,Xi, every, s) i¨ [every x: X(x)] Y(x) i¨ X J Y [standard logic meaning (E); no existential import (can be supplemented with presupposition: jX j V 1)] (36) Val (hY,Xi, every, s) i¨ [every x: X(x)] Y(x) i¨ X J Y & jXj V 1 [Aristotelian meaning; existential import] Both (35) and (36) require that whatever satis®es the restriction of every also satisfy its scope. But unlike (35), (36) further requires that the restriction of every not be empty. If it is empty, the sentence is false, whereas in

98

Chapter 4

(35) it is vacuously true or truth-valueless, depending on whether we add an independent presupposition requirement to (35) or not. The argument for the Aristotelian analysis of every requires that we look at two sets of cases, contingent and noncontingent. Let us begin with contingent cases, which readily lend themselves to (36). Consider for instance the scene described in (37). As Moravcsik (1991) points out, in this scenario neither the standard logic answer in (38) nor the presuppositional answers in (39) seem appropriate answers for the housewife. (37) A door-to-door salesman is selling toys in a neighborhood. As he makes his ``pitch'' with yet another housewife, he says to her, ``But all the other children in this neighborhood have this toy.'' (p. 427) (38) ``Well, that's true. But since there are no other children in this neighborhood, all you are saying is that there is no child around here without this toy, and that is a very weak reason for making a purchase.'' (p. 428) (39) a. ``Sir, the question of the truth-value of your assertion does not arise, since the subject expression fails to denote.'' b. ``I don't understand what you mean. There are no other children in this neighborhood.'' (p. 428) Instead, the housewife could felicitously answer with (40). The appropriateness of this answer indicates that the universal quanti®er in (37) has existential force. (40) ``You are a liar. There are no other children in this neighborhood. Our son is the only one.'' Not all contexts are contingent. In noncontingent contexts matters are more complicated. Yet I think that even here a case can be made for the analysis holding that universal quanti®ers have existential import. Let us begin with Moravcsik's observation that whereas all in contingent contexts has existential import, this does not seem to be the case in noncontingent, lawlike contexts (of which he discusses various types). Consider (41). (41) . . . MENSA, an international association for people with high I.Q.'s, sponsors an academic competition of some sort for high school students in a certain region. Circulars are sent out to every high school, and these are read by the principals to the students.

Only and Even

99

There are ``qualifying rounds'' for the competition, but not for everyone. For, as one of the sentences in the circular reads: ``all students with an I.Q. of N (where this is a very high number) automatically qualify.'' (Moravcsik 1991, 428) If it turned out that no student in the region's schools automatically quali®ed, it would be inappropriate to protest using (42) (Moravcsik 1991, 428±429) or the kind of answer in (43). (42) ``You are a liar. There are no students with I.Q.'s that high in any of the a¨ected schools.'' (43) ``I don't understand what you mean. There are no students with such high I.Q.s in any of the a¨ected schools.'' But it would be appropriate to respond with (44) (Moravcsik 1991, 429). (44) ``That is too bad. What I said is still true, butÐalasÐno student will qualify automatically.'' Moravcsik takes the felicity of (44) vis-aÁ-vis answers like (42) and (43) to show that whereas in contingent contexts universal quanti®ers have an Aristotelian interpretation, in lawlike contexts they are interpreted in the standard logic fashion, that is, without existential import. This, he goes on to argue, has far-reaching implications for our understanding of logical form. There is another way of looking at the facts. In noncontingent, lawlike statements universal quanti®ers have existential import. What makes noncontingent statements di¨erent from contingent ones is that noncontingent contexts are opaque because they are embedded in the scope of an operator, that is, ``according to the rules of . . .'' or ``according to X and its rules.''10 An empirical argument in favor of this general Aristotelian analysis over the ``mixed'' analysis is provided by the contrast between (45) and (46). It consists in the observation that whereas the lawlike statements in (45) are judged true, those in (46) are judged false. (The examples are discussed in Heim and Kratzer 1998; see below.) (45) a. Every unicorn has exactly one horn. [``true''] b. Every unicorn is a unicorn. [``true''] (46) a. Every unicorn has exactly two horns. [``false''] b. Every unicorn fails to be a unicorn. [``false'']

100

Chapter 4

We can make sense out of this fact if we assume that every unicorn is embedded in the scope ``according to mythology and its rules.'' Now even in noncontingent contexts such as (45) and (46) every has existential import, but only within the opaque context created by ``according to mythology and its rules.'' If, on the other hand, every in lawlike contexts had the standard logic meaning and were not embedded (Moravcsik), both (45) and (46) should be vacuously true, which they are not. Thus, given what was said above, it seems quite reasonable to assume that universal quanti®ers have existential importÐtheir existential force is part of their lexical meaning. What di¨ers between contingent and noncontingent contexts is not how the quanti®cation itself works, but whether it is embedded or not. In various respects this discussion has followed a view taken by Heim and Kratzer (1998), who, on the basis of (45) and (46), also argue that every is unambiguous. Though similar, the two views are not identical. One di¨erence is that Heim and Kratzer do not take Moravcsik's data into account; as a result, their analysis does not distinguish between the Aristotelian and the presuppositional views, which they combine (``existence is presupposed'') and which they contrast with the standard logic view. Every is thus said to have an ``Aristotelian/presuppositional'' interpretation in both contingent and lawlike statements. Another di¨erence is that Heim and Kratzer's approach maintains that noncontingent statements involve quanti®cation over so-called possible individuals. My approach is not committed to this particular analysis of the relevant opaque contexts.11 Going back to the main thread of the argument: I have argued that only is a universal-like quanti®er that has existential force. It appears that other universal quanti®ersÐin particular, everyÐare also not merely presuppositional but require as part of their interpretation that there be something that their restriction is true of. As usual, the restriction is backgrounded. 4.6

Downward Monotonicity: Weak and Strong

If, as I have suggested, both only and every are universal(-like) quanti®ers with existential force, then we might expect that because of the latter property they are not downward monotonic in their restrictions. This has indeed been suggested for only, which is claimed to be nonmonotonic (e.g., Atlas 1993).

Only and Even

101

What suggests that only is indeed not downward monotonic in its restriction is the fact that (47a), for example, fails to entail (47b). This lack of entailment was already observed in medieval times (see Horn 1996). (47) a. Only SOCRATES entered the race. b. Only SOCRATES entered the race early. As we have seen, a quanti®er is downward monotonic in its restriction exactly when replacing the restriction by a subset preserves truth. (48) [Q A] B is downward entailing in its restriction i¨ [Q A] B entails [Q A0 ] B, where A0 J A. One set is a subset of another if any element that is in the ®rst set is also an element of the other: (A0 J A) $ Ex ((x A A0 ) ! (x A A)). Therefore, not only is the set of events of entering the race early a subset of the set of events of entering the race; the empty set is too. Accordingly, (47a) should entail (47b) regardless of whether someone entered the race or not. But it does not. Still, to say that only is nonmonotonic is too radical a conclusion to draw and misses a generalization. The truth of (47b) is in fact guaranteed by the truth of (47a) if we can make the independent assumption that there was an early entrance into the race. In light of this, let us consider a second notion of downward monotonicity: ``weak'' downward monotonicity, as opposed to the ``strong'' downward monotonicity described in (48). For reasons that will become clear as we go along, it can be de®ned as follows: (49) Weak downward monotonicity [Q A] B is weakly downward monotonic in its restriction i¨ [Q A] B entails [Q A0 ] B, where (a) A0 J A, (b) A0 is not { }, and (c) at least for some value for A and A0 , A0 H A. Although only is not strongly downward monotonic in its restriction, it is weakly downward monotonic. Using (47) as our example, we can observe that when (49a±c) are met, the inference goes through. (49a±c) are met under these circumstances. First, whatever element is in A0 (early entrances to the race) is also in A (entrances to the race). Second, we consider a circumstance where A0 is not emptyÐa circumstance, that is, where the race was entered early. Finally, to meet condition (49c) we need to check whether there is an example where (49a) and (49b) are met and where A0 (the early entrances to the race) denotes a proper subset of A (the

102

Chapter 4

entrances to the race). Although this is not the case in (47) because Socrates picks out an individual, we ®nd that by changing Socrates to philosophers we indeed have such a case; assuming that someone entered the race early, then if only philosophers entered the race, it follows that only philosophers entered the race early. Importantly, however, it does not follow that all philosophers entered the race earlyÐsome may have entered it later than others. Thus, all of (49a±c) are met, and taking them as conditions, the inference from (47a) to (47b) indeed goes through. (The purpose of (49c) might not be obvious at this point, but it will become clear when we consider de®nite descriptions.) Only may not be alone in being only weakly downward monotonic. Every may be too, which at this point comes as no surprise. The weakly downward monotonic character of every is revealed in (50). (50) a. This year, every student passed the ®nal exam. b. This year, every student from Liechtenstein passed the ®nal exam. Unless we know for a fact that this year there was at least one student from Liechtenstein, (50b) does not follow from (50a).12 This suggests that like only, every can be viewed as being downward monotonic in its restriction only in the weak sense of (49), not in the strong sense of (48). The inference from (50a) to (50b) goes through when the following conditions are met: student from Lichtenstein denotes a subset of student (see (49a)), the subset is not the empty set (see (49b)), and there is at least one choice of A and A0 where A0 can be a proper subset of A (see (49c)). (50) itself provides such an example if it is interpreted in the pragmatically plausible circumstance where there were not only students from Liechtenstein, but also students from other countries.13 As mentioned earlier, Horn (1996) not only uses the suspension facts to argue that only does not have existential force; he also uses the fact that only can license negative polarity items (NPIs) to argue in favor of an asymmetricalist analysis.14 Taking into account that only, like every, is weakly downward monotonic, the NPI facts do not seem problematic for my analysis. They suggest that what matters for the licensing of NPIs is that there be some downward entailment; whether it is weak or strong seems irrelevant. Let us now consider the purpose of (49c): it is needed to keep singular de®nite descriptions from counting as weakly downward monotonic. On the quanti®cational view of singular de®nite descriptions that was adopted in chapter 2, the meaning of the can be rendered as the combi-

Only and Even

103

nation of a logical universal …E† and an existential (e.g., The king of France is bald is true i¨ whoever is king of France, he is bald and there is a king of France). Given the universal and existential force of singular de®nite descriptions, one may wonder whether the is not also weakly monotonic in its restriction. If so, it should license NPIs there, which it does not (e.g., *The student who had ever been to Boston liked it). The in singular de®nite descriptions is not weakly monotone decreasing in its restriction because the condition in (49c) cannot be met. If we say that the student passed the ®nal exam, then if there was a student from Liechtenstein, we can conclude that the student from Liechtenstein passed the ®nal exam. Notice, however, that student from Liechtenstein denotes only an improper subset of student, and this is so because the de®nite description the student picks out just one individual. Consequently, even though the has universal and existential force, it does not count as weakly downward monotonic in its restriction and thus di¨ers crucially from only and every.15 Having argued that the NPI facts do not support an asymmetricalist analysis of only, but that they are indeed compatible with the claim that only has existential force, we are ready to examine the assertion that only's ability to trigger negative inversion argues against a symmetricalist analysis. Horn (1996) says the negative inversion facts would be incompatible with the view that only has existential force because only here patterns together with operators that clearly do not have such force, namely, not, no, few, rarely, and so on. (51) a. b. c. d. e.

Only in stories does a dropped glass betray agitation. Not even two years ago would he have been inclined to help us. In not many years will Christmas fall on a Sunday. In no clothes does Robin look attractive. Rarely/On few occasions did Kim tell any jokes.

It turns out, however, that although only di¨ers from not, no, rarely, and few in having existential force, all these quanti®ers have something important in common, something that clearly distinguishes them from those that do not trigger negative inversion (e.g., all (ways), some (time), a few, often): in all negative inversion sentences the sister of the adverbial phrase provides a downward-entailing context (either a strong or a weak one, depending on the quanti®er in question). Evidence showing that the sister of the clause-initial PP is a downwardentailing context comes from entailment patterns. (51a), for instance,

104

Chapter 4

entails Only in stories does a dropped wine glass betray agitation (assuming that there is a dropped wine glass in stories; see above). Similarly, (51b) entails Not even two years ago would he have been inclined to help us and be generous about it, (51c) entails In not many years will Christmas fall on a Sunday and will there be snow, and so on. NPI facts provide further evidence. (52) a. Only in stories does a dropped glass ever betray anything. b. Not even two years ago would he have been the least bit inclined to help us. Interestingly, when we compare the inverted examples in (51b±d) with the minimally di¨erent examples in (53), which lack negative inversion, we ®nd that in (53) the negative element introducing the adverbial phrase no longer creates a (strongly) downward-entailing context for the rest of the sentence; it only creates a downward-entailing context within the adverbial. (53) a. In not many years Christmas will fall on a Sunday. b. Not even two years ago, he would have been inclined to help us. c. In no clothes, Robin looks attractive. Unlike the sentences in (51), those in (53) do entail that Christmas will fall on a Sunday, that he would have been inclined to help us, and that Robin looks attractive under certain circumstances, respectively. Not surprisingly, when there is no inversion, NPIs in the IP remain unlicensed (e.g., *Not even two years ago, he would have been the least bit inclined to help us). Why should only in (51a) create a (weakly) downward-entailing context for the sister of the adverbial phrase? The analysis developed here provides an answer. Because of the principle in (9), the material that contributes only to the scope of only (the focus) surfaces inside the adverbial. As a result of Q-raising and focal mapping, the nonfocused rest of the sentence ends up restricting the universally quantifying only. The ®nal consequence is that the nonfocused material, which includes the sister of the phrase only is attached to, provides a downward-entailing context. This shows that once we take into account how focus a¨ects the quanti®cational structure of only, the fact that only triggers negative inversion is not only not a surprise, it is in fact predicted. It does not con¯ict with the claim that only has existential force. An interesting corollary to this analysis is this. Since because of (9), Q-raising, focal mapping, and the uni-

Only and Even

105

versal meaning of only the nonfocused part of the sentence necessarily provides a downward-entailing context and since negative inversion correlates with the sister of the preposed phrase providing such an environment, we expect that, unlike with in not many years, not even two years ago, and in no clothes, the phrase only in stories will not just allow but in fact require negative inversion. The ungrammaticality of *Only in stories, a broken glass betrays agitation suggests that this is true. One question that of course remains is why negative inversion should have the semantic properties that I have shown it to have. I have no answer to o¨er here, but hope that future research will reveal one.16 This may be a good moment to sum up the main points of the discussion. I have claimed that in its lexical meaning only is similar to a universal quanti®er. It quanti®es over the nonfocused material that forms its restriction when it is in an adverbial position (Q-raising) and focal mapping applies. The resulting interpretation of only is conservative. Apparently like other universal quanti®ers, only has existential force; as a result, it is only weakly, rather than strongly, downward monotonic in its restriction. The weak downward monotonicity of only is enough to explain why it licenses NPIs and (together with the way focus shapes the quanti®cational structure of only-sentences) why it can trigger negative inversion. So far I have not considered exactly what only universally quanti®es over and how it does it; I take this up next. 4.7

Only Quantifying over Events

The analysis of only that I have developed assimilates it to an adverbial quanti®er in syntactic terms. It therefore seems promising to assume that only, like regular adverbial quanti®ers, quanti®es over events. Since it is a universal-like quanti®er, in meaning we might expect it to closely resemble always, which, as we saw in chapter 3, also universally quanti®es over events. Sentences with always and only indeed share a great deal in meaning, but as a closer comparison of minimal pairs reveals, their meanings do not overlap entirely. Beginning with the pair (54)±(55), we can observe that when used felicitously, always seems to be quantifying over events such that there needs to be more than one that satis®es the predicates in the restriction. Thus, (54) is taken to suggest that Gertrude quoted repeatedly. No such suggestion is made in (55). This sentence describes a single instance of Gertrude's quoting someone (unless the past tense is

106

Chapter 4

interpreted habitually, in which case only can be seen as taking narrow scope with respect to a distributing habitual operator; see (56)). (54) Gertrude always quoted SHAKESPEARE. (55) Gertrude only quoted SHAKESPEARE. (56) Gertrude always only quoted SHAKESPEARE. On the interpretation of (54) that we are concerned with here, its primary interpretation, the sentence is true when the focus on Shakespeare is exhaustive (see Krifka 1992). There is also a way of using (54) where the focus is not taken to be exhaustive, where Gertrude may also have quoted someone other than Shakespeare. Since this interpretation seems to arise when the sentence is uttered with a fall-rise contour, we can consider it an example of a nonexhaustive focus (see the appendix to chapter 2). Importantly, though, a nonexhaustive interpretation of the focus in (55) is never possible; it con¯icts with the meaning of only, which seems to explicitly impose an upper limit on events. What are the semantic interpretations of these sentences? In chapter 3 we saw that examples with always are interpreted as in (57), which serves as the input for the pragmatic principle that relates quanti®er restrictions to aboutness. (57) [all e: C(e) & Past(e) & Quote(e) & Agent(e,gertrude)] Past(e) & Quote(e) & Agent(e,gertrude) & Theme(e,shakespeare) Next, it seems fair to assume that it is for Gricean reasons that the universal quanti®er is taken to have a restriction that contains more than one event. After all, if the speaker knew that there was just one such event, he could have used something more informative (e.g., once). Assuming that always, like other universal quanti®ers, has existential force, we can spell out the lexical meaning of the metalanguage ``all'' in (57) in set-theoretic terms as in (58). According to (58), a universal quanti®cation over events is true when it is the case that the restriction is not empty and every event that is an element of the restriction is also an element of the scope. (58) [all e: F(e)] G(e) i¨ F 0 { } & F J G Given (58), (54) is ultimately interpreted with the truth-conditions given in (59). (59) {f: Past(f ) & Quote(f ) & Agent(f,gertrude)} 0 { } & {f: Past(f ) & Quote(f ) & Agent(f,gertrude)} J {e: Past(e) & Quote(e) & Agent(e,gertrude) & Theme(e,shakespeare)}

Only and Even

107

(59) requires that the set of events where Gertrude quoted not be empty and that all its members also be members of the set of events where Gertrude quoted Shakespeare. Now let us turn to only. (55) has the direct translation that is given in (60). (60) [only e: C(e) & Past(e) & Quote(e) & Agent(e,gertrude)] Past(e) & Quote(e) & Agent(e,gertrude) & Theme(e,shakespeare) How do we understand the restricted metalanguage ``only''? It seems that the truth-conditions of (55) require that whatever quoting Gertrude didÐ and she did some; this is the existential force of onlyÐit did not go beyond her quoting Shakespeare. This suggests the equivalence in (61). (61) [only e: F(e)] G(e) i¨ F 0 { } & Ef ((f A F) ! be (Part(f,of e) & e A G)) (61) states that the restriction is not empty and that each of its elements forms a part of a member of the scope. The resulting truth-conditions of (55) are represented in (62). (62) states that Gertrude quoted (®rst line) and that whatever quoting she did (second line) was part of her quoting Shakespeare (third and fourth lines). (62) be (Quote(e) & Past(e) & Agent(e,gertrude)) & Ee 0 ((Quote(e 0 ) & Past(e 0 ) & Agent(e 0 ,gertrude)) ! be 00 (Part-of(e 0 ,e 00 ) & Quote(e 00 ) & Past(e 00 ) & Agent(e 00 ,gertrude) & Theme(e 00 ,shakespeare))) (62) is very similar to the analysis proposed by Bonomi and Casalegno (1993). The authors reach this kind of analysis by examining the alternative semantics account of only, which they show to be problematic for reasons that would take us too far a®eld here. Finally, one attractive feature of the analysis that may not be immediately obvious but is worth highlighting is that it extends in a simple way to scalar uses of only (see also Bonomi and Casalegno 1993). Along with the ``regular'' only, whose meaning is spelled out in (61), there is a second, scalar one, whose interpretation can be stated as follows: (63) [only-scalar e: F(e)] G(e) i¨ F 0 { } & Ef ((f A F) ! be (Not-ranked-higher(f,than e) & e A G)) As is easy to appreciate, scalar only is very similar to nonscalar only. The only di¨erence between the two lies in the relation that holds between the events that are members of the set denoted by the restriction and the members of the set denoted by the scope. Whereas in the interpretation of

108

Chapter 4

``regular'' only this is a ``part of '' relation, in the interpretation of scalar only it is a relation of being ranked on a contextually chosen scale. Given (63), our earlier example Juliet was only DRUGGED is analyzed as having the following truth-conditions: (64) be (Theme(e, juliet) & Past(e)) & Ee 0 ((Theme(e 0 , juliet) & Past(e 0 )) ! be 00 (Not-ranked-higher(e 0 ,than e 00 ) & Theme(e 00 , juliet) & Past(e 00 ) & Be-drugged(e 00 ))) According to (64), Juliet was only DRUGGED is true exactly when Juliet was something and it was not ranked higherÐworse, that isÐthan her being drugged. This excludes that she was dead, as desired. 4.8

Only's Partner Even

As noted earlier, only is syntactically peculiar in that it is an ``admanythings.'' If we look at the overt syntax of sentences it occurs in, we ®nd that in some instances it resembles an adverb and in others a determiner. I accounted for this by saying that at LF only uniformly functions like an adverb; should it not surface in an adverbial position, it covertly moves to one by Q-raising. As a result, only is syntactically unary at LF and can receive its arguments through focal mapping. Many of the syntactic properties that only has, even shares. It therefore seems natural to extend the kind of analysis that was given for only to even. In what follows I will review some of the syntactic parallelisms between only and even, on the basis of which I propose that like only, even is an adverbial-like quanti®er that can undergo Q-raising and that receives its arguments through focal mapping. I then take up the question of what even quanti®es over and what relation holds between its arguments. I will proceed by reviewing some of the literature on this issue and gradually introducing the view that I will argue for. 4.9

Syntactic Similarities between Only and Even

Let us begin with even's surface syntactic behavior. Like only, even can be attached to a variety of phrases, and like only, it associates with a focused element in its c-command domain. Consider the examples in (65). (65) a. On his trip to Europe, Peter even went to ALBANIA. b. Even CLAIRE laughed. c. Denise even managed TO GET THE PRINTER TO WORK.

Only and Even

109

By analogy with only, even seems to violate the generalization that it needs to associate with a focus in its c-command domain only when what we are dealing with is an instance of second occurrence focus, as in (66). (66) A: Sid even tried THE OCTOPUS. B: No, DOUG even tried the octopus. Remarks made earlier about the example with only carry over to this example: because (66B) mimics (66A), the focus on the octopus in (66B) is barely realized, but it is there, and it is with this focus that even associates as a consequence of the mimicking that characterizes second occurrence focus. One di¨erence between the syntactic behavior of even and that of only is that unlike only, even in an adverbial position can associate with the subjectÐprovided the subject is focused, of course (e.g., Jackendo¨ 1972, Rooth 1985). This property, illustrated in (67), is one that even shares with negation and with adverbial quanti®ers (see p. 43). (67) THE DEAN even/*only/didn't/always congratulate(d) David. The next question that arises is whether even behaves like only with respect to Taglicht's observation. The data in (68) suggest that it does. (68) a. The freshmen were reminded to bring even THE REGISTRATION FORM. b. The freshmen were reminded to even bring THE REGISTRATION FORM. (68b) can only mean that the freshmen received a piece of advice something like this: ``Bring even THE REGISTRATION FORM, along with the tuition bill and your preliminary schedule. You'll need it.'' Although this reading is also available for (68a), it is not the only interpretation for that sentence. (68a) can also mean that even the registration form is such that the freshmen were told to bring it (even though the registration forms were impossible to ®nd, for example). As with the wide scope reading of only THE AZALEAS in (11a), this reading seems to come about most easily with a certain exclamatory intonation and/or with a pause before the even (only). The fact that in (68b) even, which is in an adverbial position, cannot take scope outside its clause parallels the behavior of only in the same position. Furthermore, it seems that even can take scope outside of its clause only by piggybacking on the scope of a quanti®ed noun phrase. Both facts together suggest that even patterns with only with respect to Taglicht's observation. Thus, except for their di¨ering behavior with

110

Chapter 4

respect to focused subjects, even and only are virtually identical in syntactic terms; the two form a natural class.17 Given the close parallels between even and only, it seems attractive to extend the earlier analysis to even. If even does not surface in an adverbial position, it moves to one by local Q-raising. It then receives an internal argument that corresponds to the nonfocused material in its scope (focal mapping). Like other quanti®er restrictions, the restriction of even encodes what the relevant part of the sentence is pragmatically about. Thus, Even CLAIRE laughed is about an event where there was laughter. 4.10

Implicatures and Being Noteworthy

Let us now consider the meaning of even-sentences. On what seems to be the standard analysis of even, it is said to give rise to two conventional implicatures, which are typically phrased in terms of likelihood (e.g., Karttunen and Peters 1979, Rooth 1985, Wilkinson 1996). The ``scalar implicature'' says that all propositions where the focused element is replaced by an alternative are more likely than the original one. The ``existential implicature'' states that there is a value other than that of the focused element that will make the sentence true. For example, Even CLAIRE laughed (65b) is said to implicate (69a) and (69b). (69) a. Claire was the least likely to laugh. [scalar implicature] b. Someone other than Claire laughed. [existential implicature] Even if (69) seems to provide an accurate analysis of the meaning of (65b), closer inspection suggests that the meaning of even may not have to do with likelihood. The following example from Kay 1990 suggests that a di¨erent notion may be involved, something Kay analyzes in detail as ``informativeness.'' (70) A: It looks as if Mary is doing well at Consolidated Widget. George [the second vice president] likes her work. B: That's nothing. Even BILL [the president] likes her work. The point of this example is that it can be used felicitously in a context where Bill may actually not have been the least likely to appreciate the work Mary is doing. Rather, given that he is the president of the company, his approving of Mary's work is more ``informative'' than George's because it ranks higher on a scale that indicates how well Mary is doing at Consolidated Widget. For ease of exposition, I would like to refer to

Only and Even

111

``more informative'' as ``more noteworthy,'' a designation I ®nd more perspicuous. In light of these observations, we can rephrase the scalar implicature in terms of noteworthiness. Instead of (69a), (65b) now has this scalar implicature: (71) Claire's laughing was more noteworthy than anyone else's. 4.11

Ambiguous Even-Sentences

One particularly intriguing property about even that has received considerable attention is that it gives rise to ambiguity when it occurs in a downward-entailing context (Karttunen and Peters 1979, Rooth 1985, Wilkinson 1996, Rullmann 1997). We can see this by considering (72). (72) a. It is hard for me to believe that Bill can understand even SYNTACTIC STRUCTURES. (Karttunen and Peters 1979) b. I doubt that Andy even knows how to calculate COMPOUND INTEREST. There is one reading of (72a), the ``easy'' reading, where Syntactic Structures is something one would expect Bill to understandÐand where the fact that he cannot may not re¯ect well on his abilities as a syntactician. The other reading, the ``hard'' reading, comes about when someone has claimed that Bill can understand something as complex as Syntactic Structures, but the speaker doubts that this is true. Similarly, we obtain an ``easy'' reading of (72b) if we think of Andy as an accountant who was recommended to us but who turned out to be much less helpful than we had hoped. We obtain the ``hard'' reading if we think of Andy as a threeyear-old boy whose parents boast that he is so gifted he can calculate compound interestÐsomething we ®nd hard to believe. The ``hard'' reading is said to be the result of the ``regular'' interpretation of even, given above, and even's taking scope over the clause it appears in. The implicatures it gives rise to are presumably passed on to the whole sentence. (73) a. It is hard for me to believe that Bill can understand Syntactic Structures. [truth-conditional meaning] b. Syntactic Structures is the most noteworthy (least likely) thing for Bill to be able to understand. [scalar implicature] c. There is something other than Syntactic Structures that Bill can understand. [existential implicature]

112

Chapter 4

Interestingly, although it seems true that the ``hard'' reading of (72a) commits one to the scalar implicature (73b), it is less obvious that this also applies to the existential implicature (73c). It seems that one can assert (72a) without conventionally implicating that Bill can understand anything else. This becomes especially clear when we consider the following, parallel example (on the ``hard'' reading): (74) I doubt that Clever Hans can even calculate COMPOUND INTEREST, because I doubt that Clever Hans can calculate ANYTHING! He is a horse, for God's sake! As the clarifying continuation makes clear, the speaker does not believe that Clever Hans can calculate something other than compound interest. This is but one problem that the existential implicature that is posited for even raises. Others become apparent when we turn to the accounts that have been given for the ``easy'' reading, the scope account (Karttunen and Peters 1979, Wilkinson 1996) and the NPI account (Rooth 1985, Rullmann 1997). Proponents of the scope account maintain that even has a single meaning, the one described above. This analysis elegantly reduces the di¨erence between the ``easy'' reading and the ``hard'' reading to a di¨erence in the scope of even; the ``easy'' reading arises when even takes wide scope over the negative element, the ``hard'' reading when even takes narrow scope. For (72b), wide scope of even over the matrix clause containing doubt generates these implicatures: (75) a. Compound interest is the least likely (most noteworthy) thing for me to doubt that Andy knows how to calculate. b. There is something other than compound interest that I doubt Andy knows how to calculate. Compound interest is the most noteworthy (least likely) thing for me to doubt that Andy can calculate if it is something that I expect Andy to be able to do without any problemÐif it is easy for Andy, for instance. In contrast with the scope account, the NPI account gives up the idea that there is one even. It posits that along with the ``regular'' even described above, there is a second even, which is an NPI and which has reversed implicatures. Its existential implicature is that there is an alternative to the focus whose semantic value makes the sentence false, and its scalar implicature states that the value of the focused element was the most likely (least noteworthy) one to make the sentence true. Like other

Only and Even

113

NPIs, NPI even is restricted to downward-entailing environments. It is licensed in the complement of doubt in (72b), for instance, where it generates the following implicatures for the ``easy'' reading: (76) a. Calculating compound interest is the most likely (least noteworthy) thing for Andy to know how to do. [scalar implicature] b. There is something other than calculating compound interest that Andy does not know how to do. [existential implicature] Comparing the two accounts, Rooth (1985) argues that the NPI account is to be preferred because, unlike the scope account, it makes the right predictions for (77b) uttered in the context of (77a). (77) a. Because they had been stolen from the library, John couldn't read The Logical Structure of Linguistic Theory or Cartesian Linguistics. Because it was always checked out, he didn't read Current Issues in Linguistic Theory. b. The censorship committee kept John from reading even SYNTACTIC STRUCTURES. On the scope account, (77b) should implicate that there is something other than Syntactic Structures that the censorship committee kept John from reading. This, however, is incorrect since (77a,b) are consistent with the censorship committee's not having interfered in any aspect of John's studying other than his reading Syntactic Structures. The existential implicature is di¨erent on the NPI analysis; here (77b) implicates that there is something other than Syntactic Structures that John did not read, which creates no problem. Although at ®rst sight this argument seems devastating for the scope analysis, Wilkinson (1996) points out that it overlooks the fact that on a natural reading of (77b) in the context of (77a) the sentence has a focus on the censorship committee. She uses this fact to e¨ectively remove the censorship committee from within the wide scope of even, which allows her to predict the right implicatures. This suggests that nothing can be gleaned from this example. Wilkinson argues further that when even is embedded under a factive predicate that can license an NPI, such as sorry or glad, the scope account actually fares better than the NPI account. (78) I am sorry that I even OPENED that book.

114

Chapter 4

According to the NPI account, (78) should implicate that there is something other than opening it that the speaker did not do with that book, where opening was the most likely (least noteworthy) thing for the speaker to do with it. But the speaker may in fact have done everything else with the book (read it, studied it, recommended it to a friend) and, even so, she can still use (78) felicitously. This, as Wilkinson points out, is a problem for the NPI account. Interestingly, however, the scope analysis does not seem to fully account for (78) either.18 The analysis predicts that (78) implicates that there is something other than opening that the speaker is sorry about having done with the book in question. Yet the speaker may use (78) felicitously without having done anything to the book other than open it. In other words, (78) is neutral as to whether the speaker did anything else with the book or not. Thus, whereas the NPI analysis wrongly predicts that there should be something other than opening that the speaker did not do with the book, the scope analysis wrongly requires that there be something other than opening the book that the speaker regrets having done with it. At this point we may simply want to give up the claim that even introduces an existential conventional implicature and try to ®nd a di¨erent explanation for those cases where something like an existential implicature seems to arise. 4.12

The Existential Implicature as Pragmatic Inference

Rullmann (1997) proposes an analysis along these lines in connection with the NPI account. The idea is that in those cases where even seems to have an existential implicature, it really reduces to a pragmatic inference. Crucially, in example (78) with sorry, where there is nothing that would correspond to the existential implicature, the relevant pragmatic inference is not available. Let us see how this works on the NPI account, starting with the compound interest example (75). If calculating compound interest is the most likely (least noteworthy) thing for Andy to be able to do (scalar implicature), and the speaker doubts he can do it, then we can pragmatically infer that the speaker will also have doubts about Andy's ability to do any harder things (e.g., ®nd a loophole in the tax law, know what to make of foreign bank statements). This is so because if we doubt that a likely thing is true, then, assuming we are rational, we are virtually certain to not believe that any less likely things are true. Or, to phrase it in terms of

Only and Even

115

noteworthiness, if the speaker says she has doubts about the least noteworthy of abilities, her saying so invites the pragmatic inference that she will also have doubts about any abilities that would be more noteworthy. Interestingly, the same idea also works for the scope account; if Andy's being able to calculate compound interest is the least likely (most noteworthy) thing for the speaker to doubt that Andy can do, but the speaker (says she) doubts it, then the hearer can pragmatically infer that the speaker will also doubt that Andy is capable of more likely (less noteworthy) things. Turning now to the example, with sorry in (78) if I am sorry I even opened the book, then, given the factivity of sorry, I opened the book. At this point there are no grounds to draw any further pragmatic inference; we cannot infer that there is something less expected (more noteworthy) than opening that I did not do with the book (ˆ wrong existential implicature of NPI account). Nor can we infer that there are other things that I did with the book and that I regret (ˆ wrong existential implicature of scope account). It seems promising, then, to give up the idea that even (whether it is the NPI even or the ``regular'' one) has an existential implicature. Getting rid of the existential component takes care of the problems that the example with sorry raises for both the scope theory and the NPI theory; it also seems attractive in light of the Clever Hans example (74). What is left to consider are simple cases like (65b), which do seem to express an existential commitment on the part of the speaker that somebody other than Claire laughed. (65) b. Even CLAIRE laughed. How does this follow? We are saying that this sentence means ``Claire's laughter was more noteworthy than any other laughter.'' Our analysis only works if this paraphrase allows for the pragmatic inference that there was someone else who laughed. If it does not, then by trying to derive the absence of an existential implicature through pragmatic inference, we may really have thrown the baby out with the bathwater. I think the inference goes through, and it does so because the predicate involved is the pragmatic ``noteworthy'' instead of the probabilistic ``likely.'' Things we consider unlikely happen all the time, and they are the things that are particularly interesting to talk about. If Claire's laughter was the most unlikely laughter, we have no pragmatic reason to infer that someone more likely to laugh than Claire laughed also; that is,

116

Chapter 4

Most unlikely though it was, Claire laughed does not suggest that someone else laughed too. On the other hand, it would be pragmatically bizarre to say that Claire's laughing was more noteworthy than any other event of laughing if there was no other event of laughing. Therefore, I think saying that the event of Claire's laughing was more remarkable than any other event of laughing strongly invites, if not forces, the pragmatic inference that there was another event of laughing.19 To summarize: I have shown that, in their own ways, the scope account and the NPI account both run into a problem with the existential implicature that each analysis posits in the ``easy'' reading. I suggested that these di½culties are resolved if we do away with the existential implicature of even and instead explain those cases where there appears to be such an implicature in terms of pragmatic inference. It seems that for this approach to work, the meaning of even needs to be phrased in terms of noteworthiness rather than likelihood. Since up to this point the scope account and the NPI account seem to fare equally well, we may want to look at other criteria to decide between them. 4.13

Syntactic and Crosslinguistic Support for the NPI Approach

Although the scope account is attractive in that it avoids positing a lexical ambiguity, it comes at the price of requiring instances of covert movement that may be doubtful. One aspect of this problem is that given the way the scope analysis accounts for the ``easy'' reading of (72b), even and the focus have to move out of their clause to take scope over doubt in the higher clause, as in (79). (72) b. I doubt that Andy can even calculate COMPOUND INTEREST. (79) [even COMPOUND INTEREST] I doubt that Andy can calculate This kind of movement is surprising if even in adverbial position can in general not take scope outside of its clause, if it obeys an analogue of Taglicht's observation for only, as suggested earlier in connection with (68). A further and no less serious problem for moving the even in ``easy'' readings is that it requires even to move out of an island (Rullmann 1997). Since (80) can have the ``easy'' reading, here even would have to move out of the complex noun phrase containing a relative clause, a very strong island indeed.

Only and Even

117

(80) They hired no linguist who had even read SYNTACTIC STRUCTURES. Also, notice that (81a) allows the ``easy'' reading along with the ``hard'' one, but (81b) does not (Rullmann 1997). (81) a. They didn't hire any linguist who had even read SYNTACTIC STRUCTURES. b. They didn't hire the linguist who had even read SYNTACTIC STRUCTURES. As Rullmann points out, this is easily explained on the NPI account; since we know that the di¨ers from any in blocking the licensing of NPIs inside nominals, it is expected that (81a) should have both the ``hard'' reading and the ``easy'' reading, and that (81b) should have only the ``hard'' reading. In contrast, (81) would not be easy to explain on the scope account, which seemingly would have to stipulate that when the determiner is any, even can raise out of a complex noun phrase, but when the determiner is the, it cannot. Finally, if even in English is indeed ambiguous between ``the most noteworthy'' and its NPI counterpart that means ``the least noteworthy,'' one might expect to ®nd other languages where it is overtly disambiguated. There are indeed such languages. German, for instance, distinguishes between the two interpretations of even with sogar (``the most noteworthy'') and auch nur (``the least noteworthy''), Dutch has zelfs and zelfs maar, Swedish tom and ens, Finnish jopa and edes (see von Stechow 1991, Rullmann 1997). Finally, Spanish contrasts incluso and hasta (Schwentner 1998) in ``hard'' readings and uses ni siquiera in ``easy'' readings. Spanish ni siquiera is of particular interest in this respect because it is a member of the n-word paradigm. As we saw brie¯y in chapter 2, n-words that appear preverbally translate as negative elements, whereas those that occur postverbally generally (though not always!) translate as NPIs. I take this to show that n-words in Spanish are lexically ambiguous between genuine negative elements and the NPI counterparts of these elements. Thus, nadie is ambiguous between ``nobody'' and the NPI ``anybody.'' For arguments in favor of this kind of analysis, see Herburger 1998. Ni siquiera ®ts nicely into the picture being drawn here. When it occurs preverbally as in (82a), it translates as ``not even.'' On the other hand, when it occurs postverbally as in (82b), it is not itself negative, but simply means ``even.'' In both instances ``even'' means ``the least noteworthy.''

118

Chapter 4

(82) a. Ni siquiera Pablo lo supo. ni even Pablo it knew ``Not even Pablo knew it.'' b. No lo supo ni siquiera Pablo. not it knew ni even Pablo (same as (82a)) Consider also (83). Here ni siquiera appears preverbally but in an embedded position where an NPI can be licensed by a matrix dudar (``doubt''). As with other n-words in analogous examples, the result is ambiguous; (83) can mean that I doubt that even the owner knows it (NPI reading), or it can mean that I doubt that not even the owner knows it. (Because it involves a logical double negation, which is pragmatically more marked, the latter reading is somewhat less salient.) (83) Dudo que ni siquiera la duenÄa lo sepa. (I) doubt that ni even the owner it knows ``I doubt that even the owner knows it.'' ``I doubt that not even the owner knows it.'' Thus, like other n-words, ni siquiera seems to be lexically ambiguous between a genuine negative reading (``not even'') and the corresponding NPI reading (``even''). In both instances ``even'' means ``the least noteworthy.'' By being a full-¯edged member of the n-word paradigm and thus being ambiguous between an NPI and its negative counterpart, ni siquiera provides direct evidence that ``even'' can be an NPI. To the extent that crosslinguistic comparisons of ``admanythings'' are signi®cant, the systematic ambiguity of ni siquiera supports the claim that English even can be an NPI. For further discussion see Barker and Herburger 1999. 4.14

Conventional Implicature versus Literal Meaning

Beginning with the syntactic properties of even, I noted that it behaves much like only, suggesting that the two form what seems to be a natural syntactic class of ``admanythings.'' I then observed that ``regular'' even probably means something like ``the most noteworthy,'' rather than ``the least likely.'' The rest of the discussion has dealt with the ambiguity that even gives rise to when it appears in a downward-entailing context (``hard'' vs. ``easy'' readings). Reviewing the two competing analyses of this factÐ the scope account, which reduces the ambiguity to scope, and the NPI account, which posits a second, NPI even (``the least noteworthy'')Ð

Only and Even

119

I noted that both run into a problem with cases where even is embedded under a factive predicate that licenses NPIs. The problem is solved if we give up the existential implicature that both accounts posit for even, and if we explain any existential commitment an even-sentence seems to express as a pragmatic inference. The paraphrase of even as ``the most noteworthy''/``the least noteworthy'' plays an important role in this argument. The decisive factor in choosing the NPI account over the scope account is that the scope account requires moving even in a way that is problematic syntactically. Crosslinguistic data also may provide independent support for the NPI account. All along I have assumed that a sentence like (84) literally says that Bill likes Mary and only conventionally implicates that Bill's liking Mary is less likely/more noteworthy than anyone else's liking her. In other words, the latter is said to play no part in the truth-conditions of the sentence. (84) Even BILL likes Mary. Karttunen and Peters (1979, 12) say that this is shown by the fact that if Bill is not the least likely to like Mary, ``the speaker can justly be criticized for having a wrong idea of how things are. Interestingly enough, though, such criticism would normally be rather mild, usually crediting the speaker with saying something that is partially correct.'' It seems we can explain this observation di¨erently if we take seriously the idea that even has to do with being noteworthy rather than with being likely. Let us say that it is part of the literal, truth-conditional meaning of non-NPI even that it involves ``the most noteworthy.'' What is now important to note is that unlike likelihood, which is a hard-to-determine yet objective, statistical matter, being noteworthy is an inherently subjective, psychological a¨air. If some event is more noteworthy than any other comparable one, we can ask, ``More noteworthy according to who?'' Let us then rephrase our analysis and say that ``regular'' even means ``the most noteworthy according to x,'' where x in a simple clause is the speaker. Going back now to (84), I think we can make sense of Karttunen and Peters's above observation while maintaining that even contributes to the literal, truth-conditional meaning of a sentence. On this view, (84) has the following truth-conditions: (85) a. Bill likes Mary and b. according to x, x ˆ the speaker, Bill's liking Mary is more noteworthy than anyone else's liking her.

120

Chapter 4

If Bill cannot stand Mary, the ®rst conjunct of (85) is false. There is no doubt then that the speaker did not speak the truth. When do we know that (84) is false because the second conjunct of (85) is false? Only, it seems, if we know that Bill's degree of a¨ection for Mary does not occupy the end of the scale that the speaker is using. If we do not know what scale the speaker is using but we only see that according to our scale Bill's degree of a¨ection for Mary does not occupy the endpoint, we can suspect one of two things. Either we can conclude that the speaker has the same scale we have, but simply does not want to be taken to be speaking the truth. Or we can conclude that he wants to be taken to be speaking the truth, but has a di¨erent scale from ours. Since the basic assumption we operate on is that a speaker wants to be taken to be speaking the truth, the second conclusion is much more plausible than the ®rst. We can observe now that if the speaker has a di¨erent scale from ours, (85b) is not false; the choice of scale is after all relativized to the speaker's belief. We may therefore not conclude that the speaker did not speak the truth, but only that her beliefÐin particular, her choice of relevant scaleÐwas misguided. If (84) is uttered in a context where Bill likes Mary but Mary is not the most surprising personÐin our viewÐfor him to like, what the speaker said is true, because it involves a claim about her beliefs. But we only partially credit her with being right about how things really are. Although she spoke the truth, one of the beliefs she reported is in our view mistaken. This, I think, captures something very similar to Karttunen and Peters's initial observation. But it does so without saying that the scalar part is a conventional implicature. Rather, it forms part of the truth-conditional meaning, but, crucially, it is relativized to the beliefs of the speaker. One ®nal descriptive point I would like to make is this. In a simple example like (84) it seems clear who holds the view that Bill's liking Mary was more noteworthy than anyone else's liking herÐit is the speaker of the sentence. However, when even-sentences are embedded in propositional attitude contexts, a de re/de dicto ambiguity arises. (86), for example, appears to have two relevant readings. (86) Naomi said that even BILL likes Mary. On the de dicto reading of even, Naomi said Even BILL likes Mary. On the de re reading, she might just have said that Bill likes Mary, and the claim that Bill's liking Mary is the most noteworthy liking of Mary is

Only and Even

121

taken to be part of the opinion of the person who utters (86), not Naomi's opinion. 4.15

Summary of the Discussion regarding Even

As the analysis of even that I would like to propose has been introduced bit by bit, it may be helpful to recapitulate the main steps to give a clear overall picture. I set out by observing that even forms a natural class with only; both can attach to a wide variety of syntactic constituents, and both require a focus in their surface syntactic c-command domain. Extending the analysis of only proposed in sections 4.3±4.7 to even, I suggested that even receives its arguments through focal mapping and Q-raising. Given a pragmatic principle relating quanti®cational structure to aboutness, the quanti®cational structure that results captures the fact that the nonfocused material provides the background information (what the relevant part of the sentence is about). Regarding the ambiguity that even-sentences exhibit in downwardentailing contexts, I adopted the view that there are two evens, a ``regular'' even and an NPI even. I was led to this conclusion not so much by semantic as by syntactic and crosslinguistic considerations. In the spirit of Kay 1990, I further argued that even relates its arguments not through the notion of likelihood, but through a more pragmatic notion, that of being noteworthy (``informative'') relative to a conversational purpose. Adopting this view made it possible to dispense with the claim that the two evens are associated with existential conventional implicatures stating that some alternative described by the nonfocused part is true (false). Instead, I explored a possibility noted by Rullmann (1997) that in those cases where an existential implicature can be thought to exist, it really reduces to a pragmatic inference. The ®nal step in the analysis consisted in attempting to dispense with the scalar implicature as well, by making it part of the literal, truthconditional meaning. In this context I argued that ``being noteworthy'' is relativized to the beliefs of a particular person (either the speaker, or the argument of a higher propositional attitude expression). Having said all of this, I can perhaps best summarize the present analysis by stating the lexical meaning of the two evens. The meaning of ``regular'' even can be captured as in (87), and that of NPI even as in (88).

122

Chapter 4

(87) [even-regular e: F(e)] G(e) i¨ be ((e A G) & Ef ((f A F & e 0 f ) ! More-noteworthy(e,than f ))) (88) [even-NPI e: F(e)] G(e) i¨ be ((e A G) & Ef ((f A F & e 0 f ) ! Less-noteworthy(e,than f ))) (87) states that an event described by the sentence without even took place and that that event is more noteworthy than all other events that satisfy the nonfocused part. (88) is the same, except that the event described by the sentence is said to be the least noteworthy.

Chapter 5 Determiners

5.1

Introduction

Common wisdom has it that adverbial quanti®ers and nominal determiners (most, no, some, etc.) di¨er drastically in how they ®nd their restrictions. As we saw in chapter 3, adverbial quanti®ers generally get their restriction through the mediation of focus. (Only sometimes, as in the case of if-clauses and preposed when-clauses, do other factors play a role.) In contrast, the quanti®cational structure of determiners is held to be shaped more than anything by syntactic structure, the standard observation beingÐas we saw in the discussion of onlyÐthat a determiner D has to be restricted by the predicate that is denoted by its internal argument, where the internal argument corresponds to the NP (if we assume that noun phrases have DP structure: [ DP D NP]). Sentences with only and even turned out to play an interesting role in the classi®cation of adverbial quanti®cation vis-aÁ-vis determiner quanti®cation. When these ``admanythings'' surface in adverbial position (e.g., adjoined to a verb phrase), they pattern semantically with adverbial quanti®ersÐtheir quanti®cational structure depends directly on focus. But, curiously enough, the same is true when only and even appear not in adverbial position but, for instance, adjoined to noun phrases. Rather than behaving in the way determiners are supposed to in analogous con®gurationsÐtaking their surface-syntactic internal argument as their restrictionÐonly and even have a quanti®cational structure that is shaped by focus. I accounted for this by positing that when they do not already surface in an adverbial position, only and even covertly raise to such a position (Q-raising). Focal mapping then straightforwardly accounts for their quanti®cational structure. In this chapter I show that, contrary to the standard view, determiners sometimes pattern with only and even. They behave like adverbial quan-

124

Chapter 5

ti®ers in that the NP that is their internal argument in the surface syntax does not have to be interpreted as their restriction; their restriction and scope depend only on focus. In line with my analysis of only and even, I will propose that these ``focus-a¨ected'' readings involve Q-raisingÐin particular, local raising of the determiner to a position that neutralizes the distinction between internal and external argument. Subsequent focal mapping produces the right quanti®cational structure. The only di¨erence between focus-a¨ected readings of determiners and the e¨ects of focus on adverbial quanti®ers and only and even lies in the kinds of things we are quantifying over, which in the case of determiners need not be events, but are often individuals. It turns out that not all noun phrases can have this focus-a¨ected reading; only those exhibiting the de®niteness e¨ect can do so. Because the existence of focus-a¨ected readings has interesting consequences for the analysis of the de®niteness e¨ect, a summary of basic facts regarding the de®niteness e¨ect seems a good place to start. The following summary, the basic description of the facts, and the claim that sometimes determiners raise on their own build to a considerable extent on Herburger 1997a. The analysis presented here di¨ers substantially, however, in that it does not adopt an alternative semantics view of focus. It also goes substantially beyond Herburger 1997a in that toward the end of the chapter I attempt a more general analysis of the weak/strong contrast and how it relates to the de®niteness e¨ect. 5.2

Focus-A¨ected Noun Phrases: The Basic Phenomenon

The well-known descriptive observation concerning the de®niteness e¨ect (due to Milsark 1974; see also Milsark 1977) is that although some noun phrases are acceptable in there-insertion contexts, others are not. Thus, we distinguish between weak noun phrases (acceptable) and strong ones (unacceptable). (1) a.

There are some/three/many/few/no children in the garden. [``weak''] b. *There is (are) every/each/all/most child(ren) in the garden. [``strong'']

Milsark further points out that weak noun phrases are not limited to there-sentences but can also appear in contexts where strong noun phrases are acceptable as well. Depending on the environment a weak noun phrase appears in, its interpretation changes. Thus, in de®niteness e¨ect

Determiners

125

(DE) contexts such as (1a) they have what he calls a cardinal reading; for instance, few children is understood as ``few in number.'' In contrast, when few children occurs outside a DE environment, few cannot be paraphrased as ``few in number'' but must be understood proportionally, as ``a small percentage of (some implicit class of ) children,'' analogous to the way it is understood in the partitive few of the children. The two interpretations of few really are di¨erent and cannot be attributed to an inherent vagueness of the determiner; in fact, as pointed out by Partee (1988), they di¨er truth-conditionally. The cardinal interpretation of few in Few children are playing in principle allows for 100% of the children to be picked out (so long as they are few in number). Consider also There appeared few egg-laying mammals in the survey because there are only few (which is fashioned after an example Partee (1988) attributes to Al Huettner). This example can be uttered felicitously if in fact all of the egg-laying mammals appeared in the survey. On the other hand, the strong(like) reading of few is proportional and thus never allows 100% to be picked out, no matter how relative ``a small proportion'' is. Since a partitive always has a strong(like) interpretation, we cannot truthfully utter Few of the egg-laying mammals appeared in the survey, when in fact all did. This truth-conditional di¨erence between cardinal few (``few in number'') and proportional few (``a small proportion'') supports Milsark's claim that there is a true semantic di¨erence between few in a noun phrase that occurs in a DE environment and few in a noun phrase that occurs outside such an environment.1 Although there-insertion provides the paradigmatic DE environment, the subject position of stage-level predicates also optionally allows for weak noun phrases with cardinal interpretations. This position is only an optional DE environment, however, because along with cardinal weak noun phrases it can also be occupied by strong noun phrases and weak noun phrases on their strong(like) interpretation (which, as we just saw, is proportional in the case of few). Finally, unlike the subject position of stage-level predicates, the subject position of individual-level predicates can only be occupied by strong noun phrases or weak noun phrases with a strong(like) reading; cardinal interpretations are impossible in this context. Consider (2). (2) a. Few children are playing. b. Few children like spinach. In (2a), with a stage-level predicate, few children can be interpreted either with the cardinal reading, along the lines of ``children who are few in

126

Chapter 5

number'' (cf. (1a)), or with a strong(like) reading, where few is proportional as in the overt partitive few of the children. Unlike (2a), (2b) has an individual-level predicate, and here we ®nd only the strong(like) reading: (2b) claims that a certain contextually small percentage of the children in question like spinach. Summarizing Milsark's distributional observations, then, we can say that there-insertion sentences constitute a DE environment, the subject position of stage-level predicates constitutes an optional DE environment, and the subject position of individual-level predicates constitutes an anti± DE environment. Now consider the following sentence: (3) Few INCOMPETENT cooks applied. (3) exhibits a weak noun phrase in the subject position of a stage-level predicate. According to the previous discussion, the subject should have exactly two readings: a cardinal, Milsark-style one and a strong(like) one. On the ®rst reading, the noun phrase means ``a small number of incompetent cooks''; on the second, it means ``a small percentage of the incompetent cooks.'' Taking into account the focus, which is as usual on the level of event quanti®cation, we obtain an interpretation of the following sort: (4) [be: C(e) & Apply(e) & Past(e) & [few x: Cook(x)] Agent(x,e)] Apply(e) & Past(e) [few x: Incompetent(x) & Cook(x)] Agent(x,e) (4) amounts to saying that some relevant event where few cooks applied was an event where few incompetent cooks applied. ``Few'' in this paraphrase is ambiguous between the Milsark-style and the proportional readings. There is, however, a third reading of (3) where focus directly reshapes the quanti®cational structure not of the event quanti®er but of the determiner. This reading can be paraphrased as ``Few cooks that applied were incompetent.'' (5) [be: C(e)] [fewprop x: Cook(x) & Apply(e) & Past(e) & Agent(x,e)] Incompetent(x) & Cook(x) & Apply(e) & Agent(x,e) (5) states that in some relevant event few of the cooks who were applicants in that event were such that they were incompetent (and applicants in that event). Here the determiner is not interpreted as saying ``few in number''; instead, it is interpreted proportionally. Unlike in the strong(like) reading, however, few quanti®es over cooks who applied, rather than over incom-

Determiners

127

petent cooks. Descriptively, in a case like (3) the focus inside the noun phrase contributes to the scope of the determiner and the entire nonfocused part restricts the determiner. The existence of focus-a¨ected readings shows that determiners are, after all, not always restricted by the material that surfaces as their internal argumentÐsometimes they behave in a way that strongly recalls the behavior of only and even when they are adjoined to a noun phrase, where their restriction seems to be entirely determined by the focus.2 The next group of cases will show that focus-a¨ected readings occur only in DE environments. As a ®rst piece of evidence let us consider the following paradigm: (6) a. Few/Many/No/Three/Some INCOMPETENT cooks applied. b. Most/All/Every/Each/Neither INCOMPETENT cook(s) applied. c. Few/Many/No/Three/Some COOKS know how to make a sou¿eÂ. In (6a) a weak noun phrase occurs as the subject of a stage-level predicate. Apart from the usual strong(like) and Milsark-style readings, it can also be interpreted in a focus-a¨ected way, and it thus has an interpretation analogous to (6), which, abstracting away from the contribution of the event semantics, we can represent in the simpli®ed form in (7). It states that few/many/no . . . cooks who applied were incompetent. (7) [D x: Applied(x) & Cook(x)] Incompetent(x) & Applied(x) & Cook(x) [ focus-a¨ected reading] Unlike in (6a), where the noun phrases are weak, in (6b) the noun phrases are strong and no focus-a¨ected readings are possible; (6b) cannot be paraphrased as ``Most/All . . . cooks who applied were incompetent.'' Finally, no focus-a¨ected readings are possible in (6c) either. This correlates with the fact that although the noun phrases are weak (their determiner is few, some, no, etc.), they are subjects of an individual-level predicate, which forces them to have a strong(like) reading. In particular, this sentence cannot mean ``Few/Many/No . . . people who know how to make a sou¿e are cooks''; it can only mean ``Few/Many/No . . . people who are cooks know how to make a sou¿eÂ.''3 Focus-a¨ected readings can also appear in the paradigmatic DE context, there-insertion. (8) a. There are many/few speakers of Basque WHO ARE CITIZENS OF SPAIN.

128

Chapter 5

b. There are many/few citizens of Spain WHO ARE SPEAKERS OF BASQUE. With many, it is possible to judge (8a) true and (8b) false (and conversely with few). Since the truth-conditions of (8a) and (8b) di¨er, we can conclude that many and few do not have their Milsark-style reading here. As we will see, if they did, switching speakers of Basque and citizens of Spain should not have any e¨ect on the truth-conditions, but it does. If the speakers of Basque who are citizens of Spain are relatively many, so are the citizens of Spain who are speakers of Basque, however contextsensitive many may be. But (8a) and (8b) can have di¨erent truthconditions. (8a) asserts, truthfully it seems, that many speakers of Basque are citizens of Spain (see (9a)). (8b), on the other hand, says that many citizens of Spain are speakers of Basque (see (9b)). AndÐdespite the relative vagueness of manyÐthis seems false to most people; after all, most Spanish citizens do not speak Basque, not even most Basques.4 (9) a. [manyprop x: Speaker-of-Basque(x)] Speaker-of-Basque(x) & Citizen-of-Spain(x) b. [manyprop x: Citizen-of-Spain(x)] Citizen-of-Spain(x) & Speaker-of-Basque(x) Further evidence that focus-a¨ected readings arise only with weak noun phrases comes from scrambling in German. German scrambled positions constitute an anti±DE environment (see Diesing 1992). Unlike scrambled positions, nonscrambled ones pattern with the subject position of stage-level predicates in English in being optional DE environments. As (10) shows, noun phrases in scrambled positions behave as we expect them to. Since they pattern with strong(like) noun phrases, the noun phrases in (10a) do not license focus-a¨ected readings. On the other hand, in (10b) the nonscrambled noun phrases do have such an interpretation (provided they are weak). (10) a. weil viele/einige/wenige SCHWALBEN ja doch because many/some/few swallows prt prt SuÈden ¯iegen south ¯y b. weil ja doch viele/einige/wenige SCHWALBEN because prt prt many/some/few swallows SuÈden ¯iegen south ¯y

in den to the

in den to the

Determiners

129

(10a), where the weak noun phrase has scrambled across two particles, cannot be interpreted in a focus-a¨ected manner as saying that Q things that are ¯ying south are swallows. It only allows for a stronglike reading of the weak noun phrase, along the lines of ``Some event where Q-many things are ¯ying south is an event where Q-many swallows are ¯ying south.'' But in (10b), where the weak noun phrase has not scrambled, both readings are possible. Finally, another respect in which focus-a¨ected weak noun phrases behave like Milsark-style ones involves their scope properties: they do not like to take inverse scope, as is shown by the way (11) is interpreted. (11) Every restaurant owner interviewed many HIGHLY RECOMMENDED chefs. On its focus-a¨ected reading, (11) can be paraphrased as ``Every restaurant owner is such that many chefs that she or he interviewed were highly recommended.'' Under this reading, every must take scope over many. (11) cannot have a focus-a¨ected reading where many takes wide scope, along the lines of ``Many chefs that every restaurant owner interviewed were highly recommended,'' even though this reading would be pragmatically as plausible as the reading it actually has. In sum, we have seen that contrary to standard assumptions, it need not be the case that the surface syntactic complement of a determiner always contributes to its semantic restriction. In some casesÐnamely, when weak noun phrases that contain a focus occur in DE environmentsÐwe ®nd a focus-a¨ected reading, where the restriction contains all the nonfocused material but not the focus inside the noun phrase, which contributes only to the scope. In these cases weak determiners behave analogously to instances of only and even that are adjoined to noun phrases.5 5.3

Focus-A¨ected Readings and Q-Raising

I would like to extend what I said earlier about only and even in adverbial position to focus-a¨ected readings of noun phrases. In particular, let us assume that the noun phrases' determiners do not take scope by (an equivalent of ) QR, which involves moving the entire phrase, but by Qraising, which applies to the quanti®er at hand, in this case a determiner. (In Herburger 1997a, which was concerned only with how focus a¨ects determiners, not with how it a¨ects quanti®ers in general, I called this process D-raising.)

130

Chapter 5

(12) [Di [[ XP ti NP] YP]] After the determiner has Q-raised, it is syntactically unary, resembling an adverbial quanti®er and a raised instance of only and even. It is now ready to receive its quanti®cational structure through focal mapping. The interpretation of an example like (13), for instance, is thus derived as shown in (14) through (16). After Q-raising, focal mapping applies, resulting in the structure in (15), which, abstracting away from the event semantics, translates as (16). (13) [[Few INCOMPETENT cooks] [applied]] [surface syntax] (14) [Few [INCOMPETENT cooks applied]] [Q-raising] (15) [[Few cooks applied] [INCOMPETENT cooks applied]] [ focal mapping] (16) [[few x: Cook(x) & Applied(x)] Cook(x) & Incompetent(x) & Applied(x) Consider next the following pair of sentences, due to Westerstahl (1985): (17) a. Many Scandinavians have won the Nobel Prize in literature. b. Most Scandinavians have won the Nobel Prize in literature. As Westerstahl observes, (17a) can have a reading where the restriction of the determiner D is not its internal argument NP; it can mean ``Many who are winners of the Nobel Prize in literature are Scandinavians.'' When we replace many with most, such a ``switched'' reading is not possible. (17b) can only be interpreted as saying that the majority of Scandinavians are such that they have won the Nobel Prize in literature; it cannot be interpreted as saying that most of those who have won the Nobel Prize in literature are Scandinavians. Following standard views, Westerstahl holds the equivalent of saying that the surface syntactic sister of the determiner D provides its restriction, whereas the external argument provides the scope. In light of the ``switched'' reading of (17a), he can therefore only conclude that many is a nonconservative determiner in this example. The argument is very similar to the one generally given in connection with only that was discussed in chapter 4. By analogy to what has been said about only, many in (17a) need not be seen as ¯ying in the face of Conservativity. This is so because the switched reading arises when Scandinavians is focused.

Determiners

131

(18) a. Many SCANDINAVIANS have won the Nobel Prize in literature. b. Many Scandinavians have won THE NOBEL PRIZE in literature. (18a) contrasts with (18b) in that only (18a) can mean that many who have won the Nobel Prize in literature are Scandinavians; we do not ®nd this reading for (18b), which means that what many Scandinavians have won is the Nobel Prize in literature. This contrast suggests that the switched reading is really an instance of focus-a¨ected quanti®cation. If so, it is conservative, in the same way that the interpretation of only is conservative. As a result of Q-raising and focal mapping, the nonfocused part forms the internal argument of many and the focus (together with the nonfocused part) forms its external argument. The mapping from this to the quanti®cational structure is entirely transparent; in other words, it is conservative. By analyzing the switched reading of (17a) as a focusa¨ected reading, we can not only hold on to the view that many is always conservative, but also predict that for (17b), where the determiner is strong (most), no switched reading is available. 5.4

Unary Weak Determiners and Q-Raising

The account just given of focus-a¨ected readings of noun phrases parallels the earlier analysis of instances of only and even that surface in nonadverbial position and posits that the determiner is syntactically unary and adverblike at one point in the derivation. Why, one may ask next, are focus-a¨ected readings possible only with noun phrases that exhibit the de®niteness e¨ect (DE)? The idea I would like to explore is that all DE noun phrasesÐboth ones that have Milsark-style semantics and focusa¨ected onesÐtake scope through Q-raising. In contrast, strong(like) noun phrases take scope by (an equivalent of ) QR, which involves movement of the entire noun phrase. As a result, DE determiners are unary after they take scope, but strong(like) determiners are binary. Of course, this analysis makes sense only to the extent that weak determiners on the Milsark-style reading can be analyzed as being unary. Among the various semantic characterizations of the DE (see, e.g., Milsark 1977, Barwise and Cooper 1981, Keenan 1987, Diesing 1992), the following one is of particular interest in this respect (cf. Higginbotham 1987, Dobrovie-Sorin 1993, Musan 1995). It starts with the observation that in

132

Chapter 5

Milsark-style readings the determiners are symmetric. In other words, on the relevant, Milsark-style reading, (19a) is equivalent to the Milsark-style reading of (19b). (19) a. Few children are playing in the garden. b. Few that play in the garden are children. A determiner is symmetric when its restriction and scope can be interchanged without a¨ecting truth-conditions, which happens to be the case if the determiner is intersective (e.g., Barwise and Cooper 1981). All the determiner cares about is the intersection of its restriction and its scope; it does not care how this intersection relates to the restriction. Since in symmetric/intersective interpretations the restriction plays no privileged semantic role, nothing prevents us from relieving the NP of its special status as the internal argument of the determiner D and having the determiner take everything in its scope, treating it as unary (or ``absolute'' in Higginbotham's (1987) terms). (20b) is exactly the kind of quanti®cational structure that results from transparently interpreting a Q-raised structure in (20a). (20) a. [Few [children play in the garden]] b. [few x] Children(x) & Play-in-the-Garden(x) Thus, what characterizes DE noun phrases is that they take scope by Qraising, resulting in a structure where the determiner behaves in relevant respects like an adverbial quanti®er and like only and even. What further distinguishes Milsark-style readings from focus-a¨ected ones is that in the latter a focus inside the NP induces focal mapping. If weak determiners are unary on their Milsark-style reading, what does this say about their lexical meaning? It seems that each weak determiner has two meanings, a binary one (for the strong(like) and focus-a¨ected readings) and a unary one (for the Milsark-style reading). In the cases of some and ®ve the di¨erence between the binary version of the weak determiner and the unary version is small. (21) a. [some x: F(x)] G(x) i¨ jF X Gj V 1 [``binary''] b. [some x] G(x) i¨ jGj V 1 [``unary''] (22) a. [®ve x: F(x)] G(x) i¨ jF X Gj ˆ 5 [``binary''] b. [®ve x] G(x) i¨ jGj ˆ 5 [``unary''] Binary ®ve, for instance, requires that the intersections of the sets corresponding to the restriction and the scope contain ®ve elements. Unary ®ve

Determiners

133

requires that the set corresponding to the scope have ®ve members.6 With the determiners few and many, however, the di¨erence between the unary and binary interpretations seems to be bigger. Whereas on both the strong(like) and focus-a¨ected readings few and many are binary and proportional, on the Milsark-style reading they are unary and ``cardinal''; all that seems to matter is that the sole argument of few (many) contain a relatively small (large) number of things. The two readings also di¨er in that the proportional one of course says something about a proportion, whereas the Milsark-style one does not; the latter only says something about a contextually determined number. (23) a. [fewprop x: F(x)] G(x) i¨ jF X Gj U n%jFj b. [fewMilsark-style x] G(x) i¨ jGj U n It may ultimately be possible to treat the Milsark-style reading of few and many as a special case of the proportional one. The idea is that although no lexical material enters into the restriction, few and many in this kind of reading quantify over a proportion of contextually relevant things. Thus, There are few children playing in the garden would mean that few of the contextually relevant things are children who are playing in the garden. Although this kind of analysis would be attractive in assimilating as much as possible the intransitive to the transitive interpretation, it admittedly puts a considerable burden on the context, which may or may not be the right strategy. In what follows I will adopt the ®rst, more conservative version, where Milsark-style few and many are not proportional. 5.5

Unary Weak Determiners Are Not Adjectives

The unary interpretations of few and ®ve are reminiscent of the interpretations of adjectival few and ®ve (e.g., the few problems, the ®ve brothers). Because of this, it might seem more economical to analyze the unary reading as involving adjectival uses of few and many instead of a second, unary quanti®cational reading that coexists with the binary one. Even though the idea is appealing and popular (see Milsark 1977, Partee 1988, Diesing 1992), I think it is problematic and ultimately cannot be maintained (see Herburger 1997a). Since on the adjectival analysis, few, many, and the like, are not quanti®ers on the Milsark-style reading but instead adjectives, the quanti®cational force of the relevant noun phrases must come from elsewhere. To

134

Chapter 5

this end, a process of existential closure for free variables is posited. On this view, (24) is analyzed as in (25), which states that there is a set such that it has few, at most three, or no members; and whatever members it has, they are children in the garden. (24) There are few/at most three/no children in the garden. (25) bX (few/at-most-three/no(X) & Ex (x A X ! (Child(x) & In-the-garden(x)))) The problem now is that the existence of the empty set makes this sentence true; it has few, at most three, or no elements, and it is vacuously true that none of its members is not a child in the garden. Since the empty set automatically makes (24) true on the analysis in (25), (24) would be predicted to be true regardless of how many children are in the garden. This clearly does not accord with our intuitions. Diesing (1992) notes this problem and proposes to remedy it by imposing a maximality condition on X to the e¨ect that it contains all and only the children in the garden (Diesing 1992, n. 9). Thus, (24) is interpreted not as in (25), but as in (26). (26) bX (few/at-most-three/no(X) & Ex (x A X $ (Child(x) & In-the-garden(x)))) (26) states that there is a set that has few (at most three, or no) members and it contains all the children who are in the garden. This seems to capture the meaning of (24) well. But, although the maximality condition works for decreasing determiners, it creates a problem elsewhere, namely, with nondecreasing weak determiners. Even if it is not maximally informative, (27) is true if there are ®ve children in the garden (see note 6). (27) There are three children in the garden. An interpretation that imposes a maximality condition on the set of children in the garden (see (28)) wrongly predicts that (27) should literally be false in such a context because it requires that there be a set with three members and that it contain all and only the children in the garden. Another way of putting it is that (28) assimilates the truth-conditions of (27) to those of There are only three children in the garden. But the two do not mean the same thing. (28) bX (Three(X) & Ex (x A X $ Child(x) & In-the-garden(x)))

Determiners

135

In light of these problems, it seems that the only way we could maintain that the determiners on Milsark-style readings really are adjectives rather than quanti®ers would be to lexically decompose decreasing weak determiners into a negation and an existential cardinality predicate and then assign the negation wide scope over the existential quanti®er that provides the existential closure of free variables. Apart from violating the lexical integrity of the determiners, this kind of analysis would require stipulating that what is in the scope of the negation is false precisely because the cardinality predicate does not hold of the set, and not for any other reason. Another di½culty the existential closure analysis encounters is that of constraining the locus of existential closure. This was not an issue under the classical Milsark-style paradigm. Since DE noun phrases correlated with Milsark-style readings, it was claimed that existential closure necessarily applies to DE contexts (the verb phrase in Diesing's (1992) terms). The existence of focus-a¨ected readings makes this no longer tenable since they, too, appear in DE contexts; but in light of the clearly binary, proportional interpretation of few and many in such readings, they cannot be reduced to adjectival readings. At this point it seems more reasonable to treat one-place weak determiners as unrestricted quanti®ers and not try to assimilate them to adjectives. 5.6

Back to the De®niteness E¨ect: QR versus Q-Raising

Summarizing Milsark's observations concerning the de®niteness e¨ect (DE): certain kinds of predicates allow only for certain subjects. (29) There-clauses DE noun phrases (Milsark-style) Stage-level predicates DE noun phrases (Milsark-style) strong(like) noun phrases Individual-level predicates strong(like) noun phrases One kind of explanation o¨ered for this pattern is that individual-level predicates are in a relevant sense about the individuals that the noun phrase describes; stage-level predicates can, but need not, be about those individuals; and, ®nally, there-clauses cannot be. There are di¨erent versions of this line of thinking. Milsark says, for instance, that individuallevel predicates have to have a noun phrase ``topic'' whereas stage-level

136

Chapter 5

predicates can (but need not) have one, and there-sentences cannot. Ladusaw (1994) spells out in detail what I think is a related intuition by saying that individual-level predications involve categorical judgments, stage-level predications can give rise to categorical or thetic judgments, and there-clauses involve thetic judgments only. Once focus-a¨ected readings are taken into account, the picture in (29) changes, so much so that it is no longer clear that the above line of thinking can really be maintained. The reason is this. Although the Milsark-style reading and the focus-a¨ected reading share the trait that they are not about the individuals described by the surface syntactic complement of D, the NP, they di¨er in another respect. Unlike the Milsark-style reading, the focus-a¨ected reading is about the individuals described by the lexical material; intuitively, it is about the individuals described by the nonfocused material that ends up restricting the determiner. Using the notion ``topic'' in a sense similar to Milsark's, this would suggest that the nonfocused part denotes the topic of the focus-a¨ected reading. It would presumably also mean that a focus-a¨ected reading involves a categorical rather than a thetic judgment. If so, however, we no longer have a correlation between DE environment and lack of topic/ thetic judgment, on the one hand, and anti±DE environments and presence of topic/categorical judgment, on the other. The new picture that emerges is schematized in (30). (30) There-clauses DE noun phrases DE noun phrases Stage-level predicates DE noun phrases DE noun phrases strong(like) noun phrases Individual-level predicates strong(like) noun phrases

Milsark-style focus-a¨ected about nonfocused part Milsark-style focus-a¨ected about nonfocused part about NP complement of D about NP complement of D

How can we make sense of this new pattern? One question concerns what exactly we should make of ``about'' in (30). Rather than thinking of it as involving a categorical judgment, in the present context we can explore a di¨erent possibility. It has been central in the analysis of focus and quanti®cation developed here that the relevant

Determiners

137

portion of the sentence is about the restriction of the quanti®er that takes scope over that portion. We have seen this mostly in the context of (covert) adverbial quanti®cation, but it extends to determiner quanti®cation. Assuming that the part of a sentence over which a nominal quanti®er takes scope is about the elements denoted by the restriction of the quanti®er, let us consider individual-level predications ®rst. As noted earlier, these predications are about the individuals picked out by the NP complement of D (strong(like) reading). Interestingly, they cannot be about the individuals described by the nonfocused restriction of a focus-a¨ected reading. This is signi®cant because it suggests that it is not the semantics of individual-level predicates that forces a strong(like) reading of the subject; after all, an individual is an individual regardless of whether it is described by the NP complement of D or by nonfocused predicates. Rather, the pattern indicates that the subject of an individual-level predicate has to move as a whole for independent reasons. In other words, if the subjects of individual-level predicates must undergo QR and cannot undergo Q-raising, it follows that the subject of an individual-level predicate must be a strong(like) noun phrase and that the predication can only be about the individuals described by the NP that is the sister of D in the surface syntax and not about those described by the restriction of a focusa¨ected quanti®er. In a similar vein, it seems that the fact that a noun phrase has scrambled also precludes its taking scope by Q-raising. This would explain why scrambled noun phrases can only receive strong(like) readings. Why individual-level subjects and scrambled phrases take scope only by QR and not by Q-raising is obviously an important syntactic question, which deserves more discussion than I can give it here. One possibility is that individual-level subjects obligatorily receive a special kind of ``topic'' Case marking (e.g., Japanese wa); for a detailed proposal along these lines see Raposo and Uriagereka 1995. Given the logic of this way of looking at the facts, stage-level predicates di¨er from individual-level predicates in not requiring that their subjects undergo QR; they can undergo QR but they can also Q-raise. Consequently, stage-level subjects can receive all three types of readings, strong(like), Milsark-style, and focus-a¨ected. What about there-sentences? Pushing our line of argumentation further, the reason that strong(like) noun phrases are impossible as the inverted subjects of there-sentences has little to do with the semantics of theresentences. One might think that there-sentences assert the (lack of ) exis-

138

Chapter 5

tence of certain individuals and therefore cannot be about individuals (see, e.g., Ladusaw 1994). Yet, as the existence of focus-a¨ected readings reveals, when there-sentences involve focus-a¨ected readings, such sentences are actually about individuals, namely, those described by the nonfocused part that restricts the determiner D. The conclusion that seems hard to avoid at this point is that strong(like) noun phrases are impossible as the inverted subjects in the there-construction because QR is syntactically barred in there-insertion sentences and only Q-raising is possible. Recapitulating, we have seen that the existence of focus-a¨ected readings shows that the di¨erence between obligatory anti-DE contexts (subjects of individual-level predicates, scrambled positions), optional DE contexts (subjects of stage-level predicates), and obligatory DE contexts (there-sentences) does not primarily have to do with topichood, the distinction between thetic and categorical judgments, or aboutness. Instead, I proposed that anti-DE contexts require scope taking by QR, optional DE contexts allow either QR or Q-raising, and obligatory DE contexts require Q-raising. The semantic/pragmatic di¨erences in aboutness schematized in (30) follow from these syntactic di¨erences. All this having been said, ®nally we are faced with the question of why strong determiners di¨er from weak ones in requiring QR. Put di¨erently, why can weak but not strong determiners be separated from their NP sisters at LF? There may be a number of answers to this question. One possibility that suggests itself in the present context is to assimilate determiners that Q-raise to the instances of only and even that surface in nonadverbial position (e.g., where they surface adjoined to DP). More speci®cally, we can say that whereas strong determiners are always heads of DPs (31a), weak determiners can be heads of DPs (31a), or they can be adjoined to NP (31b) in the same way that only and even can be adjoined to DP (31c). In strong(like) readings of weak noun phrases, the determiner projectsÐresulting in a DP. In contrast, in the other readings of weak noun phrases (Milsark-style, focus-a¨ected), the noun phrase starts out as an NP with an adjoined determiner:7 (31) a.

DP D NP every, most . . . few, some, three . . .

b.

NP D NP few, some, three . . .

c.

DP Adv only, even

DP

Determiners

139

What happens in (31b) and (31c)Ðas opposed to (31a)Ðis that the D/Adv moves on its own. The narrow scope that DE noun phrases take suggests that the movement is rather local, to a position below the scope position of QR-ed DPs. As with Q-raising of only, this could be either adjunction to some XP (e.g., VP, AGRO P) or movement to the speci®er of a relatively low functional projection. Again, in either case the movement will be licit in the sense of Chomsky (1995); even if the movement involves a head adjoining to an XP or a head ®lling the speci®er of an XP, neither in its base position nor in its landing site does the head project further, which means that the movement results in a uniform chain. If such an analysis is on the right track, then the di¨erence between weak and strong determiners is not so much a semantic one; rather, it boils down to whether the determiner necessarily has to head its own DP or whether it can also function as an ``adNP.'' The typology of determiners that emerges can be summarized as follows. Strong determiners behave uniformly. Given that they syntactically always project a DP and consequently only move together with the NP (ˆ QR), they only ever have one kind of interpretation, a binary one where the NP denotes the restriction of D and the sister of the DP its scope. Unlike strong determiners, weak determiners can behave in two ways. They can either project their own DP, in which case they undergo QR as strong noun phrases do, resulting in the strong(like) reading of a weak noun phrase. Alternatively, weak determiners can surface adjoined to NP, much as only and even can adjoin to noun phrases. In this case weak determiners take scope through Q-raising. A Q-raised structure can be interpreted as isÐwith the determiner lacking an internal argument, which results in a Milsark-style reading, where the quanti®er is unary. Or a Q-raised determiner can receive an internal argument through focal mapping, as a result of which the nonfocused material in the domain of D is interpreted as the restriction of a binary reading of D.

Notes

Chapter 1 1. Davidson (1967) does not treat adverbs like quickly as predicates of events. His reason is that (for example) in a situation where Susan crossed the Channel by swimming and where she undertook no other crossing, (i) and (ii) are true simultaneously. Yet (iii) and (iv), where quickly is added to the verbs, need not be true simultaneously; if Susan took eighteen hours, it would be quick for a swimming, but slow for a crossing. (i)

Susan crossed the Channel.

(ii) Susan swam across the Channel. (iii) Susan crossed the Channel quickly. (iv) Susan swam across the Channel quickly. On the operator analysis, this problem does not arise if quickly is considered an intensional operator. Since the possible worlds of crossing and swimming are not the same, the meaning of quickly will come out di¨erently in (iii) and (iv). Parsons (1990) argues that it is indeed possible to give a Davidsonian treatment of attributive operators: attributive adverbials have to be relativized, just like attributive adjectives. Since ``quickly for a crossing'' need not be the same as ``quickly for a swimming'' (it typically is not), (iii) and (iv) come out as having divergent truth-conditions, even on the event analysis, as required. 2. In general, it is quite di½cult to ®nd genuinely optional direct objects. (Stab may have more internal structure than other transitive verbs.) Genuinely optional dative and prepositional arguments are considerably easier to ®nd. In this context consider also Parsons's (1990, 98¨.) example of the ``agentless'' passive of stab, which involves a PP. (i) In a dream last night, I was stabbed, although in fact nobody had stabbed me, and I wasn't stabbed with anything. The argument for separation that is given here is that at least some arguments are genuinely optional. Why direct objects should tend to be less genuinely optional than other arguments is an important but separate issue.

142

Notes

3. Although the essential point, which is what is relevant here, is simpleÐnamely, that what the video games do needs to be distinguished from the rest of the event description and treated as a separate though related eventÐthe actual logical form that achieves the relevant result looks a bit daunting. (i) be ([bX: 3(x) & Ex(Xx ! Video-game(x))] Ez(Agent(e,z) $ Xz) & Teach(e) & [every y: Quarterback(y)] [be 0 : e 0 U e] (Ez(Goal(e 0 ,z) $ z ˆ y) & [bZ: 2(Z) & Ez(Zz ! New-play(z))] Ez(Theme(e 0 ,z) $ Zz)))) (i) translates roughly as ``There is an event such that a group of three things, each of which is a teacher in that event, is such that for every quarterback there is a related event, which is part of the ®rst event and where the quarterback is the goal of that second event and where two new plays are the theme.'' For details and for arguments why no polyadic analysis of verbs will succeed in capturing the relevant reading, see Schein 1993, 57¨. Chapter 2 1. Failure to distinguish focal presuppositions from lexical presuppositions has led some authors to reject any analyses of focus that aim to account for its presuppositional e¨ect (e.g., Rochemont 1986). This criticism seems unfair to me. Even if the complement of know ``presupposes'' truth, it can still be focused. (i) John knows THAT MARY SPEAKS ARABIC. But this is not a problem. It is consistent with the idea that focal presuppositions operate over and above the lexical presuppositions of a verb like know. In particular, the focally nonpresupposed part of (i), the that-clause, can contain ``lexically presupposed'' material without any con¯ict arising. 2. I owe this observation to Paul Portner (personal communication). 3. An element is monotone decreasing and creates a downward-entailing context if it licenses an inference from superset to subset. For example, a quanti®er Q with restriction A and scope B is monotone decreasing in its restriction exactly when [Q A] B entails [Q C] B where C is a subset of A. Thus, no is monotone decreasing in its restriction, as shown by the valid inference from (i) to (ii). (i) No man came. (ii) No tall men came. I will say more on downward monotonicity in chapter 4. 4. This problem was brought to my attention by Sylvain Bromberger (personal communication). 5. One phenomenon I haven't mentioned that quickly comes to mind when discussing aboutness is sentence topics. (i) As for Hector, I don't know if he will be there tomorrow. (ii) Cathy, I haven't met yet. As Vallduvõ (1990) shows, sentence topics (which he calls ``links'') and the nonfocused part of the sentence both encode some aboutness. But in addition to say-

Notes

143

ing what the sentence is about, a sentence topic establishes a link to a previously introduced discourse referent. VallduvõÂ's analysis, which is mostly based on data from Catalan, invokes a separate information-theoretic component to encode this. A related but semantically based analysis of sentence topics that concentrates on wa-topics in Japanese is o¨ered by Portner and Yabushita (1998). 6. The matter is discussed in Ladusaw 1992. In Herburger 1998 I argue that n-words are lexically ambiguous. 7. Except, it seems, in the King's opinion in Alice in Wonderland. (i) ``I see nobody on the road,'' said Alice. ``I only wish I had such eyes,'' the King remarked in a fretful tone. 8. This example is from Bosque 1980, 41. It is not used as a linguistic example there, but is part of the main text. 9. There are particular instances where the focus falls on a postverbal negative quanti®er and where the sentence actually lacks a backgrounded focal entailment because the negative quanti®er is interpreted as having wide scope. These are cases of what I would like to refer to as ``mimicking focus.'' ?

(i) A: A doÂnde vas? ``Where are you going?'' B: Voy exactamente A NINGUNA PARTE. go-I exactly TO NO PLACE `I'm not going anywhere.' B's answer to A's question is understood to be denying the presumption that B is going anywhere. Moreover, the nonfocused part of the question is repeated word by word in the answerÐin a sense, mimicked. (The answer has a decidedly unfriendly edge that would be absent had B said, De hecho, no voy a NINGUNA parte ``Actually, I'm not going anywhere.'') Since such sentences are limited to very speci®c discourse environments, it seems they can safely be set aside as an independent phenomenon. 10. It is sometimes noted that the scope of a quanti®er depends in some way on intonation and that di¨erent stress assignments can a¨ect the scope order between certain quanti®ers. However, it seems di½cult to pin down the correct generalizations. Remaining agnostic about the relevance of focus for scope, I claim here only that a focused quanti®er has to bind into the matrix of the event operator. If the quanti®er is decreasing and it takes scope over be, the sentence lacks a backgrounded focal entailment; conversely, if the event quanti®er takes wide scope (or the quanti®er is nondecreasing), such an entailment resurfaces. The same holds for nonfocused quanti®ers. At this point I have no insights to o¨er about how di¨erent stress patterns might a¨ect quanti®er scope. 11. It may seem that negative descriptions such as ``an event of not visiting'' are vacuous in that they are true of a very large number of things indeed (reading the newspaper, crossing the street, etc.). As the following examples show, however, we are typically not so literal-minded as to take negative descriptions as holding vacuously of a huge number of irrelevant things. Rather, the context of utterance

144

Notes

and the desire to make sense out of an utterance are su½ciently strong to make the descriptions sensible. (i) Once, no one knows when, nobody arrived. (Schein 1993) (ii) Nothing I never said ever got me into any trouble. (attributed to Calvin Coolidge) Taken literally, the number of unknown times when nobody arrived and the number of things that were never said are both immensely large. Still (i) and (ii) can be used sensibly; (i) is not automatically true if a dog barks, nor is (ii) automatically false if a restaurant bill got Coolidge into trouble. Even though, pedantically speaking, the barking of a dog is an event of not arriving, and a restaurant bill is something that never gets said, they do not count. 12. Thanks to Anna Szabolcsi, Ricardo Echepare, Gorka Elordieta, and Itziar Laka for both data and judgments. 13. It is worth noting in this context that the focus on ANYTHING in (42B) is contrastive in that it contrasts with Montmartre. Also consider (i). (i) Sascha didn't visit JEAN/HIM yesterday. He didn't visit ANYONE. In (i) the contrasting foci are not of the same semantic type; one is a quanti®er (anyone), the other a referring expression (Jean, him). Since the present account makes no claim concerning the contrastiveness of focus (see the appendix to chapter 2), it does not face any particular problem with the contrast between Jean/him and anyone. On the other hand, the type mismatch is unexpected on Jackendo¨ 's analysis and on the alternative semantics analysis, which, as we have seen, take focus to signal contrast among elements of matching semantic type. 14. I have no account of why stress on the negation results in the unstructured wide reading. I suspect that it may be a strategy to mark focus on everything except the negation. Although it seems ironic that if we want to focus everything but the negation we should stress the negation, it is not easy to think of more obvious alternatives. 15. As noted in connection with the ®rst example of a structured wide reading, (42), here too we can observe a mismatch in semantic type between contrasting foci, bald and anything. This is not a problem, though, for the reasons given earlier. 16. When John in (i) is focused, it cannot be interpreted as coreferential with the preceding pronoun he. If John is assumed to move, then the unacceptability of this interpretation can be assimilated to other crossover constellations. (i) *The woman hei loved betrayed JOHNi . Adopting a movement analysis is not the only way to account for (i), however. For alternative proposals see Rooth 1985, Rochemont 1986, Ferro 1994. 17. It is not problematic to have an entire sentence that is focused, even when some material it contains counts as ``familiar'' because its referent was introduced in the previous discourse. The distinction between ``familiar'' material (e.g., def-

Notes

145

inite description) and ``novel'' material (e.g., inde®nite descriptions) is very different from the distinction between being backgrounded and being focused. Although ``novel'' material is often focused, this is by no means necessary; see (iB). Nor is it required that familiar material not be focused, as the familiar ROSALIA wrote a poem shows. (i) A: How did you solve the problem? B: I sent a message TO THE ORGANIZERS. 18. In its present form structured Davidsonian decomposition does not account for focus in topicalized phrases, which, as just noted, literally fall outside of its scope. It is an interesting question, one that I need to leave for future research, whether the present analysis can be extended to account for focus in sentence topics by assuming that topicalization involves some covert quanti®cation. 19. Perhaps the mismatch between QR and overt movement is not as big as I have indicated. According to a line of research exempli®ed in Szabolcsi 1997, Hungarian has overt movement that mirrors a plausible hierarchy of functional projections that in a language like English provide the landing sites for covertly moved quanti®ers. 20. Type-shifting a generalized quanti®er (a set of sets) into a (plural) individual is a somewhat complex two-step procedure. First, all the sets that are contained in the set of sets that is denoted by the quanti®er are intersected. The members of this intersection are then taken together to form a ``plural individual.'' The intersection operation works for certain quanti®ers (principal ®lters), but not for others. For (e.g.) every man it works: The extension of every man contains every set that forms the extension of the predicates that every man combines with truthfully, namely, x is mortal, x is not a woman, and so on. Intersecting these sets gives a set containing all the men. From this a plural individual is formed, which can then be type-shifted into a predicate. The analysis works analogously for quanti®ers like exactly twelve apostles, the children. However, many focusable quanti®ers cannot be type-shifted in this way: for example, some men, most men, many men, no men. Since the extension of some men, for example, contains both men who are short and men who are tall, its intersection will be empty. No plural individual can be formed from its members, since there are none. For the other quanti®ers the intersection operation is not well formed either. This is easy to see in the case of no man. No man truthfully combines with x is a woman and with x lacks DNA. Clearly, though, the intersection between the set of women and the set of things lacking DNA is empty, so no plural individual can be formed out of no man. For a detailed demonstration that the same holds of proportional quanti®ers like most and many, see Ogihara 1987. 21. If I were to extend the present analysis to cleftsÐsomething I will not attempt hereÐI would say that the clefted element in a pseudocleft like (79) represents the main predicate (in the relevant sense) and that the nonfocused part represents the subject of that predicate. An analysis of clefts along these lines is suggested in passing in Higginbotham 1987.

146

Notes

Chapter 3 1. It might appear that determiners quantify over (regular) individuals and adverbs quantify over events (sort of special individuals in the ontology). But in one direction this generalization does not hold. As (i) shows, a determiner like most quanti®es not only over regular individuals, but also over events, as long as the event is denoted by its sister NP. (i) Most destructions of cities during wartimes could have been avoided through skilled diplomacy. The de®ning property of determiners then is a syntactic one, namely, that they take nominal complements. Whereas determiners can quantify both over regular individuals and over events, depending on the noun, adverbial quanti®ersÐbecause they are adverbsÐalways seem to quantify over events because verbs are always predicates of events and never predicates of regular individuals. 2. Unlike the examples discussed so far, (10) does not contain adverbial quanti®ers in the strict sense. This does not matter since in relevant respects only behaves like one, (except that it can't associate with a subject, see p. 43). 3. There is a technical issue here, relating to the question of how the variable replacing the focus is bound (see chapter 2). Assuming that adverbs take sentential scope, they are held to quantify over propositions (sets of intervals in Rooth's (1985) view, sets of situation for von Fintel (1994); see below). Yet the focussemantic value of a sentence is not a proposition but a set of propositions. To remedy this mismatch, it is proposed that it is not the focus-semantic value that de facto restricts the adverb but its union, which amounts to the existential closure of the analysis replacing the focus (see section 2.10). 4. This issue is also discussed in more detail in Rooth 1996. Unfortunately I became aware of this paper too late to incorporate a discussion of it here. 5. Rooth (1995b) uses the following example to argue that second occurrence focus does not always require word-by-word repetition of the preceding utterance: (i) The dean isn't taking any candidates except for Susan, Madilyn, and Harold seriously. So what. Even JOHN is only considering younger candidates. Claiming that younger has the phonetic properties of a second occurrence focus, he says this example shows that second occurrence focus does not require repeating part of a previous utterance, but can be licensed by ``implicational bridging''; the hearer is invited to infer that Susan, Madilyn, and Harold are younger candidates. It is not entirely clear what the facts are. Rooth mentions that Krifka (1995) thinks that examples such as (i) involve a secondary stress on the alleged second occurrence focus. If experiments con®rm that younger is phonetically like other second occurrence foci, then this would seem problematic for the present account, but it would not necessarily be devastating; it could be taken to show that second occurrence focus involves mimicking, where what is mimicked need not always be a previous utterance, but may be an inference licensed by a previous utterance. 6. This example has the structure OP1 FOC1 , OP2 FOC2 . What happens in a slightly di¨erent constellation, one where both foci occur within the scope of the

Notes

147

lower operator, as in Rooth's original example, where there is a nesting pattern OP2 OP1 FOC1 FOC2 ? Similarly, what happens if we have a crossing pattern? Although the judgments are subtle, it seems that a similar restriction holds in these cases as well. In particular, it seems that an inherited reading is clearly best when the second sentence mimics the ®rst word by word. (i)

They only introduced AFRICAN-AMERICAN job candidates to John.

(ii) a.

So what? They also only TO MARY. b. ??So what? They also only candidates TO MARY. c. ??So what? They also only d. ?So what? They also only MARY.

introduced African-American job candidates introduced AFRICAN-AMERICAN job introduced Black job candidates TO MARY. introduced BLACK job candidates TO

(iii) They only introduced JOHN to African-American job candidates. (iv) a.

So what? They also only introduced candidates. b. ??So what? They also only introduced AMERICAN job candidates. c. ??So what? They also only introduced d. ?So what? They also only introduced candidates.

MARY to African-American job MARY to AFRICANMARY to Black job candidates. MARY to BLACK job

7. Having already encountered two e¨ects of mimickingÐsecond occurrence focus and the power of mimicking to circumvent the scope rigidity of postverbal negative quanti®ers in Spanish (see chapter 2)Ðwe might look for other e¨ects as well. A possible candidate involves positive polarity items (some, already, sometimes, etc.) in the scope of not; Horn (1989) adopts the view that such items are acceptable ``when they represent, word by word, an emphatic denial of a preceding speaker's assertion'' (p. 397). Typically, these denials are cases of what I have called wide readings (either structured or unstructured); see chapter 2. (Horn analyzes them as involving a special metalinguistic use of negation.) Here are some examples he cites, which I annotate for a likely focus: (i)

A: The Sox have already clinched the pennant. B: The Sox HAVEN'T already clinched the pennant.

(ii) Chlamydia is not ``SOMETIMES'' misdiagnosed, it is FREQUENTLY misdiagnosed. (iii) You ate some mushrooms.ÐI did NOT ``eat some mushrooms.'' In all these examples a positive polarity item appears in the scope of a negation, and it does so because the utterance in which it appears mimics a preceding one. Further investigation is beyond the scope of this monograph, but it seems possible to analyze many of Horn's examples of metalinguistic negation as involving the mimicking that is typical of wide readings of negation. 8. In studying this issue further, it may be interesting to take into account the data in (i) and (ii). Although I share Johnston's intuition concerning (30) and (31), I

148

Notes

think that (i) and (ii), which parallel (30) insofar as they have a postverbal whenclause that contains a focus and an atelic predicate in the main clause, actually may have an interpretation where the matrix contributes to the restriction of the adverb and not to the scope. (i) It is striking that Joe is always at the cafe when BETH is waiting on the tables. (ii) The study claimed that pregnant women usually feel nauseous when THEIR BLOOD SUGAR IS LOW. (i) can be interpreted as saying that it is striking that all events where Joe is at the cafe when the tables are being waited on are events where Joe is at the cafe when Beth is waiting on the tables. Similarly, (ii) can mean that the study claimed that most instances in which pregnant women feel nauseous are instances in which they feel nauseous and their blood sugar is low. 9. Rooth points out the following as one of the major problems. Inde®nites are claimed to be novel in that an inde®nite introduces a new discourse referent. This is the respect in which they di¨er from de®nites (Heim 1982). Yet, because of the nonlocal nature of focus, in the full-¯edged tripartite structure all the nonfocused material that e¨ectively forms part of the restrictive clause also has to appear in the nuclear scope. Now, if the nonfocused part contains an inde®nite, this inde®nite appears twice: once in the restriction, and once in the scope. To ensure that the same referent is picked out in both cases, which is clearly what is needed, somehow the novelty requirement on the second occurrence has to be blocked. 10. Interestingly, (i) is strongly degraded (at least in my judgment; cf. BuÈring 1994, however). È LTLICH waren (i) *?weil selten Bananen ERHA It seems to be generally the case that when focus is assigned VP-internally in German, there cannot be any phonologically realized nonfocused material between the adverb and the focus. 11. Not only are the two analyses similar in their treatment of adverbial quanti®ers, they also both employ an E-type analysis of donkey anaphora in order to overcome the overly strong uniqueness claims of the traditional E-type account. An event-based analysis is proposed by Ludlow (1994), a situation-based one by Heim (1990) and von Fintel (1994). Chapter 4 1. In certain contexts (e.g., signs) only can also appear in postposition: Left Turn Only. I will leave this interesting syntactic aspect undiscussed. 2. At least, this is what I think is going on in (7). Examples like this were brought to my attention by Ernie Lepore (personal communication). Another special use of only may be its temporal use as ``not until,'' ``no sooner than'' (cf. German erst, Spanish recieÂn). (i) They realized the problem only this morning, when it was too late.

Notes

149

3. Personally I ®nd this reading less preferred than the second one. This was also reported by the native speakers I consulted. The interpretation seems easiest when (11a) is read as an exclamative and/or there is a pause before the DP. 4. One instance where an adverbial quanti®er may not appear to take its surface scope involves modals. Thus, on the salient reading of (i), never outscopes may. (i) Students may never attend faculty meetings. This is reminiscent of the inverse scope that modals generally take with negation (must not being an exception). Consider for example (ii), which negates the possibility, necessity, and so on, of Mary's stopping with her project. (ii) Mary {cannot, may not, should not, will not} stop with her project now. 5. The material in this section was presented at the 1999 meeting of the Linguistic Society of America (Herburger 1999). 6. Recall from chapter 2 that the nonfocused part has to be part of the scope because of sentences where the relevant focus-a¨ected quanti®er is in a higher clause and therefore does not directly bind into the conjunct interpreting the focus. (i) John only said that MARY passed the exam. Here only associates with Mary (contra Drubig 1994), resulting in the interpretation that no one but Mary was such that John said that she passed the exam. To derive this interpretation, it is essential that the scope of only contain not only the focus but also the nonfocused material, as in ``All instances where John said that the exam was passed were instances where he said that Mary passed the exam.'' 7. In sorting out these issues I was greatly helped by discussion with Paul Portner. 8. As mentioned earlier, Atlas also argues that the nonfocused part is what the sentence is about, whereas the focused part marks the assertion. Distinguishing between assertion and entailment, Atlas states: From the point of view of a theory of speech-acts, in asserting Only a is F, we do not thereby assert a is F, the way we would if the statement were to consist of a conjunction a is F & F. Rather, what we do assert entails a is F, but it does not ``say'' it. (Atlas 1993, 306, citing Atlas 1991, 139)

He further appeals to the distinction between assertion and entailment, saying that it shows how the positive proposition (here, someone trusts Bill) can fall outside the scope of the assertion (here, if, discovered, too bad, etc.) and yet be entailed, and so be a truth-conditional aspect of meaning (Atlas 1993, 139). 9. Horn also cites an intuitive di¨erence between (i) and (ii) as providing an argument in favor of an asymmetrical analysis and against a symmetrical one. (i) Only Muriel voted for Hubert. (ii) Muriel and only Muriel voted for Hubert. Unlike the negative polarity item facts (which Atlas denies) and the inversion facts (which he ignores), this is an argument that Atlas (1991, 1993) addresses. He argues that although the sentences are truth-conditionally equivalent, they receive di¨erent logical forms, which capture their di¨erence in ``assertion structure.'' 10. I am grateful to Paul Pietroski for helpful discussion on these issues.

150

Notes

11. If every indeed has existential force, as I have argued, then this fact has interesting consequences elsewhere in the analysis of quanti®ers. For one thing, it assimilates noun phrases with every to noun phrases with the by making them proper principal ®lters. A principal ®lter is a restricted quanti®er such that when all the sets (classes) that correspond to the scope of the quanti®ers in various true sentences are intersected, that intersection corresponds exactly to the set (class) denoted by the restriction. Every woman and the woman are principal ®lters, no woman and some woman are not. A principal ®lter is proper if the restriction cannot be empty. Thus, on the standard logic view, the NP counts as a proper principal ®lter, but every NP does not. Barwise and Cooper (1981) use the notion of proper principal ®lter to explain why de®nite descriptions, but no other restricted quanti®ers, can occur in partitives like many of the books. If every has existential force, then it is like the in that a noun phrase that it introduces is not just a principal ®lter, but a proper principal ®lter. If this is correct, the account of the Partitive Constraint has to be rethought. Another aspect of Barwise and Cooper's (1981) classi®cation of quanti®ers that is a¨ected by the notion that every has existential force is their way of distinguishing between strong and weak noun phrases. They maintain that a noun phrase X is strong if its value is either always true or always false when substituted into the schema in (i). Weak noun phrases, on the other hand, show varying results, depending on how things are in the world. (i) X is an X. Every NP is a paradigmatic strong noun phrase, as we will see in chapter 5. And adopting the standard logic view whereby every has no existential force (and is not presuppositional either), every NP indeed gives a noncontingent result when substituted into (i). On this view, (ii) is true in virtue of the fact that there are no unicorns in the real world; that is, it is vacuously true. (ii) Every unicorn is a unicorn. This is not so on the view adopted here (or on the one that Heim and Kratzer (1998) argue for). The truth of (ii) is contingent; it depends on whether the sentence is understood to be embedded under ``in the realm of the mythology that claims the existence of unicorns'' or not. If it is uttered relative to the real world, it is not true; rather, it is false (or truth-valueless on Heim and Kratzer's view). Di¨erent ways of distinguishing weak from strong noun phrases are explored in chapter 5. 12. As argued above, when (50b) is used in a context where there were no students from Liechtenstein, the sentence is not only not true (where a sentence can be not true because it is either truth-valueless or false), it is in fact false. 13. There seems to be some di¨erence between the examples with only and every, as pointed out by Paul Portner (personal communication). When the context does not entail that there is something that satis®es the quanti®er's restriction, it seems easier to accommodate the existence of such a class with every than with only. A possible explanation is that it is rather straightforward to determine what the

Notes

151

restriction of every is, for the surface syntax determines it to a large extent (see chapter 5). For only the determination is less direct, since this is a matter that is mediated by focus and the covert syntax. 14. Horn gives a number of examples to show that only indeed licenses NPIs in the nonfocused part. (i)

(Of all her friends,) Only PHIL would lift a ®nger to help Lucy.

(ii) Only YOUR WIFE gives a hoot about what happens to you. (iii) There was only ever ONE GLOVE there that I saw. (iv) I only eat any meat WHEN I'M DEPRESSED. There are what look like curious gaps in the paradigm, some of which Atlas (1993) has used to argue that only, which he considers to be nonmonotonic, never licenses NPIs. This seems an odd conclusion to draw in light of the data in (i)±(iv). 15. One may wonder about plural de®nite descriptions, which I have excluded from the discussion in the text. The in plural de®nite descriptions is not weakly monotonic. Although this is not obvious with distributive predicates, it becomes clear when we consider predicates that combine with essential plurals (see Schein 1993). For instance, (i) does not entail (ii). (i) The children formed a circle around the tree. (ii) The boys formed a circle around the tree. 16. Although every shares many properties with only, it does not trigger negative inversion. (i) *On every Sunday did they play cards. The reason has to do with the quanti®cational structure of every, which is very di¨erent from that of only. In particular, the IP in (i) must contribute to the scope of every rather than to its restriction degree and hence cannot provide a downward-entailing environment. The quanti®cational structure of every will be discussed in more detail in chapter 5. 17. In connection with the claim that adverbial even cannot take scope outside its clause, it should be noted that Wilkinson (1996) discusses an example that seems to suggest the opposite. First, she notes that a continuation of (i) is pragmatically best if it includes tried in the scope of even. (i) Everyone was trying to nominate somebody. (The rules of protocol only allow one nomination per person.) Bill tried to nominate Hillary, Warren tried to nominate Jim, and . . . The crucial claim, then, is that (ii) is felicitous in such a context, which would suggest that even takes scope over the matrix predicate even though it surfaces in the embedding. (ii) Someone tried to even nominate HIMSELF. Although I believe the judgments are subtle, the speakers I have consulted actually prefer (iii) over (ii) in the above context. (iii) Someone even tried to nominate HIMSELF.

152

Notes

This would not be surprising if adverbial even is clause bound, after all, as suggested in the text. Unlike in (ii), in (iii) even c-commands tried and can thus include it in its scope. (iii) means that nominating himself was the least likely thing for somebody to try, but he did it (see below). This ®ts pragmatically very well with (i). I suppose the reason that (ii) also seems acceptable, but less so, is that it means that someone tried to nominate the least likely personÐhimself; here tried is not interpreted in the scope of even. This also ®ts into the context in (i), but not as well as the interpretation of (iii). In light of these observations, I would like to conclude tentatively that adverbial even does pattern with adverbial only in not being able to take scope outside its clause. 18. This is something that Wilkinson does not observe, but that is pointed out by Rullmann (1997). 19. Rullmann hints at this possibility when he says the following (1997, 59): If instead we assume that the relevant ordering is one of pragmatic entailment (Fauconnier, 1975; Kay, 1990), we may not even need a special condition on the use of even to derive the existential presupposition, because by the very nature of the scalar presupposition the asserted proposition will always pragmatically entail at least one alternative proposition. I will leave these issues aside however, and continue to assume Karttunen and Peters' (1979) version of the scalar presupposition.

If I understand correctly, in addressing this issue Rullmann (1997) suggests that even has a conventional, non-truth-conditional part of its meaning such that a speaker can use even only if he intends the hearer to draw a scalar inferenceÐfor example, to infer from the fact that Claire was the least likely to laugh, that someone else also laughed. This sounds very similar to saying that even has an existential implicature; if so, we would be back where we started. Chapter 5 1. It may appear that the lack of a 100% interpretation of few is due to the Gricean maxim of quantity (``Say as much as you can''), which would suggest that there is in fact no ambiguity between proportional and nonproportional few, few being always proportional (see Musan 1995). On this view Few children like spinach would be assimilated to Most children like spinach, which we know is not normally uttered in a context where all children like spinach because such an utterance would violate the maxim of quantity. Taken literally, however, it still would be true in such a situation, even though uninformative. Notice, however, that if Few children like spinach were argued to lack a 100% interpretation for Gricean reasons, one might wonder why this is not also the case for a sentence like Few children are playing. Although one might be able to come up with an answer for this, it would be di½cult to explain under this view why few and most in the following context do not pattern alike: (i) Most children like spinach, in fact all do. (ii) aFew children like spinach, in fact all do.

Notes

153

By adding in fact all. . . , we can defeat the maxim-of-quantity implicature of most. But, as the incoherence of (ii) shows, with few this is not possible. This strongly suggests that it is a matter of semantics and not of Gricean implicatures that few in Few children like spinach does not allow 100% of the children to be picked out. 2. The existence of focus-a¨ected readings is also pointed out in Geilfuû 1993. It is not observed there, however, that they are limited to DE environments. 3. (i), due to David Pesetsky (personal communication), may be even a more illustrative example. (i) Few SALVADOREANS speak Spanish. (i), with an individual-level predicate, clearly does not have a focus-a¨ected reading, which would be ``Few of those who speak Spanish are Salvadoreans.'' This is true even though this reading would be more consonant with our knowledge of the world, and thus favored if at all possible, than the strong (like) reading that (i) actually has: ``A small percentage of the Salvadoreans speak Spanish.'' 4. It may appear that in (8) the focused predicate (the relative clause) is not inside the NP in the overt syntax but instead is its sister. This could suggest that the relative clause is the scope not because it is focused, but because of its syntactic position. Notice, however, that focus-a¨ected readings arise equally in (i) and (ii), where the focused predicate is unequivocally the NP in the overt syntax. (i) There are many SPANISH Basque speakers. (ii) There are many BASQUE-SPEAKING Spaniards. 5. It may be that there are other factors besides focus that can cause a reordering of quanti®cational structure. This is argued by de Hoop and Sola (1996) and has also been suggested to me by Richard Larson (personal communication). In this context, consider also the discussion in chapter 3 of the e¨ects of telicity that Johnston (1994) observes in connection with adverbial quanti®ers. Clearly, taking into account whether and how phenomena other than focus a¨ect quanti®cational structure is very important to reach a complete understanding of how quanti®cation works. My concern here is limited to the e¨ect of focus, which, even if it turns out to not be the only factor, is certainly a major one. 6. (i) is judged true in a context where seven children are playing in the garden. (i) There are ®ve children in the garden. One might take this to suggest that ®ve means ``at least ®ve.'' This would be problematic, however. It would suggest that only ®ve and exactly ®ve mean ``only at least ®ve'' and ``exactly at least ®ve,'' which they clearly do not. Why then is (i) true in the context described above? It is true because of how the world is. Every event where seven children are in the garden contains a subevent where ®ve children are in the garden. It is because of this, and not because ®ve means ``at least ®ve,'' that (i) is judged true in our scenario. Five means ``®ve''; see (22). By similar reasoning, we cannot utter (ii) truthfully if pulling the train up the slope actually took two locomotives.

154

Notes

(ii) One locomotive pulled the train up the slope. Whereas an event where seven children are playing contains a subevent where ®ve children are playing, an event where two locomotives pull a train up a slope does not contain a subevent where one locomotive does the job. 7. This kind of analysis has been inspired by conversations with Norbert Hornstein and Juan Uriagereka, who are exploring closely related issues (see Hornstein and Uriagereka 1999).

References

Atlas, Jay David. 1991. Topic/comment, presupposition, logical form and focus stress implicatures: The case of focal particles only and also. Journal of Semantics 8, 127±147. Atlas, Jay David. 1993. The importance of being ``Only'': Testing the neo-Gricean versus neo-entailment paradigms. Journal of Semantics 10, 301±318. Barker, Laura, and Elena Herburger. 1999. Even even can be an NPI: Supporting evidence from Spanish. Ms., Georgetown University, Washington, D.C. Barwise, Jon, and Robin Cooper. 1981. Generalized quanti®ers and natural language. Linguistics and Philosophy 4, 159±219. BaÈuerle, Rainer, and Urs Egli. 1985. Anapher, Nominalphrase und EselssaÈtze. Papier 105 des Sonderforschungsbereichs 99, University of Konstanz. Berman, Stephen. 1987. Situation-based semantics for adverbs of quanti®cation. In University of Massachusetts Occasional Papers 12, edited by Juliette Blevins and Anne Vainikka. GLSA, University of Massachusetts, Amherst. Boer, Steven, and William Lycan. 1976. The myth of semantic presupposition. Indiana University Linguistics Club, Bloomington. Bonomi, Andrea, and Paolo Casalegno. 1993. Only: Association with focus in event semantics. Natural Language Semantics 2, 1±45. Bosque, Ignacio. 1980. Sobre la negacioÂn. Madrid: CaÂtedra. BuÈring, Daniel. 1994. The interaction of focus, phrase structure, and quanti®cation. In SCIL VI, edited by Chris Giordano and Daniel Ardron. MITWPL, Department of Linguistics and Philosophy, MIT, Cambridge, Mass. Carlson, Lauri. 1984. Focus and dialogue games: A game-theoretical approach to the interpretation of intonational focusing. In Cognitive constraints on communication, edited by Lucia Vaina and Jaakko Hintikka. Dordrecht: Kluwer. Chomsky, Noam. 1971. Deep structure, surface structure, and semantic interpretation. In Semantics: An interdisciplinary reader in philosophy, linguistics and psychology, edited by Danny D. Steinberg and Leon A. Jakobovits. Cambridge: Cambridge University Press. Chomsky, Noam. 1976. Conditions on rules of grammar. Linguistic Analysis 2, 303±351.

156

References

Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24, 239±298. Davidson, Donald. 1967. The logical form of action sentences. In The logic of decision and action, edited by Nicholas Rescher. Pittsburgh, Pa.: University of Pittsburgh Press. de Hoop, Helen, and Jaume SolaÂ. 1996. Determiners, context sets, and focus. In Proceedings of the Fourteenth West Coast Conference on Formal Linguistics, edited by Jose Camacho, Lina Choueiri, and Maki Watanabe. Stanford, Calif.: CSLI Publications. [Distributed by Cambridge University Press.] de Mey, Sjaak. 1991. ``Only'' as a determiner and a generalized quanti®er. Journal of Semantics 8, 91±106. de Swart, Henriette. 1991. Adverbs of quanti®cation: A generalized quanti®er approach. Doctoral dissertation, Rijksuniversiteit Groningen. [Published 1993, New York: Garland.] Diesing, Molly. 1992. Inde®nites. Cambridge, Mass.: MIT Press. Dobrovie-Sorin, Carmen. 1993. Referentiality, scope and the ECP. Ms., CNRS, Paris. Dowty, David. 1989. On the semantic content of the notion ``thematic role.'' In Properties, types and meaning. Vol. 2, Semantic issues, edited by Gennaro Chierchia, Barbara Partee, and Raymond Turner. Dordrecht: Kluwer. Drubig, Hans Bernhard. 1994. Island constraints and the syntactic nature of focus and association with focus. Arbeitsbericht Nr. 51, Sonderforschungsbereich 340, University of TuÈbingen. Dryer, Matthew. 1994. The pragmatics of focus-association with only. Handout from the 1994 annual meeting of the Linguistic Society of America, Boston. Eckardt, Regine. 1998. Review of ``In the event of focus.'' GLOT International 3:9/10, 13±15. Fauconnier, Gilles. 1975. Pragmatic scales and logical structure. Linguistic Inquiry 6, 353±375. Ferro, Lisa. 1994. Interpreting coreference in sentences with focus: A discoursebased alternative to weak crossover. Ms., University of Connecticut, Storrs. Fintel, Kai von. 1994. restrictions on quanti®er domains. Doctoral dissertation, University of Massachusetts, Amherst. Fintel, Kai von. 1997. Bare plurals, bare conditionals, and only. Journal of Semantics 14, 1±56. Geach, Peter Thomas. 1962. Reference and generality. Ithaca, N.Y.: Cornell University Press. Geilfuû, Jochen. 1993. Nominal quanti®ers and association with focus. In Proceedings of the Semantic and Syntactic Analysis of Focus, edited by Peter Ackema and Maaike Schoorlemmer. (OTW Working Papers.) Research Institute for Language and Speech, Utrecht University.

References

157

HajicÏovaÂ, Eva. 1984. Presupposition and allegation revisited. Journal of Pragmatics 8, 155±167. Heim, Irene. 1982. The semantics of de®nite and inde®nite noun phrases. Doctoral dissertation, University of Massachusetts, Amherst. Heim, Irene. 1990. E-type pronouns and donkey anaphora. Linguistics and Philosophy 13, 137±177. Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Malden, Mass.: Blackwell. Herburger, Elena. 1997a. Focus and weak noun phrases. Natural Language Semantics 5, 53±78. Herburger, Elena. 1997b. In the event of focus. Doctoral dissertation, University of Southern California, Los Angeles. Herburger, Elena. 1998. Spanish n-words: Ambivalent behavior or ambivalent nature? In The interpretive tract, edited by Uli Sauerland and Orin Percus. (MIT Working Papers in Linguistics 25.) MITWPL, Department of Linguistics and Philosophy, MIT, Cambridge, Mass. Herburger, Elena. 1999. Only, conservativity, and the syntax semantics interface. Paper presented at the 1999 annual meeting of the Linguistic Society of America, Los Angeles. Higginbotham, James. 1987. Inde®niteness and predication. In The representation of (in)de®niteness, edited by Eric Reuland and Alice ter Meulen. Cambridge, Mass.: MIT Press. Horn, Laurence. 1989. A natural history of negation. Chicago: University of Chicago Press. Horn, Laurence. 1996. Exclusive company: Only and the dynamics of vertical inference. Journal of Semantics 13, 1±40. Hornstein, Norbert. 1995. Logical Form. Oxford: Blackwell. Hornstein, Norbert, and Juan Uriagereka. 1999. Labels and projections: A note on the syntax of quanti®ers. Ms., University of Maryland, College Park. Jackendo¨, Ray. 1972. Semantic interpretation in generative grammar. Cambridge, Mass.: MIT Press. Jacobs, Joachim. 1991. Negation. In Semantics: An international handbook of contemporary research, edited by Arnim von Stechow and Dieter Wunderlich. Berlin: Walter de Gruyter. Johnston, Michael. 1994. The syntax and semantics of adverbial adjuncts. Doctoral dissertation, University of California, Santa Cruz. Kadmon, Nirit. 1987. On unique and non-unique reference and asymmetric quanti®cation. Doctoral dissertation, University of Massachusetts, Amherst. Karttunen, Lauri, and Stanley Peters. 1979. Conventional implicature. In Syntax and semantics 11: Presuppositions, edited by Choon-kyu Oh and David A. Dinneen. New York: Academic Press.

158

References

Katz, Jerrold. 1973. On de®ning ``presupposition.'' Linguistic Inquiry 4, 256±260. Kay, Paul. 1990. Even. Linguistics and Philosophy 13, 59±111. Keenan, Edward. 1987. A semantic de®nition of ``inde®nite NP.'' In The representation of (in)de®niteness, edited by Eric Reuland and Alice ter Meulen. Cambridge, Mass.: MIT Press. Krifka, Manfred. 1992. A framework for focus-sensitive quanti®cation. In Proceedings from SALT 2, edited by Chris Barker and David Dowty. Department of Linguistics, The Ohio State University, Columbus. Krifka, Manfred. 1995. Focus and/or context: A second look at second occurrence expressions. Ms., University of Texas, Austin. Ladd, D. Robert. 1980. The structure of intonational meaning. Bloomington: Indiana University Press. Ladusaw, William. 1992. Expressing negation. In Proceedings from SALT 2, edited by Chris Barker and David Dowty. Department of Linguistics, The Ohio State University, Columbus. Ladusaw, William. 1994. Thetic vs. categorical, stage and individual, weak and strong. In Proceedings from SALT 4, edited by Mandy Harvey and Lynn Santelmann. CLC Publications, Cornell University, Ithaca, N.Y. Laka, Itziar. 1990. Negation in syntax. Doctoral dissertation, MIT, Cambridge, Mass. [Published 1995, New York: Garland.] Larson, Richard, and Clair Lefebvre. 1991. Predicate clefting in Haitian Creole. In NELS 21, edited by Tim Sherer. GLSA, University of Massachusetts, Amherst. Larson, Richard, and Gabriel Segal. 1995. Knowledge of meaning. Cambridge, Mass.: MIT Press. Lewis, David. 1975. Adverbs of quanti®cation. In Formal semantics, edited by Edward Keenan. Cambridge: Cambridge University Press. LoÈbner, Sebastian. 1990. Wahr neben Falsch. TuÈbingen: Max Niemeyer Verlag. Ludlow, Peter. 1994. Conditionals, events, and unbound pronouns. Lingua e Stile 29, 165±183. May, Robert. 1985. Logical Form: Its structure and derivation. Cambridge, Mass.: MIT Press. Milsark, Gary. 1974. Existential sentences in English. Doctoral dissertation, MIT, Cambridge, Mass. [Published 1975, New Yrok: Garland.] Milsark, Gary. 1977. Toward an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 3, 1±29. Moravcsik, Julius. 1991. ``All A's are B's'': Form and content. Journal of Pragmatics 16, 427±441. Musan, Renate. 1995. On the temporal interpretation of noun phrases. Doctoral dissertation, MIT, Cambridge, Mass. [Published 1997, New York: Garland.] Neale, Stephen. 1990. Descriptions. Cambridge, Mass.: MIT Press.

References

159

Ogihara, Toshiyuki. 1987. ``Obligatory focus'' in Japanese and type-shifting principles. In Proceedings of the Sixth West Coast Conference of Formal Linguistics, edited by Megan Crowhurst. Stanford, Calif.: CSLI Publications. [Distributed by Cambridge University Press.] Parsons, Terence. 1990. Events in the semantics of English: A study in subatomic semantics. Cambridge, Mass.: MIT Press. Partee, Barbara. 1986. Ambiguous pseudoclefts with unambiguous be. In Proceedings of NELS 16, edited by Stephen Berman, Joe-Woong Choe, and Joyce McDonough. GLSA, University of Massachusetts, Amherst. Partee, Barbara. 1988. Many quanti®ers. In ESCOL '88: Proceedings of the Fifth Eastern States Conference on Linguistics, edited by Joyce Powers and Kenneth de Jong. Department of Linguistics, The Ohio State University, Columbus. Partee, Barbara. 1991. Topic, focus and quanti®cation. In Proceedings from SALT 1, edited by Steven Moore and Adam Zachary Wyner. CLC Publications, Cornell University, Ithaca, N.Y. Pierrehumbert, Janet, and Julia Hirschberg. 1990. The meaning of intonational contours in the interpretation of discourse. In Intentions in communication, edited by Philip R. Cohen, Jerry Morgan, and Martha E. Pollack. Cambridge, Mass.: MIT Press. Portner, Paul, and Katsuhiko Yabushita. 1998. The semantics and pragmatics of topic phrases. Linguistics and Philosophy 21, 117±157. Raposo, Eduardo, and Juan Uriagereka. 1995. Two types of small clauses. In Small clauses, edited by Anna Cardinaletti and Teresa Guasti. San Diego, Calif.: Academic Press. Rochemont, Michael. 1986. Focus in generative grammar. Amsterdam: John Benjamins. Rooth, Mats. 1985. Association with focus. Doctoral dissertation, University of Massachusetts, Amherst. Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics 1, 75±116. Rooth, Mats. 1994a. Association with focus or association with presupposition? In Focus and natural language processing. Working paper of the Institute for Logic and Linguistics 7, vol. 2, edited by Peter Bosch and Rob van der Sandt. Heidelberg: IBM Germany. Rooth, Mats. 1994b. Focus. In The handbook of contemporary semantic theory, edited by Shalom Lappin. Oxford: Blackwell. Rooth, Mats. 1995a. Evidence for focus in second occurrence focus. Handout from a talk given in Bad Teinach, Germany, May 1995. Rooth, Mats. 1995b. Inde®nites, adverbs of quanti®cation, and focus semantics. In The generic book, edited by Gregory Carlson and Francis Je¨ry Pelletier. Chicago: University of Chicago Press.

160

References

Rooth, Mats. 1996. On the interface principles for intonational focus. In Proceedings from SALT 6, edited by Teresa Galloway and Justin Spence. CLC Publications, Cornell University, Ithaca, N.Y. Rullmann, Hotze. 1997. Even, polarity, and scope. Ms., University of Alberta, Edmonton. Russell, Bertrand. 1905. On denoting. Reprinted in The philosophy of language, 2nd ed., edited by A. P. Martinich. Oxford: Oxford University Press. Schein, Barry. 1993. Plurals and events. Cambridge, Mass.: MIT Press. Schwentner, Scott. 1998. Two types of scalar particles: Evidence from Spanish. Ms., Ball State University, Muncie, Ind. Selkirk, Elisabeth. 1984. Phonology and syntax. Cambridge, Mass.: MIT Press. Selkirk, Elisabeth. 1995. Prosodic re¯exes of focus. Talk given at UMass Focus Workshop, University of Massachusetts, Amherst, December 1995. Stalnaker, Robert. 1974. Pragmatic presupposition. In Semantics and philosophy, edited by Milton K. Munitz and Peter K. Unger. New York: New York University Press. Stalnaker, Robert. 1979. Assertion. In Syntax and semantics 9: Pragmatics, edited by Peter Cole. New York: Academic Press. Stechow, Arnim von. 1991. Current issues in the theory of focus. In Semantics: An international handbook of contemporary research, edited by Arnim von Stechow and Dieter Wunderlich. Berlin: Walter de Gruyter. Szabolcsi, Anna, ed. 1997. Ways of scope taking. Dordrecht: Kluwer. Uribe-Etxebarria, Myriam. 1994. Interface licensing conditions and negative polarity items: A theory of polarity and tense interactions. Doctoral dissertation, University of Connecticut, Storrs. VallduvõÂ, Enric. 1990. The informational component. Doctoral dissertation, University of Pennsylvania, Philadelphia. [Published 1992, New York: Garland.] Westerstahl, Dag. 1985. Logical constants in quanti®er languages. Linguistics and Philosophy 8, 387±413. Wilkinson, Karina. 1996. The scope of even. Natural Language Semantics 4, 193± 215. Williams, Edwin. 1983. Semantic and syntactic categories. Linguistics and Philosophy 6, 423±446. Zanuttini, Ra¨aella. 1991. Syntactic properties of sentential negation: A comparative study of Romance languages. Doctoral dissertation, University of Pennsylvania, Philadelphia. Zubizarreta, Maria Luisa. 1998. Prosody, focus, and word order. Cambridge, Mass.: MIT Press.

Index

Aboutness, 11, 19, 29, 110, 138 and assertion, 20 expressed by the restriction of a quanti®er, 19, 29, 38, 105 and quanti®er structure, 19 as a relative notion, 20 Adverbs, 59 as arguments, 4±6 bound reading, 61 c-command domain of, 63, 69 free reading, 62, 63 iteration of, 6, 7 and the Mapping Hypothesis, 59 as operators, 6, 7 optionality of, 6, 7 polyadic analysis of, 4±6 polyadicity problem, 7 quanti®cational structure of, 59, 60, 73 (see also Quanti®ers, adverbial) scopal ambiguity of, 7 as selective quanti®ers, 82 situation-based account of quanti®cational structure of, 59, 82±84 (see also Situations) tripartite structure of, 74, 80 Adverbial quanti®ers. See Adverbs; Quanti®ers Alternative semantics approach. See Focus Arguments optional, 8 realized as datives, 8 as separate conjuncts (see Decomposition; Separation) Aspect, 80 Assertion, 44, 47, 51, 53 and aboutness, 20 Atlas, J. D., 86, 93, 94, 96, 100 Backgrounded focal entailment, 20±23, 26, 27, 29, 31, 33±35, 37, 39, 40, 46±49, 51, 52, 56, 96

Barker, L., 118 Barwise, J., 131, 132 Basque, 32, 62 BaÈuerle, R., 77 Berman, S., 82 Bivalent semantics. See Semantics, twovalued Boer, S., 16, 17, 96 Bonomi, A., 86, 107 Bound reading adverbials (see Adverbs) negation (see Negation) Carlson, L., 51±53 Casalegno, P., 86, 107 Categorical judgment vs. thetic judgment, 136, 138 and individual- vs. stage-level predicates, 136 Chomsky, N., 12, 40, 139 Cinque, G., 12 Cleft, 41, 50 Conservativity, 2, 89, 90±93, 130, 131 de®nition, 89 and e¨ect of focus on weak determiners, 93 neoconservativity, 90, 91, 93 of only, 2, 89, 97 (see also Only) right-conservativity, 90 Context variable, 65, 66 Contrastive focus. See Focus Cooper, R., 131, 132 Davidson, D., 4, 7±9 Davidsonian decomposition. See Decomposition Decomposition, 8 and entailment patterns, 4, 6 event decomposition, 4 structured Davidsonian decomposition, 2, 9, 17, 18±21, 26±30, 42, 47±49, 59±61, 68, 71, 73, 75, 79, 80

162 Decomposition (cont.) of verbs, 3 DE contexts. See De®niteness e¨ect Decreasing quanti®ers. See Quanti®ers De®nite descriptions, 35, 38, 102 and scope of negation, 36±38 existential descriptions of, 38 Fregean view of, 48 quanti®cational analysis of, 35±38 Russellian view of, 35 universal and existential force of, 103 De®niteness e¨ect, 1, 124±127, 131, 135 anti-DE environments, 128, 136, 138 optional DE environments, 128, 138 and Q-raising, 3 QR vs. Q-raising, 135 (see also QR; Q-raising) Deising, M., 59, 80, 97, 128, 131, 133±135 de Mey, S., 90 De re/de dicto ambiguity, 120 de Swart, H., 60, 75, 92 Determiners, 49 decreasing vs. nondecreasing, 134 de®nite, 35 internal vs. external arguments of, 124, 127, 131, 139 (non)conservativity of, 130 (see also Conservativity) relational view of, 35 strong, 2, 105, 131, 138, 139 symmetric, 132 weak, 1, 93, 105, 131, 132, 135, 138, 139 Doborvie-Sorin, C., 131 Donkey anaphora, 77 Downward monotonicity, 100±102, 105 de®nition, 101 every, 102 weak vs. strong, 101, 102 Dowty, D., 4 D-raising, 129 Drubig, H. B., 41 Dryer, M., 63, 65, 73, 80 Dutch, 117 Eckhardt, R., 16 Egli, U., 77 English, 26, 31, 47, 49, 118, 128 Entailment, 17 backgrounded (see Backgrounded focal entailments) patterns, and decomposition, 4, 6 and presupposition, 16 Even, 2, 85, 86, 108, 131 ambiguity of, 86 di¨erences from only, 109

Index even-sentences in propositional attitude contexts, 120 existential commitment as pragmatic inference, 119 meaning of even-sentences, 110 and noteworthiness, 111, 114±118, 121, 122 and Q-raising, 110, 130 scope account vs. NPI account, 111±118 and second occurrence focus, 109 summary of properties, 121, 122 syntactic behavior of, 108, 118 two kinds of, 121, 122 Event decomposition. See Decomposition Event quanti®er. See Quanti®ers Events, as individuals, 82 Exhaustive focus. See Focus Existential closure, 74, 134, 135 Existential force, 2 of every, 97±100 and only, 93, 97 (see also Only) Existential implicature, 110±116 Existential presupposition. See Only; Presupposition Experiencer, 3, 4 Factive predicate. See Predicates Fall contour. See Intonation Fall-rise contour. See Intonation f-a noun phrases. See Noun phrases Finnish, 117 Focal entailment. See Backgrounded focal entailment Focal mapping, 2, 43, 45, 47, 104, 132, 139 and conservativity, 89 and even, 110 and LF syntax, 46 and nonlocal e¨ect of focus, 44 and only, 92, 97, 108 and QR, 46 and Q-raising, 85, 89, 92, 97, 104, 123, 130, 131 Focal presupposition. See Presupposition Focus alternative semantics approach, 13, 38, 53, 63±65, 75, 82, 124 and constituency, 44 and embedded clauses, 71 and exhaustiveness, 53, 54, 57, 106 and LF syntax, 42, 47 and negation, 2, 29, 31, 36, 38, 50 and negation, free vs. bound readings, 30 and scrambling in German, 31, 32 (see also German; Scrambling) and stress, 12 (see also Intonation) and types of quanti®ers, 15

Index contrastive vs. noncontrastive, 2, 50, 53 discontinuous, 45 function-under-discussion analysis, 16, 21, 28 in-situ analysis of, 42 meaning e¨ects vs. phonological realization, 11, 12 (see also Intonation) movement, 32, 40, 41, 65±67, 116 multiple, 45 nonlocality of, 40, 41, 79 phonological realization, 47 phrase, 40 pragmatic e¨ects of, 50 predicational analysis of, 48, 49 presuppositional analysis, 16, 27, 28, 38, 39, 49 primary function of, 12 second occurrence focus, 2, 48, 59, 63, 64, 67±69, 80, 87±89, 109 structure imposed on quanti®cation, 18 tripartite mapping analysis of, 42 versus VP-external/-internal, 80 Geach, P., 93 Generalized quanti®ers. See Quanti®ers Generic sentences, 19 German, 31, 45, 62, 80, 81, 87, 117, 128 Goal, optional speci®cation of, 25 Gricean maxims, 58, 106 of quantity, 53 of relevance, 57 Haitian Creole, 49 HajicÏovaÂ, E., 30, 36 Heim, I., 74, 77, 78, 80, 82, 99, 100 Herburger, E., 16, 24, 42, 117, 118, 124, 129, 131, 133 Higginbotham, J., 131, 132 Hirschberg, J., 51 Horn, L., 57, 90, 93±95, 97, 101±103 Hornstein, N., 47 Huettner, A., 125 Hungarian, 32, 62 if-clauses, 59, 69±72, 77±79, 123 Implicature conventional implicature, 93, 118 conversational implicature, 17, 53, 93 existential (see Existential implicature) scalar implicature, 110±112, 114 (see also Even; Noteworthiness) Inde®nites nonquanti®cational, 80 as quanti®cational, 82 quanti®cational variability of, 2, 73±79 (see also Quanti®cational variability)

163 Individual-level predicates. See Predicates Intonation, 1, 30, 57, 109 fall contour, 30, 50, 51, 53, 54, 56, 57 fall-rise contour, 30, 32, 50±57, 106 and focus, 50, 51 and mimicking, 67±69, 88 pragmatic e¨ect of, 54, 58 Inversion negative inversion, 97, 104, 105 Islands and focus movement, 41, 65±67, 116 Jackendo¨, R., 12±15, 29, 30, 32, 50, 53, 87, 109 Jacobs, J., 29, 30, 36 Japanese, 49, 137 Johnston, M., 70, 72, 80 Kadmon, N., 77 Karttunen, L., 110±112, 119, 120 Katz, J., 16 Kay, R., 110, 121 Keenan, E., 131 Kratzer, A., 99, 100 Krifka, M., 60, 67±69, 77, 106 Ladd, D. R., 54±57 Ladusaw, W., 23, 138 Laka, I., 23, 24 Lambda-abstraction, 13 and focus-semantic value, 13 Larson, R., 19, 42, 49 Lefebvre, C., 49 Lewis, D., 73±75, 80 LF, 1, 2, 34, 42±44, 46, 47, 108 and LF0 , 46 and the Mapping Hypothesis, 80 movement of focused elements, 40, 42 as scopally disambiguating, 34 LoÈbner, S., 48 Lycan, W., 16, 17, 96 Mapping Hypothesis, 59, 80, 81 May, R., 46 Milsark, G., 124±129, 131±133, 135±139 Mittefeld, 31. See also German; Scrambling Modi®ers adverbial, as predicates of events, 3 (see also Adverbs) polyadic analysis of adverbs (see Adverbs) Moravscik, J., 97±100 Monotonicity. See Downward monotonicity Movement. See also Focus; Islands A-movement, 47 A0 -movement, 46 Case, 47 and chain uniformity, 92, 139

164 Movement (cont.) covert, 40, 47, 123 focus, 32, 41, 66, 67, 116 Musan, R., 131 Neale, S., 35 Negation bound reading as paradigmatic, 34, 35 bound vs. free readings and intonation, 51±54 decreasing quanti®ers (see Quanti®ers, decreasing) and focus, 2, 29, 32, 36, 38, 50 (see also Focus) scope of, 11, 29, 30 (see also Scope) scope of, and bound vs. free readings, 29± 31, 34, 37 and stress, 33, 34 structured wide reading, 32±37, 52 as a univocal propositional operator, 29, 31 unstructured wide reading, 33±37 Negative inversion. See Inversion Negative polarity, 24, 100±104. See also Even Neoconservativity. See Conservativity Neo-Davidsonian event quanti®er, 1 (see also Quanti®ers, event) Noteworthiness, 111, 114±118, 121, 122. See also Even Noun phrases bare plurals, 90 DE noun phrases, 135, 139 f-a noun phrases, 124±129 focus assigned to part of, 91, 127 inde®nite, 73 only and even adjoined to the NP, 127, 129 strong, 2, 124 theta-related noun phrases and individuation of events, 82 weak, 2, 124±126 Noncontrastive focus. See Focus NPI. See Negative polarity N-words, 23±25, 118 Ogihara, T., 48, 49 Only, 2, 85, 86, 88, 107, 131 as adjectival vs. quanti®cational, 87 asymmetrical, presuppositional approach, 93±96, 103 di¨erences from even, 109 and existential presupposition, 93, 94, 105 as having a focused element in its c-command domain, 87

Index interface assumption about, 90 as lexically similar to a universal quanti®er, 92 and negative polarity licensing, 97 and negative preposing, 97, 103, 104 and (neo)conservativity, 93, 105 and (non)monotonicity, 100±102, 105 as presuppositional vs. having existential force, 93 Q-raising of, 91, 130 (see also Q-raising) resemblance to always, 105 scalar uses, 107 and second occurrence focus, 87±89 syntactic distribution of, 86, 87, 105, 108 as universal-like, 100, 105 Parsons, T., 3, 4, 8 Partee, B., 42, 50, 63, 75, 87, 125, 133 Paul, H., 48 Peters, S., 110±112, 119, 120 Pierrehumbert, J., 51 Polyadic analysis of adverbs. See Adverbs Portner, P., 45 Pragmatic aboutness. See Aboutness Predicates adverbial modi®ers as predicates of events, 3 context predicate, 19 individual-level, 81, 125, 126, 135±138 psychological, 48 stage-level, 125±128, 135±137 Predication traditional, 3 Presentational sentences, 44 Presupposition, 2 and assertion, 13 existential, 93, 94 (see also Only) failure of, 16 Fregean view of, 16 focal, 13, 15, 16, 22, 27, 28, 38, 39 (see also Backgrounded focal presupposition) in terms of entailment, 16 suspendability of, 94±97 and three-valued logic, 17 (see also Threevalued semantics) Presuppositional analysis (of focus). See Focus p-set, 13 Pseudocleft, 50 Q-raising, 2, 131 and the de®niteness e¨ect, 3 of even, 110, 130 and focal mapping, 85, 89, 92, 97, 104, 123, 130, 131 of just D, 129, 130

Index of only, 91, 130, 139 vs. QR, 135, 137, 138 QR, 3, 46, 47, 129, 135, 137±139 Quanti®cational variability, 59, 73±79 Quanti®ers adverbial, as present in every sentence, 19 adverbial, overt, 1, 59±63, 71, 76, 78, 79 adverbial, tacit, 19, 43, 59, 137 adverbial, vs. nominal determiners, 123 Aristotelian interpretation of universal quanti®ers, 98±100 decreasing, 15, 21, 22, 26±28, 38, 47, 48 event, 1, 18, 20, 22, 27, 29, 40, 43, 47, 48, 50, 54, 59, 76 generalized, 90 inde®nite, 48 internal vs. external argument of, 42, 43, 90, 110, 124 negative, 21, 23, 24, 46 nondecreasing, 48 possible natural language quanti®er, 89 quanti®er structure and aboutness, 19 unary quanti®ers association with focus, 69 QVE. See Quanti®cational variability Raposo, E., 137 Right-conservativity. See Conservativity Rochemont, M. Rooth, M., 13, 38±40, 53, 59, 63±66, 68±70, 75, 76, 88, 109±113 Rullman, H., 111±113, 116, 117 Russell, B., 35 Schein, B., 4, 8, 18, 56, 83 Schwentner, S., 117 Scope, 1 bound vs. free readings (see Adverbs; Negation) of decreasing quanti®er relative to event operator, 21, 26, 28 inverse, 9, 129 of negation, 11, 30, 31 of negative quanti®er, 21, 23 nuclear, 81 and n-words, 23, 25 of only, (see Only) and Q-raising vs. QR, 2, 3 and scrambling, 45 Scrambling, 31, 45, 80, 128, 137 Second occurrence focus. See Focus and mimicking, 67, 68, 88 Segal, G., 19, 42 Selkirk, L., 12 Separation, 4, 8, 9, 30. See also Decomposition

165 Situation-based account. See Adverbs Situations minimality requirement of, 82±84 part-whole structure of, 82, 84 Spanish, 23±26, 87, 117 n-words, 23±26 quanti®er system, 26 Stage-level predicates. See Predicates Stalnaker, R., 16, 20 Structured Davidsonian decomposition. See Decomposition Stress. See also Intonation on negation, 33, 34 shift of, and shift in interpretation, 1 Strong determiners. See Determiners Structured wide reading. See Negation Subatomic semantics. See Semantics Suspension of presupposition. See Presupposition Swedish, 117 Taglicht's observation, 88, 109, 116 Theme, 3, 4, 7 optional, 8 there-insertion, 124±127, 135±138 and categorical vs. thetic judgements, 136 Theta-roles, 3, 4, 8, 76, 79 Thetic judgment. See Categorical vs. thetic judgment Three-valued semantics, 16, 17. See also Two-valued semantics Topic discourse topic, 65, 66, 135 marking, 137 topicalized phrases, 45 Tripartite structure. See Adverbs; Focus Two-valued semantics, 2, 20. See also Three-valued semantics Unselective binding, 73±75, 77, 80 Unstructured wide reading. See Negation Uriagereka, J., 137 Uribe-Etxebarria, M., 23 VallduvõÂ, E., 63, 65, 73, 80 Verbs types and decomposition, 3 von Fintel, K., 13, 59, 60, 65, 66, 75, 76, 82, 86, 87 von Stechow, A., 40, 48, 88 VP-external vs. VP-internal. See Focus; Mapping Hypothesis Weak vs. strong determiners. See also De®niteness e¨ect; Determiners Westerstahl, D., 130 when-clauses, 59, 69, 70, 72, 123

166 Wilkinson, K., 110±114 Williams, E., 50 Yabushita, K., 45 Zanuttini, R., 23, 24, 26 Zubizarreta, M., 12

Index