University of Southern California Dissertations and Theses

In the event of focus

(USC Thesis Other)

In the event of focus

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Infomiation Company 300 North Zeeb Road, Arm Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IN THE EVENT OF FOCUS by Elena Herburger A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSTIY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Linguistics) May 1997 Copyright 1997 Elena Herburger Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number; 9733067 UMI Microform 9733067 Copyright 1997, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CAUFORNIA THE GRADUATE SCHOOL UNIVERSmr PARK LOS ANGELES. CALIFORNIA 90089 This dissertation, written by b' leaja u£u under the direction of Dissertation Committee, and approved b y all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean ïllLl Date DISSERTATION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Acknowledgments : I am glad to finally have a chance to acknowledge, however briefly and incompletely, the often vital help I have gotten while I was writing this thesis. My biggest gratitude is to Barry Schein. His deep understanding of linguistics, intellectual generosity and circumspect advice have helped me tremendously and they made it exciting to be a student at ÜSC. He has put many hours into reading various drafts, and needless to say, his comments have improved it drammatically. Thanks also zo the other members of my committee, Joseph Aoun, Jean-Roger Vergnaud, and earlier Brian Loar from Philosophy, whom Janet Levin replaced after he left for Rutgers. Their probing questions forced me to clarify many issues that were important and which would never have occurred to me, or which I would have "set aside for sometime later." Were it not for the many discussions I had with my colleagues Filippo Beghelli, Robin Belvin, Gerhard Brugger, José Camacho, Gorka Elordieta, Dan Jackson, Liliana Sanchez, and Patricia Schneider-Zioga I would know much less— and I would have had much less fun; in fact, I think I would have found it unbearable to be a graduate student. I am also greatful for the solidarity and helpful advice of the "older guys" Utpal Lahiri, Hamida Demirdache, and Myriam Uribe- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. i l l Etxebarria. Laura Reiter has been a great friend throughout. She and Linda Culver have made the main office always be a pleasant place to go to (which, for someone who was used to Central European burocracy, seemed like a small miracle.) Despite the daunting LA traffic, I enjoyed all my cross-town trips to UCLA to meet with Anna Szabolcsi, Tim Stowell and also Ed Keenan, It was during Tim's visit to Vienna that I made up my mind to become a linguist and it was Tim's hospitality in LA that made me feel at home when I first arrived. In the 1992 Girona summerschool, I took a class with Richard Larson. Ever since, he has generously given me advice whenever I asked for it, ranging from comments on my work, to help in applying for jobs. I am also glad I visited MIT in 1994/1995, which was an intellectually very stimulating experience. Particularly helpful for my thesis project were the meetings I had with Sylvain Bromberger, Kai von Fintel, Renate Musan, Orin Perçus, Hubert Truckenbrodt Roger Schwarzschild and especially Irene Heim, whose comments have helped me rethink the analysis in chapter 4. In Boston I also met Esther Torrego, Esthela Trevino and Paco Ordonez. We had long conversations about Spanish syntax— and many other things as well. Thanks also to Ricardo Echepare, Gorka Elordieta, and Itziar Laka for patiently explaining Basque focus to me, and to Martin Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IV Haiden for exchanging long chains of thoughts on the German "Mittelfeld" over E-mail. During my stays in Washington DC, the people in College Park have been very hospitable. I cannot list them all, but I especially want to thank Norbert Hornstein for passionately discussing with me various issues related to events and quantifiers and for explaining to me many angles of the big picture which for me had been shrouded with fog. Not until I had the chance to co-teach with him a class on events did I realize that teaching a class you learn more than taking it. I also appreciate the helpful input I received from Paul Fortner on various occasions. Finally, gracias mil to Juan Uriagereka, who has done more things for me while I was writing this thesis than I can say here. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. table of CONTENTS: ACKNOWLEDGEMENTS..........................................11 CHAPTER 1: INTRODUCTION....................................1 1.1. Semantic background: an event-based semantics........ 3 1.1.1. Adverbial modification............................. 6 1.1.2. Why so much decomposition?.........................13 1.2. Overall structure the dissertation...................16 CHAPTER 2: FOCUS AND PRESUPPOSITIONS......................21 2.0. Introduction........................................ 21 2.1. Focus and Discourse................................. 24 2.1.1. An analysis à la Jackendoff (1972)................24 2.1.2. "Under discussion" is too weak.................... 28 2.2. Structured Davidsonian Decomposition................. 32 2.2.1. The basic idea.................................... 32 2.2.2. Wide scope decreasing quantifiers.................36 2.2.3. Narrow scope decreasing quantifiers...............38 2.2.4. Postverbal Spanish n-words as a diagnostic..... 2.3. Focus and negation.................................. 53 2.3.1. Different readings :bound, free, and wide.......... 53 2.3.2. (Focal) presuppositions as direct entailments from logical form.................................65 2.3.3. Definite descriptions, focus and negation......... 70 2.3.4. Other negative contexts 7 6 2.4. Revising and situating the SDD.......................81 2.4.1. Focus on elements too small or too distant........ 81 2.4.2. The SDD vis-a-vis similar analyses................ 90 2.5. Summary of chapter 2.................................96 Appendix to chapter 2 : Different kinds of intonation: fall-rise and fall....... 100 1. The fall-rise intonation of bound readings........... 100 2. Contrast: an effect of the fall-rise contour......... 104 3. More complex examples.................................10' 4. Summary...............................................115 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. VI CHAPTER 3: ADVERBIAL QUANTIFIERS........................ 11’ ’ 3.0. Introduction........................................117 3.1. Association with focus............................. 117 3.1.1. A subcase of the SDD.............................117 3.1.2. Not every adverb associates......................121 3.1.3. Second occurrence focus..........................124 3.1.4. Second occurrence foci and mimicking.............132 3.1.5. Restricting embedded clauses.....................138 3.2. Adverbs and quantificational variability...........144 3.2.1. The QVE and unselective adverbs..................145 3.2.2. Adverbs as selective quantifiers.................14 9 3.2.3. The QVE in preposed clauses: asymmetric readings..153 3.3. Events or situations?.............................. 161 3.5. Summary of chapter 3............................... 166 CHATPER 4: FOCUS AND NOUN PHRASES....................... 169 4.0. Introduction........................................169 4.1. The Definiteness Effect.............................170 4.1.1. Milsark's and Diesing's account...................170 4.1.2. Problemsrmonotonicity and cardinality/symmetry.... 176 4.2. F-a readings........................................181 4.2.1. The basic phenomenon..............................181 4.2.2. The DE on f-a readings............................184 4.2.3. F-a readings and Conservativity...................18 9 4.3. Determiners that work like adverbs..................192 4.3.1. Semantically unary vs. binary.....................192 4.3.2. D-raising vs. QR..................................194 4.3.3. Focus inside the VP...............................202 4.3.4. Focus in strong (-like) DPs....................... 205 4.5. Summary of chapter 4............................... 207 BIBLIOGRAPHY.............................................209 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 1: INTRODUCTION 1 It is a basic linguistic fact that the meaning of a sentence does not only depend 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 different, and so are the meanings. Although fundamental, syntactic structure is not alone in shaping meaning. We know that focus also has a considerable semantic effect. Thus, by shifting the main stress within a sentence, the interpretation shifts as well, Compare, for instance, Oliver doubts that OSWALD killed Kennedy, where Oswald is stressed, with Oliver doubts that Oswald KILLED Kennedy, where the stress falls on killed: Oliver's doubt is different in each case. Whereas in the first example he is not sure about who killed Kennedy, in the second example he is uncertain about what Oswald did to Kennedy. The concern of this thesis is how focus affects meaning. More specifically, it investigates how focus affects quantificational structure, which will be argued to be rather ubiquitous and to also be present in cases where we have no overt indication that it is there. One of the main phenomena to be discussed is this: when a quantifier has only one syntactic argument, then its quantificational structure is entirely shaped by the focus. This is shown tc Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. be true not only of adverbial quantifiers, but also of certain types of nominal quantifiers (determiners), where it has interesting consequences for the analysis of the Definiteness Effect. Couched within the view that all sentences can be analyzed as descriptions of events, well-known effects of focus on adverbial quantifiers reduce to a simple phenomenon: the nonfocused part within the scope of the relevant event quantifiers restricts the quantifier, and it does so directly as a matter of logical form (semantics). This process is referred to as Structured Davidsonian Decomposition. It is further proposed in this thesis that the structuring effect of focus is not only present in sentences with overt adverbs of quantification such as usually^ rarely, always, etc.; it can also be found in sentences which lack an audible quantificational adverb; but which, on our assumptions, contain a tacit event quantifier. Recognizing the presence of Structured Davidsonian Decomposition in such cases makes it possible to explain how focus affects the logical presuppositions of a sentence. As for the consequences that focus has on the felicity of a sentence in a given stretch of discourse, these are argued to be derivative on the way focus shapes quantificational structure and logical presuppositions. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. By making the semantic and pragmatic effects of fccus follow directly from quantificational structure the present analysis tries to be parsimonious. It sets out to account for focus without taking recourse to an information structure component or an alternative semantic valuation of the sentence. The claim is not only that it is indeed feasible to give such a purely semantic account, but that it offers stronger empirical predictions than other approaches, allowing for a rather fine-grained account of the distribution of focal presuppositions and the interaction between focus and adverbial quantifiers. 1.1. Semantic background: an event-based semantics This thesis assumes as its semantic background that sentences are best thought of as descriptions of events (where "event" is meant in a generous sense which also includes states.) Developing an analysis of focus that uses event-semantics, it tries to provide a further argument for the general claim. Traditionally, we think of sentences like (1) as predications where the verb love is a two-place predicate taking Mary and John as its arguments, as shown in (2): (1) John loves Mary. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) love(John, Mary) While (2) is simple and gives an intuitively correct interpretation, recently several arguments have been proposed suggesting that sentences should be accorded more complex logical forms than assumed in (2); in particular, the verb is said to decompose into a "sub-atomic" 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) 3e (love(e) & lover(e, John) & loved-one(e, Mary)) All verbs on this view translate as one-place predicates of events, regardless of whether they are syntactically unaccusative, intransitive, transitive, or ditranstive. Like verbs, adverbial modifiers are also direct predicates of events (see below). Verbs being one- place predicates, what are traditionally considered the arguments of the verb (e.g. John and Mary in (2)) are now 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) (of. Parsons 1990) : Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (4) 3e (love(e) & Experiencer(e, John) & Theme(e, Mary)) Even though (4) makes use of theta-roles, it is worth noting that a full-fledged event semantics is not committed to the existence of a general definition of what exactly the individual theta-roles agent, experiencer, theme, etc. consist in. All that is needed is that the event participants can be distinguished relative to each verb (of. Parsons 1990:101ff, Schein 1993:331ff). This is just as well, because a general definition of the individual theta- roles is notoriously difficult, and perhaps even impossible (cf. Dowty 1989) . Why should we buy into this more elaborate way of analyzing John loves Mary when the traditional "love (John, Mary)" seems to be doing the job just fine? Part of the aim of chapters two and three is to show that the semantics of focus and adverbial quantification makes an investment in event decomposition well worth the trouble because it allows for a simple and empirically detailed analysis of the effects of focus. In what follows here, I would like to summarize some independent reasons which have led to the idea that sentences are interpreted as descriptions of events. This includes both the classic argument (Davidson 1967), as well as more recent arguments for treating the arguments of the verbs as forming their own conjuncts ("separation") . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1.1.1. Adverbial modification: Davidson's 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 which apply to an event introduced by the verb. The empirical motivation driving this proposal is that it makes it possible to explain certain entailment patterns that we find with adverbs. For instance, it explains why (4a) entails (4b), (4c), and also (4d). It also explains why both (4b) and (4c) entail (4d). (4) a. Brutus stabbed Caesar in the back with a knife. b. Brutus stabbed Caesar in the back. c. Brutus stabbed Caesar with a knife. d. Brutus stabbed Caesar. (5) (4a) (4b) (4c) (4d) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 All of this follows directly if verbs describe events and adverbs are predicated of these events because, simply put, the longer conjunctions entail the shorter ones: (6) a. 3e (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. 3e (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. 3e (stab(e) & Agent(e,Brutus) & Theme(e,Caesar)) "There was a stabbing whose agent was Brutus and whose theme was Caesar." (cf. (4d)) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 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 in (4b) and (4c) we may be talking about two different events respectively, one where Caesar was stabbed in the back (say with an icepick), and one where he was stabbed with a knife (say in the thigh). But while (4b) and (4c) may be describing different events, (4a) requires there to be one single event of stabbing Caesar such that the instrument that was used was a knife and he was stabbed in the back. The argument for treating adverbials as predicates of events that are denoted by the verb is presented against the background of an analysis that accords to each adverb its own argument place in the verb. Compared to the Davidsonian account, this kind of analysis is unattractive because it is committed to the claim that the verb in (4a) [Brutus stabbed Caesar in the back with a knife) is a four-place predicate, while the verb in (4d) (Brutus stabbed Caesar) , for instance, is a different (but homophonie) two-place predicate. Multiplying verb meanings, this analysis obscures the logical entailment between (4a) and (4d). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 Actually, there is one way that would make it possible to treat adverbs as arguments of the verb and to still capture the entailment between (4a) and (4d): if we assume that (4d) contains a four-place predicate, the same as in (4a) and if we further say that, unlike in (4a), in (4d) the predicate is elliptical. Even though (4d) contains no visible modification for time or place, it would mean "Brutus stabbed Caesar somewhere, with something." Existentially quantifying instead of a "missing" adverb would get the meaning right in (4d), and, moreover, it 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.). This, however, makes it impossible to state how many adverbial argument places the verb stab, for instance, 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 quantifies into the place of the "missing" modifier, this is unexpected; if one stabbing takes place next to a guard, for instance, then every instance of stab Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. id has to be analyzed as taking place next to someone. But this is clearly not what we want because stabbings can take place next to noone, in fact, they typically do. On the other hand, if verbs introduce event arguments and if adverbials are analyzed as predicates of these events, it is no problem that adverbials can, in principle, be iterated ad infinitum, nor is it troublesome that many adverbials are genuinely optional in the sense that when they are left out, no implicit existential quantification can be assumed to take the place of the argument within the adverbial. On a third possible analysis of adverbs— one which treats them neither as arguments nor as predicates— adverbials are analyzed as operators that apply to VP meanings and return VP meanings (i.e. as a function of the type <<e,t>, <e,t>>). Since on this account adverbials are not treated as arguments of the verb, there is no problem in capturing that they can be iterated so easily. In other words, there is no polyadicity problem. Nor is it difficult to do justice to the fact that they can be genuinely optional. The difficulty with this kind of analysis is a rather subtle one. Although it captures some of the entailments that we observed in (4), it turns out that it does not capture all of them (cf. Parsons 1990:55ff). Treating both with a knife and violently as operators, the prediction that is made is that one modifier will be able to take scope over the other. (7), for instance. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 should be scopally ambiguous between (8a) and (8b): in one case with a knife would take wide scope, in the other case violently would: (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) only entails that x stabbed y with z, but not that x stabbed y violently. The problem now is that (7) entails both, namely that x stabbed y violently and also that z 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) vs. (8b). Yet, speakers do not perceive (7) as being ambiguous in that way. On the other hand, the two entailments follow directly on Davidson's analysis because each of the adverbials forms its own conjunct. Leaving out the knife conjunct, it follows that Brutus stabbed Caesar violently; leaving out the violent conjunct, it is entailed that Brutus stabbed Caesar with a knife: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 (9) 3e (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 which treats adverbials as predicates of events explains all the logical entailments that we find in examples like (4) and it fits in naturally with the fact that adverbs can be iterated freely and that many of them are genuinely optional.- -Davidson (1967) does not treat attributive modifiers as predicates of events. His reason is essentially that in situation where Susan crossed the Channel by swimming and undertook no other crossing, (i) and (ii) are true simultaneously. Yet, (iii) and (iv), where the same attributive modifier is added to the verbs, need not be true simultaneously: (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. As Parsons demonstrates, however, it is also possible to give a Davidsonian treatment of attributive operators: attributive modifiers have to be relativized. Since "quickly for a crossing" is need not be the same as "quickly for a swimming" (it typically is not), (iii) and (iv) come out as having divergent truth-conditicns, as required. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 1.1.2. Why so much deconposition? All the entailments in (4) would also be explained on a less radical account: sentences are descriptions of events but only the adverbial modifiers form their own conjuncts and the traditional arguments of the verb are not separated, but we leave the verb polyadic instead. This is in fact the kind of analysis proposed originally by Davidson (1967). Thus, on this view, (4a) has a logical form like this: (10) 3e (Past(e) & stab(Brutus, Caesar) & with a knife(e) & in the back(e) Why should decomposition be more radical than that, as I have been assuming here and as will be assumed throughout this thesis? One reason for the full-fledged decomposition involving the separation of the arguments into their own conjuncts is that there are theta-marked arguments which are genuinely optional. These arguments do not have to be lexically realized, and moreover, they are not semantically "implicit" when they are absent. Thus, in a relevant sense, these optional arguments behave like the optional adverbial modifiers that were discussed above. Parsons argues that the theme of stab is genuinely optional because we can coherently say Brutus stabbed and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 missed. There is nothing- that he stabbed (Parsons 1990:97ff). 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 stab, write, and say, etc. are ambiguous, which seems certainly 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 % to y.- In addition to capturing the entailment patterns we find with genuinely optional arguments, a further, and entirely independent argument for the separation comes from sentences like (11) (Schein 1993, ch.4): -In general, it is quite difficult to find genuinely optional direct objects. (In the case of stab there may be more internal structure to the verb than with other transitive verbs.) On the other hand, genuinely optional dative and prepositional arguments are considerably easier to find. Cf. in this context also Parsons' example of the "agentless" passive of stab, which involves a PP (p. 98ff): (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. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 (11) Three video games taught every quarterback exactly two new games. This sentence has various readings. On one interpretation, for instance, it is construed as saying that each of three video games is such that it taught every quarterback two new games, where they games may vary according to quarterbacks. On a similar, but inverse scope reading, (11) says that for every quarterback it holds true that three possibly different video games taught him two possibly different new games. Here, the video games can vary relative to each quarterback, and the new games, in turn, may vary with each the video games. The most salient reading of (11), however, seems to be this: as in the first reading, the quantifiers take the scope that corresponds to their surface order. Unlike the first reading though, in this reading it is left vague how the video games relate to the quarterbacks. The reading only distributes new plays with respect to quarterbacks, but does not commit 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 an event where three video games were the teachers, and that teaching resulted in every quarterback learning two new plays each." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 As the paraphrase makes clear, what the video games do, namely teach, and what happens to every quarterback— their being taught two new games— have to be treated as separate (though related) matters. The point now is that we can keep apart what the video games do from what the quarterback experience if the subject is separated out from the rest and has its own event quantifier. But in order for the subject to describe its own little event, the subject cannot be a direct argument of the verb. Thus, the third interpretation of (11) offers an independent argument for the separation of arguments.- Thus, while 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 modifiers 'While the essential point, which is what is relevant here, is a simple one— namely that what the video games do needs to be separated out from the rest and treated as a separate though related event— the actual logical form that achieves the relevant result looks a bit daunting: (i) 3e ([3X: 3(x) & Vx(Xx - video game(x)] Vz (Agent(e,z) ~ Xz) & teach(e) & [every y : quarterback(y)] [3e'r e'^e] (Vz (Goal(e',z) - z=y) & [3Z: 2(Z) & Vz(Zz - new play(z)] Vz (Theme(e',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 first event and where the quarterback is the goal of that second event and where two new plays are the theme." For details and arguments why no polyadic analysis of verbs will succeed in capturing the relevant reading, the interested reader is referred to Schein (1993:57ff). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 form their own conjuncts, what are traditionally considered the arguments of the verb should also be interpreted as separate conjuncts. The analysis of focus that will be developed in the next two chapters will make extensive use of this kind of event-semantics. Hopefully, it will also provide a further argument showing that the traditional polyadic analysis of verbs is simple-minded and that investing in a somewhat more complex Davidsonian decomposition has various empirical pay-offs which make the initial investment well worth-it. Finally, one thing worth pointing out in connection with separation is this. To make sure that our logical forms get the truth-conditions right, theta-roles have to seen as being assigned exhaustively (cf. e.g. Larson and Segal (1995: 485). (12) for instance receives the logical form (13): (12) Romeo kissed Juliet. (13) 3e (Agent(e, Romeo) & kiss(e) & Past(e) & Theme(e, Juliet) We want this logical form to insure that Romeo was the one kissing, on the one hand, and that Juliet, on the other, was the one being kissed by him. The worry is that if nothing else is said, (13) does not deliver this result. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 Concentrating on Roneo, it only claims that there was a kissing and Romeo was an agent in it. But this could be true if Romeo paid someone else to kiss Juliet and he himself did not kiss her. Clearly though, 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 its all and only bearers. Now that Romeo is the sole agent in the kissing event, it correctly follows that he— and not somebody else— kissed Juliet. 1.2. Overall structure of the dissertation: Chapter 2 introduces certain basic facts about focus and discusses how focus influences the presuppositions of a sentence. The aim of this chapter is to show that we can account for the distribution of focal presuppositions rather simply if we assume that focus structures or restricts the type of event quantification that has been introduced here. Those cases which lack focal presuppositions (and which have led various authors to conclude that focus has no serious presuppositional effect, but is pragmatic instead) will be reduced to a scope interaction between the event quantifier Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 and other elements. In an appendix to this chapter, I discuss the pragmatics of intonation and its interaction with negation and contrastiveness. Chapter 3 applies the analysis introduced in chapter 2 to sentences which overtly quantify over events, namely sentences containing adverbs of quantification like usually, often, never, etc. An intriguing phenomenon that is discussed in this context is the Quantificational Variability Effect. Finally, the analysis which treats adverbial quantifiers as selective quantifiers over events is compared with a proposal that treats adverbs as quantifiers over situations. It is argued that in spite of their important similarities the two analyses do not always make the same empirical predictions. It is generally assumed that what makes adverbs of quantification so susceptible to the effects of focus is that the syntax does not mark which part of the sentence has to contribute to the restriction of the adverb. In this respect adverbs are held to fundamentally differ from nominal quantifiers (determiners), where the syntax tells us that they have to be restricted by the noun phrase complement of the determiner. Chapter 4 shows that certain determiners in fact behave like adverbs, giving rise to "focus-affected" readings. Strikingly, focus-affected readings are only possible with weak noun phrases. The particular semantic properties exhibited by these readings Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. are shown to have consequences for the analysis of the Definiteness Effect. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 2: FOCUS AND PRESUPPOSITIONS 21 2.0. Introduction: It would make pragmatic sense to utter (la) in a stretch of discourse where we are discussing who wrote poetry. On the other hand, it is clear that it would be less felicitous to use this sentence if what were at issue was the kind of writing Rosalia engaged in. In that case, it would be appropriate to use the minimally different (lb): (1) a. ROSALIA wrote poetry. b. Rosalia wrote POETRY. Intuitively, the reason (la) and (lb) are felicitous in different discourse environments is that they presuppose different things: whereas (la) presupposes that someone wrote poetry, (lb) presupposes something different, namely that Rosalia did some kind of writing. This chapter deals with the phenomenon that is illustrated in the contrast between (la) and (lb). The claim put forth is twofold: first, focus has a genuine presuppositional effect (this has been controversial) . Second, the presuppositional effect of focus is captured if we analyze focus as imposing a certain amount of structure on the event quantification that was introduced in chapter 1. The resulting Structured Davidsonian Decomposition (SDD) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 22 offers a simple, scope-based analysis of the distribution of focal presuppositions, treating them directly as logical presuppositions. It is not denied that focus has pragmatic consequences. These, however, are argued to be derivative on the semantic effects of focus. I would like to make two preliminary points here to delimit the scope of the discussion in this chapter and to avoid possible misunderstandings. First, what is meant by "focal presuppositions" are those presuppositions whioh are due to focus, for instance, the presuppositions that we encountered in connection with (la) and (lb). As pointed out in Jackendoff (1972), focal presuppositions need to be distinguished 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 and in what follows I will have nothing to say about them.* -Failure to keep apart focal presuppositions from lexical presuppositions has led some authors to reject any analyses of focus which aim to account for its presuppositional effect (e.g. Rochemont 1986). This criticism seems unfair to me. Thus, although the complement of know "presupposes" truth it can still be focused: (i) John knows THAT MARY SPEAKS ARABIC. But this is not a problem. All it shows is that focal presupposition are something that operates over and above the lexical presuppositions of a verb like know. In particular, the focally non-presupposed part of (i), namely the that-clause, can contain "lexically-presupposed" material without there arising any conflict. Thus, lexical Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9-5 The second point is that I will concentrate on the meaning effects of focus, leaving as a separate and undiscussed issue its phonological realization through stress. Thus, throughout, capital letters are used to mark the semantic domain of focus, rather than the main stress. Even though my concern lies not with stress, the following fact is helpful to be aware of in order to evaluate the data: while focus is marked through stress, there is not always a one-to-one correlation between a word carrying the main stress of the sentence and it being the focus. This is so because in certain cases a single stress assignment can in fact signal a number of foci. Thus, in English focus on any given phrase is generally realized as increased stress on its right-most and most deeply embedded word (of. Chomsky 1971, Cinque 1993). This, however, has the consequence that stress on shirt in (2), for instance, can be interpreted as giving rise to all the different foci that are represented in (3a)-(3e) since in all of the relevant phrases stress falls on the right-most and most deeply embedded word of the phrase: (2) The police arrested an ex-convict with a red shirt. presuppositions and the presuppositions that are due to focus need to be kept apart. Something can be lexically presupposed but not form part of the focal presupposition of the sentence (of. also Jackendoff 1972). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 24 (3) a. The police arrested an ex-convict with a red SHIRT. b. The police arrested an ex-convict with A RED SHIRT. c. The police arrested AN EX-CONVICT WITH A RED SHIRT. d. The police ARRESTED AN EX-CONVICT WITH A RED SHIRT. e. THE POLICE ARRESTED AN EX-CONVICT WITH A RED SHIRT. In order to evaluate the data that are presented here it will be helpful to keep in mind this relation between stress and focus. For more detailed discussion of how focus is phonologically realized and how this relates to syntactic structure, I would like to refer the reader to e.g. Selkirk (1985, 1995), Cinque (1993), and Zubizarreta (1996). With these two preliminary points made, let us now turn to the way focus affects meaning. 2.1. Focus and Discourse: 2.1.1. An analysis à la Jackendoff (1972) : Among the various 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. Conversely, the absence of focus is held to signal that the information encoded by the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 ncn-fccused material is "under discussion." This characterization of focus is prominent in Jackendoff's (1972) discussion and it also underlies the "alternative semantics" approach to focus (cf. Rooth 1985, 1992, von Fintel 1994 a.o.). Since Jackendoff's analysis can be considered to be the mother of subsequent analyses, which it has influenced in a number of crucial aspects, this analysis seems a particularly good place to start.- Concretely, Jackendoff's analysis works like this. Holding that focus serves to introduce alternatives, the focused element Rosalia in (la), for example, is predicted to contrast with other people salient in the discourse of whom we might have said that they wrote poetry (Emily, William, Federico, etc.) To capture this contrastive effect of focus, it is proposed that the focused element is abstracted over and replaced by a variable that matches it in semantic type, thus creating a "p-set": (4) Ax (x wrote poetry) p-set of (la) This set-abstraction also provides the mechanism generating the "focus-semantic" value for the sentence, which in the -The view that the non-focused part has a topic-like function is also inherent in certain pragmatic or information theoretic treatments of focus. Following work by E. Prince, Vallduvi assumes that the non-focused part of the sentence expresses not an existential presupposition but "shared knowledge," which constitutes information that the speaker assumes the hearer believes, see Vallduvi 1990. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 26 alternative semantics account enists alongside with the regular semantic value, and which is used to account for various effects of "association with focus" (see below). Originally, Jackendoff considered a somewhat different claim, namely the possibility that the type-matching variable that replaces the focus is not bound by a lambda- operator, as in (4), but by an existential quantifier instead: (5) 3x (x wrote poetry) =existential presupposition of (la) (5) actually captures a focal presupposition for (la) ("Someone wrote poetry"), something which (4) does not do. Nonetheless, (5) was ultimately discarded and (4) was adopted instead. To see the reasoning that motivated the adoption of (4), we can first note that examples like those in (1) suggest the generalization in (6), where focus splits the sentence into presupposition and assertion (Chomsky 1971) : (6) The non-focused part expresses an existential presupposition, the focused part expresses the assertion. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 (6) describes the effect of focus in (1) well enough; and it is this generalization which the analysis sketched in (5) aims to account for. There is a problem with (6), however. As Jackendoff observes the generalization in (6) is not fully general. For instance, it is falsified by examples like (7) (Jackendoff 1972:246): (7) NOBODY likes Bill. Contrary to what (5) would lead us to expect, (7) does not presuppose that somebody likes Bill, only to then go on to assert that it was nobody. If it did, (7) would be contradictory, but the sentence is perfectly sensible. In light of this, Jackendoff concludes that (6) is too strong and that it needs to be replaced by the weaker generalization in (8): (8) The non-focused part of a sentence expresses the property that is "under discussion." More concretely, making use of the p-set that is generated by abstracting over the focused element, Jackendoff argues for (9). The variable x here stands in for the focus, presupp stands for the non-focused material of the sentence: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 28 (9) is a coherent set is well-defined in the present discourse Ax Presupp. (x) is amenable to discussion is under discussion Even though the non-focused part is coded as "presupp," the purpose of (9) is to relieve the non-focused part of any real presuppositional work. Whether something instantiates the property that is "under discussion" or not, is crucially left open; after all, the intent of (9) is to encompass both cases where focus in fact denotes an existential presupposition, as in (1), as well as examples like (7), where no focal presupposition is present. 2.1.2. "Under discussion" is too weak: We can observe now that while (9) makes no false predictions, it faces certain empirical limitations, suggesting that we should explore a different way of analyzing focus. To begin with, (9) amounts to a generalization to the worst case. By being so general that it can encompass boti (7) (where no focal presupposition is present) as well as the examples in (1) (which indeed do have focal Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 presuppositions), (9) offers no explanation of the presuppositional difference we find between these two kinds of examples. In addition, it turns out that— despite Jackendoff's conclusion to the contrary— (7) does not provide an argument in favor of (9). If we take the claim that the focus is replaced by a type-matching variable on its face-value, then by existentially binding this variable as in (5) it would actually be possible to account for the existence of a focal presupposition in (1) and the absence thereof in (7);- since what is focused in (7) is a quantifier (nobody), a type- matching variable replacing it would be a variable over quantifiers (rather than a variable over individuals). Consequently, existentially binding this variable will not have the unwelcome consequence of generating the presupposition that someone likes Bill. Instead, it will only give us the "presupposition," if we can call it that way, that some quantifier applies to "likes(x,b)." Since the quantifier in question might be decreasing cf. nobody,, few, if any, of bis colleagues, etc, this does not entail that someone likes Bill.’ Thus, (7) provides no motivation for 'Thanks to Paul Fortner for raising this important point. ’A quantifier is monotone decreasing (in its restriction) if it licenses an inference from superset to subset, such that [Q A] B entails [Q C] B where C is a subset of A. A quantifier introduced by no, for instance, is decreasing because it licenses the inference from (i) , which Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 (9), and instead we could keep the view underlying (5), which is summarized here in (10): (10) The presuppositional effect of focus is captured by existentially binding the type-matching variable that replaces the focus. Clearly, (10) has more empirical bite than (9) because, unlike (9), (10) captures that the examples in (1) have focal presuppositions, and that (7), on the other hand, lacks such a presupposition. This suggests then that rather than assuming that the non-focused part of a sentence merely expresses what is "under discussion" (cf. (9)), we should keep the claim that the non-focused part can express a focal presupposition, deriving focal presuppositions by existentially closing off the variable that replaces the focus (cf. (10)). An analysis like (10) is contemplated in Rooth (1994a) (see section 2.3.4). Yet, although (10) is empirically stronger than (9), ir is not strong enough. As a next step in the argument we can observe that looking at a wider range of examples than have been considered so far, it turns out that (10) does not quantifies over men, to (ii), which quantifies only over a subset of the men, namely the tall ones: (i) No man came. (ii) No tall men came. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 predict the distribution of focal presuppositions in all the cases. Crucially, (10) suggests that (7) lacks a focal presupposition because the focus is on a quantifier, rather than on a referring expression. Since the variable that replaces the focus is that of a quantifier, existentially binding this quantifier only derives a "trivial" presupposition. It turns out, however, that the relevant factor is not so much that focus is on a quantifier, but that the focus is on a decreasing quantifier. In other words, the semantic nature of the focused quantifier matters, something which (10) cannot capture. Compare (7) with (11), for instance, where, unlike in (7), the focus is on a non-decreasing quantifier: (11) MANY OF HIS COLLEAGUES like Bill. (11) can be smoothly used in a context where it counts as previously established that Bill is liked. In contrast, this cannot be said of (7); if we use (7) in this kind of context, we create a conflict with what was said before. This contrast between (7) and (11) indicates that while (7) does not carry the focal presupposition that someone likes Bill, (11) in fact does. So contrary to what the analysis in (10) predicts, the nature of the focused quantifier matters, in particular it matters whether it is decreasing as in (7) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. or non-decreasing as in (11); only when focus is on a decreasing quantifier like nobody is there no focal presupposition, when focus falls on a non-decreasing quantifier {many of his colleagues], a focal presuppositions resurfaces again. Thus, although (10) successfully distinguishes between (1) and (7), it lumps (7) together with (11), which does not seem right. 2.2. Structured Davidsonian Decomposition: 2.2.1. The basic idea: At this point, I would like to introduce a different analysis of the distribution of focal presuppositions, an analysis which differentiates not only between (1) and (7), but also between (7) and (11). 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 quantification : (12) All the non-focused material forms the restriction to an existential event operator. The focused material forms (or binds into) the event quantifier's second argument, (to be revised) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 Along with the event quantifier which is restricted by the non-focused material there may be other event quantifiers whose quantificational structure is not directly shaped by focus. In the case of (1), the Structured Davidsonian Decomposition (SDD) in (12) will give us a logical forms that are shown in (13): (1) a. ROSALIA, wrote poetry, b. Rosalia wrote POETRY. (13) a. [3e: 0(e) & write(e) & Past(e) & Theme(e, poetry)] Agent(e, ROSALIA) b. [3e: C(e) & Agent(e, rosalia) & write(e) & Past(e)] Theme(e, POETRY) The meaning that this derives for (la) is as follows :"Some relevant event of writing poetry in the past was such that its agent was Rosalia." In contrast, (lb) is interpreted along these lines: "Some relevant of writing whose Agent was Rosalia was such that the theme of that writing was poetry." Regarding C(e), following standard views, every quantifier is assumed to come with a context predicate C whose value is fixed by the context of utterance. In the case of the event operator, this kind of context predicate Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 ensures that we are only looking at those events that are relevant in the context in which the sentence is used. Thus, the context predicate C(e) allows the speaker/hearer to relate the event described in the sentence with the previous discourse. 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 logical form 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. Given the SDD, the presuppositional effect of focus can now be directly attributed to the existential import of the event quantifier and to the asymmetric status that we find independently between the restriction of a quantifier and its matrix; even though both the restriction and the matrix of 3e have existential import, at the very core of restricted quantification lies the fact that the restriction of a quantifier has a scene-setting status which its matrix or second argument lacks. Exploiting this asymmetry, in the case of (la), for example, the SDD derives the focal presupposition that some event of writing poetry took place. In contrast, (lb) is predicted to presuppose that there was some event of Rosalia writing. To say that (la) presupposes that some event of poetry writing took place is not exactly the same as to say that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 someone wrote poetry— the description of focal presupposition that was given initially. But it is close enough. Concretely, I would like to suggest that the existential commitment that someone wrote poetry need not follow from the logical form itself, but can instead be attributed to our understanding of the nature of poetry writing; the way the world is, events of poetry writings typically require agents. As a result, even though the logical form of a sentence only tells us that there was an event of poetry writing, we can still infer that someone wrote poetry. This inference is not semantic in the strict sense but due to our knowledge of the world because we know that whenever something gets written, there is an agent. Clearly, if we would conceive of the writing differently, then, leaving language the same, this inference would not follow. Independent support for the "weak" way of semantically characterizing focal presuppositions that is offered by the SDD will be given below in section 2.2.4. So far, then, it has been proposed that event quantification is not always unrestricted, as assumed in chapter 1. Rather, it can be restricted by the Davidsonian conjuncts that contain the non-focused material. The focal presuppositions in examples like (1) follow from the combination of the existential import of the restricted event quantifier and its special status relative to the quantifier's matrix. It is because focus affects the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 quantificational structure of a sentence that it has an effect on the felicity of the sentence in a given discourse. Thus, a sentence with the (focal) presupposition p can only be felicitously used in a given context if p is either already established (Common Ground) or if it can be "accommodated" without too much trouble (Stalnaker 1978). But what kinds of focal presuppositions p a sentence has is not a pragmatic matter, but it follows directly from its logical form, where the logical form is shaped by the SDD. In essence then, the overall claim is this: focus is a semantic (logical form) phenomenon which has pragmatic consequences. 2.2.2. Wide scope decreasing c[uantifiers : Having seen how the SDD deals with the examples in (1), what about (7), the example that Jackendoff observed not to have any focal presupposition? (7) NOBODY likes Bill. On the present account, the absence of a focal presupposition in (7) follows from scope: if the decreasing quantifier nobody takes scope over the event operator, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 existential event quantifier need not pick out any event at all and the sentence can still be true.' (14) [NOBODY X] [3e: C(e) & like(e) & Theme(e. Bill))] Experiencer(e,x) "Nobody is such that at some (relevant) event of liking Bill had him or her as its experiencer." It is because the wide scope of the decreasing quantifier in (14) deprives the event operator of its existential import that (7) has no focal presupposition. By the same token, although what is focused in (11) is a quantifier, the sentence is correctly predicted to presuppose that Bill is liked. The reason is that the focused quantifier that takes scope over the event is non-decreasing, leaving the existential import of 3e intact: (11) MANY OF HIS COLLEAGUES like Bill. (15) [MANY X: OF HIS COLLEAGUES(x)] [3e: C(e) & like(e) & Theme(e. Bill)] Experiencer(e,x) In (14) everything gets interpreted within the scope of nobodyf robbing the sentence of a focal presupposition. 'Thus, in relevant respects this analysis of focal presuppositions parallels the Russellian analysis of definite descriotions (see section 2.3.). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 Nonetheless, the non-focused and the focused part still behave differently in that the non-focused part restricts the event of focus whereas the focused part enters only into the matrix. This preserves the interpretive asymmetry between focus and non-focus that we still find in (7), despite the fact that the sentence does not carry a focal presupposition. 2.2.3. Narrow scope decreasing quantifiers: Nothing in what has been said so far forces a decreasing quantifier to take wide scope over the event operator. The question is now whether there are any cases where we can see that a decreasing quantifier in fact takes narrow scope. By the logic of our analysis of (7), we expect that under such a scope constellation focal presuppositions will resurface; since (7) lacks a focal presupposition precisely because the decreasing quantifier nobody takes wide scope over the event, when an analogous decreasing quantifier takes narrow scope the sentence should have a focal presupposition. 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"? Given that this verges on the incoherent, it is hard to tell because Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 the coherent wide-scope interpretation of nobody will always be in the foreground. A better example may be this one: (16) John sang a song. But he sang it FOR NOBODY. In accordance with what the first sentence states, we can say that the second sentence presupposes that an event where John sang a song took place. Of this event it is asserted that it was not performed for anybody's benefit. The focused dative here being genuinely optional (of. chapter 1), we can argue that in (16) the negative quantifier takes narrow scope with respect to the event operator. As a result, the sentence is correctly predicted to have the focal presupposition that John sang a song. That (7) and (16) should differ in terms of their focal presuppositions is expected under an analysis where focus structures quantification over events. Since focal presuppositions directly derive from a restricted event operator with existential import, their distribution is sensitive to the scope which that operator takes, so that when the event quantifier occurs within the scope of a decreasing element, no focal presuppositions are entailed, and, conversely, when it takes wide scope over any negative element, a focal presupposition will resurface. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 0 There is a problem with the example in (16), however. One could argue that in the second sentence the negative quantifier in fact takes scope over the entire sentence, including the event operator, and that the reason we think that this sentence presupposes that there was a singing has only to do with the content of the first sentence, which explicitly states that a song was sung. There are examples which are immune to this kind of objection. In these examples a decreasing quantifier unmistakably takes narrow scope and, in line with our prediction, these examples do indeed exhibit focal presupposition. Particularly clear cases of this can be found in Spanish because in this language the narrow scope of a negative quantifier can be read off from the surface syntax of the sentence. Showing this will require a little detour through the quantifier system of Spanish, which will be undertaken next. 2.2.4. Postverbal Spanish n-words as a diagnostic: It is generally assumed that when so-called n-words ir 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 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 postverbal position (cf. e.g. Laka 1990, Zanuttini 1991, Ladusaw 1993, Uribe-Etxebarria 1994): (17) a. Nadie vino, n-body came b. No vino nadie. Mot came n-body c. +Vino nadie. came n-body "Nobody came." (18) a. Yo nunca habia estado en Côrdoba. I n-ever had been to Côrdoba b. Yo no habia estado en Côrdaba nunca. I not had been to Côrdoba n-ever c. *Yo habia estado en Côrdoba nunca. I had been to Côrdaba n-ever "I had never been to Côrdoba." Descriptively, the standard paradigm illustrated in (17) and (18) suggests that preverbal n-words pattern with negative quantifiers {nobody, never etc.) whereas postverbal n-words correspond to NPIs {anybody, ever, etc.). There has been an extensive debate in the literature about whether n- words should be considered as being fundamentally negative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 42 quantifiers (e.g. Zanuttini 1991) or whether they should be treated as underlying NPIs (e.g. Laka 1990). The issue is not important here.' What is important is that the standard paradigm is in fact incomplete: under some conditions it is actually possible to find postverbal n-words that are not c-commanded by any negative element. As originally pointed out by Zanuttini, this is possible when the n-word is interpreted as a negative quantifier that takes narrow scope with respect to the event described by the verb. Cf. for instance (19): (19) Mariana se fué con nada. Mariana left with nothing. Since nada in (19) can be interpreted as a (narrow scope) negative quantifier, when (19) 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 quantifier) : - The matter is discussed in Ladusaw (1993), see also Herburger (1996b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 3 (20) Mariana no se fué con nada. Mariana didn't leave with anything.(simple negation) Mariana didn't leave with nothing, (i.e. She left with something.) (double negation) In the examples discussed so far, the postverbal negative quantifiers occur either in adjuncts. However, these quantifiers are not limited to such environments. (21) provides an instance where the postverbal negative quantifier is an indirect object: (21) No se movia ni una brizna de hierba, ni una triste hoja. Not a strand of grass moved, not a sad leave. Todo era tan tierno que no tenia bastantes ojos para mirar. Everything was so touching that I didn't have eyes enough to see. Al final, 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 mio. said-I softly to n-body that everything was mine I said softly to nobody that everything was mine. (from "Parecia de seda" by Mercè Rodoreda, translation and emphasis mine) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 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 specification. As (21) illustrates, however, this is not necessarily so; the last sentence states not only that the narrator of the story said something, namely that everything was hers, but that this event of saying was not directed towards anyone. In this context, consider also the contrast in (22), which makes use of the difference between mirar a vs. ver: (22) El pobre Juan se ha vuelto loco. Se pasa los dias 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). Clearly, though, it is not possible to literally see and to see nothing." This difference between see nothing and look at nothing explains the contrast that is illustrated in (22); a postverbal n-word without a licensor is only then acceptable 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. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 5 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). Finally, an example that brings out the narrow scope properties of postverbal negative quantifiers in Spanish very nicely is (23):' (23) Hemos leido en alguna parte que la meditacion para los budistas "zen" "no consiste en 'no pensar en nada' sino en 'pensar en nada'". We have read somewhere that for Zen Buddhist 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 (which would be'expressed by no pensar en nada, where the negation takes wide scope over the event operator). Rather, the sentence asserts that the mediation of Zen Buddhists involves a particular kind of thinking, a thinking which is not directed towards anything. This type of thinking is described by pensar en nada. Here, again, the negative quantifier clearly takes narrow scope with respect to the event described by the sentence. 'This if from the text discussion in Bosque (1980:41; Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 6 The point of this excursus into the Spanish quantifier system is that it gives us examples where we can tell directly from the surface syntax of the sentence whether a negative quantifier 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.' Going back to the main thread of the argument, the Spanish facts allow us to test the prediction that if a negative quantifier takes narrow scope, the sentence has a focal presupposition. The prediction is indeed born out: (24) presupposes that Mariana left, (25) that the person denoted by the subject said that everything was hers, and (26) that Juan spends his days watching. Last but not least, (27) carries the focal presupposition that for Zen Buddhists meditation consists in thinking: (24) Mariana se fué CON NADA. focal presupposition: "Mariana left." assertion: "It was with nothing." (25) ..dije bajito A NADIE que todo era mio. focal presupposition: "I said that everything was mine." assertion: "It was said to noone." 'English is much less helpful in this respect because to the extent that speakers like postverbal negative quantifiers (many find it rather stilted except with have, cf. Zanuttini 1991), they freely assign them wide scope. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 (26) ...se pasa les dias mirando A NADA. focal presupposition: "He spends his days watching." assertion: "It is nothing." (27) ...la meditacion para los budistas "zen"[...] consiste[...] en pensar EN NADA. focal presupposition: "For Zen Buddhists meditation consists of thinking." assertion: "It is thinking of nothing." The logical forms generated by the SDD derive these focal presuppositions straightforwardly. Unlike in the logical form of (7) (=(13)), in (28) and (29) the decreasing quantifiers are interpreted inside the scope of the event operator. Consequently, the event operator that is restricted by the non-focused material has existential import and the focal presuppositions of (24) and (25) follow as a matter of logical entailment: (28) [3e: C(e) & leave(e) & Past(e) & Agent(e, Mariana)] [NOTHING x] WITH(e,x) "Some (relevant) event of leaving by Mariana was a leaving with nothing." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 (29) [3e: C(e) & say(e) & Past(e) & Agent(e, I) & Theme(e, that everything was mine)] [NOBODY x] Goal(e,x) "Some (relevant) event of me saying that everything was mine had nobody as its goal."- Earlier it was noted that the focal presuppositions that are derived by the SDD are somewhat weaker than those that are traditionally assumed. Thus, ROSALIA wrote poetry is predicted to presuppose that an event of poetry writing took place, not that someone wrote poetry. I further suggested that the deficit of the focal presuppositions that are derived by the SDD is made up by our world knowledge; since we know that events of poetry writing have an agent, we can infer that if there was poetry writing then there also was somebody writing it. -There are particular instances where the focus falls on a postverbal negative quantifier and where the sentence actually lacks a focal presupposition because the negative quantifier is interpreted as having wide scope. These are cases of what I would like to refer as "mimicking focus": (i) A: lA donde vas? "Where are you going?" B: Voy exactamente A NINGUNA PARTE. Go-I exactly TO NO PLACE B's answer to A's question is understood to be denying the presupposition that B is going anywhere. Moreover, the non focused part of the question is repeated word-by-word in the answer, in a sense, mimicked. (The answer has a decidedly unfriendly edge which would be absent it had been De hecho, no voy a NINGUNA parte "Actually, I'm not going anywhere".) Since the occurrence of such sentences is limited to very specific discourse environments, it can be safely be set aside as an independent phenomenon. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 (25) provides support for the claim that focal presuppositions have to be as weak as the SDD claims they are. As the SDD predicts, (25) presupposes that in an event of saying, something got said by the individual that is 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 focal presupposition included that something was said to someone, the sentence would be contradictory; something was said to someone and it was to noone. The fact that the sentence is not contradictory shows then that the weak focal presuppositions that are derived by the SDD, namely that there was an event of saying by the subject, are exactly right and anything stronger would get the meaning wrong. While the focal presuppositions of the examples in (24)-(27) are straightforwardly predicted on the SDD account, they are not expected on the second version of the Jackendoff-style analysis stated in (10) (see section 2.1.2). As we saw, on this proposal focal presuppositions are derived by existentially binding the type-matching variable that is said to replace the focus. It was already noted that this kind of analysis fails to tell apart the behavior of (wide-scope) decreasing and non-decreasing Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 focused quantifiers; it does not capture the fact that the • former undermine focal presuppositions, but that the latter do not. We can observe now that it is also unclear that an analysis like (10) is sensitive to the scope of a focused decreasing quantifier and that it can capture that it matters whether the quantifier takes wide scope over the event or narrow scope; presumably, in both cases the quantifier is replaced by a type-matching variable q, which is then existentially bound, giving us a "presupposition" 3q q(F), regardless of the scope of the quantifier. On the other hand, since the SDD analysis derives focal presuppositions directly from the existential import of the restricted event operator, it correctly captures that when a focused decreasing quantifier takes wide scope there will be no focal presupposition, and conversely, that when a focused decreasing quantifier takes narrow scope a focal presupposition will resurface. A last point that is worth noting is this. While I have concentrated on the behavior of quantifiers that are focused— because I was presenting the SDD analysis against the background of the Jackendoff-style analysis— it does in fact not matter for the distribution of focal presuppositions whether a quantifier is focused or not. Thus, regardless of whether the quantifier is focused or not, as long as a decreasing quantifier takes scope over the event operator, the sentence is predicted to lack a focal Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 presupposition. This is indeed what we find. For instance, when we shift the focus in (7) away from the quantifier to the Bill, for instance, the sentence still lacks a focal presupposition, as is shown by the felicity of (30) : (30) Nobody likes BILL. In fact, nobody likes ANYONE. IT'S A LOVELESS WORLD! The lack of a focal presupposition of Nobody likes BILL follows since it is interpreted along the lines of "Nobody is such that some relevant event of involving him or her and Bill is a liking event." Thus, although in addressing t have concentrated on how focused quantifiers affect focal presupposition, the SDD analysis captures that non-focused quantifiers have analogous effects. On the other hand, this is not the case with the analysis in (10). Since the way this analysis is set up it is only concerned with the effects that focused quantifiers have on focal presuppositions, it offers no predictions concerning the effects of non-focused quantifiers.- "It is sometimes noted that the scope of a quantifier depends in some way on intonation and that different stress- assignments can affect the scope order between various quantifiers. It seems difficult, however, to pin down the correct generalizations. Remaining agnostic as to the relevance of focus for scope, all that is claimed here is this: a focused quantifier has to bind into the matrix of the event operator. If the quantifier is decreasing and ic scopes over 3e, the sentence lacks a focal presupposition, and conversely, if the event quantifier takes wide scope (or the quantifier is non-decreasing) a focal presupposition Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 In summary, what has been argued in this chapter sc far is that it is empirically not enough to say that the non- focused part merely expresses what is "under discussion" at the given point in the discourse where the sentence occurs. Rather, focus can have a genuine presuppositional effect. The first possibility that was considered for accounting for these effects was by existentially binding a type-matching variable that replaces the focus. We saw, however, that this way of doing things does not give as fine-grained results as an alternative view where the effects of focus reduce to the non-focused material restricting a quantifier over events (Structured Davidsonian Decomposition). Only on this second view does it follow that it matters for the presence of focal presupption whether a quantifier is decreasing or increasing and furthermore that it matters in the case of a decreasing quantifier whether the quantifier takes wide scope or narrow scope relative to the event described by the verb. Focal presuppsoitions are then systematically absent when a decreasing quantifier (focused or not) takes wide scope over the restriction of the event quantifier. resurfaces. The same holds for non-focused quantifiers. I have no insights to offer at this point as to how different stress patterns might affect quantifier scope. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 2.3. Focus and negation: 2.3.1. Different readings: bound, free, and wide I would now like to extend the empirical scope of the SDD, while sticking with negative contexts, which are the most interesting for my purposes. The overall claim that is explored in this and the following sections is that the SDD offers a simple account of the interaction between focus and negation. On this account, negation can remain univocally a prepositional operator. Thus, whether not effectively negates a verb, a quantifier, or an entire sentence e.g., it always functions the same way. Moreover, negation also remains univocal in the sense that no lexical ambiguity is posited between a "regular" not which lets presuppositions project and a presupposition-denying so-called metalinguistic negation (Horn 1989) . Rather, the entailments of presuppositions— and the lack thereof— follow directly from the scope of the negation. One basic observation is this. As many linguists have noted, there is more than one way to interpret a sentence like (31) (e.g. Jackendoff 1972, Jacobs 1991): (31) Sascha didn't visit THE EIFFEL TOWER yesterday. On its most salient interpretation we understand (31) along the lines of (32). Here not negates the focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 letting the rest of the sentence escape from its effect. I will refer to this as the bound reading. (32) "What Sascha visited yesterday wasn't the Eiffel Tower." bound reading Along with the bound reading, (31) also has the "free reading" paraphrased in (33). Intuitively, in this reading the effect of the negation is restricted to the verb: (33) "What Sascha didn't visit yesterday was the Eiffel Tower." free reading The two readings are quite different and they arise in different pragmatic contexts. The bound reading appears when we are presupposing that Sascha engaged in some site seeing and we are wondering what in particular he saw. The free reading becomes readily available if we imagine uttering (31) in a different 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 yesterday. There is also a simple way to force the two different readings without construing pragmatic context, namely adding different continuations to the sentence. While continuing (31) with but... forces the bound reading, adding a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 continuation beginning with and not, but not... forces the free reading: (34) a. Sascha didn't visit THE EIFFEL TOWER yesterday, but THE LOUVRE. bound reading b. Sascha didn't visit THE EIFFEL TOWER yesterday, and not THE LOUVRE. (He DID in fact visit the Louvre.) free reading Curiously, intonation also helps tease apart the two readings (Jackendoff 1972) . Thus, a "fall-rise" contour, where the sentence melody falls at the end, typically signals a bound reading. A "fall" contour, on the other hand, forces the free reading. (Why this should be so will be investigated in the appendix to this chapter.) It is clearly inviting to make the difference between bound, free and wide reading follow directly from the scope of the negation (cf. HajiCovâ 1984, Jacobs 1991). The point now is that a scopal account of the various different readings is particularly easy to obtain with the SDD; because of the Davidsonian decomposition, the various readings reduce to scope-interactions where negation can remain a prepositional operator regardless of the semantic nature of the element that is effectively negated (e.g. visit vs. the Eiffel Tower) . Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 Beginning with the bound reading, this interpretation results when the negation effectively takes scope over the conjunct containing the focus: (35) [3e: C(e) & visit(e) & Agent(e, Sascha) & Past(e) & yesterday(e)] NOT Theme (e, THE EIFFEL TOWER) bound reading "Some (relevant) visit by Sascha yesterday did not have the Eiffel Tower as its theme." Since in (35) the negation takes scope only over the matrix, but not over the restriction, it follows by the SDD that (31) on its bound reading will presuppose that Sasoha did some visiting yesterday. This correctly predicts that the focal presupposition of the bound reading of (31) is the same as that of the non-negated counterpart of (31) [Sascha visited THE EIFFEL TOWER yesterday.) Whereas in the bound reading the negation takes soope over the foous, the free reading arises when the negation only takes scope over the verbal conjunct. As a result of this scope constellation, we have a negative description of an event as a "not visiting." The focal presupposition of this reading, namely some not visiting by Sascha took place, is clearly different from that of the bound reading (some visiting by Sascha took plaoe). This in turn explains why Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 the two readings are found in different discourse environments. * - (36) [3e: C(e) & NOT visit (e) & Agent(e, Sascha) & Past(e) & yesterday(e)] Theme(e, THE EIFFEL TOWER) free reading "Some (relevant) event of not visiting by Sascha yesterday had as its theme the Eiffel Tower." Thus, as (35) and (36) illustrate, the difference between the bound and the free reading is one of scope in the logical form: in the bound reading the negation takes scope over the focus, in the free reading it takes scope over the verb. Crucially, though, since the verb and the focus both "It may seem that negative descriptions such as "an event of not visiting" are vacuous in that they are true of a large number of things (e.g. skiing, reading, etc.). As the following examples show, however, we are typically, not so literal-minded as to take negative descriptions as holding vacuously true of a huge number of irrelevant things. Rather, the context of utterance and the desire to make sense out of an utterance is sufficiently strong to make the descriptions sensible: (i) Once, noone knows when, nobody arrived. (Schein 1993) (ii) Nothing I never said ever got me into any trouble, (attributed to Calvin Coolidge) The number of unknown times where nobody arrived and the things that never got said are, when taken literally, immensely large. Still (i) and (ii) can be used sensibly; (i) is not automatically true if a dog barks, nor is (ii) automatically false some picture got Coolidge into trouble. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 form conjuncts in the logical form, negation remains a propositional operator in both cases. Support for making the contrast between the bound and the free reading 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 the Eiffel Tower follows the negation only a bound reading is possible, and a but continuation seems necessary, cf. (37). Conversely, if the focused noun phrase precedes the negation, as in (38), the sentence only has the free reading: (37) daI5 der Sascha gestern nicht DEN EIFFELTURM besichtigt hat, sondern DEN LOUVRE/#und nicht DEN LOUVRE. only bound reading that Sascha yesterday not the Eiffel Tower visited has, but the Louvre/#and not the Louvre. (38) daB der Sascha gestern DEN EIFFELTURM nicht besichtigt hat, und nicht DEN LOUVRE/#sondern DEN LOUVRE. only free reading that Sascha yesterday the Eiffel Tower not visited has, and not the Louvre/#but the Louvre. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 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) below 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:*' (39) a. Nem MARIAT veri Peter. 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." -^Thanks to Anna Szabolcsi, Ricardo Echepare, Gorka Elordieta and Itziar Laka for both data and judgments. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 60 Given that the overt syntax of Hungarian and Basque is more revealing than that of English, it is not surprising that the difference between bound and free reading in these languages should be manifested in word order if the difference is indeed scopal, as the SDD account suggests.-" Next, along with the bound and the free reading that are observed in the literature (e.g. Jackendoff 1972), there also seems to be a third way in which (31) can be interpreted. 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 effect of the negation is not restricted to the focus here, but extends across the entire sentence: (41) "It is not the case that what Sascha visited yesterday was the Eiffel Tower." The structured wide reading can be found in a scenario such as (42), for example: ^Strictly speaking, the evidence from German, Hungarian, and Basque only supports the claim that scope matters in disambiguating the various reading. It does not necessarily show that in English the relevant scope relations are expressed at LF (although they may, of course, particularly if some independent evidence for an LF process can be found). Thus, I am assuming that although LF constellation is the paradigmatic way to mark scope there may in principle be other scope disambiguating mechanisms feeding the semantics (logical form). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 61 (42) A: Yesterday, Sascha visited THE EIFFEL TOWER. B: No, Sascha didn't visit THE EIFFEL TOWER yesterday, because he in fact didn't visit ANYTHING! He FELT SICK AND STAYED IN THE HOTEL ALL DAY. The first sentence in (42B) states that it is not the case that some visit of Sascha yesterday was a visit of the Eiffel Tower. And, as the second sentence makes clear, this is so because Sascha made no visit at all. Thus, B claims that A is mistaken because A assumes a false focal presupposition, namely that Sascha ventured out of the hotel to take in some sights.- - As it negates a presupposition, the structured wide reading in (42) could be thought of as involving a special kind of negation, namely a "metalinguistic" negation (of. Horn 1989). This kind of negation is held to be different from the "regular" negation. In particular, whereas regular "It worth noting in this context that the focus on ANYTHING is contrastive in that it contrast with THE EIFFEL TOWER. Cf. also (i): (i) Sascha didn't visit JEAN/HIM yesterday. He didn't visit ANYONE. In (i) the contrasting foci are not of same semantic type; one is a quantifier {anything) the other a referring expression [Jean, him). Since the SDD account makes no claim concerning the contrastiveness of focus (see appendix) it does not face any particular problem with the contrast between Jean/him and anything. On the other hand, they type mismatch is unexpected on the Jackendoff-style analysis, which, as was shown in section 2.1.1., takes focus to signal contrast among elements of matching semantic type. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 62 negation lets presuppositions "project" (see below), one of the primary functions of metalinguistic negation is to deny presuppositions. This is not the view that is taken here. Rather, negation remains unambiguous even in the structured wide reading. What makes this reading different— and what causes there to be no focal presupposition— is nothing else but the wide scope of the negation: (43) NOT [3e: C(e) & visit(e) & Agent(e, Sascha) & Past(e) & yesterday(e)] Theme(e, THE EIFFEL TOWER) structured wide reading "It is not the case that some (relevant) visit by Sascha yesterday had the Eiffel Tower as its theme." Focus in (43) does have an effect here in that it structures the event description. Because of the wide scope of not, however, no focal presupposition is entailed by the restricted event quantifier. This means that not in (43) functions in the same way as a wide scope decreasing quantifier does in the example Nobody likes Bill; by taking wide scope, it undermines the entailment of a focal presupposition. On this view then, not only does not not need to type-shift depending on the element it associates with (verb, focused element, entire sentence), not is also not ambiguous between a "presupposition-preserving" notl and a metalinguistic "presupposition-denying" not2. Rather, the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 presupposition-denying effect of negation follows from the negation taking wide scope over the entire sentence, including the restricted event quantifier. Apart from parsimony, what speaks in favor of an analysis that treats negation as unambiguous in this way is that if not were actually lexically ambiguous, we might expect to find languages that lexically disambiguate between a presupposition-preserving notl and a presupposition- denying not2. Yet, after an extensive review of the typological literature on this subject, Horn (1989) concludes that there is little, if any, evidence that there are such languages. In light of this, Horn suggests that the difference should not be considered semantic but pragmatic instead. On the SDD analysis, presupposition denying negation is not different from any other negation and there is no semantic or pragmatic ambiguity. It is not surprising then that languages do not seem to overtly distinguish between a presupposition-preserving and a presupposition denying negation. The idea is that there is no such difference, neither in term of lexical meaning, nor in pragmatic use; negation is univocal and the only differences that there are among the various interpretations are due to scope. 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 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 structured wide reading, in this reading the negation takes scope over the entire sentence. With the stress being on the negation itself, however, there is no additional focus effect. I will call this fourth kind of reading the unstructured wide reading: (44) Sascha DIDN'T visit the Eiffel Tower yesterday. (45) NOT [3e: C(e) & visit(e) & Agent(e, Sascha) & Past(e) & yesterday(e) & Theme(e, the Eiffel Tower) unstructured wide reading "It is not the case that there is some (relevant) event where Sascha visited the Eiffel Tower yesterday," The unstructured wide reading can appear in a context like this : (46) A: I bet SASCHA VISITED THE EIFFEL TOWER YESTERDAY! THAT'S why there was such a traffic jam! B: No, Sascha DIDN'T visit the Eiffel Tower yesterday. HOW COULD HE! HE IS STILL IN MOSCOW.-' -®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 the everything except the negation. While 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 any more obvious alternatives. If this is on the right track, then the SDD would directly predict the logical form; it would look exactly like (45), Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 2.3.2. (Focal) presuppositions as direct entailments from logical form: Having seen four ways in which the sentential negation not can be interpreted (bound, free, wide structured, and wide unstructured), we are faced with the following question. Do all four ways have equal status, or should we view one of them to be the unmarked one, as the one which counts as the paradigmatic "negation of a sentence"? I think the answer is yes, there is indeed reason to say that "negation of a sentence S" corresponds to one particular reading, namely the bound reading. First, we already noted that a bound reading of a negated sentence and the non-negated version of the sentence share the same focal presupposition. For example, the bound reading of (31) and its non-negated version in (47) both presuppose that Sascha paid a visit yesterday: (31) Sascha didn't visit THE EIFFEL TOWER yesterday (but...) (47) Sascha visited THE EIFFEL TOWER yesterday. That (31) on the bound reading and (47) should carry the same focal presupposition follows directly from logical form since in the bound reading the negation takes scope only only that everything in the scope of 3e would be in capital letters, showing that it is there because it is focused. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 66 over the matrix, letting the restricted event quantifier escape its effects. Given now that presuppositions have the defining property that they are entailed by a sentence S and the negation of S (e.g. Soames 1991), then by negation of S we must mean its bound reading. For only this construal of a negated sentence has the same focal presuppositions as its non-negated counterpart; the wide readings do not (they have no focal presupposition at all), nor does the free reading (it has a focal presupposition, but a different one). Once we equate the negation of S with its bound reading, the projection problem for (focal) presuppositions evaporates: since we do not assume that the negation of a sentence amounts to having the negation take scope over the entire sentence, we need no longer postulate a process which allows the focal presuppositions of the sentence to "project" across the negation in order to make sure that a sentence and "its negation" share the same focal presupposition; it is directly entailed from the logical form. Thus, if we take the negation of S to correspond to its bound reading, then, given the SDD, it follows directly that a sentence and its negation share the same focal presupposition. Nothing else needs to be said, in particular, no mechanism of presupposition projection needs to be posited. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 This is a welcome result. Not only does presupposition projection in itself not seem very explanatory, there is also a rather general reason against it; ultimately, presupposition projection renders the notion of presupposition useless. The argument runs as follows.-" Since a presupposition is entailed both by a sentence and its negation, if we— assuming presupposition projection— take the negation of S to simply mean -iS, where we prefix the negation in front of the entire sentence, we get (i), and also the equivalent (ii). But now if it turns out that a presupposition p is false (of. (iii)) , a contradiction follows, namely S and -tS (cf. line (iv)), which cannot be. If so, however, a presupposition p can never be false, it always has to be true (cf. line (v)). This is not the case though: if Sascha did not leave the hotel room yesterday and did not visit any site whatsoever, then the focal presupposition of the bound reading of (31), for instance, (that he paid a visit yesterday) is clearly not true. The upshot of this is that if we assume presupposition projection we wind up wrongly predicting that all logical presuppositions are true. - This argument was suggested to me by Sylvain Bromberger. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 (48) (i) (S - p) & (-S - p) (ii) (-iS V p) & ( - i - i S V p) (iii) 2E_____________ (iv) -iS & S (V) p One way to go at this point is to simply abandon the concept of presupposition as basically useless. After all, if it forces us to the conclusion that all presuppositions are true, it is really not very interesting. But, on second thought, this is too rash. We can in fact maintain a useful concept of presupposition— we only have to get rid the premise that the "negation of S" amounts to simply prefixing a negation to S, -,S. Instead, not only takes scope over the "assertion" material, but not over the presupposition material. In other words, the negation of a sentence corresponds to its bound reading. It now follows directly that both a sentence and its negation share the same focal presupposition, no projection mechanism is needed. Of course, focal presuppositions are only one kind of presupposition among various and it would be interesting to see whether such a scopal solution to the projection problem can be given for other types of presuppositions. That this is possible as far as the presuppositions of definite descriptions are concerned is suggested by a certain reading of the Russellian analysis of such expressions, which is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 discussed below. Whether this kind of analysis can be extended to other cases of presupposition projection, e.g. with if, is something that at this point I would like to leave for future discussion. Recapitulating a bit, we began by noting that negation and focus can interact in a number of ways. First, in the bound reading, what is effectively negated is the focus. Second, in the free reading, the effect of negation is limited to the verb, giving rise to a "negative" description of an event (a not visiting, e.g.). Thirdly, in the structured wide reading, the event description is structured by focus, but no focal presupposition is entailed. And, lastly, an unstructured wide reading becomes available when the stress itself is on the negation (or on the auxiliary supporting it). It was argued that the different readings and the varying distribution of focal presuppositions are due to the scope of the negation and the way focus structures an event description. Since in each of the readings negation takes scope over a Davidsonian conjunct, negation remains a univocal prepositional operator. Moreover, no presupposition-denying metalinguistic negation was invoked. Rather, the presuppositional denying character of the wide readings follows from the wide scope of the (regular) negation. Finally, the "negation of a sentence" is the negation of its main assertion, i.e. its bound reading. Understanding "negation of a sentence" this way Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 avoids the projection problem for focal presuppositions, and explains why both a sentence and its negation share the same focal presupposition as a matter of logical form. This, in turn, makes it possible maintain a useful notion of presupposition, one where not all presuppositions are predicted to be true. 2.3.3. Definite descriptions, focus and negation: With the SDD analysis of focal presuppositions relying so heavily on scope interactions between a description with existential import (3e) and a negative element (a decreasing quantifier, the negation not), it is interesting to consider the presuppositions of descriptions which are not verbal but nominal instead. Concretely, this section discusses, first, how the presuppositions of nominal descriptions like Che king of France are accounted for on a Russellian analysis, and, second, it shows how the presuppositions of definite description depend on the interaction between focus and negation. On the classic Russellian view, definite descriptions like the king of France are quantifiers rather than referring expressions (e.g. Russell (1905), Neale (1990)). On a generalized quantifier version of this view, we can say that the definite determiner the functions like other binary Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 determiners e.g. most, all in that it denotes a relation between the NP complement of D and the scope of the DP: (49) a. [the A(sg.)] B iff A c B and |A| = 1 b. [the A(pl.)] B iff A c B The lexical meaning of the states that everything that satisfies the restriction of D, which is denoted by the NP complement of D, also satisfies the matrix B, where the B equals the scope of the DP. When the NP is in the singular, there is exactly one such thing, but not when the NP is plural, in which case there are several such things. Thus, I am assuming two things here as far as definite descriptions are concerned; first, that the is quantificational, and second, that it gives rise to a restricted quantification where the NP complement of D is the restriction and the scope of the DP is the matrix.- - Given this kind analysis, we now expect to find the following distribution of presuppositions. When a definite description like the king of France appears in a non negative environment, then— because of the existential force of the determiner the— the definite description will presuppose the existence of a French king. Conversely, when the description appears in the scope of a negative element. -'A detailed discussion of these notions will be offered in chapter 4. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 no existential presupposition that there is a king of France is entailed and the sentence can be true if there is no French king, i.e. if the presupposition fails (cf. Russell 1905). Taking what we know about how negation interacts with focus, we can now make quite subtle predictions as to when a definite description in a negated sentence will induce an existential presupposition (of. Jacobs (1991) and HajiCova (1984)) : when the sentence has a wide reading, the definite description will not induce a presupposition. This is so because under such a reading everything winds up inside the scope of the negation. Similarly, no presupposition will be entailed if the definite description is part of the focus and the sentence is interpreted as having a bound reading because, again, the quantifier will end up in the scope of the negation. On the other hand, if a sentence has a bound reading, and if the definite description is not (part of) the focus, then the description will in fact carry an existential presupposition, so that if nothing satisfies the description the sentence will automatically be false. The next examples show that these predictions are indeed true. (50) below represents the paradigm example of so-called presupposition failure— except it makes clear that there is in fact stress on the negation in this example. As was shown in the previous section, this type of focus assignment with stress on the negation causes the sentence to have an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 unstructured wide reading. As a result of this kind of reading, the definite description winds up being interpreted inside the scope of the negation, and, consequently, the sentence does not entail the presupposition that France has a king: (50) The king of France ISN'T bald. There IS no king of France ! (51) NOT 3e C(e) [the x: king-of-France (x)] & bald(e) & Theme(e,x) unstructured wide reading Whereas (50) is an instance of an unstructured wide reading, in (52) below, the first 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 first clause of (52), the definite description is interpreted inside the scope of the negation, the sentence is correctly predicted not to carry the existential presupposition that there is a king of France: (52) The king of France isn't BALD, because the king of France isn't ANYTHING. There IS no king of France! (52) is interpreted such that the first clause states that it is false that something can be attributed to the king of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 France is baldness (structured wide reading). Then, the jbecause-clause denies that there is any property at all that can be attributed to him. This now explains the coherence of the third 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 not tall, not short, not a hemophiliac, and, in particular, he is not bald either." Next, a case where the definite description itself is focused and where the sentence has a bound reading is given in (53). The prediction is that here, too, no existential presupposition will be entailed because the definite description winds up in the scope of not: (53) THE KING OF FRANCE isn't bald. There IS no king of France. Who is bald is the PRESIDENT OF THE REPUBLIC! (54) [3e: C(e) & bald(e)] NOT [the x: king of France(x)] Theme(e,x) bound reading It is only because the first sentence does not presuppose the existence of a French king that the second sentence, which explicitly denies the existence of such a king, makes ^®As noted in connection in the first 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 (cf. fn. 15). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 sense and offers an explanation of the first sentence, rather than contradicting it. The third sentence then takes up another aspect of the first sentence, namely its focal presupposition that somebody other than the French king is bald, cf. (54). 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 definite description is not part of the focus, we find that, consonant with our predications, the definite description has to denote, or else the sentence is false; (55) Clinton didn't give A MEDAL to the king of France, but A COLLECTION OF EARLY AMERICAN PAINTINGS rather. As its but-continuation ensures, in (55) the negation takes only scope only over the focused theme, but not the rest of the sentence (=bound reading). In contrast with the other examples we have seen, here the definite description finds itself outside of the scope the negation. Consonant with what we expect, (55) is indeed false, and it is false because one of its presuppositions is false: throughout Clinton's life France has been a republic and there has not been any French king. In sum, both verbal descriptions of events and nominal descriptions of individuals have parallel presuppositional properties: they only induce existential presuppositions to Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 the extent that they appear in a non-negative environment. Once they occur inside the scope of a negative element, nothing needs to satisfy the description and the sentence can still be true. (56: wide reading, unstruc. (51) no pres, wide reading, struct. (52) no pres, bound reading NP focused (53) no pres, bound reading NP unfocused (55) presupp, In the overall picture, the SDD adds predictive power to the Russellian analysis of definite descriptions in that it shows how whether or not a definite description is interpreted in a negative context systematically depends on the interaction between focus and negation and on whether the description is focused or not. 2.3.4. Other negative contexts : The negative elements that I have considered in connection with focal presuppositions are decreasing quantifiers (in particular nobody, nothing) and the negation not. What about other negative elements? By the logic of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 the argument, the SDD should also explain the distribution of focal presuppositions in negative or downward-entailing contexts that are created by other elements. Concretely, this means that when an existential event quantifier appears in the scope of, for instance, unlikelyr deny, if— which also create negative environments— the sentence should lack a focal presuppositions. In this context, I want to consider (57) as an illustrative example. (This example is due to Rooth (1994a)): (57) A: Did anyone win the football pool this week? B: Probably not, because it's unlikely that Î4ARY won it, and she is the only person who ever wins. Clearly, (57B) does not presuppose that the football pool was won. How come? As was mentioned earlier, Rooth (1994a) considers the possibility of binding the type-matching variable that on his view replaces the focus existentially rather than with a lambda operator. He thus contemplates an analysis like which is in relevant respects similar to the one we encountered in (10): (10) The presuppositional effect of focus is captured by existentially binding the type-matching variable that replaces the focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 Apart from capturing (some) of the presuppositional effects of focus (see 2.1.), there is also a second reason Rooth (1994a) considers existentially binding the variable, which has to do with the account of how adverbs associate with focus. This will be discussed in chapter 3. With the equivalent of (10) in the background, Rooth then points out that if presuppositions project— which is the standard claim and which he also assumes— then (57B) is problematic because it in fact lacks a focal presupposition. To account for this, he suggests that there is no focal presupposition there after all. Concretely, if the focused proper name is not treated as a referring expression, but rather as a quantifier, this would then "trivialize" the focal presuppositions of (57) beyond the point of having any real existential force; no focal presupposition would predicted because all that is predicted is that some, possibly decreasing quantifier applies to what is denoted by the non-focused part of the sentence. Put differently, the lack of focal presupposition in (57) is handled exactly in the same way that (10) would handle the lack of focal presupposition in (7) {NOBODY likes Bill); an existential commitment to there being some quantifier that applies to a predicate is not a commitment to their being an individual which satisfies the predicate. Yet, if we trivialize the presuppositions in (57) by treating Mary as a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 quantifier, we may ask what steps us from doing the same in other cases, in (la) for instance: (1) a. ROSALIA wrote poetry. Since we do not want to give up the claim that (la) presupposes that some poetry was written, it seems that by "trivializing" focal presuppositions we would immediately lose at one side what we have gained on the other— we would gain the lack of focal presupposition in (57B), but we would lose predicting that (1) indeed does have a focal presupposition. The SDD account can explain the lack of focal presuppositions in (57) in a different way. Much like in the structured wide reading of a negated sentence, unlikely takes scope over the entire that-clause, which is internally structured by focus: (58) UNLIKELY [3e: C(e) & win(e) & Past(e) & Theme(e, it)] Experiencer(e, MARY) "It is unlikely that some (relevant) event of winning it 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 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 it is unlikely that it will rain hard. Given the semantic nature of unlikely, it then follows that the event description which appears in its scope in (58) need not pick out any event of winning. Hence, (57) is correctly predicted to not carry the focal presupposition that the football pool was won and it is not necessary to trivialize focal presuppositions. The claim then is this: the SDD account of when and where focal presuppositions occur is not limited to negation and quantifiers like nobody, but can be extended to other negative contexts. (57) provides an example of this. Clearly, there are a number of other elements that would be interesting to also investigate in this connection, e.g. predicates like deny, doubt, the conjunction if, etc. I will leave it as a matter for future consideration. The prediction is clear, however; as soon as the restricted event operator that denotes the event of focus appears in a negative (downward-entailing) environment, then, no matter what element creates this environment— be it negation itself or some other element, e.g. unlikely— the sentence is predicted to lack a focal presupposition. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 2.4. Revising and situating the SDD : Having shown how the SDD accounts for the effects that focus has on the presuppositions of a sentence, I would now like to revise the SDD somewhat because in its present form it only accounts for focus on the immediate constituents of the clause, which, however, is not enough. The second aim of this section is to situate the SDD in a somewhat broader context, showing how it owes to previous analyses, and also how it differs from them. 2.4.1. Focus on elements too small or too distant: In light of the fact that languages like Hungarian and Basque have focus movement, one might consider attractive a covert movement analysis of focus for English, where focus undergoes syntactic movement at LF. What would add to the appeal of such a movement analysis is that it would find some empirical support in the classic cross-over argument for focus movement (Chomsky 1976).- The SDD does not have -'If John in (i) is focused, then it cannot be interpreted as being coreferent with the preceding pronoun he. If John is assumed to move, then the unacceptability of this interpretation can be assimilated to other cross-over constellations. (i) *The woman he^ loved betrayed JOHN_ Movement is not the only possibility to account for (i), Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 the property of being a movement analysis; much like Rooth's "alternative semantics" proposal (cf. Rooth 1985, 1992) it is an account of focus that does not rely on focus movement at LF. The motivation for an in-situ theory is a simple one. First, focus can affect all kinds of elements, some of which are not generally assumed to undergo syntactic movement e.g. the adjective red in (59a) and the suffix ex in (59f): (59) a.The police arrested an ex-convict with a RED shirt. b.The police arrested an ex-convict with a red SHIRT. c.The police arrested an EX-CONVICT with a red shirt. d.The police ARRESTED an ex-convict with a red shirt. e.THE POLICE arrested an ex-convict with a red shirt. f.The police arrested an EX-convict with a red shirt. A second motivation for an in-situ account of focus is that the semantic effects of focus are notoriously non-local and are not subject to syntactic island constraints (of. Rooth 1985 and references cited therein). The non-locality of focus effects can be shown in many ways. For present purposes, it is enough to consider the various possible interpretations we can assign to (60): however. For alternative proposals see Rooth (1985), Rochemont (1986), Ferro (1994). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 (60) [-p^Mary told me about ^^the rumor [-p^that Bill had said [jpithat SUE was going to India] ] ] ] This sentence can have three different readings. One the first reading, the focus on Sue is interpreted within the most deeply embedded clause CP3, as in (61a) . But in addition, (60) has two other, and more interesting readings: one where the event of focus is the event introduced in CP2, that is one clause higher up than the focus, and a third reading, where the focus structures the event quantifier of the matrix clause, CPI. The readings are paraphrased through clefts in (61b) and (61c), respectively. (61) 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 (61b), for example, we find that the event description which is structured by focus is the one introduced by the verb say in CP2 (=event of focus). Since Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 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 affects (the event description introduced in CP2) because movement out of the subject position of a tensed clause is illicit. More drastically even, in (61c), where the focus affects the event description introduced in the matrix clause, CPI, the focus would presumably have to be moved out of the tensed clause and then across complex NP. In its present form, the SDD can only handle focus that falls on constituents that form a Davidsonian conjunct where the conjunct directly relates to the event of focus. This includes focus on the subject, the various objects, the verb, and adverbial modifiers. It also captures focus on the verb phrase, as in (62), focus on the entire sentence, as in (64), and it can also account for the "multiple focus" that is illustrated in (66): (62) Rosalia WROTE POETRY. (63) [3e: C(e) & Agent(e, Rosalia) & Past(e)] WRITE(e) & Theme(e, POETRY) "Some (relevant) event in the past whose agent was Rosalia was a writing of poetry." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 (64) ROSALIA WROTE POETRY. (65) [3e: 0(e)] Agent(e, ROSALIA) & Past(e) & WRITE(e) & Theme(e, POETRY) "There was a (relevant) event in the past whose agent was Rosalia and which was a writing of poetry." (66) ROSALIA wrote POETRY, (EUGENE wrote PLAYS.) (67) [3e: C(e) & write(e) & Past(e)] Agent(e, ROSALIA) & Theme(e, POETRY) "Some (relevant) event of writing in the past had Rosalia as its agent and its theme was poetry." What the SDD cannot yet account for is focus on sub constituents (cf. red in (59a)) and the non-local effects of focus, the second and third interpretation of (60) (of. (61b) and (61c)), for instance. The problem is that in neither of these cases does the focus form a conjunct that directly relates to the relevant event description; in (59a) the focus is on something too "small" (an adjective inside the object), in the relevant readings of (60) on something too "far away" (an element that is tied to an embedded event description). The problem in size and distance can be bridged if we revise the SDD in such a way that the non-focused material Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 appears not only inside the restriction, but also in the matrix: (68) Structured Davidsonian Decomposition (revised version): All the non-focused material in the scope of an event quantifier is copied into its restriction. Which event description is structured in this way is left open; it may be the very description that directly binds into the conjunct provided by the focus, or it may be any description that takes scope over that description. With the revised version of the SDD stated in (68) we can now capture the sub-constituent focus on red in (59a) as in (69), Repeating all the non-focused material in the matrix overcomes the lack of a direct relation between the event and the modifying adjective red: (69) [3e: C(e) arrest(e) & Past(e) & Agent(e, the police) & [a X: ex-convict(x) & [a y : shirt(y)] With(x,y)] Theme(e,x)] arrest(e) & Past(e) & Agent(e, the police) & [a X: ex-convict(x) & [a y: shirt(y) & RED(y)] With(x,y)] Theme(e,x)] "Some (relevant) arrest by the police of an ex-convict with a shirt was an arrest by the police of an ex convict with a red shirt." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 87 The logical form winds up being a little cumbersome, but, as the gloss makes clear, it gets the meaning intuitively right. Assuming a certain amount of lexical decomposition, this analysis can also be extended to account for focus on the prefix ex in ex-convict. The various readings of (59) can be paraphrased as follows: (The brackets highlight the place where the SDD applies.) (70) "There was an event of Mary telling me a rumor that there was an event where Bill said that [some event of going to India] [had as its agent Sue]." (of. (60)) (71) "There was an event of Mary telling me a rumor that [some event of Bill saying that there was a going to India] [was a saying that there was a going to India whose agent was Sue]." (cf. (61a)) (72) "[Some event of Mary telling me a rumor that there was an event of Bill saying that there was a going to India] [was an event of Mary telling me a rumor that there was an event of Bill saying that there was a going to India whose agent was Sue]." (cf. (61b)) In a case where the focus forms a conjunct that relates directly to the event description that is structured Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 8 8 by focus, repeating the non-focused material in the matrix does not do any semantic work. This is so because to say that some relevant writing of poetry was done by Rosalia is equivalent to saying that some relevant writing of poetry was a writing of poetry and done by Rosalia. The second version is somewhat redundant than the first but the two are equivalent (of. chapter 4). Therefore, in cases where the focus conjunct directly relates to the focus event I will continue to leave out the non-focused material from the matrix so as to keep the logical forms shorter and more readable. They will continue to look as they have in the previous sections. However, on the general version of the analysis the non-focused material in the restriction of the event operator must also appear in the matrix. Otherwise, the non-local nature of focus effects is not captured.-- --There is one case that is not captured even on this revised form of the SDD, namely focus on a determiner as in (i) : (i) MOST candidates passed the driving test. It is not obvious how to separate the focused determiner from the nonfocused restriction. Given the SDD, this is necessary, however, because the restriction of the quantifier needs to be interpreted as restricting the event of focus, while its determiner has to either be part of the matrix (narrow scope) or bind into it (wide scope). One possibility to explore is that most candidates has more internal structure than normally assumed. In particular, candidates denotes not a predicate but a (quantificational) bare plural, as a result of which it can be interpreted in the restriction. The interpretation of (i) would then be something like (ii): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 The overall picture that emerges looks like this. In languages like English focus on a syntactic phrase has a phonological reflex in that focus on any given phrase is generally realized as stress on its right-most, most deeply embedded word. Focus has an effect on the semantic interpretation of the sentence, it is not just pragmatic. In particular, the logical forms of sentences are derived through Davidsonian decomposition where focus affects this decomposition by putting a certain amount of structure on it. Specifically, the non-focused material is copied into the restriction of an event quantifier (SDD). The event quantifier that can be restricted in this way can either be the one which directly binds into the conjunct containing the focus (local cases), or any "higher" event quantifier (non-local cases). A sentence entails a focal presupposition if the restricted event operator is interpreted in a non-negative (non downward-entailing) context. Negative contexts are created by wide scope decreasing quantifiers (e.g. nobody), negation, and words like unlikely, for instance. Even in a case where a restricted event quantifier appears in a negative context the basic asymmetry between a quantifier's restriction and (ii) "Most candidates are such that in some relevant event where candidates passed the driving test, they passed the driving test." At this point, this is only an idea. More research is needec here to see how this can be worked out in detail. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 matrix preserves an asymmetry between the focused and the non-focused material. The pragmatic effects of focus are consequences of the way focus shapes semantic interpretation. 2.4.2 The SDD vis-a-vis similar analyses: Naturally, the SDD owes in various ways to previous accounts. The literature on focus being vast and multi faceted, I am not able to give a comprehensive account of how the present account fits in with all other analyses of focus. I will therefore limit myself to those which I think it is most indebted to. As we already noted, in being an in-situ account the analysis follows the proposal in Rooth (1985, 1992). Also, it is interesting to observe that the interpretation the SDD accords to (la), for instance,— "some relevant of writing poetry is such that its agent is Rosalia"— is effectively equivalent to saying that some relevant poetry writing was done by Rosalia. As the shortened paraphrase makes clear, the focus on this view marks the main semantic predication in the sentence (in a particular sense of main predicate) and the non-focused part forms an event quantifier which combines with the focused conjunct(s) much like an indefinite quantifiers combines Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 with a verb phrase on the traditional view. There are actually analyses which maintain that the focus is literally the main predicate of the sentence, while the non-focused part forms the argument to which that predicate applies. In spirit, at least, the SDD analysis is 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 last century to the work of Hermann Paul, who termed the focus the "psychological predicate" and contrasted it with the non-focused part which he thought of as the "psychological subject." More recently, a predicational approach has been pursued by Ogihara (1987) and Ldbner (1990). On their accounts, the focused element is type-shifted into a predicate and the non-focused part is type-shifted into a definite description which combines with the predicate formed by the focus. Thus, ROSALIA writes poetry is interpreted as in (73): (73) Ax [x=Rosalia] (ix: write-poetry(x)) Analyzing the non-focused part as denoting a definite description captures the presuppositional effect of focus. Although this analysis works very well when the focused element is a referential expression, as in (73), it has problems with cases where focus falls on a quantifier. One Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 possibility is to say that since a quantifier takes the rest of a sentence as its argument focus has no effect when it falls on a quantifier. NOBODY likes Bill would then be interpreted like this (cf. Ldbner 1990, von Stechow 1991): (74) [Nobody x] (likes-Bill(x)) If so, however, it is no longer the case that the non- focused part denotes a definite description. This does not only cut into the generality of the analysis, it also does not predict the existence of focal presuppositions in examples where focus falls on a non-decreasing quantifier (cf. (11)) and in examples where focus falls on a decreasing quantifier but that quantifier takes narrow scope (cf. (25)- (28), section 2.1.2.). As a result, this version of the predicative analysis encounters empirical limitations similar to those of the second version of the Jackendoff- style analysis. On a different version of the predicational analysis developed in Ogihara (1987), it is proposed that a focused quantifier should be assimilated to a focused individual denoting expression (e.g. a proper name). Thus, when focus falls on a quantifier, the quantifier is first type-shifted into an individual denoting expression. Consequently, as in the case with a proper name, this expression is shifted to a predicate. So when a quantifier is focused something of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 quantifier type <<e,t,>,t> goes to individual type <e>, and then to predicate type <e,t>. Clearly, this analysis predicts that only those quantifiers that can be type-shifted into individuals can also be focused. Ogihara argues that this is by and large true as far as the Japanese "obligatory focussing" phenomenon is concerned, which is the phenomenon that he is concerned with (Ogihara 1987). The predication, however, does not extend to English. Although quantifiers 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 already saw this in the case of no (body), and many:-~ —Type-shifting a generalized quantifier (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 quantifier are intersected such that the intersection contains all and only those elements that belong to each of the sets that make up the denotation of the quantifier. The members of this intersection are then taken together to form a plural individual of the kind that was briefly discussed in chapter 1 (cf. Link 1983). The intersection operation works for certain quantifiers (so-called principle filters), but not for others. For every man e.g. 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, etc. Intersecting these sets will give us 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 quantifiers like exactly twelve apostles, the children. Many quantifiers cannot by type-shifted in this way, e.g. some men, most men, many men, no men. Since the extension of some man, for example, contains both men that are short and men that are tall, its intersection will be empty. Hence no plural individual can be formed from its Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 (7) NOBODY likes Bill. (11) MANY OF HIS COLLEAGUES like Bill. And the same holds true of other quantifiers that cannot be type-shifted into individuals, of. (75): (75) In order to decide in how a corporation is run you have to own MOST/SOME OF THE SHARES. Thus, although the SDD resembles the predicational analysis of focus in its basic spirit, the means it uses to derive the result are quite different. This difference offers an empirically more fine-grained analysis of the distribution of focal presuppositions. Whereas in its grouping of focus vs. non-focus the SDD analysis resembles predicational approaches of focus, in its reliance on event semantics it resembles Larson and Lefebvre's (1991) analysis of focus in Haitian Creole. Essentially, the SDD differs from Larson and Lefebvre's analysis only in that they hold that it is the focus, rather members, because there are none. On the other quantifiers 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 a DNA.' Clearly, though, the intersection between the set of women and the set of things lacking a DNA is empty, so no plural individual can be formed out of no man. To see that the same also holds true of proportional quantifiers like most and many I would like to refer the interested reader to Ogihara (1987) who orovides a detailed discussion of this. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 than the non-focused part, which provides the restriction of the event operator. And, conversely, it is the non-focused part (rather than the focus) which provides the second argument of the event operator. On this view (la) {ROSALIA wrote poetry] comes out not as in (7 6), but as its mirror image in (77) instead: (76) [3e: C(e) & write(e) & Past(e) & Theme(e, poetry)] Agent(e, ROSALIA)] SDD (77) [3e: C(e) & Agent(e, ROSALIA)] write (e) & Past(e) & Theme(e, poetry) Larson and Lefebvre If the SDD parallels the predicational analysis, Larson & Lefebvre's analysis illustrated in (77) parallels the "structured meaning" approach, where the focus is interpreted as an argument that is formed by the non-focused part (cf. Stechow 1991) . Distributing the labor of focused as in (77), rather than as in (76), can be motivated by the fact that it fits in with a widely held analysis of clefts. This analysis maintains that in a (pseudo)-cleft like (78), for instance, the clefted (and focused) element forms a semantic argument to a predicate denoted by the non-focused part (e.g. Williams 1983, Partee 1986): (78) Who John saw was MARY. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 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 that in some cases the non-focused part denotes a focal presupposition. On the other hand, as has been argued in quite some detail, I believe, this is not necessary on the SDD analysis- Piggy-backing on the workings of restricted quantification, on this analysis the distribution of focal presuppositions follows directly as a matter of logical form; in particular, it follows from the existential import of the restricted event quantifier.-* 2.5. Summary of chapter 2 : Recapitulating the main points of this chapter, first it was argued that the concept existential focal presupposition is useful after all because the alternative theory that the non-focused part generally only expresses what is "under discussion" is empirically too weak; although this pragmatic notion makes no false predictions, it does -'If I were to extend the SDD analysis to clefts — something which I will not attempt here— I would say that the clefted element in a pseudo-cleft like (73) represents the main predicate (in the relevant sense) and that the non- focused part represents the subject to that predicate. Although this runs counter to the predominant analysis of clefts, there do not seem to be any fundamental reasons against this. An analysis of clefts along these lines is suggested in passing in Higginbotham (1987). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 not distinguish between sentences that have focal presuppositions and those that lack them. Second, we considered the possibility of accounting for the presuppositional effect of focus by assuming a version of the Jackendoff-style analysis where focal presuppositions are derived by existentially binding a type-matching variable which replaces the focus. We saw, however, that even though this analysis makes better predictions than the one which relies on the pragmatic notion of the non-focused part denoting what is "under discussion," it nonetheless fails to predict that only wide-scope decreasing quantifiers will rob a sentence of its focal presuppositions; narrow scope decreasing quantifiers do not have that effect, nor do non-decreasing quantifiers, whether they take wide or narrow scope. In contrast, and this is the third step of the reasoning, the facts are accounted for on the SDD account. Making use of a full-fledged Davidsonian decomposition, on this view focal presuppositions derive directly from the existential import of an event quantifier that is restricted by the non-focused conjunct. Consequently, only if a decreasing quantifier takes scope over the event operator will the sentence lack a focal presupposition; if the quantifier takes narrow scope, or if it is not decreasing, a focal presupposition will resurface. Even when no focal presupposition is entailed, however, focus does have an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 effect in that the quantificational structure that the SDD assigns to the sentence preserves a basic asymmetry between the focused and the non-focused part. The pragmatic effect that focus has on the discourse felicity of a sentence is argued to follow from its semantic effect. Next, the SDD was shown to offer a simple analysis of the interaction between focus and negation and the direct effect this interaction has on the presupposition of definite descriptions. In this context, we saw that the SDD account does not require the assumption that focal presuppositions project, rather, the negation of a sentence corresponds to its bound reading and it directly follows as a matter of logical form that both a sentence and its negation share the same focal presupposition. This is a welcome result because by not having to rely on presupposition projection we are not forced to the wrong conclusion that all focal presuppositions are true. Also, since all the different readings follow from scope we can dispense with the difference between a "presupposition- denying" metalinguistic negation and a "presupposition- preserving" descriptive negation. Support for this was shown to come from the typological observation that languages do not lexically distinguish between these kinds of negations. Finally, in the last section the SDD was revised so that it can handle subconstituent foci and long-distance Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 focus effects. In particular, the matrix of a restricted event quantifier contains not only the focused material but also the non-focused material that appears in the restriction of the quantifier. This section also tried to highlight some of the ways in which the SDD owes to previous accounts. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 Appendix to chapter 2 : Different kinds of intonation: fall-rise and fall 1. The fall-rise intonation of bound readings: In discussing the interaction between focus and negation, a striking correlation appeared between the sentence's intonation contour and the type of reading it receives. The observation is that in a bound reading the utterance ends in a high tone, whereas in a free reading it (typically) ends with a fall-tone (cf. Jackendoff 1972): (1) a. Sascha didn't visit -THE EIFFEL TOWER yesterday, "fall-rise" contour "It wasn't the Eiffel Tower that Sascha visited yesterday (but something else)."(bound reading) b. Sascha didn't visit 'THE EIFFEL TOWER yesterday, "fall" contour "It was the Eiffel Tower that Sascha didn't visit yesterday." (free reading) Two questions arise at this point. First, why do we see this pattern, and, second, what does it mean for the analysis of focus? In principle, the relation that is found between intonation and type of reading could be taken to suggest Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 that there are simply two kinds of focus; there could be a fall focus which induces a free reading, and a fall-rise focus, giving rise to a bound reading. This would create a conceptual problem for the SDD account of the contrast between (la) and (lb), which, after all, attributes the meaning difference solely to scope, letting focus itself function the same way in both readings. It turns out, however, that the correlation between intonation contour and type of reading is not real. The fall-rise intonation is not limited to bound readings, but can also in non-negated sentences. (2), for instance, can be pronounced with a fall-rise intonation even though there is no negation around with which the focus could interact in a bound reading: (2) Sascha visited ~THE EIFFEL TOWER. Whatever the interpretation of (2) (see below), the fall- rise contour is clearly possible, which indicates that there is no strict correlation intonation contour and type of reading; although a fall intonation is typical of the free reading of a negated sentence, there are free readings that have a fall-rise intonation contour. What is needed then is an independent explanation for the relation between intonation and type of reading that is shown (1); in particular, the question is why bound readings require a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 2 fall-rise contour, but why is a fall-rise contour also possible elsewhere. There not being a one-to-one correlation between type of reading and intonation contour, we cannot attribute it to there being two different types of focus. An important observation is offered in L. Carlson (1984). He observes that a fall-rise contour creates the sense that there is a continuation to the sentence . The fall contour, on the other hand, creates a sense of informational completeness (cf. also Hirschberg and Pierrehumbert 1990).- The relation between intonation and type of reading that is illustrated in (1) now becomes less of a mystery. Because in the bound reading the assertion is negated in that negation effectively takes scope over the focus, when we come across such a reading, we know that focused element does not provide the element which could form the matrix for the event operator and which would satisfy the focal presupposition. This, however, leaves us wondering what— if not the focused element— fulfills this role. It is this very information gap that creates the need for a continuation of the sentence in order to make it -Carlson's general approach is game-theoretical; he analyzes the effects of focus in terms of how it affects the way discourse evolves. As a result, in his terms the generalization is captured as follows: "A sentence with nonfinal intonation cannot constitute an end point of a well-formed dialogue game." (Carlson 1984: 554) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 informationally complete. The idea is then that it is nhe need for a continuation in a bound reading which explains why this kind of reading requires a fall-rise contour. All things being equal, we do not require a continuation in the free reading. Since with the negation forming part of the presupposition the assertion is not negated, the free reading— unlike the bound reading— does not create an informational gap which needs filling. Therefore, as the sentence in itself is informationally complete, a fall-rise intonation is not necessary. In connection with the structured wide reading, we also noted briefly that it parallels the bound reading in also requiring fall-rise intonation, of. (3) (=(42)). (3) A: Yesterday, Sascha visited THE EIFFEL TOWER. B: No, Sascha didn't visit "'THE EIFFEL TOWER yesterday, because he in fact didn't visit 'ANYTHING! He FELT SICK AND STAYED IN THE HOTEL ALL DAY. This also fits in with the observation that a fall-rise contour signals that there is going to be a continuation. Intuitively, the first two sentences of B's answer are meant to say that although Sascha's doing some visiting yesterday may well be what's in the background (it restricts the event), contrary to A's focal presupposition, it did not take place. For it to be clear, however, that the first Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 sentence in 3's reply is meant to deny all of A's utterance on grounds that its focal presupposition is mistaken, a specific continuation is necessary, namely He didn't visit ANYTHING. It is the need for this continuation which explains the fall-rise contour on the first sentence, the one having the structured wide reading. 2. Contrast: an effect of the fall-rise contour To say that sentences with a "free" focus (nonnegated sentences or negated sentences with a free reading) do not require a fall-rise contour does not mean that they cannot ever be pronounced with a fall-rise contour, only that they do not have to. Clearly, (2), which involves a free focus, can be pronounced with a fall-rise contour. As with the bound and the structured wide reading, pronouncing a sentence with a free focus with a fall-rise contour suggests that the sentence has a continuation. There seem to be two types of continuations. One kind of continuation that can be signalled by a fall-rise contour in case of a free focus is that of a possible list (of. Carlson's 1984, who calls this the "at least" interpretation): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 105 (4) A: What did Sascha see? B: Well, he visited ~THE EIFFEL TOWER. (And maybe some other places too.) Because a fall-rise contour in an utterance with a free focus can suggest that the focus is part of continuing list, it has the effect of making the focus "non-exhaustive;" it suggest that what is denoted by the focused phrase might not be the only thing that makes the sentence true, there may in fact be others (the Louvre, Montmartre, etc.) This shows that exhaustiveness is not a characteristic property of focus itself. Rather, a free focus with a fall contour is understood as being exhaustive 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 can." If I say Sascha visited 'THE EIFFEL TOWER yesterday, the fall contour signals that this is all there is to be said concerning his visits yesterday. This, in turn, implicates that he visited no site other than the Eiffel Tower. On the other hand, saying Sascha visited “THE EIFFEL TOWER yesterday leaves room for the possibility that he visited something else as well. Besides signalling a list-continuation, the fall-rise contour on a free focus can also signal a continuation of the type: and not... or but not... (cf. Carlson 1984): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 106 (5) Sascha visited ~THE EIFFEL TOWER yesterday, (and/but not THE LOUVRE). The result of this is a focus that is contrastive. To give another example, in (6), C's answer, which is uttered with a fall-rise contour, has a distinctive contrastive flavor. It can be understood as something like "It was Karin and not X that she loaned it to," where X is the denotation of some other phrase that could be the object of loan and which we might have expected instead of Karen. This contrastive flavor is absent in (SB), which is uttered with a regular fall intonation. Consequently, B's answer is taken a simple answer to A's question. Although it does implicate that the focus is exhaustive (see above), it does not suggest that we might have thought that she loaned it someone else (Peter). (6) A: Who did Hilda loan her bike to? B: She loaned it 'TO KAREN. C: She loaned it -TO KAREN (but not TO PETER.) Thus, along with creating a "list" context, which undermines the pragmatic exhaustivity of focus, a fall-rise contour in a sentence with a free focus can have also have the effect of excluding alternatives to the focus element that are in some way implicit in the discourse. The result is a contrastive focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 If this characterization of contrastive focus is on the right track, then, contra Jackendoff (1972) and the alternative semantics approach, contrastiveness is not a general characteristic of focus. Rather, contrastive focus is really just a subcase of a more general phenomenon, focus. It has the same basic interpretive properties as regular, non-contrastive focus, forming the main assertion of the sentence by contributing to the matrix of a restricted event operator. On top of that, the fall-rise contour adds that there is going to be a continuation which makes the sentence informationally complete, e.g. by excluding likely alternatives that are somehow 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 on a signalling that the sentence is informationally complete. When a fall-contour on a free focus is changed to fall-rise, this can be taken to suggest a list continuation, where the resulting "at least" interpretation undermines the exhaustivity of focus. 3. More confier examples : Some interesting, and rather complex examples showing how a fall-rise intonation can also occur in non-negated sentences are provided in Ladd (1980). Let us see how the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 analysis of the pragmatic effect of different intonation contours works in this case: (7) A: Did you feed the animals? Ladd's (16) B: I fed -THE CAT. (8) A: You have a VW, don't you? Ladd's (18) B: I've got -AN OPEL. (9) A: Do you want a glass of water? Ladd's (19) B: I'll have -A BEER. For Ladd, the fall-rise intonation marks "something like focus within a given set." It picks something out of a set of possibilities and focuses on it, but, according to him, it specifically notes the connection of the set of possibilities to the context. Furthermore, Ladd's discussion suggests that the discourse antecedent which the focus element is related to tends to appear in the same (or 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 (7). Here the focused element the cat denotes a subset of the noun phrase the animals, to which it is said to be anaphorically related. Even if the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 subset relation is not entirely obvious in (8) and (9), Ladd argues that it is nonetheless there. Thus, the focus in (8) and (9) is interpreted to be anaphoric to what amounts to a discourse antecedent which in these cases are derived through some kind of accommodation on part of the hearer, namely "something to drink" and "a German/small car," respectively. Finally, it is the anaphoric connection of the fall-rise tone which is held responsible for the polite softening that distinguishes B replies that are uttered with a fall-rise from the same replies with a simple fall intonation. The latter are felt to be rather abrupt rejections of an offer. Although this account recognizes the existence of fall- rise intonation in non-negated sentences, it is not clear that the analysis also explains that the fall-rise intonation can be found with bound readings in negated sentences (see above). Ladd offers the following suggestion, which, as far as I can see, does not shed much light on the matter. Using John doesn't drink -BECAUSE HE IS UNHAPPY as an example, he states this: ...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). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 Yet, the way in which the hecause-clause winds 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. It is furthermore not clear why a subset understanding of the because-clause should force the negation to associate with the focus. Apart from this consideration, there is another worry, which relates to the analysis of fall-rise in non-negated sentences. Even though a subset relation is present in (7), and arguably in (8)-(9), such a relation is not what characterizes a fall-rise in the general case. For instance, (10) and (11) would presumably satisfy all of Ladd's the requirements for a fall-rise melody: the focus is related through a subset relation to its antecedent, and, furthermore, focus and antecedent share the same focus- frame. Still, a fall-rise tone is not needed here. What's more, it is felt to be odd. The normal intonation in these examples is a fall one: (10) A: What would you like to drink? B: I'll have 'A BEER, please vs. #... ~A BEER (11) A: You drive a small car, don't you? B: Yes, I drive 'AN OPEL. vs. #... -AN OPEL Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill It seems then that Ladd's (1980) account of the 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 (10) and (11) . The account of the fall-rise intonation given earlier can be argued to carry over to Ladd's examples in the following way. What crucially distinguishes (7)-(9) from (10)-(11) is that in the latter, B's response does not conflict with A's question, but constitutes an affirmative answer. On the other hand, in the (7)-(9) B does not agree with A. Rather, preserving the focal presupposition 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 (7), but it need not be, as is shown in (8) and (9), The point now is that whereas (10) and (11) do not require any continuation, (7) and (9) are intended to be continued along the lines of ...hut 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 which explains the fall-rise contour in (7)-(9), The "polite softening" that Ladd observes with a fall-rise intonation in (7)-(9) can thus be attributed to having left room for a continuation which answers the question, without having had to say "no" explicitly. Once Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 the continuations are pronounced, B's replies actually sound considerably less polite: (12) A: Do you want a glass of water. B: I'll have BEER, but I don't want any water. Also, if (7)-(9) are pronounced using the fall contour, no continuation is implicit, and the result is a bit rude; the answer has the pragmatic impact of saying "no" but it does not in any way acknowledge that the question that was asked was not answered directly, thus violating the turn taking compliance that regiments polite discourse. The claim that fall-rise intonation signals the need for a continuation also helps explain the different intonation patterns that appear in (13) (which is also due to Ladd): (13) A: Harry's the biggest fool in the State of New York. B: In ~ITHACA, maybe. C: In 'THE WHOLE WORLD, maybe. B shows skepticism about A's claim. The fall-rise intonation suggests an implicit continuation along the lines of ...but not in the State of New York. C's answer, on the other hand, shows that C (emphatically) agrees with A's claim. A must be right because something even stronger Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 which entails A's claim holds true; if Harry is the biggest fool in the whole world, he surely is also the biggest fool in the State of New York. Since C very much agrees with A, nothing further needs to be said and there is no need for a fall-rise contour. Finally, consider the striking contrast between (14) and (15) (also due to Ladd): (14) A: I suppose it was pretty rough meeting all the linguists, wasn't it— there were probably a few who wouldn't talk to you. B: Hah! 'MOST OF THEM wouldn't talk to me! (15) 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 (14) B agrees with A's assessment; after all, not only were there just a few who would not talk to B, most of the guests actually ignored B. In (15), in contrast, A and B do not coincide in opinion. 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 in fact justified in his worry because Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 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." Since there is no disagreement in (14) there is no disagreement, there is no need for a continuation, and the fall contour is sufficient. As a last point, interestingly, what is being disagreed with is not necessarily the literal meaning of the original sentence. It can in fact be its conversational implicature (of. Horn 1989). When we assert that someone was able to solve the problem we implicate, by the Gricean maxim of relevance, that he solved it. As the following example of Ladd's shows, in cases like these the disagreement that is signaled by a fall-rise-intonation can in fact be a disagreement with the sentence's conversational implicature: (16) a. He was -ABLE to solve the problem (but he didn't solve it.) b. She was -CLEVER ENOUGH to figure out the solution (but she didn't do it.) Again, the fall-rise contour here signals that the sentence has a disagreeing continuation. Curiously, though, what is disagreed with is not the literal meaning of the sentence, but its conversational implicature instead. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 4. Summary: In summary, I have adopted the view that the meaning effect of different intonational contours is this: whereas the fall-rise contour suggests that the sentence should be understood as having a continuation, the fall contour conveys a sense of informational completeness. In a case where the assertion is negated— in a bound reading of a negated sentence— a fall-rise intonation is obligatory because the sentence, as it stands, is not informationally complete. (After all, if it is not the focus that provides the assertion to the presupposition, what does?) 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 non-negated sentences). There are two ways of a continuation may be needed in such a case. One is to hint at that there might be a whole list of elements that could be added to the focus, thereby undermining the exhaustivity of focus, which was explained in Gricean terms. The other kind of continuation that is signalled by the fall-rise contour is one that explicitly rules out elements which are present in the discourse in which one could have expected to take the place of the focused element. It is this kind of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 116 continuation which makes certain foci to be understood as contrastive. On this view, exhaustivity and contrast are effects of the pragmatic import of intonation contours, they are not properties of focus itself. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 3: ADVERBIAL QUANTIFIERS 117 3.0. Introduction: Analyzing sentences as descriptions of events, we have seen how the effect that focus has on the presuppositions of a sentence and on its discourse felicity are explained in a rather simple way if focus imposes a certain amount of structure on event quantification (Structured Davidsonian Decomposition). This chapter extends the empirical scope of the SDD analysis to sentences where the quantifier over events is not tacit (3e), but where there is an overt quantifier, namely an adverb of quantification, e.g. usuallyr alwaysr rarely, never, etc. 3.1. Association with focus: 3.1.1. A subcase of the SDD: One striking property of adverbs of quantification is that the syntax seems to remain rather noncommittal as to their quantificational structure. Instead, it appears that the task is left to a considerable degree up to the focus assignment. This fact about adverbs is pointed out in Rooth (1985, 1995a) and has been discussed in a number of places since, e.g. de Swart (1991), Krifka (1992), von Fintel ^994). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 To see how strong the effect of focus on adverbial quantifiers is, we can compare the interpretations of the (la) with that of (lb) . The two interpretations do not. only not cut up the sentence in the same way, they in fact result in different truth-conditions: (1) a. MEMBERS OF MINORITIES rarely trust the police, b. Members of minorities rarely trust THE POLICE. (la) can be paraphrased as "rarely when someone trusts the police is that person a member of a minority." (lb), in contrast, means "rarely when a member of a minority trusts someone do they trust the police." The same kind of contrast that we see in (la) vs. (lb) 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 towards Albert. (2b), however, requires that all greetings that were directed towards Albert were initiated by Louise. If it so happens that Louise greeted someone other than Albert, (2b) can still be true (it is if he was not greeted by anyone else), but (2a) is automatically false. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 The simple generalization that emerges from these cases of "association with focus" is this: the non-focused part of a sentence restricts the adverb, the focused part enters into the adverb's matrix. Since adverbs of quantification can be viewed as phonologically realized quantifiers over events, the SDD as introduced in chapter 2 captures association with focus of adverbs in a straight-forward way: (3) Structured Davidsonian Decomposition: The non-focused material inside the scope of the event quantifier is copied into its restriction. The only difference with the examples considered up to this point is that here the determiner is not tacit (3e), but in fact overt always, usually, seldom, often, never, etc. In every other respect, however, the two sets of cases are identical. Thus, the SDD interprets (2) as in (4). (As before, since the event operator that is affected by focus binds into the conjunct containing the focus, I have left out the non-focused material in the matrix, which is redundant here, cf. 2.4.1). Strictly speaking (4a) is interpreted as "all events of Louise saying hi were events of Louise saying hi and directed toward Albert.") Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 120 (4) a. [All e: C(e) & say-hi(e) & Past(e) & Agent(e, Louise)] TO(e, ALBERT) "All (relevant) events of Louise saying hi were directed to Albert." b. [All e: C(e) & say-hi(e) & Past & To(e, Albert)] Agent(e, LOUISE)] "All (relevant) events of saying hi to Albert had Louise as their agent." Since everything in (4) is analogous to the examples of focus discussed in chapter 2— except that the quantifier is overt and not existential— it seems fair to say that the SDD as presented in chapter 2 and 3 provides a particular account of a well-known fact about overt adverbial quantifiers and extends it to cases where there is no overt quantifier, but where on our assumptions there is in fact a covert one (3e). To underline how similar these adverb are to nominal determiners when we assume quantification over events, we can render the adverbs as "all" {always), "most" (usually) , "few" (seldom), "many" (often), "no" (never), etc.' 'It might appear that nominal 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 nominal determiner like most, for instance, quantifies not only over regular individuals, but also over events, as long as the event is denoted bv its sister NP: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 3.1.2. Not every adverb associates: Along with the kind of reading illustrated in (1) and (2), where the adverb "associates with focus," there is also a second kind of interpretation. Here the adverb is not restricted by the non-focused part but is instead interpreted within the non-focused part. Consequently, it winds up contributing to the focal presupposition 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] . But it can also be understood to be saying that the one who usually comes late is Friedrich, which is a reading that arises when we are talking about who has the habit of showing up late. As in (4), in the first reading of (5) the non-focused part restricts the overt adverb of quantification {usually) : (i) Most destructions of cities during war times could have been avoided through skilled diplomacy. The defining property of determiners then is a syntactic one, namely that they take nominal complements. While determiners can quantify both over regular individuals and events, depending on the noun, adverbial quantifiers— because of being ad-verbs— always seem to quantify over events because verbs are always predicates of events, and never predicates of regular individuals. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 (6) [Most e: C(e) & come-late (e)] Theme(e, FRIEDRICH) "Most (relevant) events of arriving late have Friedrich as their theme." But in the second reading it is not usually which is restricted by non-focused material. Rather, the adverb is embedded within a higher event quantification, one of the tacit kind, and it is this latter description which is structured by the SDD: (7) [3e: C(e) [Most e': C(e)]come-late(e') & R(e',e)] Theme(e, FRIEDRICH) The resulting interpretation states that some event to which most relevant late-comings are related had Friedrich as its theme. This interpretation makes sense if we conceive of the event quantified over by Be as a containing a number of other events, namely the frequent late comings. Of this event it is asserted that its theme is Friedrich.- -As with the difference between the bound and the free reading of negation, the two readings are syntactially disambiguated in German, and also Hungarian and Basque. For instance, the German (i) only has the reading where Friedrich associates with the overt adverb, (ii), on the other hand, only has the other reading: (i) weil meistens DER FRIEDRICH zu spat kommt because the Friedrich usually too late comes (ii) wiel DER FRIEDRICH zu spat kommt Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 In light of this analysis of the second reading of (5), it would now be interesting to find examples where one overt adverb of quantification co-occurs with another one. This would provide an instance where what we have said corresponds to a tacit 3e in (5) would actually be realized overtly; one adverb would be unaffected by focus— analogous to usually in (7)— while the other one would be structured by focus (analogous to 3e in (7)). (8) offers an example of this sort: (8) Rarely did BRIAN always interrupt. Usually, JACK would do that (that is, always interrupt). The first sentence in (8) contains two overt adverbial quantifiers, rarely and usually. 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, given in (9), says that few events such that on all relevant occasions there was an interruption related to that event had as their agent Brian,— few instances of always interrupting involved Brian: (9) [Few e: C(e) [all e': C(e')] interrupt(e') & R(e,e')] Agent(e, BRIAN) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 As the second interpretation of (8) and the second reading of (5) show, although focus structures quantification over events, not every overt quantifier over events (i.e. every adverb of quantification) is structured by focus. Association with focus of an adverb is only an option. This is consistent with the SDD, which states that focus structures the quantification of some adverbial quantifier in whose scope the focus occurs, but it does not say which. 3.1.3. Second occurrence focus : Next, a complex and interesting example of how focus and adverbs can interact is (10), which is adapted from Partee (1991): (10) A: Eva always gave xerox copies to THE GRADUATE STUDENTS. B: (No,) PETR always gave xerox copies to the graduate students. The interest of this example lies in the faot that even though in B's reply the graduate students is not focused, curiously enough, always is interpreted as if it were associating with it— just as in (lOA). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 What should we make of the existence of this kind of reading? Vallduvi (1992, ch:7) and Dryer (1994) argue that this poses a problem for the view that adverbs generally associate with focus (cf. Rooth 1985). In an alternative semantics account, association with focus takes place in two stages. First, the sentence is interpreted in as if there were no focus present, whereby it receives its "ordinary" semantic value. At the same time, it also has a second semantic value, namely the "focus- semantic" one, which is formed by replacing the focus with a type-matching variable, as in Jackendoff's account. The effects 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 which, the non-focused part de facto restricts the adverb, even if the process is indirect. 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 quantify over propositions (sets of intervals in Rooth's (85) view, sets of situation for von Fintel (1994), see below). Yet, the focus-semantic 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 which de facto restricts the adverb but its union, which amounts to the existential closure of the analysis replacing the focus (see 2.3.5). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 Applying this kind of analysis blindly, the problem that Vallduvi and Dryer point out is that in (lOB) always does not seem to be associating with focus, because the graduate students, which is what it associates with, is seemingly unfocused. This poses a challenge to the alternative semantics account of association with focus. Addressing this issue, Rooth (1994b, 1995) disputes that there is no focus on elements like the graduate students in (lOB) and that hence there is no association with focus between always and the graduate students. He claims, instead, that there a phonologically realized "second occurrence focus" on the noun phrase. It is not marked the way a regular focus is, because— unlike a regular focus— it does not carry a pitch accent. But he cites experimental data showing that it does have a longer vowel than a regular unstressed word, and it has a greater waveform amplitude as well (Rooth 1995b). The graduate students in (lOB) being focused after all, it can be argued to associates with always 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. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill So if its is possible to phonologically realize two pitch accents, why does this not happen in cases like (lOB)? Examples like (12) raise a further issue. (This point is due to Rooth.) The problem here is that only in (12B) is necessarily understood to be associating with Marilyn (as in (12A)). But it is not immediately obvious why the "real" focus BOBBY KENNEDY does not associate with only as well, which would have the consequence that only quantifies over introductions all of which are said to be introductions of Marilyn to Bobby; instead, though, only seems to associated with Marilyn, leaving Bobby Kennedy to associate with also: (12) A: We only introduced MARILYN to John. B: We also only introduced Marilyn to BOBBY KENNEDY.^ Similarly, why does always in (lOB) have to associate with the graduate students and why can it not associate with the focused PETR as well? Rooth suggests that in a case like (12), the problem can be solved by moving BOBBY KENNEDY out of the reach of only, and similarly in (108) we can remove PETR from the scope of always. But he also points out that the same strategy would be necessary to account for (13), where it ^Unlike the examples that have been discussed so far, (12) does not contain adverbs of quantification in the strict sense. This does not matter since in all relevant respects only behaves like usually, never, etc. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 seems less plausible: in order to make sure that BOBBY does not associate with only, it, too, would have to move out of the reach of only. Yet, such movement would blatantly violate a syntactic island (a complex NP 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]] 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. Ironically, however, 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 non-local focus relations, and need not have focus moves across syntactic islands (cf. chapter 2). Another way of tackling the problem is discussed in von Fintel (1994) (of. Vallduvi 1990, Dryer 1994). This line of reasoning maintains that the effects of focus on adverbs are rather indirect and pragmatically mediated; in some cases it happens to be determined by focus, but that is not necessary and other factors may play a role as well. Von Fintel follows Rooth's analysis over wide stretches. Where his differs from Rooth's is in that he Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 claims that the relation between the focus-semantic value to the context variable C is even more indirect than just described. In particular, the idea is that the focus- semantic value of a sentence, which denotes what is "under discussion" (see chapter. 2), 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 the very discourse topic, which also serves as the antecedent for the focus-semantic value, in which case we get the illusion, so the claim, that the focus-semantic value directly helps restrict the adverb. Thus, the association of MARY with always is accounted as follows. (14) MARY always took John to the movies. Always takes sentential scope. The focus-semantic value of the sentence is ^ (x took John to the movies). It is anaphoric to a discourse topic, e.g. the question Who took John to the movies? This discourse topic, in turn, provides an anaphor to the context variable of always: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 (15) Discourse topic Who took John to the movies? / \ Always C Ax (x took John to the movies) focus-semantic value ("focus anaphor") Crucially, now, it is not necessary for the context variable of the anaphor to hook up to the discourse topic, that is just one option, it can also take other things as its antecedent. The idea is that this is what happens with always in (lOB); it appears to be associating with the graduate students, not because the graduate student is in any way focused, but because the context variable of always in (103) is anaphorically linked to the context variable of always in (IDA). Assuming that always in (lOA) 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 always in (103). Consequently, this occurrence of always also quantifies over giving of xerox copies, and not over giving xerox copies to graduate students, as desired. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 (10) A: Eva always gave xerox copies to THE GRADUATE STUDENTS. B: (No,) PETR always gave xerox copies to the graduate students. While this conspiratorial analysis of association with focus is attractive because it would seem flexible enough to account for examples like (lOB), von Fintel (1994) shows that it ultimately faces a similar difficulty as the proposal considered by Rooth; it requires movement of focus out of islands. To get the anaphoric relations between the context variables of two adverbs off the ground, it is essential that the two adverbs have to contain the same non-focused material in their scope. This, however, requires movement in examples like (12), where John has to be removed from the reach of always in (12A); if not, it does not follow that always in (12B) quantifies only over introductions— which is what would be needed— and not introductions to John; if it did all introductions of Marilyn to John would be to Bobby, which is not right. (12) A: We only introduced MARILYN to John. B: We also only introduced Marilyn to BOBBY KENNEDY. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 It would not be problematic to move John in (12A), the problem is that the exact same argument can be made for (13); to make sure that only in (13B) quantifies 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, while the conspiratorial analysis of association with focus initially seems to have exactly the right kind of flexibility to account for examples like (lOB), a closer look reveals that it faces a serious problem in that it requires that non-focused element in a previous sentence sometimes move out of syntactic islands. 3.1.4. Second occurrence foci and mimicking: What can we say then about second occurrence foci? I think they have two striking properties which we should take into account. One, which we have already seen, is that they Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 are phonologically realized, but barely so. Recall Rooth's observation that they have longer vowels than regular unstressed words and a greater waveform amplitude. The second striking property of second occurrence foci is that their distribution is very limited. In particular, they only seem possible when one utterance mimics the preceding one word-by-word. This is noted in Krifka (1995), for instance, who offers the following paradigm: (16) A: Mary only supports AFRICAN-AMERICAN job candidates. B: a) So what? Even JOHN only supports African-American job candidates. b) ?So what? Even JOHN only supports AFRICAN- AMERICAN job candidates. c) ??So what? Even JOHN only supports black job candidates. d) So what? Even JOHN only supports BLACK job candidates.^ ’This example has the structure OPl FOCI, 0P2 F0C2. What happens in a slightly different constellation, one where both foci occur within the scope of the lower operator, as in Rooth's original example, where there is a nesting pattern 0P2 OPl FOCI F0C2. 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 first word by word: (i) They only introduced AFRICAN-AMERICAN job candidates to John. (ii) a. So what? They also only introduced African-American Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 The contrast between (a) and (c) shows that a second occurrence focus is only possible when the second sentence is a word-by-word rendition of the first, when it mimics it. Furthermore, when there is a second occurrence focus, it cannot be realized as a regular focus with a pitch accent, cf. (a) vs. (b). (Contra Rooth, Krifka actually argues that there is no phonological marking on African-American at all, but in light of Rooth's experimental observations, I do not think this is right. Thus, I am assuming that African- American in (b) is pronounced with the particular reduced marking of second occurrence focus.) Given the SDD and the two characteristics of second occurrence foci just noted— its phonological weakness and its being restricted to mimicking contexts— we can now job candidates TO MARY. b. ??So what? They also only introduced AFRICAN- AMERICAN job candidates TO MARY. c. ?? So what? They also only introduced black job candidates TO MARY. d. ?So what? They also only introduced BLACK job candidates TO MARY. (iii)They only introduced JOHN to African-American job candidates. (iv) a. So what? They also only introduced MARY to African- American job candidates. b. ??So what? They also only introduced MARY to AFRICAN- AMERICAN job candidates. c. ??So what? They also only introduced MARY to black job candidates. d. ?So what? They also only introduced MARY to BLACK AMERICAN job candidates. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 analyze the interpretation of the second occurrence focus in (lOB) in the following manner: (10) A: Eva always gave xerox copies to the GRADUATE STUDENTS. B: (No,) PETR always gave xerox copies to the graduate students. First, PETR is a focus that does not associate with the overt adverb, of. the second reading of (5) analyzed in (7). Second, there is indeed a focus on the graduate students in (1GB), but because (lOB) mimics (IDA), the second occurrence focus is barely realized; it seems that the internal structure of the mimicked chunk {always 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 first and the second sentence also explains why the quantification structure remains parallel, that is, why always in (IQB) associates with the graduate students and not with PETR. In sum, we get a reading where the "real" focus on PETR does not associate with always, but where always instead associates with the weakly realized second occurrence focus on the graduate students. The event description that is structured by the focus is a tacit one which embeds an overt one in its restriction. So far, everything is the same as in (7). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 What is different is that in (17)— the logical fern for (lOB)— the embedded event quantifier always is itself structured by the second occurrence focus the graduate students: (17) [3e: C(e) [all e': C(e') & gave(e') & Theme(e', xerox copies] Goal(e', the graduate students) & R(e',e)] Agent(e, PETR) "Some (relevant) event where all instances of giving copies to graduate students were related to that event had as its agent Petr." Thus, because the SDD analysis does not require that every adverb associate with a focus, it can explain why PETR in (lOB) does not associate with always. And because second occurrence foci are the result of utterance mimicking a previous one, we can explain why a second occurrence focus is phonologically reduced and also why its association properties are preserved i.e. why always associates with the graduate students. The same kind of account that we have given for the second occurrence focus in (lOB) also 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 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 forces the quantification 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 rather similar analysis in that he argues that second occurrence foci are a so far unrecognized kind of anaphor. I think the intuition behind "anaphor" and "mimicking" is essentially the same. His analysis substantially differs from this one in that he assumes that second occurrence foci have no phonological reflex. Instead, I argue that the weak phonological marking of second occurrence foci not only exists (of. Rooth), but that it is fact intimately tied to the mimicking.- -Raving already encountered another effect of mimicking in chapter 2, namely its power to circumvent the scope- rigidity of postverbal negative quantifiers in Spanish, it would be interesting to see if there are further effects to be found, apart from this and second occurrence focus. A possible candidate is the positive polarity items {some, already, sometimes etc.) in the scope of not; citing Baker (1970), Horn (1989) observes that such items are acceptable "when they represent, word by word, an emphatic denial of a preceding speaker's assertion" (Horn 1989: 397). Typically, these denials are cases of what I have called wide readings (either structured on 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.— I 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 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 3 8 3.1.5. Restricting embedded clauses: Up to this point, we have only considered examples where the focus and the adverb it associates with are members of the same clause. What happens when embedded clauses, e.g. when-clauses, enter into the picture? One could think that when-clauses necessarily serve to restrict adverbial quantifiers and that they in this respect behave like if-clauses (see below). As Rooth (1985) shows, this is actually not the case. Postverbal when-clauses that are focused they do not restrict the adverb, but enter its matrix instead. Comparing (18a) and (18b), for instance, we find that in the first sentence, where the when-clause is focused, we are quantifying over events of Mary reading, of which it is claimed that most of them take place when she is in the subway. In the second example, in contrast, where the matrix predicate is focused, it is the when-clause that restrict the adverb, and we are quantifying over Mary's being in the subway. Thus, the first sentence tells us something about Susan's reading habits, whereas the second one informs us about what Susan does on the subway. utterance in which it appears mimics a preceding one. It is beyond the scope of the present work to investigate this further but it seems possible to analyze many of the examples of metalinguistic negation that Horn gives as involving the mimicking which is typical of wide readings of negation. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 (18) a. Susan usually reads WHEN SHE IS ON THE SUBWAY, b. Susan usually READS when she is on the subway. (19) a. [Most e: C(e) & read(e) & Agent(e, Susan )] read(e) & Agent(e, Mary) Ve' R(e,e') & Theme(e') & be on the subway(e') "Most (relevant) events where Susan reads are such that Susan reads and on all related events she is on the subway." b. [Most e: C(e) & Theme(e, Susan) & Ve ' R(e,e') & be on the subway(e')] Agent(e, Susan) & Ve ' R(e,e') & Theme(e') & be on the subway(e') & READ(e) "Most (relevant) events involving Susan such that on all related events she is on the subway are events where Susan reads such that on all related events she is on the subway." I am assuming here that when introduced universally quantified events which are related one-to-one to the events quantified over in the main clause. Cases where only subpart of a postverbal when-clause is focused behave analogously; in (20), for instance, all the non-focused material, including that of the when-clause, winds up restricting the adverb: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 (20) 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 clauses introduced by when behave just like the that-clauses we saw in chapter 2; when they are focused, or contain a focus, the focus goes into the matrix of the event quantifier. Strikingly, unlike postverbal when-clauses, preverbal ones can never contribute to the matrix of the adverb (Rooth 1985). Regardless of where we assign the focus, the focus in the when-clause in the following example cannot enter into the matrix of usually: (21) When Mary is on the subway, she usually reads. In this respect a preverbal when-clause behaves like an if- clause, which, as mentioned before, necessarily contribute to the restriction of an adverb, rather than to its matrix; in (22), for instance, usually must quantify over events of snowing, but cannot quantify over me staying inside: (22) a. If it snows I usually stay inside, b. I usually stay inside if it snows. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 We can make sense out of this if we assume that it is inherent in the meaning of if that it introduces a topic like element (cf. von Fintel 1994). Because of this, they effective function as restrictions to the adverbial quantifier. As for the behavior of when-clauses, the preverbal ones— because of their syntactic position— also seem to function as a special kind of sentence-topic elements and it is because of this that they seem to be restricting the quantifier (cf. Booth 1985, von Fintel 1994). The idea is that if-clauses and preverbal when- clauses provide an event description. Because they are topics, this event description is interpreted as if it were the restriction of the adverb of quantification. (23) Most events where every related event is a snowing are such that I stay inside. (24) [Most e: C(e) & Ve' C(e') & R(e,e') & snowing(e')] stay-inside(e) & Theme(e. I) To make sure that usually effectively counts snowing, which is what the truth-conditions require, we have to assume that the relation R is one that does not relate one e to many e'; rather for every e, there has to be exactly one different e '. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 The interesting question now is what happens when focus appears in a preverbal clause (or in an if-clause)? Since preverbal clauses cannot contribute to the matrix but necessarily enter the restriction of the adverb, we expect that the effects of focus will be limited to within the description in the restriction; the focus will only shape the quantificational structure of that description, but not enter into the matrix of the overt adverbial quantifier. Consider (25): (25) When THE 1RS calls, Donald usually worries. Here, the direct effects of focus are indeed limited to the embedded clause. Given the SDD, we can render its effect as in (26) : (26) Most events where every related event of calling had the 1RS as its theme are such that they (the first events) resulted in Donald's getting worried. The effect of focus here is limited to the embedded clause restricting the adverb; this event description is structured in the regular way, as predicted by the SDD: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 (27) [Most e; C(e) [Ve': C(e') & R(e,e') & call(s')] Agent(e', THE 1RS)]] worry(e) & Experiencer(e, Donald) Getting rid of the center-embedding, this is interpretation is equivalent to saying that most events where somebody called and the caller was the 1RS are such that they resulted in Donald getting worried. On this paraphrase, the non-focused material within the clausal topic functions like a relative clause to the events that are being quantified over by the overt adverb. At this point we can observe than more natural even than the focus assignment in (25) is one where there is also a focus inside the matrix clause: (28) When THE 1RS calls, Donald usually WORRIES. This alters the quantification structure only in so far as the matrix of the adverb usually is itself a structured description. The resulting interpretation states that most events where a related calling had the 1RS as its agent, and which are events where Donald is involved are events of worrying. Here, the SDD applies twice— once in the restricting embedded clause, and once in the main clause: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 (29) [Most e: C(e) [Ve': C(e') & call(e')] Agent(e', THE 1RS) & Experiencer(e, Donald)] WORRY(e) "Most (relevant) events where a related calling event had the 1RS as its agent and which are events experienced by Donald, are events of worrying." Thus, the SDD offers a simple account of the effects of focus in clauses which necessarily have to form part of the restriction of an overt adverb: focus structures the embedded event quantification in the regular way, and, consequently, it affects only indirectly affects the main event quantification of the sentence, the one introduced by the overt adverb. With the effects of focus in restricting clauses in mind I would now like to turn to a different phenomenon exhibited by adverbs of quantification, which has received a great deal of attention in the literature, namely the quantificational variability effect. 3.2. Adverbs and quantificational variability: So far we have seen how the quantificational structure of adverbs like always^ rarely, usually, etc. interacts with the way focus is assigned and to what extent this directly follows from the SDD. One phenomenon which depends on the quantificational structure of adverbs is the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 Quantificational Variability Effect (QVE) exhibited by indefinite noun phrases. Given this, we now expect that there will be a direct relation between focus assignment and the QVE. This is what is shown next. 3.2.1. The QVE auid unselectxve adverbs : Lewis (1975) famously notes that in a sentence like (30) the interpretation of the indefinite noun phrase riders is linked to the interpretation of the adverb seldom: (30) Riders on the Thirteenth Avenue line seldom find seats. This sentence can be true when in the few peak hours most of the riders in these hours do not get seats. It is irrelevant if on most other hours (namely the off-peak hours, where there are only few passengers) all of the passengers do in fact get seats. In other words, (30) is similar to Few riders on the Thirteens Avenue line ever find seats. Lewis takes the dependence between riders on seldom to show that seldom here is quantifying not over times (a standard view of adverbs), but unselectively over both times and riders. To this end, he proposes that indefinites like riders are devoid of any quantificational force, but only Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 introduces a free variable that can unselectively be bound by seldom. (30) is now interpreted as in (31): (31) Few <x,y> where x is a rider on the 13th. Ave. line and y a stretch of time are such that x finds a seat at y In an extension of the Lewis-style analysis, Heim (1982) views adverbs as having a tri-partite structure with a restrictive clause and a nuclear scope. Following Lewis, indefinites are not quantificational but only introduce a free variable which needs to be bound somehow. All the indefinites that get interpreted in the restrictive clause of the adverb get unselectively bound by it and hence show quantificational variability. As for the indefinites that only appear in the nuclear scope, they receive their quantificational force through existential closure; consequently, indefinites which only appear in the nuclear scope do not show quantificational variability. So far Lewis's observation and the basic point of the unselective binding account. We know now that which part of a sentence goes in the restriction of an adverb and what part marks its matrix depends crucially on focus. And this also applies to his example given in (30); as (32) shows, by changing the focus assignment, we change the intuitive meaning of the sentence and, in particular, the locus of quantificational variability: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 (32) a. RIDERS ON THE THIRTEENTH AVENUE BUS seldom find seats (unlike people on THE LA SUBWAY.) b. Riders on the Thirteenth Avenue bus seldom FIND seats (they often GRAB them from somebody.) o. Riders on the Thirteenth Avenue bus seldom FIND SEATS (they usually have to stand.) (32a) is roughly equivalent to saying that few of those that find seats are bus riders on the Thirteenth Avenue bus, beoause, perhaps, the Thirteenth Avenue bus is notoriously overcrowded. In contrast, (32b) asserts that few match-ups between a Thirteenth Avenue bus rider and a seat come about in a civilized way— by the seat being found. (Instead, most riders on that bus are really aggressive and simply grab the seat from someone.) Finally, (32c), which has the "unmarked" focus on the VP, has the reading that Lewis observes, where few riders on the Thirteenth Avenue bus are such that they find seats. Here the interpretation of riders seems to directly depend on seldom, that is riders shows quantificational variability. Consider also (33). Varying with the assignment of focus, the adverb in the first example quantifies over avoiding elevators, in the second example the adverb seems to count claustrophobies, and in the third example pairs of claustrophobies and elevators: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 (33) a. A CLAUSTROPHOBIC usually avoids an elevator. b. A claustrophobic usually avoids AN ELEVATOR. c. A claustrophobic usually AVOIDS an elevator. The Lewis-style analysis gives as logical forms of the following sort: (34) a. Most <x,y> where x is a time, y an elevator and y is being avoided are such that the one avoiding at time % is a claustrophobic b. Most <x,y> where x is a time, y a claustrophobic and y avoids something, what is being avoided at x is some elevator c. Most <x,y,z> where x is a time, y a claustrophobic, and z an elevator are such that y avoids z at % What this illustrates then is that the QVE is tied to the quantificational structure of the adverb, and this structure, in turn, depends on focus. On the unselective binding account, the natural idea to explore is that the effects of focus on the QVE that have just been illustrated can be handled by mapping the non-focused indefinites into the restrictive clause and the focused indefinite into the matrix. I will not pursue this here. For a relevant Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 proposal see Partee (1991), see also Rooth (1995a), who discusses some of the intricacies that are involved in combining an unselective binding account of adverbs with an alternative semantics treatment of focus.- 3.2.2. Adverbs as selective quantifiers : All along we have been treating quantificational adverbs as selective quantifiers over events, not as unselective quantifiers over n-tuples. Indefinites, on this view, are genuine quantifiers, they do not introduce free variables (cf. also chapter 4). How can the QVE be explained on this view? It is not difficult; it falls out from the fact that when an indefinite, which is quantificational, is interpreted in the restriction of an event quantifier, the assignments to the indefinite's variable vary with the assignments to the event variable of the adverb (of. e.g. de Swart 1991). ®He points out the following as one of the major problems. Indefinites are claimed to be novel in that an indefinite introduces a new discourse referent. It is in this respect that they differ from definites (Heim 1982). Yet, because of the non-local nature of focus, in the full- fledged tri-partite structured all the non-focused material that effectively forms part of the restrictive clause also has to appear in the nuclear scope. Now, if the non- focused part, contains an indefinite, this indefinite appears twice. To insure that the same referent gets picked out in both cases, which is clearly what is needed, somehow the novelty requirement of the second occurrence has to be blocked. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 Considering the examples in (33) again, the quantificational structure that the SDD assigns now predicts which indefinite will exhibit quantificational variability; since only the non-focused material enters the restriction of the adverb only a non-focused indefinite, but not a focused one will show quantificational variability. This follows straightforwardly from the SDD: (33) a. A CLAUSTROPHOBIC usually avoids an elevator. b. A claustrophobic usually avoids AN ELEVATOR. c. A claustrophobic usually AVOIDS an elevator. As (35) shows, in (33a) the interpretation an elevator is tied to the interpretation of the event quantifier usually because the indefinite appears in the restriction of the event quantifier such that in most cases of an elevator being avoided, the one avoiding it is a claustrophobic. As a result, to the extent that in each case of avoiding an elevator a different elevator is involved, the interpretation of an elevator will directly vary with the assignments to e and will hence show quantificational variability: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 (35) [Most e: C(e) & avoid(e) & [an x: elevator(x) Theme(e,x)] avoid(e) & [an x: elevator(x)] Theme(e,x) [A y: CLAUSTROPHOBIC y] Experiencer(e, y) "Most (relevant) events of avoiding an elevator are events of avoiding the elevator that are experienced by a claustrophobic." The fact that the full-fledged loigcal form in (35) repeats the non-focused material in the matrix is not problematic. It does not matter that the indefinite an elevator appears twice. Since theta-roles are assigned exhaustively (see chapter 1), it correctly follows that the same elevator is picked out in both cases. Hence, there is no problem of "requantification" here (cf. Rooth 1995, von Fintel 1994). To highlight this, I use the in the paraphrases of the logical forms. Analogous to (33a), in (33b) the events quantified over are those of a claustrophobic avoiding (something). It comes out as with a logical form like this: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 (36) [Most e: C(e) & [a x: claustrophobic(x)] Experiencer(e, x) & avoid(e)] [a X: claustrophobic(x)] Experience(e,x) & avoid(e) [AN y: ELEVATOR(y) Theme(e,y) "Most (relevant) events of avoidance that a claustrophobic experiences are events of avoidance experienced by the claustrophobic which involve an elevator." Finally, in the logical form of (33c) both an elevator and a claustrophobic are in the restriction of the event quantifier, as shown in (37), Hence, this interpretation counts events involving a claustrophobic and an elevator: (37) [Most e: C(e) & [a x: claustrophobic (x)] Experiencer(e, x) & [an y: elevator(y)] Theme(e,y)] [a X: claustrophobic (x)] Experiencer(e, x) & [an y : elevator(y)] Theme(e,y) & AVOID (e) "Most (relevant) experiences a claustrophobic has with an elevator are experiences of avoidance the claustrophobic has with the elevator." As these examples show, analyzing adverbial quantifiers as selective quantifiers over events, rather than unselective quantifiers over n-tuples, poses no problem for the analysis of the QVE. On the contrary, keeping the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 traditional analysis of indefinites as quantifiers, their quantificational variability falls out directly from their getting interpreted in the restriction of the event quantifier, as a result of which the assignments to the variable of the indefinite vary relative to the assignments of the variable of the event quantifier. Whether or not an indefinite winds up in the restriction of an event quantifier depends on whether it is unfocused or focused; if it is unfocused it goes into the restriction, if it is focused it goes into the matrix. This is predicted by the SDD. 3.2.3. The QVE in preposed clauses: asymmetric readings One well-known challenge for the unselective binding analysis of adverbs are the cases of asymmetric quantification in if-clauses, which give rise to the proportion problem. If-clauses are considered the paradigmatic restrictive clauses of adverbs. Consequently, all the indefinites that appear in such a clause should be unselectively bound by the adverb. Examples like (38) (at least under a certain interpretation) show that this prediction is not born out: (38) If a farmer buys a donkey, he is usually poor. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 154 Here usually does not seem to be quantifying unselectively over both farmer-donkey pairs, but only over farmers (or farmers buying something) . This is evident from the fact that (38) can be true if the only rich farmer buys the majority of donkeys and all the poor farmers, of whom there are many, just buy one donkey each. What forces the quantification of usually to be asymmetric in this case? Or, put differently, why does a farmer show quantificational variability, but not a donkeyl One possible answer is that it is the presence of the pronoun (cf. Heim 1990 attributes this to Bauerle and Egli 1984). Notice, for instance, the contrast (38) versus (39): (39) If a farmer buys a donkey, it is usually in good hands. While (38) says something pertaining to farmers (or farmers buying something) (39) quantifies over donkeys (or donkeys being bought) . The difference between the two, so the idea, lies in the anaphor he (=the farmer) vs. it (=the donkey). Somehow, it seems, the quantificational variability seems to lie on that indefinite which is the antecedent of a pronoun later in the sentence. It may came as little surprise at this point that there is another difference between (38) and (39), namely a difference in focus (Kadmon 1987, Heim 1990, Krifka 1992); Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 whereas in (38) a donkey is focused (or part of the focus), and a farmer is not, in (39) it is the other way around, a farmer is focused, but not a donkey. (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. The following paradigm indicates that pronouns prefer taking non-focused noun phrases as their antecedents, rather than focused ones. (42) a. If a farmer buys A DONKEY, he is usually poor. b. #If A FARMER buys a donkey, he is usually poor. c. #If a farmer buys A DONKEY, it is usually in good hands. d. If A FARMER buys a donkey, it is usually in good hands. It is an interesting question why in general pronouns take non-focused noun phrases as their antecedents, rather than focused ones, but is not our primary concern here. In fact, it turns out that not always when an indefinite provides the antecedent to an anaphor will the adverb quantify over Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 156 material containing that indefinite. As the following example from Heim (1990) shows, when the focus on a noun phrase is legitimized through the preceding discourse (as contrast, see chapter 2, appendix), a pronoun can easily be anaphoric to a focused noun phrase. And, crucially, in such a case the adverb's restrictions contains the non-focused indefinite— and not by the indefinite which serves as an antecedent to the pronoun. (43), for instance, would be false if 200 out of 299 drummers that live in apartment complexes that are fully occupied (Heim 1990:152): (43) Drummers mostly live in crowded dormitories. But if a drummer lives in an APARTMENT COMPLEX, it is usually half empty. Similarly, in (44) it is the non-focused a donkey that shows quantificational variability, rather than the antecedent to the pronoun, A FARMER: (44) Donkey 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.) Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 157 This shows then that it is really focus that forces the asymmetric quantification giving rise to asymmetric quantification and hence to the proportion problem— and not the presence of a pronoun. On the analysis of adverbial quantifiers developed here the asymmetric readings of if-clauses are expected. The asymmetric quantification of the adverb is predicted by the way focus structures quantification over events. Recall (25), which was interpreted along the lines of (26): (25) When THE 1RS calls, Donald usually worries. (26) Most events where every related event of calling had the 1RS as its theme are such that they resulted in Donald's getting worried. (27) [Most e: C(e) [Ve': C(e') & R(e,e') & call(e')] Agent(e', THE 1RS)]] worry(e) & Experiencer(e, Donald) Since the events quantified over by usually relate one-to- one to those quantified over by Ve in the restricting clause, in terms of truth-conditions it is as if the adverb in an example like (25) directly quantified over the events quantified over in the when-clause. From this it follows that an indefinite in the non-focused part of the when- clause will show quantificational variability, but not a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 158 focused one. The logical form of (38), with focus on the VP, now looks like this. (45) [Most e: C(e) [Ve': C(e') & R(e,e') & [a x: farmer(x)l Agent(e',x)] [Ay: DONKEY(y)] Theme(e',y) & BUY(e')] poor(e) & Theme(e, he) "Most (relevant) events where every related event involving a farmer was a buying of a donkey was an event of him being poor." This is equivalent to saying that most events involving a farmer where the farmer buys a donkey are such that they are related to an event where the farmer is poor, which corresponds to the paraphrase we started out with. In light of this consider also (46): (46) a. If AN INDUSTRIALIST hires a bodyguard, the bodyguard is usually very well-paid. b. If an industrialist HIRES A BODYGUARD, the bodyguard is usually very well-paid.' ’Out of context, (46b) might seem a bit odd. But consider it as a sentence following an assertion to the effect that when an industrialist subcontracts cleaning staff for his factory, he typically tries to pay them as little as he possibly can. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 159 These sentences have different interpretations. Whereas in the first one, we quantify over hirings of bodyguards, the second one effectively counts events involving industrialists. As a result, the first sentence is true in situation where there are, say, four industrialists and one of them, the most worried one, hires 10 bodyguards and pays them very well, while the other three, who are less worried, only hire one each and pay them just average. Unlike the first sentence, the second one is false in this scenario. As in (38), the interpretations follows from how structures quantification over events. The if-clause has to be in the restriction of the adverb; it itself is structured by focus. Since the events quantified over by the adverb are related one-to-one to those quantified over in the if- clause, it seems as if the adverb were directly quantifying over them. Given that the first if-clause quantifies over events of hiring a bodyguard, we get the impression that usually in this case also quantifies over hirings of bodyguards, of. (47a). In contrast, in the second if-clause focus is not on the subject, but on the verb phrase. As a result the event description is restricted by the conjunct containing the subject. From this it follows, that the adverb seems to be quantifying over events involving a bodyguard as an agent, of. (47b): Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 160 (47) a. [Most e: C(e) [Ve'; (C(e') & R(e,e') & hire(e') & [a x: bodyguard(x)] Theme(x,e')] [AN y: INDUSTRIALIST(y)] Agent(e ', y) ] [the z: bodyguard(z)] Theme(e,z) & be well-paid(e) "Most events such that each related hiring of a bodyguard was done by an industrialist were such that the bodyguard was well-paid." b. [Most e: C(e) [Ve': C(e') & R(e,e') [an x: industrialist(x)] Agent(e',x)] HIRE(e') & [A y : BODYGUARD(y)] Theme(e',y)] [the z: bodyguard(z)] Theme(e,z) & be well-paid(e) "Most events where every related event involving an industrialist was a hiring of a bodyguard were such that the bodyguard was well-paid." Summarizing, we have seen that the QVE on indefinites initially observed by Lewis and which motivated the unselective binding analysis of adverbs is a function of the assignment of focus. In particular, non-focused indefinites show quantificational variability because they appear in the restriction of an adverb, whereas focused ones do not, given that they appear only in the matrix. The interaction between focus and the QVE follows straightforwardly on the SDD. Even though the non-local nature of focus effects makes it necessary to interpret the non-focused material Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 161 both in the restriction and the matrix of the relevant adverb, no problem of requantification arises because theta- roles being exhausitive, they force the same individuals to be picked out in both instances. The SDD analysis also offers a simple account of the notorious asymmetric readings of if-clauses and preposed when-clauses. Since these clauses necessarily restrict the over adverb, the SDD only structures the event quantification within the embedded clauses, which, in turn, accounts for the indirect way focus here affects the overt adverbial quantifier. 3.3. Events or situations? There is a treatment of adverbial quantifiers that is in many ways similar to the event-based analysis given here and which is sometimes held to have the status of a notational variant. On this view, adverbial quantifiers are selective as well, letting indefinite noun phrases be genuinely quantificational. As a result, the analysis of the QVE is similar. Not only are the two analysis similar in their treatment of adverbial quantifiers, they have also both been employed to account for donkey anaphora, where they are used to overcome the overly strong uniqueness claims that the traditional E-type account of donkey anaphora made; an event-based analysis is proposed by Ludlow Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 162 (1994), a situation based one is developed in Heim (1990). And similarly, a situation-based analysis has also been proposed to account for association with focus of adverbs, cf. V. Fintel (1994), who provides an extensive discussion of how a situation-based analysis can be wedded with a Rooth-style alternative semantics. Given these similarities, it is worth comparing the two to see if they are empirically equivalent. Although on both kinds of analysis adverbs are treated as selective quantifiers, what they quantify over is different in each case. In particular, whereas events are individuals (albeit special ones), situations are parts of worlds. Being parts of worlds, they are ordered with respect to each other in a part-whole relation. As a result, John fed his two cats at seven denotes a whole number of situation; the bare, or minimal situation where he just feeds his two cats at seven, but also a situation, for instance, where everyone in the village (including John) feeds their cats at seven. On the other hand, if we use events, John fed his two cats at seven describes just that, an event of feeding which took place at seven where John is the sole feeder and this two cats are the only once being fed. If the other people in the village John lives in also happen to feed their cats at seven, that is a different event, one that is not picked out by John fed his two cacs at seven. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 163 The onion-like structure of situations has an important consequence for the analysis of quantification. It has to be made sure somehow that the only situations counted are the minimal ones in a case situations are quantified over, or else too many events will enter the restriction of an adverb. The general schema for adverbial quantification then looks like this, where adverb quantify over minimal situations (of. Berman 1987, Heim 1990): (48) [Q A] B iff 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 which is also in the set of situation which support the proposition. It is this minimality requirement which insures that a sentence like John always feeds his two cats at seven WHEN IT RAINS will come out true as long as all of John's cat feedings at seven take place when it rains— regardless 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 kind of example suggests that although the minimality requirement works in the case just discussed, it does not work in the general case.- It involves the "This argument was suggested to me by Barry Schein. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 164 interpretation of quantifiers such as two or more women, most guests in the scope of an adverbial quantifier; because the situations that are quantified over have to be minimal, the situation-based account of adverbs has trouble accounting for these cases. Imagine a situation where a cook prepares twelve barbecued chickens for a reception. The host thinks that these are too many chickens and says the following to the cook: (49) Never, even if most people eat, WILL THEY EAT MORE THAN TEN CHICKENS ALTOGETHER. There is a lot of other meat. As it turns out, the host is wrong. All the 23 guests arrive very hungry and 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 quantifies over in (49) are those were most of the guests eat and which furthermore do not contain any other situation where most of the guests eat. This, however, means that never has to quantifying over situations of exactly twelve people eating; 12 is more than half of 23, thus counting as most. Moreover it contains no other group which would also count as most, because a group of 11 would not do. The problem now is that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 165 since each of the twelve eats half a chicken, it is in fact true that in no such eating where most (i.e. 12) guests eat are more than ten chickens consumed. Consequently, we would predict that the host was right. As a matter of fact, though, we confidently judge that the host was mistaken and that the cook was right. Thus, when a determiner like most is interpreted within the restriction of an adverb as in (49), the situation-based account given an incorrect result because it forces to limit the domain of quantification to the minimal situations where most people eat, which are not all the cases where most people eat. There are other, non- minimal situations which also matter for the truth- conditions of the sentence. If we quantify over events, we are not forced to look at cases where only twelve people eat and hence we do not run into this problem. As (50) shows, (49) gets interpreted as meaning that no eating is such that if most people are eaters in that eating will it be the case that they each eat more than ten chickens: (50) [No e: C(e) & eat(e) [most x: people(x)] Agent(e,x)] 3e' EAT(e') & Agent(e',THEY) & [MORE THAN TEN y: CHICKENS(y)] Theme(e', y) There is however such an eating; it is the eating where all the twenty-three guests eat and there they do eat more than Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 166 ten chickens altogether {namely eleven and a half). Nothing in the analysis prevents this event from being taken into account, and, given this, it correctly follows that the cook was right and that the host was mistaken. On the other hand, exactly the opposite would be predicted on the situation-based approach. Thus, while it seems similar to say that adverbs quantify over events or they quantify over situations, the two analyses do not make the same empirical predictions. The event-based analysis straightforwardly accounts for the interpretation of quantifiers like most guests, two or more women. In contrast, the situation-based approach fails here, because given the part-whole structure of situations it has to impose a minimality requirement on situations that are quantified over which turns out to be too restrictive. 3.5. Summary of chapter 3 : Starting with focus in simple clauses, we saw that "association with focus" exhibited by an adverb of quantification is just a special case of the SDD where the event quantifier is not the tacit 3e, but in fact overt (always, usually, never, etc.). We then observed that along with the regular association reading, there is a reading where an overt adverb is in fact not restricted by the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 focus, but where the adverb contributes to the focal presupposition of the sentence. This second is accounted for if we view the focus as restricting a covert quantifier which embeds the over one in its restriction. Special attention was given to the interpretation of second- occurrence foci. They were argued to be the result of the mimicking between the two sentences, which causes the relation between adverb and focus to be carried over from one sentence to the next and which is responsible for the very weak phonological marking that characterizes second- occurrence foci. On this analysis, no focus movement out of island needs to be posited and the interpretation of focus works as usual. Turning to the QVE, we saw in particular how the locus of QV is a function of the assignment of focus and how the SDD accounts for this. The quantificational structure that focuses imposes on adverbial quantification was shown to provide a solution to the asymmetric quantification that gave rise to the proportion problem. With the selective treatment of adverbs as a background, we then considered how donkey anaphora fare under an analysis where adverbs are selective quantifiers over events. Two arguments were reviewed showing that such an account is quite successful; the sage-plant example, and the example of symmetric predicates. Finally, comparing the event-based analysis of adverbs with a situation-based account, it was argued that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 168 while the two analyses seem superficially rather similar, only the event-based account captures the behavior of quantifiers that are interpreted with the scope of an adverb. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER 4: FOCUS AND NOUN PHRASES 169 4.0. Introduction: Common wisdom has it that adverbs of quantification [usuallyr neverr sometimes.. .) and nominal determiners [most, no, some...) differ in how they find their restrictions. As we saw in the previous chapter, adverbial quantifiers 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 quantificational structure of determiners is known to be shaped by syntactic structure, the standard observation being that a determiner D has to be restricted by the predicate denoted by its complement NP. (This holds on the assumption that noun phrases really have a DP structure [^p D [.- ]] .) What this chapter aims to show is that— curiously enough— sometimes a determiner behaves like an adverbial quantifier in that its quantificational structure is entirely determined by the focus. These are what I call "f(ocus)-a(ffected)" of DPs. F-a readings have the interesting property that they only appear in so-called Definiteness Effect environments. Since their existence has -A version of this chapter will appear in Natural Language Semantics. It has benefitted from the comments of the reviewers. The MLS version is somewhat different in that it uses an alternative semantics account for focus. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 repercussions for the analysis of Definiteness Effect, let us begin by looking at it first. 4.1. The Definiteness Effect 4.1.1.Milsark's and Diesing's account: The classic place for a description and analysis of the DE is Milsark's dissertation and his (1977) article. The well-known descriptive observation he makes there is that only some DPs are acceptable in there-insertion contexts, others are not. In particular, he distinguishes between "weak" noun phrases, which are acceptable in there-insertion contexts, and "strong" ones, which are barred: (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 In addition to uncovering the basic contrast in (1) , Milsark also points out that weak noun phrases are not limited to DE-environment. They can also appear in contexts where strong noun phrases are acceptable as well. However, he observes that depending on the environment a weak noun phrase appears in, its interpretation changes. Thus, in DE Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 contexts such as (la) they have what he calls a cardinal reading; for instance, few in few children in the DE- environment in (la) means something like "few in number." In contrast, when few children occurs outside of a DE- environment, few cannot be paraphrased as "few in number," but must be understood proportionally, as "a small percentage of children." It is worth noting that the two interpretations of few are really different, and cannot be attributed to an inherent vagueness of the determiner; in fact, they differ truth-conditionally (cf. Partee 1988, for detailed discussion on this issue). Essentially, the difference is this; the cardinal interpretation of few in (2a) 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 A1 Huettner). This example can be uttered felicitously if in fact all of the egg-laying mammals appeared in the survey. On the other hand, unlike few in the cardinal reading, few in a strong-like reading is proportional and thus never allows 100% to be picked out, no matter how relative "a small proportion" is. Given that a partitive DP always has a strong(-like) interpretation, we cannot utter truthfully Few of the egg-laying mammals appeared in the survey, when Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 in fact all did. This truth-conditional difference between cardinal few ("few in number") and proportional few ("a small proportion") supports Milsark's claim that there really is a true semantic difference between few in a DP that occurs in a DE environment and few in a DP outside such an environment While there-insertion provides the paradigmatic DE- environment, the subject position of so-called Stage Level predicates optionally also allows for weak noun phrases with cardinal interpretations. SL subjects are only an optional DE-environment, however, because along with cardinal weak DPs they also allow for strong DPs and weak DPs on their -It may appear that the lack of a 100 % interpretation of few is due to Gricean Maxim of Quantity ("say as much as you can"), which would suggest that there is in fact no ambiguity between proportional and non-proportional few, few is always proportional. 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 if would violate Grice's Maxim of Quantity. Taken literally, however, it still would be true in such a situation, even it if is uninformative. Notice, however, that if the lack of a 100% interpretation in Few children like spinach were argued to be due to Gricean reasons, one might wonder why this is not also the case in a case like Few children are playing. While it might be possible to come up with an answer for this, it would be difficult to explain under this view why few and most in the following context do not pattern the same: (i) Most children like spinach, in fact all do. (ii) #Few children like spinach, in fact all do. By adding in fact all ... we can defeat the Quantity implicature of most. But, as the incoherence of (ii) shows, with few, this is not possible. This strongly suggests that it is as 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. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 Strong-like interpretation. Unlike the subject position of SL predicate, the subject position of IL predicates only allows for strong noun phrases or weak noun phrases with a strong- like readings; cardinal interpretations are impossible in this context. Consider (2): (2) a. Few children are playing, b. Few children like spinach. In (2a), where the predicate is SL, few children can be interpreted either in with the cardinal reading along the lines of "children who are few in number" (cf. (la)), or in a strong-like interpretation, where few is proportionally as in the overt partitive few of the children. In (2b) , on the other hand, the predicate is IL. And here we only find the strong-like reading: (2b) claims that a certain, contextually small percentage of the children in question, like spinach. Summarizing Milsark's distributional observations, we can say then that there-insertion sentences constitute a DE-environment, the subject position of SL- predicates an optional DE-environment, and the subject position of IL-predicates an anti-DE-environment. Turning to Milsark’s analysis of the DE, it crucially relies on a correlation he draws between a determiner being cardinal and the DP appearing in a DE-environment. The claim is that the determiners of strong and strong-like noun Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 phrases remain truly quantificational— as standardly assumed. On the other hand, the determiners of weak noun phrases in DE-environments are non-quantificational cardinality predicates (adjectives). Thus, cardinal few will mean something like "not numerous," "few in number." With the determiners of weak noun phrases in DE-contexts lacking quantificational force, the quantificational force for the noun phrase is said to be provided by an existential quantifier hidden in the expletive there, which existentially binds the free variable of the DP. Strong and strong-like noun phrases are ruled out in there-insertion by saying that their presence there would give rise to vacuous quantification on the part of there, which is deemed illicit. It is left as an open question how weak DPs find their quantificational force when they do not appear in there-insertion sentences, but as the subjects of SL predications, (cf. (2a)). Finally, as to why weak non- quantificational DPs are barred as IL-subjects and why only strong(-like) DPs are possible in this context, Milsark suggests that only quantificational DPs make good "topics," and being a topic is a prerequisite for IL predication. Adopting various important aspects of Milsark's analysis, Diesing (1992) takes up the question of how cardinal weak noun phrases in DE-environments receive their quantificational force when they are the subject of SL predicates. To this end, Diesing assumes, first, that weak Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 noun phrases are ambiguous in the sense Milsark proposes: some are genuinely quantificational (the strong-like ones), others introduce free variables. She then proposes that the latter all appear inside the VP. This optionally includes subjects of SL predicates, which— unlike the subject of an IL-predicate— are argued to optionally reconstruct into the VP. The VP, in turn, is held to be a locus of existential closure in the sense of Heim (1982) (see chapter 2). Thus, Diesing extends Milsark's analysis of weak DPs from those occurring in there-insertion contexts to a wider set of cases, including the subject of SL predicates. In addition to accounting for SL subjects, Diesing's theory also captures another, not yet mentioned observation of Milsark's; in contrast with strong(-like) noun phrases, cardinal weak noun phrases do not like to take inverse scope. This follows directly from the view these noun phrases are necessarily inside the VP at LF and, as a result of that, existentially closed in that very domain. Being quantificational, strong(-like) DPs,, are held to undergo Quantifier Raising (QR) at LF, whence they take wider scope." "In addition to the scope facts, Diesing uses the overt syntax of German as further evidence for the claim that the VP is the domain of DE-weak DPs and their existential closure. She observes that scrambled DPs in German are obligatorily strong or strong-like, whereas non-scrambled DPs can have Milsarkian readings. It should be noted in this context, however, that while the overt syntax of German is disambiguating with respect to scrambled DPs, it is not disambiguating with respect to non-scrambled DPs, for although Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 4.1.2. Problems: monotonicity and cardinality/symmetry: The existential closure of the Definiteness Effect is not without its limitations. I faces what we can monotonicity problem, on the one hand, and the cardinality/symmetry problem, on the other. The problems are quite different in nature. The monotonicity problem is intrinsic to the existential closure analysis and can perhaps be solved by resorting to certain, non-standard assumptions. The cardinality/symmetry problem, however, is an independent empirical limitation, which only becomes apparent once the effects of focus on weak DPs are taken into account. Unlike with the monotonicity problem, it does not seem that the cardinality/symmetry problem can be overcome without giving up the existential closure analysis of the DE. they can have Milsarkian readings, they do not have to. Also, strong DPs are in fact felicitous in non scrambled positions in German: (i) weil gestern ja doch [alle Vogel in den Siiden geflogen sind] because yesterday prt. prt. all birds in the south flown have "because yesterday all birds flew to the south" This fact is acknowledged in Diesing (1992, p. 52, 108.) What I take it to suggest is that although scrambled DPs might provide direct evidence for Diesing's Mapping Hypothesis, non- scrambled DPs in German do not lend themselves to this purpose. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill The monctcncity problem can be briefly stated as follows:- If the determiners of DE weak DPs are non- quantificational and the DPs are bound by a process of existential closure, then all DE weak DPs are in effect existentially quantified. This will provide a sentence like (la) with the logical form that is given in (3). The determiner here is treated as a group-denoting cardinality predicate, where the group is expressed in terms of a second order variable. (3) 3X (C(X) & some/many/few/no (X) & Vx 3x Xx & (Xx - child(x) & in the garden(x))) The problem with this is that by existentially binding the variables of the DP, this analysis predicts that weak elements should behave like other existentially quantified elements. In particular, they should be monotone increasing and should license inferences from the set denoted by NP to a superset of that set, e.g. an inference from D tall children did V to D children did V. Clearly though, not all DE weak DPs are monotone increasing and clearly, some are decreasing— those introduced by no, few and at most five, etc. The examples in (4), for instance, do not entail those in (5). Rather, the entailment goes in the opposite -The problem is also noted in a footnote 9 in Diesing (1992:143), cf. also Higginbotham (1987). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 direction, showing that the determiners in question are in fact decreasing: (4) There are no/few/at most five tall children in the garden, (5) There are no/few/at most five children in the garden. It might seem that the problem with decreasing quantifiers can be overcome if one assumes that the what the existential quantifier picks out in (3) is the empty set, of which it certainly holds true that whatever members it has, if any, are children in the garden. But if we appeal to the empty set then we would also predict that There are no children in the garden should be true when there are in fact children in the garden. Now, if we imposed a maximality condition on sets denoted by weak NPs such that the set has to contain all and only the children in the garden, for instance, we would correctly rule out that the sentence in question comes out true when no children are in the garden (cf, Diesing 1992, fn, 9), Still, the problem now is that by using maximality we would predict that the weak noun phrase like three children would virtually mean the same as the three children. But there is a clear difference; the three children suggests that these are all the children, but three children does not. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 One possible route to salvage the existential closure account would be to decompose a decreasing quantifier into a wide scope negation and an increasing quantifier, e.g. few would be not + many, and no would be not + some, and at most five would be no + more than five. Depending on how much importance is given to the lexical integrity of determiners, however, this move may or may not be possible. Moreover, notice that even if we give up the lexical integrity of determiners, we would have to find some way to ensure that the sentence under the wide scope negation is false precisely because the cardinality predicate does not hold— and not because any other predicate does not hold. Some additional, ad-hoc mechanism would have to be devised to ensure this. The easiest way to avoid the monotonicity problem is then to simply give up the claim that weak determiners are ambiguous between a non-quantificational reading and a quantificational one. Rather, weak determiners are always quantificational. And in a case like few the ambiguity lies in whether the determiner is symmetric or proportional (cf. Higginbotham 1987) . A determiner is then symmetric if all that matters is how many things satisfy both the restriction and the matrix. Here, the restriction and matrix can be interchanged without affecting the truth-conditions: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 180 (6) Symmetry: [DA] B * • [0 B] A Thus, a symmetric reading of few only requires that the number of things that satisfy both the restriction (A) and the matrix (B) be relatively small; whether these amounts to all things that are A (=100% scenario) is irrelevant. On the other hand, proportional few is not symmetric. It demands that only a contextually determined small proportion of the things that satisfy the restriction also satisfy the matrix. Here what has to be small is the number of things that are A vis-a-vis the number of things that are both A and B, which rules out the 100% scenario. Since this account treats all weak determiners as quantificational and does not make use of existential closure, it has no problem monotonicity. What still has to be explained on this view is the correlation between determiner types and environments of the DP, and perhaps also why we should find this kind of ambiguity. Setting these questions aside here, it now turns out that even if we replace the property of cardinality with symmetry, thereby avoiding the monotonicity problem, we still have not yet picked out the right DPs that are possible in a DE-environment; once we look at focus inside the noun phrase, we find that it is not a defining property of DE weak DPs that their determiners be symmetric. Some in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 fact turn out to have proportional determiners, namely the focus-affected readings. 4.2. F-a readings : 4.2.1. The basic phenomenon Consider the following example: (7) Few COOKS applied. In (7) we see a weak DP in the subject position of a SL predicate. Disregarding focus for the time being, according to the description that was summarized above, this sentence should have exactly two readings: a Milsarkian one, where few is symmetric, and a strong-like one, where few is proportional : (8) a. [FeW;.__ x: C(x) & cook(x)] 3e apply (e) & Agent (e,x) Milsarkian b. [Few_„„_ X: C(x) & cook(x)] 3e apply (e) & Agent (e,x) strong-like Taking focus into account, there is a third reading of (7), which can be paraphrased a follows: "few that applied were cooks :" Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 (9) [Fevjp„j,. x; C(x) 3e apply (e) & Agent(x,e)] COOKS (x) focus-affected Here the determiner of the DP is not interpreted as symmetric but as proportional instead. Unlike in the strong-like reading, however, (cf. (8b)) few quantifies over those that applied, rather than over cooks; it is claimed that most of these applicants where cooks. It is possible to confirm that the f-a reading is really available by verifying that (7) can be judged true in a scenario like the one outlined in (9). (10) The fellowship committee is sorting through the applications for funding to Paris. Without knowing how many applications there are, at an early point during their review, they observe that on average only every twentieth application is by a cook, which is lower than they expected. (7) can be true under circumstances like (10), which supports the claim that (7) can have a reading where it means that few that applied were cooks. First, the committee does not know the total number cooks or what overall percentage of cooks applied. This precludes any analysis of the DP as a strong-like proportional; a small percentage of the cooks applying is not compatible with all Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 the cooks applying. Second, given that the committee does not know the overall number of applicants, it will be hard for them to see if their overall number amounts to few, which rules out the reading where the DP is introduced by a symmetric few, as in (8a). On the other hand, the committee does have all the necessary information to establish the truth conditions of the f-a reading: they know the ratio of applying cooks to applicants as such, and they know it is small. The fact that (7) can be uttered felicitously in a context like (10) thus supports the claim that it has a f-a interpretation. Descriptively, in a case like (7) the focus inside the NP in a f-a reading functions as the matrix of the determiner and the non-focused part functions as the restriction. The existence of f-a readings shows that— contrary to standard assumptions— nominal determiners do not always have their restriction determined by the syntax. Rather, sometimes they behave in the way adverbs of quantification do in that the restriction seems to be entirely determined by the focus. How the two can be assimilated will be discussed in 3. Before that I want to show that f-a readings are restricted in their distribution in that they arise exactly in DE-environments. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 4.2.2. The DE on f-a readings : Going through various cases, this section elaborates on the distributional properties of f-a readings, showing how they are only possible in weak DPs that occur in DE- environments. The relevant examples are taken from English and German. As a first piece of evidence let us consider the following paradigm: (11) a. Few/many/no/three/some COOKS applied. b. Most/all/every/each/neither COOK(S) applied. c. Few/many/no/three/some COOK(S) know how to make a soufflé. (11a), where a weak DP occurs in an optional DE-environment, namely in the subject position of a SL predicate, can be interpreted in f-a manner, and can be paraphrased as "few/many/none....that applied were cooks:" (12) [D x: C(x) 3e apply(e) & Agent(x,e)] COOK(x) f-a reading In addition to its f-a reading, I think (11a) also has a strong-like interpretation (Q that were cooks (rather than something else) applied) and a Milsarkian reading. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 Unlike in (11a), in (11b), where the DPs are strong, no f-a readings are possible, for (11b) cannot be paraphrased as "most, all, etc. that applied were cooks." Finally, in (11c) the DPs are weak, but they occur in an anti-DE environment, namely as the subjects of an IL predicate. Just as in (lib), in (11c) no f-a readings are available. In particular, this sentence cannot mean few/many/most people that know how to make a souffle are cooks but only that few/many/most people that are cooks know how to make a souffle.' Next, notice that f-a reading can also appear in the paradigm case of the DE, namely there-insertion contexts: (13) a. There are many/few speakers of Basque THAT ARE CITIZENS OF SPAIN. b. There are many/few citizens of Spain THAT ARE SPEAKERS OF BASQUE. 'Another good example to consider is (i), which is due David Pesetksy (personal communication): (i) Few SALVADOREANS speak Spanish. The predicate being IL-level, the sentence does clearly not have a f-a reading, which would be "few of those that speak Spanish are Salvadoreans." This is so despite the fact that this reading would be more in accordance with our knowledge of the world, and thus favored if at all possible, than the strong-like reading it actually has: A small percentage of the Salvadoreans speak Spanish. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 Taking many, for instance, it is possible to judge the first one true and the second one false, (and conversely with few). Since the truth-conditions of (13a) and (13b) differ, we can conclude that many and few cannot be considered symmetric here, (If they were, switching speakers of Basque and citizens of Spain should, by definition, not have any effect on the truth conditions whatsoever, but it does). Rather, it seems the reason that (13a) and (13b) have different truth-conditions is that they have the f-a readings given in (14a) and (14b), respectively: (14) a. [Manyprop. x: C(x) 3e (be speaker-of-Basque(e) & Theme(e,x))] CITIZEN-OF-SPAIN(x) b. [Manyprop. x: C(x) 3e (be citizen-of-Spain(e) & Theme(e,x))] SPEAKER-OF-BASQÜE(x) (14a) asserts, truthfully it seems, that many speakers of Basque are citizens of Spain. (14b), on the other hand, means that many citizens of Spain are speakers of Basque. And— despite the relative vagueness of many— this seems false to most people; after all, most Spanish citizens do not speak Basque, not even most Basques.' It may appear that in (13) the focused predicate (the relative clause) is not inside the NP in the overt syntax but a sister of it. If so, one could suggest that the relative clause is the matrix not because it is focused, but because of its syntactic position. Notice, however, that f-a reading arise equally in (i) and (ii), where the focused predicate is unequivalcally the NP in the overt syntax: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 Further evidence that shows that f-a readings only arise with weak DPs comes from scrambling in German. As Diesing (1992) discusses, in German scrambled positions pattern with subjects of IL predications in English, in so far as scrambled positions constitute an anti-DE environment. On the other hand, we note that non-scrambled positions pattern with the subjects of SL predications in English in being optional DE-environments (cf. fn 2). As (15) shows, DPs in scrambled positions behave as we expect them to: since they pattern with strong(-like) DPs, they do not license f-a readings. On the other hand, provided they are weak, clearly do have a f-a interpretation: (15) a. weil viele/einige/wenige SCHWALBEN ja doch in den Siiden fliegen because many/some/few swallows prt.prt.to the south fly b. weil ja doch viele/einige/wenige SCHWALBEN in den Siiden fliegen because prt.prt.many/some/few swallows to the south fly (15a), where the weak DP has scrambled cannot be interpreted in a f-a manner as "Q things that are flying to the South ;i) There are many SPANISH Basque speakers. ;ii) There are many BASQUE-SPEAKING Spaniards. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 are swallows." It only allows for a strong-like reading of the weak DP "Q things that are swallows fly to the South." But in (15b), where the weak DP has not scrambled, both readings are possible. Finally, another respect in which f-a weak DPs behave like Milsarkian weak DPs is their scope properties. Like Milsarkian DPs, f-a DPs do not seem to like to take inverse scope. Consider for instance (16): (16) Every teacher flunked many POOR kids. On its f-a reading, (16) can be paraphrased as "Every teacher is such that many kids that s/he flunked were poor." Under this reading, every must take scope over many. It is not possible to have a f-a reading where many takes wide scope, along the lines "Many kids that every teacher flunked were poor," even though this reading would be pragmatically as plausible as the reading that it actually has. What these data suggests then is that focus on weak DPs in DE-environments can give rise to a f-a reading, where the focus assumes the role of the matrix of D (rather than the IP) and the non-focused part the role of restriction (rather than the NP that is the sister of D). Crucially, f-a readings are only possible in DE-contexts. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 4.2.3. F-a readings and Conservativity; In the context of f-a readings, it is interesting to consider the following pair of sentences, which Westerstahl (1985) shows to contrast in an interesting way in their semantic interpretations: (17) a. Many Scandinavians have won the Nobel prize in literature. b. Most Scandinavians have won the Nobel prize in literature. What Westerstahl observes is that the first sentence can have a reading where the restriction of the determiner is not the NP, and where its matrix does not correspond to the VP. This reading corresponds to: "many that are winners of the Noble prize in literature are Scandinavians." He also points out that when we replace many with most, as in the second example, such a "switched" reading is not possible; (17b) can only be interpreted as saying that the majority of Scandinavians is such that they have won the Nobel prize in literature, but not along the lines of "most of those that have won the Nobel prize in literature are Scandinavians." Following standard views, Westerstahl holds that the first argument of D is necessarily provided by the predicate denoted by the NP and where the second argument is given by Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 the open sentence that denoted by the IP after QR. This coupled with the assumption that nothing further is going on, he concludes that many is a non-conservative determiner in this example. All natural language determiners are normally held to be conservative in that they obey the principle in (18): (18) Conservativity: [Q A]B iff [Q A]B n A (Keenan and Stavi 1986) Counting Scandinavians as the first argument (A) and have won the Nobel Prize in literature as the second one (B), we find that, contrary to (18), it does not hold that many that won the Nobel Prize in literature are Scandinavians (= switched interpretation of (17a)) if and only if many that are Scandinavian have won the Nobel Prize in literature and are Scandinavian. Westerstahl takes this suggests that there is one many, which occurs in a "switched" interpretation, and which unlike the other natural language determiner, and a homophonous many, is not conservative. Many in (17a), however, need not be seen as flying in the face of Conservativity. Notice that the switched reading arises only when Scandinavians is focused: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 191 (19) a. Many SCAMDINAVIAiiS have won the Nobel prize in literature. b. Many Scandinavians have won THE NOBEL PRIZE in literature. Only (19a) can mean that many that have won the Nobel prize in literature are Scandinavian, but not (19b), which means something along the lines of "what many Scandinavians have won is the Nobel prize in literature." In light of this, it seems fair to say that the switched reading is really an instance of f-a quantification, rather than the result of a non-conservative interpretation of many. Thus, (19a) really has a logical form like in (20), where the non-focused part forms the restriction and the focus the matrix: (20) [Manyp„p.x: C(x) & have-won-the-NP-in-literature(x)] SCANDINAVIAN(x) f-a Given the logical form in (20), it is now the non-focused part, and not the NP, that corresponds to "A" in (18), and, furthermore, it is the focus that corresponds to "B", not the IP. Under this way of partitioning the lexical material into restriction and matrix Conservativity is in fact obeyed; it does hold that many that have won the Nobel Prize in literature are Scandinavian if and only if many that have Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 192 won the Nobel prize in literature are Scandinavian and have won the Nobel prize in literature. Furthermore, if we analyze the switched reading of (17a) as a f-a reading, we can not only maintain that many is always conservative, but we also correctly predict that (17b), where the determiner is strong {most), will not have such a reading; here focus on Scandinavians does not affect the quantificational structure of the determiner (see 4.3.3.). 4.3. Determiners that work like adverbs 4.3.1 Semantically unary vs. binary In light of the three different readings of weak DPs, the strong-like, Milsarkian and the f-a, how can we analyze weak DPs? We noted that since f-a readings share the same distribution as Milsarkian weak DPs (modulo focus), the determiners of weak DPs cannot be generally characterized as symmetric. This is so because the determiners of f-a weak DPs have the same basic meaning as the determiners of strong-like weak DPs and the only difference between a strong-like reading and a f-a one lies in what part of the sentence makes up the restriction and what part gives the matrix and the scopal property (i.e. strong-like DPs can take wider scope than f-a ones). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 Beginning with the interpretation of the determiner on the Milsarkian reading, I think it is best to assume that there are quantificational few's, a proportional one which occurs in f-a and strong-like readings, and also a symmetric one, which appears in the Milsarkian reading (see above). It may be ultimately possible to reduce one to the other, but it is not obvious how and I will not attempt to develop such an analysis here, but rather assume that there is a lexical ambiguity.' Given the symmetric meaning of the determiners in Milsarkian reading, we can further assume that in the Milsarkian reading the determiners are unary, or "absolute," cf. Higginbotham (1987), Dobrovie-Sorin (1991), Musan (1995). What makes this possible is that the determiners are always symmetric. Therefore, we can analyze it as unary as much as we can analyze it as binary; by definition, commuting or collapsing the arguments has no effect on the truth-conditions of the sentence. Under this analysis, a sentence like (21) now has the logical form in (22): the 'In Herburger (1995) I attempt to show that cardinal few can be treated as a unary proportional few, meaning 'few of the contextually relevant things'.Clearly, the problem is how to establish what the set of contextually relevant things is. I try to appeal to the general notion of context that is needed for every quantifier, but it is not clear that this will work. One problem is, for instance, that (i) can be uttered when I am comparing the number of children in the garden with the number of fathers in the garden. In such a case, it is not clear what the set of things would be few of whose members are children in the garden: (i) There are few children in the garden. Reproduced with permission of the oopyright owner. Further reproduction prohibited without permission. 194 determiner is only restricted by the context predicate C(x), but not by any lexical material; we can say, it is "lexically" unary: (21) There are some/no/three/few children in the garden. (22) [Some/no/three/fewj..^^ x: C(x)] 3e be(e) & child(x) & Theme(e,x) & in the garden(e) "Some/no/three things that are relevant are such that there is an event where they are children in the garden." Whereas in the Milsarkian reading the determiner remains semantically unary and unrestricted (except for the ever-present context predicate), in the strong-like and f-a reading the determiner is semantically binary— with different constituents of the sentence providing their respective arguments in each case, as we have seen. 4.3.2. D-raising and QR: How do these logical forms derive, in particular how do they relate to the syntactic structure (LF) and what work does the focus do? Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 Various ways have been proposed in the literature in which a determiner can take its scope at LF. Classically, the determiner is assumed to raise together with its sister NP by Quantifier Raising (QR), thereby adjoining to IP. Alternatively, it has also been proposed (in various theoretical contexts and for various reasons) that only the determiner raises, stranding the NP (cf. e.g. Heim 1982, Hornstein and Weinberg 1990).* Let us assume that both ways of taking scope exist, but with a division of labor between them, cf. Higginbotham (1987), Herburger (1995), cf. also Chomsky (1995). While strong determiners relate to their NP complement such that they have to take it along when they move to take scope (=QR), weak determiners are not glued to the NP in the same way. Rather, weak determiners have the option of raising without their NP complement, which results in D-raising. At the same time, weak determiners also can function like the strong ones do, in which case they take scope by QR: *Both these views share the assumption that DPs take scope by creating operator-variable structures at LF, which is what I will be assuming too. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 IP / IP QR b. TP D-raising ,T VP [t, NP] QR and D-raising bridge different distances; because 0- raising involves more local head movement, it is predicted that this way of taking scope will result in narrower scope Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 than will QR. Given the LFs in (23), when there is QR the NP will always be the internal argument of D, analogous to the object of a transitive verb, which provides its internal argument. Because of this, it will necessarily function as the restriction of the determiner (cf. Larson and Segal 1995). On the other hand, assuming that all determiner must have a matrix, when the determiner has only one argument, as in the case of D-raising, this argument will necessarily correspond to its matrix. The idea now is this: since strong and strong-like noun phrases take scope by QR, they are always restricted by the NP, and they can take wide scope. On the other hand, both Milsarkian and f-a weak readings result when a weak determiner takes scope by D-raising. Consequently, the determiner is syntactically unary. In relevant respects, in looks like a typical adverb of quantification. Being syntactically unary is just what is needed for the Milsarkian reading, given our earlier argument that these determiners are semantically unary (see above). As for the f-a reading, here— like an a case of adverbial quantification— the determiner is unary syntactically, but it is not semantically: it is restricted by the non-focused material in its scope. Analogous to the SDD, the effect of focus on nominal quantification can now be stated as follows : Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 9 8 (24) Structured quantification by determiners: The non-focused material within the scope of a unary determiner is copied into its restriction. This semantic process will provide us with the logical form in (25) for our example (7) (7) Few COOKS applied. (25) 3e C(e)([Few x: C(x) & Agent(e,x)& apply(e)] COOK(x) & Agent(e,x)) & apply(e)] "There is a (relevant) event such that few who were agents in the applying that it was were cooks (and agents in the applying that it was.)" Leaving out the redundant material in the matrix, this corresponds to (26): "“in an earlier version, I accounted for f-a reading by invoking movement of focus at LF (Herburger 1993) . Irene Heim (p.c.) dissuaded me of this and suggested pursuing an analogy with adverbs, based on the following examples, which do have f-a readings but where the requisite focus movement would be non-local : (i) Few cooks from JOHN'S school applied. "Few of the cooks from some school that applied were cooks from John's school." (ii) [Few INCOMPETENT cooks from FRANCE] applied. "Few of the cooks that applied were incompetent French cooks." Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 (26) 3e C(e) [few x: C(x) & apply(e) & Agent(e,x)] COOK(x) "There is a (relevant) event such that few that were agents in the applying that it was which was an applying were cooks." F-a readings are not limited to cases where focus falls on an immediate constituent of a weak DP (the NP, that is). They also arise when subconstituents of the DP are focused. This poses no problem: (27) Few INCOMPETENT cooks applied. This sentence can mean that few cooks that applied were incompetent, i.e. it can have a f-a reading, where the non- focused part makes up the restriction of few and the focused part makes up its matrix: % (28) 3e (C(e) [few x: C(x) & apply(e) & Agent(e,x) & cook(X)] INCOMPETENT(x)) "There was an event such that few of the cooks that were applying in it were incompetent." Given the logical forms in (25)-(28) one might wonder why the event operator takes wide scope over the focus affected determiner. Prima facie it would seem better to assign it narrow scope because assigning wide scope runs the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 risk of wrongly predicting that the sentence is true vacuously: notice that pretty much any event counts as an event where few (and thus possible no) cooks applied, a car washing does, as does a thunder storm, and somebody buying a newspaper, etc. The problem is only apparent though; as long as C(e) ensures that a contextually relevant event is picked out, the truth-conditions will not be vacuous. Thus, although the wide scope of the event over the f-a decreasing determiner might seem problematic at first.-' One reason to assign the event quantifier the very local nature of D-raising, which also captures the lack of inverse scope exhibited by Milsarkian and f-a weak readings. Another argument in favor of having the event operator takes scope over a f-a determiner is suggested by the following example:'- (29) Few COOKS ever applied. (29) does not have a f-a reading, that is, it can not mean that few of those that ever applied were cooks. If we consider ever as a negative polarity event operator here. "Recall the following example,already cited earlier in chapter 2 (fn. 12), which shows that event descriptions that are phrased in negative terms need not be vacuous and can be used sensibly. (i) Once, noone knows when, nobody arrived. "I owe this argument to Anna Szabolcsi. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 this follows; if the event takes scope over a f-a determiner, few will not take scope over ever, which, as a result, will remain unlicensed. In the context of (29), it is also interesting to note that it provides an argument against accounting for f-a readings in terms of DRT-type tri-partite structures. As proposed in Partee (1991), in such an analysis, the focused predicate is mapped into the nuclear scope and the non- focused material goes into the restrictive clause. The problem is that in these tri-partite structures the determiners are assumed to take scope over the entire sentence. This does not capture the narrow scope of f-a weak DPs and it would also wrongly predict that (29) has a f-a reading. I would also like to emphasize the important role that D-raising plays on the present analysis. It is because the determiners of DE weak DPs undergo D-raising (and are hence syntactically unary) that focus inside the NP can have such a drastic effect on their interpretation. Thus, if the determiner of f-a DPs were assumed to take scope not through D-raising but through DP-raising instead, the effect of focus would not be adequately captured because the focused predicate within the NP would necessarily have to be part of the restriction of the determiner. But, crucially, in a f-a reading he focused element within the NP does not contribute to the determiner's restriction, but to its matrix instead. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 2 Assuming that DPs always take scope via QR, Geilfufl (1993) proposes an analysis where the union of the focus- semantic value is always intersected with the domain variable of a quantifier. While he is mostly concerned with focus inside the VP, which is captured under this analysis, he also briefly notes the existence of f-a reading. To these, however, the analysis that assumes QR (or DP- raising) , rather than D-raising, does not extend, for the reasons just discussed. This is also observed in de Hoop and Sola (1995) .^ 4.3.2. Focus inside the VP : In this last section, I would like to consider how focus affects strong-(like) QPs. How this can be dealt with on the present analysis? Let us begin with the examples (30): -"Maintaining a general QR analysis, de Hoop and Sola attempt a different description of the facts where f-a readings are said to not require any focus inside the NP and where f-a readings are attributed to a general context- sensitivity of few and many. Thus, the authors propose that (i) and (ii) have f-a readings even when there is no focus on the noun, but when main stress falls somewhere else, e.g. the VP in (ii). I have not been able to replicate this claim. (i) There are few linguists in the pub. (ii) Many Scandinavians have won the Nobel prize in literature. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 203 (30) a. Many Scandinavians won THE NOBEL PRIZE IN LITERATURE. b. Most Scandinavians won THE NOBEL PRIZE IN LITERATURE. These sentences mean roughly what many (or most) Scandinavians won was the Nobel prize in literature— and nor the Tour de France, for instance. Since the focus does not fall within the NP, it does not matter whether the noun phrase is weak or strong(-like); given that the focus here is not on a constituent of an NP but on an immediate constituent of the sentence, the structuring effect that focus has in these examples will be at the level of quantification over events. When many Scandinavians is interpreted as a Milsarkian DE-DP many D-raises. As a result, many only has one syntactic argument, Scandinavians have won the Nobel prize in literature. The SDD provided the following logical form (leaving out redundant material in the matrix); (31) [3e: C(e) [many x: C(x)] Scandinavian(x) & won(e) & Experiencer(e,x)] Theme(e, THE NOBEL PRIZE IN LITERATURE) -’In the full-fledged logical from, the quantifier many Scandinavians also appears in the matrix. This does not give rise to a problem of requantification because, given the exhaustivity of theta-roles, both in the restriction and the matrix the same Scandinavians will be picked out (see 3.2.2). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 204 It comes out as meaning that there is an event were many are are Scandinavians winning and the theme of that event is the Nobel Prize in literature. This is a somewhat round-about way of expressing the meaning of (30), but it gets the truth-conditions right. On the other hand, when many is strong-like, or when the determiner is most, the determiner it takes scope by QR, as a result of which it is restricted by the NP Scandinavians. Assigning focus to the VP in such a case gets an interpretation were for many (or most) Scandinavians some (relevant) event of winning where they were the agent had as its theme was the Nobel Prize in literature: (32) [Many/most x: C(x) & Scandinavian(x)] [3e: C(e) & won(e) & Experiencer(e,x)] Theme(e, THE NOBEL PRIZE IN LITERATURE) Thus, focus on the VP does not pose any particular problem. It is accounted for by the SDD. The effects of such focus on the relevant noun phrase are only indirect in that focus here structures the material the noun phrase takes scope over. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 205 4.3.3. Focus in strong(-like) DPs : Finally, let us also consider the effects of focus within a strong(-like) DP. They are somewhat harder to explain and what I have to offer at this point are some suggestive remarks. We saw earlier that focus in the NP of a strong(-like) DP does not alter the quantification structure of the determiner. The reason is that here the syntax already establishes that the NP has to restrict D. Focus can shape quantification structure in the absence of any syntactic structure. But when the syntactic structure directly affects the quantificational structure, focus cannot override that. Now even though in a strong(-like) DP the NP has to be part of the restriction of D— for the reasons just given— this does not mean that focus has no interpretive effect here. We noted that in a case like (33), for instance, the focus on SCANDINAVIANS can contrast Scandinavians with other people, e.g. Italians: (33) Most SCANDINAVIANS have won the Nobel prize in literature. Intuitively, (33) means something like "most such that what they are is Scandinavians (and not Italians) are such that they have won the Nobel prize in literature." The NP Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 206 necessarily being part of the restriction of D, the effect that the focus on Scandinavians has seems to be limited to within the NP. One possibility to explore is that there is more internal structure to the DP than meets the eye. We can say that most Scandinavians contains a covert verb be, it would then be equivalent "most that are Scandinavians." Taking this one step further, we can assume that the be, like every verb, introduces an event operator. This event operator, like others, can be structured by focus; the result of this would be something of the sort "most of those such that there being is a being Scandinavian," which comes close to my initial paraphrase ("most such that what they are is Scandinavian.") On this view then, the effects of focus within an NP which, because of the syntax, necessarily restricts D, are analogous to the effects of focus in if- clauses and preverbal when-clauses described in chapter 3, which also necessarily restrict their quantifier, in this case an adverb: in both cases there is an embedded quantification and this quantification in the restriction of the overt quantifier is structured by focus in a regular way. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 207 4.4. Summary of chapter 4: The point argued for in this chapter is that the overt syntax of quantificational noun phrases is in a certain set of cases less indicative of their quantificational structure than is standardly assumed. Focus inside the NP of noun phrases that exhibit the Definiteness Effect gives rise to f-a readings where all the non-focused material inside the scope of the determiner is interpreted as a restriction and the focus contributes to the matrix. Thus, in relevant respects f-a determiners behave very much like the typical adverb of quantification. It was proposed that when a weak DP exhibits the Definiteness Effect it takes scope by D- raising, which accounts for their narrow scope. In the case of the Milsarkian reading, the determiner is both syntactically and semantically unary; this is possible because the determiners in this interpretation are all symmetric. On the other hand, in the f-a reading, the determiner is syntactically unary, but not semantically: as with adverbs of quantification, focus here has a structuring effect in that the non-focused material provides a restriction for the determiner. Finally, as far as the Definiteness Effect is concerned, the existence of f-a reading indicates that the determiners are clearly not cardinality predicates, but genuinely quantificational. This raises a number of questions about the Definiteness Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 8 Effect. Hopefully, at some other time and place they can be answered. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 209 BIBLIOGRAPHY : Berman, Stephen, 1987, "Situation-based Semantics for Adverbs of Quantification" UMOP 12: 45-58, Bauerle, Rainer and Urs Egli, 1985, "Anapher, Nominalphrase und Eselssatze", Papier 105 des Sonderforschungsbereichs 99, Universitat Konstanz, Bosque, Ignacio, 1980, Sobre la Negacion. Madrid: Catedra, Carlson, Lauri, 1984, "Focus and Dialogue Games: A Game- Theoretical Approach to the Interpretation of Intonational Focusing" in Vaina, L, and J, Hintikka, eds,, Cognitive Constraints on Communication. Kluwer, Chomsky, Noam, 1971, "Deep Structure, Surface Structure, and Semantic Interpretation," In Steinberg D, and L, Jakobovits, eds,. Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology. Cambridge: Cambridge University Press, Chomsky, Noam, 1995, The Minimalist Program. Cambridge: 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 Rescher, Nicholas (ed,) The Logic of Decision and Action. Pittsburgh: Univ. of Pittsburgh Press, Diesing, Molly, 1992, Indefinites. Cambridge: MIT Press, Dobrovie-Sorin, Carmen, 1993, "Referentiality, scope and the ECP" ms, CRNS, Dowty, David, 1989, "On the Semantic Content of the Notion 'Thematic Role'," In Chierchia, Gennaro, Barbara Partee and Ray Turner, eds,. Properties, Types and Meaning vol 2, Dordrecht: Kluwer, Dretske, Fred, 1972, "Contrastive Statements" Philosophical Review 81: 411-437, Dryer, Matthew, 1994, "The Pragmatics of Focus-Association with Only”, LSA Meeting, Boston, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 210 Ferro, Lisa. 1994. "Interpreting Coreference in Sentences with Focus: A Discourse-Based Alternative to Weak Crossover" ms. University of Connecticut at Storrs. Geilfufi, Jochen. 1993. "Nominal Quantifiers and Association with Focus' Proceeding on the Semantic and Syntactic Analysis of Focus OTW Working Papers, Research Institute for Language and Speech, Utrecht University, 33-41. Fintel, Kai von. 1994. "Restrictions on Quantifier Domains" PhD dissertation, Univ. of Massachussetts, Amherst. HajiCova, Eva. 1984. "Presupposition and Allegation Revisited" Journal of Pragmatics 8, 155-167. Heim, Irene. 1982. "The Semantics of Definite and Indefinite Noun Phrases" PhD dissertation, Univ. of Massachussetts, Amherst. Heim, Irene. 1990. "E-type Pronouns and Donkey Anaphora" Linguistics and Philosophy 13: 137-177. Herburger, Elena. 1995. "On the quantificational nature of indefinite Noun Phrases" Langue et Grammaire I, 169-183. Herburger, Elena. 1996a. "Focus and Weak Noun Phrases" Natural Language Semantics, in press. Herburger, Elena. 1996b. "'The Road to Nowhere' and the Distribution of Spanish N-words" LSRL 26. Higginbotham, James. 1987. "Indefiniteness and predication" in Reuland Eric and Alice ter Meulen, eds.. The Representation of (In)definiteness. Cambridge: MIT Press. de Hoop, Helen and Jaume Sola. 1996. "Determiners, Context Sets, and Focus" WCCFL 14. Horn, Laurence. 1989. A Natural History of Negation. Chicago: The University of Chicago Press. Hornstein, Norbert and Amy Weinberg. 1990. "The necessity of LF" The Linguistic Review 7: 129-167. Jackendoff, Ray. 1972. Semantic Interpretation in Generative Grammar. Cambridge: MIT Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 211 Jacobs, Joachim. 1989. "Negation". In Arnim von Stechovj and Dieter Wunderlich, eds., Semantik: Ein internationales Handbuch der zeitgenossischen Forschung' Berlin: Walter de Gruyter. Kadmon, Nirit. 1987. "On Unique and Non-unique Reference and Asymmetric Quantification" PhD dissertation, Univ. of Massachusetts, Amherst. Kadmon, Nirit. 1989. "Uniqueness" Linguistics and Philosovhv 13: 273-324. Krifka, Manfred. 1992. "A Framework for Focus-sensitive Quantification". SALT 2. Krifka, Manfred. 1995. "Focus and/or Context: A Second Look at Second Occurrence Expressions" ms, Univ. of Texas at Austin. Ladd, D. Robert. 1980. The Structure of Intonational Meaning. Bloomington: Indiana University Press. Ladusaw, William. 1993. "Expressing Negation". SALT 2. Laka, Itziar. 1990. "Negation in Syntax", PhD dissertation, MIT. Larson, Richard and Clair Lefebvre. 1991. "Predicate Clefting in Haitian Creole." Larson, Richard and Gabriel Segal. 1995. Knowledge of Meaning. Cambridge: MIT Press. Lewis, David. 197 5. "Adverbs of Quantification." In Ed Keenan, ed.. Formal Semantics, Cambridge: Cambridge University Press. Ldbner, Sebastian. 1990. Wahr neben Falsch. Tübingen: Max Niemeyer Verlag. Ludlow, Peter. 1994. "Conditionals, Events, and Unbound Pronouns" Lingua e Stile 29: 165-183. Milsark, Gary. 1977. "Toward an Explanation of Certain Peculiarities of the Existential Construction in English" Linguistic Analysis 3: 1-29. Musan, Renate. 1995. "On the Temporal Interpretation of Noun Phrases" PhD dissertation MIT. Neale, Stephen. 1990. Descriptions. Cambridge: MIT Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 212 Ogihara, Toshiyuki (1987) "'Obligatory Focus' in Japanese and Type-Shifting Principles" Proceedings from WCCFL 6. Parsons, Terence. 1990. Events in the Semantics of English: A Study in Subatomic Semantics. Cambridge: MIT Press. Partee, Barbara. 1988. "Many Quantifiers." ESCOL 5. Partee, Barbara. 1991. "Topic, Focus and Quantification",SALT 1. Rochemont, Michael S. 1986. Focus in Generative Grammar. Amsterdam: John Benjamins. Rooth, Mats. 1985. "Association with Focus" PhD dissertation, Univ. of Massachussetts, 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 Bosch, Peter and Rob van der Sandt (eds). Focus and Natural Language Processing Working Paper of the Institute for Logic and Linguistics 7, Vol 2, 389-398, IBM Germany. Rooth, Mats. 1994b "Focus" in Lappin, Shalom, ed. The Handbook of Contemporary Semantic Theory, Blackwell Publishers. Rooth, Mats. 1995a. "Evidence for Focus in Second Occurrence Focus" handout. Rooth, Mats. 1995b. "Indefinites, Adverbs of Quantification, and Focus Semantics" in Carlson, Gregory and Francis Jeffry Pelletier The Generic Book. Chicago University Press. Russell, Bertrand. 1905. "On Denoting" reprinted in Martinich, A.P. ed The Philosophy of Language 2nd Edition, Oxford: Oxford University Press. Selkirk, Lisa. 1984. Phonology and Syntax. Cambridge: MIT Press. Selkirk, Lisa. 1995. "Prosodic Reflexes of Focus" talk given at UMASS Focus Workshop, December 1995 Schein, Barry. 1993. Plurals and Events. Cambridge: MIT Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 213 Soames, Scott. 1989. "Presuppositions". In Gabbay, D. and ; Guenthner, eds., Handbook of Philosophical Logic 4, Reidel. Stechow, Arnim von. 1989 "Current Issues in the Theory of Focus". In Arnim von Stechow and Dieter Wunderlich, eds., Semantik: Ein internationales Handbuch der zeitgenossischen Forschung’ Berlin: Walter de Gruyter de Swart, Henriette. 1991. "Adverbs of Quantification: A Generalized Quantifier Approach" PhD dissertation, Rijksuniversiteit Groningen. Uribe-Etchevarria, Myriam. 1994. "Interface Licensing Conditions and Negative Polarity Items: A Theory of Polarity and Tense Interactions". PhD dissertation, Univ. of Connecticut at Storrs. Vallduvi, Enric. 1990. "The Informational Component". PhD dissertation, Univ. of Pennsylvania. Westerstahl, D. 1985. "Logical Constants in Quantifier Languages". Linguistics and Philosophy 8: 387-413. Zanuttini, Raffaela. 1991. "Syntactic Properties of Sentential Negation: A Comparative Study of Romance Languages". PhD dissertation, Univ. of Pennsylvania. Zubizarreta, Maria Luisa. 1996. "Prosody, Focus and Word Order", ms. USC. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

On the nature of particles in Japanese and its theoretical implications

PDF

Syntactic structures in nominals: A comparative study of Spanish and Southern Quechua

PDF

Evaluating heaviness: Relative weight in the spoken production of heavy-NP shift

PDF

The phonemics and morphology of Hokkaido Ainu

PDF

Representation of focus and presupposition in Japanese

PDF

The syntax of focus and topic in Mandarin Chinese

PDF

Correspondence and faithfulness constraints in optimality theory: A study of Korean consonantal phonology

PDF

Preview of the vanishing hero: a study of the protagonists in Jacobean drama

PDF

The syntax of the Chinese BA-constructions and verb compounds: A morphosyntactic analysis

PDF

The syntax of clitic doubling in modern Greek

PDF

The erudite romanticism of Juan Eugenio Hartzenbusch in 'Los Amantes de Teruel'

PDF

The act and the word: A study of abstraction versus the concrete in the work of Albert Camus

PDF

On the representation and licensing of Q and Q-dependents

PDF

Nation, race, and heterogeneity in the literature of the Americas

PDF

Painting as text: Developments in Russian art during the second half of the nineteenth century

PDF

Identity in crisis: Modernism and the texts of adolescence

PDF

Emotions: Linguistic representation and cultural conceptualization

PDF

Frequency in sentence comprehension

PDF

Women and the unspeakable: Rape in Ovid's "Metamorphoses"

PDF

Countercultural responses to the crisis of masculinity in late republican Rome

Asset Metadata

Creator Herburger, Elena (author)

Core Title In the event of focus

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-263344

Unique identifier UC11352881

Identifier 9733067.pdf (filename),usctheses-c17-263344 (legacy record id)

Legacy Identifier 9733067.pdf

Dmrecord 263344

Document Type Dissertation

Rights Herburger, Elena

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA