Modification, Summaries of English Language

Download Modification and more Summaries English Language in PDF only on Docsity!

DRAFT

Modification

Marcin Morzycki

Michigan State University

June 12, 2013 Latest revision: June 2014

This is a draft of a book in preparation for the Cambridge University Press series Key Topics in Semantics and Pragmatics. It’s something between a textbook for people with a basic background in semantics and an introduc- tory survey of work on the semantics of adjectives, adverbs, and degrees. For a fuller explanation, see section 1.2. Comments would be extremely helpful , so please don’t hesitate to contact me if you have any, even very minor ones.

DRAFT

Acknowledgments

I ran into a Polish expression recently: nie mój cyrk, nie moje małpy (‘not my circus, not my monkeys’). It means ‘ain’t my problem’. Well, the thing about any published piece of work is that it’s always your monkey. You’re responsible, no matter how much of a mess it has made or how much it has failed to adequately address measure phrases across categories in chapter 6. And a book is a particularly big monkey, so there’s no telling what disasters it might have wrought that you’ll only discover once the monkey appears in print. For that reason, I am indebted to the people below, who have helped keep my monkey in line. I received particularly extensive, helpful, and insightful written com- ments about the entire book from Alan Bale and Tom Ernst. Many others provided helpful feedback on various scales, both specific and general. Among them were (alphabetically) Alexis Wellwood, Claudia Maienborn, Dan Lassiter, Galit Sassoon, Jason Merchant, Jessica Rett, Joost Zwarts, Luca Sbordone, Martin Schäfer, Muffy Siegel, Pasha Koval, Peter Klecha, Rajesh Bhatt, Ryan Bochnak, Sebastian Löbner, Taehoon Kim, and Vladislav Poritski. This variously shaped how I thought about sections, pointed out relevant work that I had overlooked, set me straight about factual points, influenced the presentation, and spared me some embarrassment. Many parts of the book would have taken a very different form or omitted im- portant elements had it not been for conversations over the years with Angelika Kratzer, Barbara Partee, Chris Kennedy, Lisa Matthewson, and Roger Schwarzschild among others. The whole enterprise would have seemed less worthwhile had it not been for the encouragement of a surpris- ing number of people who I won’t attempt to name them individually, but I’m grateful to them all. Among my principal influences in shaping this book have been cur- rent and former students at Michigan State University. They provided not merely a proving ground for various modes of explanation and pedagogical strategies, but also theoretical insights, empirical observations, pointers to literature, and perspectives about how to conceptualize big-picture ques-

viii ACKNOWLEDGMENTS

tions that that changed my thinking about many issues. In addition to countless undergrads, these include especially the semanticists—Adam Gobeski, Adam Liter, Ai Matsui Kubota, Ai Taniguchi, Alex Clarke, Chris O’Brien, Curt Anderson, David Bogojevich, E. Matthew Husband, Gabriel Roisenberg Rodrigues, Karl DeVries, Kay Ann Schlang, Olga Eremina, Peter Klecha, and Taehoon Kim—and others working in syntax, semantics, and acquisition, including Adina Williams, Greg Johnson, Hannah Forsythe, Joe Jalbert, Kyle Grove, Marisa Boston, Ni La Lê, and Phil Pellino. More gener- ally, our local syntax-semantics community has provided an entertaining and intellectually fulfilling environment, and writing this would have been far more painful without them. My faculty colleagues, particularly Alan Munn and Cristina Schmitt, warrant special mention in this regard. At Cambridge University Press, Helen Barton was a pleasure to work with, and far more tolerant of me than I had any right to expect. In some- thing that is partly a textbook, it doesn’t seem inappropriate to acknowledge the influence of my own teachers, including Angelika Kratzer, Barbara Par- tee, Kyle Johnson, Lisa Matthewson, Sandy Chung, Jim McCloskey, Donka Farkas, and Bill Ladusaw among others. It was a 1999 seminar of Lisa’s on modification that first turned my interest in that direction. With her book Reference (Abbott 2010), Barbara Abbott (my semantics predecessor at Michigan State, as it happens) reaffirmed my sense that one need not adopt a somber impersonal tone in a book such as this. Finally, none of this would have been possible without the unfailing support, encouragement, feedback, and advice of Anne-Michelle Tessier and my dog, Hildy. Hildy’s advice and feedback largely focused on the importance and pressing urgency of belly rubs. I regret that I was unable to fully incorporate this into the text. It nevertheless bears directly on keeping one’s priorities in order. Anne-Michelle’s assistance was more intellectually wide-ranging and involved less shedding. I couldn’t hope to adequately express my debt to either of them.

2 PRELIMINARIES

something in the real world to which it refers? Before we can address this question, there is some practical business to attend to.

1.2 What this book is and isn’t

This book is about formal linguistic semantics. That said, I really hope it might prove useful to people approaching it from other theoretical and methodological perspectives as well—if nothing else, in its characterization of the facts and of various particular puzzles. It has two primary target audiences. One is grad students and advanced undergrads who have under- gone the initial rites of passage into formal semantics and have (at least) survived with their will to continue intact. Another is researchers in related fields, who sometimes find themselves in a distinct though not entirely dissimilar situation. They may have a longstanding familiarity with work in semantics, but a passive one, as spectators but not practitioners. If they would like to play a more active role, neither general introductory texts nor handbook articles are ideally suited to their needs. This is intended as something between an advanced textbook and a top- ical survey of research in a broad area, a bridge between the basic orderly framework-building of textbooks and the sophisticated, cacophonous, and often formally challenging to-and-fro of the primary literature. The aim is to present some analytical tools and concepts that can serve as a starting point for the reader’s own research. It’s to provide a way of thinking about a particular set of problems and a sense of where to look to find out more. A number of things follow from that. First, I have tried to emphasize problems over particular solutions and analytical strategies over partic- ular instances of them. That said, the most interesting problems often emerge only against the background of some theoretical assumptions. It’s impossible to be surprised if you have no expectations. Second, there is no attempt here to be comprehensive. ‘Modification’ is a topic so broad that it could encompass virtually all of semantics. There may be no area of the field in which some class of modifiers hasn’t been a major concern. So, in the interests of keeping the book a reasonable length— in fact, finite—there are many interesting topics of potential discussion that I will forgo. Discussion of adverbials other than adverbs in the strict sense will be conspicuously absent, as will discussion of relevant work in psycholinguistics and language acquisition. The focus will be on the grammar of adjectives, adverbs, and degrees. Third, I have tried to maintain a consistent theoretical framework throughout. When encountering the literature for the first time, people

BACKGROUND ASSUMPTIONS 3

are sometimes struck with a kind of intellectual vertigo. They have a few hard-won analytical tools in hand, but soon discover that work in semantics varies widely in formalism, style of analysis, and theoretical assumptions. It’s as though they had just learned Italian only to find, upon visiting Italy, that people freely switch between Italian, French, Portuguese, Latin, and for some reason Japanese—and a handful of people seem to be saying really interesting things in Klingon. There is no solution to this in the long term other than to learn to deal with it. Nevertheless, I have enforced an artificial consistency on the discussion, translating various ideas into a single analytical and representational language. (Italian, one is tempted to say, taking the analogy too far.) This of course entails making many small adjustments to the original proposals, and a few larger ones. I call attention to the latter. The book presupposes familiarity with the essential tools of formal se- mantics. I’ve tried to keep things relatively accessible, but engaging most of the content fully will require some previous background. Having absorbed the first few chapters of Heim & Kratzer (1998), Chierchia & McConnell- Ginet (1990)—or the relevant parts of volume two of Gamut (1991) or certain other semantics textbooks—should be sufficient. That should in- clude a general understanding of quantification, lambda abstraction, and semantic types. I have attempted to make the chapters of the book as independent of each other as possible. There are some dependencies that are difficult to avoid, though—you will get more out of chapter 4 (on comparatives) if you have first read chapter 3 (on vagueness, degrees, and the lexical semantics of gradable predicates). Chapter 6 (on crosscategorial phenomena) is best read in light of all preceding ones. But, on the other hand, if you wanted to skip past further preliminaries now and dive right into chapter 2, you would not suffer unduly for having done so.

1.3 Background assumptions

1.3.1 Glossing logical notation

Some introductory courses and textbooks develop a sophisticated semantics without recourse to logical notation other than lambdas, so I should briefly gloss the symbols I’ll rely on. Many readers will want to skip this section. Obviously, one shouldn’t mistake it for the shortest introduction to logic ever written. It just provides a way of mapping symbols onto familiar concepts or natural-language paraphrases. First, some connectives, which make new propositions out of old ones:

BACKGROUND ASSUMPTIONS 5

will follow Heim & Kratzer in treating them as components of a metalan- guage that might at times include bits of English as well. I won’t adopt an indirect interpretation system of the classically Montagovian sort, in which much of the semantics resides chiefly in how expressions of natural lan- guage are translated into expressions of a logic. One additional peculiarity is that I’ve systematically curried/schönfinkeled all logical predicates for consistency—that is, I will write eat ( y )( x ) rather than eat ( x , y ). As for the syntactic assumptions, they are conventionally generative but with a minimum of theoretical commitments. For the most part, only the shape of trees—the constituency, not syntactic category—will matter, and I’ll often omit syntactic category labels entirely. Where a neutral term like ‘nominal’ becomes inappropriate, I assume DP is the category of e.g. the monkey from Cleveland (Abney 1987), and NP as the category of the next maximal projection down ( monkey from Cleveland ). I assume that the syntax has movement, and that quantified nominals usually take scope by undergoing Quantifier Raising. The way I’ll represent movement will diverge in a notational way from Heim & Kratzer’s. In their standard treatment, a moved expression such as a generalized quantifier leaves behind an individual-denoting trace in the position it previously occupied. This trace receives a numerical index. By moving, the quantifier creates next to its landing site a binder for this index. This is represented as a number that occupies a node in the tree, which branches from the node to which the displaced quantifier attached. Thus, for them, QR looks like this:

(5) a. Floyd deloused every monkey. b. t

h e , t i

t

h e , t i

e

t (^) 1

h e , et i

deloused

e

Floyd

1

h et , t i

every monkey

The trace is then interpreted as a variable, over which the binding node triggers lambda abstraction. In contrast, I’ll represent movement in this

6 PRELIMINARIES

way:

(6) t

h e , t i

t

h e , t i

e

x (^) 1

h e , et i

deloused

e

Floyd

x 1

h et , t i

every monkey

This simply replaces the trace with the corresponding variable and the binder with the corresponding lambda. (It’s a little easier to read than the original when there are both individual- and degree-denoting expressions moving.) In a somewhat more unusual move, I will use variables with numerical indexes whenever they are associated with movement as a subtle reminder of the more standard indexed-trace representation and of their connection to movement. The purist is free to disregard the non-subscripted material, which will render the representation virtually identical to the original. As (6) reflects, I will occasionally place variables directly into the object language—that is, hang them from trees or from expressions in trees—in, again, a relatively standard fashion. Variables introduced this way and left free are assumed to get their value from the context(ually-supplied assignment function). An example of how a computation might run (I’ll generally skip more steps than I do here):

(7) a. π every ∫ = P h e , t i Q (^) h e , t i. 8 x [ P ( x )! Q ( x )] b. π every monkey^ ∫ = π every^ ∫ ( π monkey^ ∫ ) = Q (^) h e , t i. 8 x [ π monkey^ ∫ ( x )! Q ( x )] = Q (^) h e , t i. 8 x [ monkey ( x )! Q ( x )] c. π deloused ∫ = x y. deloused ( x )( y ) d. π Floyd deloused x (^) 1 ∫ = π deloused ∫ ( π x (^) 1 ∫ )( π Floyd ∫ ) = [ x y. deloused ( x )( y )]( x (^) 1 )( Floyd ) = deloused ( x (^) 1 )( Floyd )

8 PRELIMINARIES

• R for relations
• G for gradable degree predicates, any type with both a d and a e in it: h e , d i, h e , dt i, h d , et i
• D for properties of degrees, type h d , t i
• p , q for propositions, type h s , t i
• f , g ,... for other functional types
• e , e^0 ,... for events, type v
• d , d^0 ,... for degrees, type d
• w , w^0 ,... for possible worlds, type s

1.4 What, if anything, is modification?

With that out of the way, we can return to the substantive question at hand: What precisely is modification? Does it constitute a single grammatical phenomenon? The easiest answer to give—and, after some reflection, simultaneously the more obvious and more surprising one—is no. We think of the grammar largely in terms of predicates and their arguments. ‘Modifier’ is simply a term for linguistic expressions that don’t fit neatly into either conceptual box. If this is right, construing modification as a unified phenomenon is dou- bly mistaken. First, it’s uselessly broad. Writing a book about modification would be like writing a book about arguments: essentially an impossibility. One can talk coherently of argument structure , of course, but this isn’t evidence that all expressions that happen to be arguments have something essential in common. Second, on this understanding, modifiers would be the complement of a natural class—that is, a meaningless set defined in re- verse, like non-Bolivian non-dermatologists. If you had encountered a class like this in a phonology problem set, you would be justified in suspecting you had taken a wrong turn somewhere. But there is another way of looking at the question, even if it’s harder to perceive. One place to start is consulting one’s intuitions about the use of the term, however inconsistent or precarious they may be. An adjective is a modifier, except for when it isn’t. An adverb is almost always a modifier, though adverbs might really be just glorified adjectives in any case. A prepositional phrase is sometimes a modifier and sometimes it isn’t, depending perhaps on whether it’s an adjunct. A noun or noun phrase isn’t a modifier, but what about in, say, died last night? Functional elements like tense morphemes, modal auxiliaries, and most determiners clearly aren’t modifiers. Clauses are modifiers in various adjoined positions, but

WHAT, IF ANYTHING , IS MODIFICATION? 9

not elsewhere. In this meandering litany, one can discern something about the nature of the conceptual struggle. The categories most readily at hand are syntactic, but we seem to be groping for something semantic. The references to syntactic category seem to be a clumsy proxy for an adequate language to talk about the lexical semantics of expressions, one that might ultimately express an intuition about their distribution too. Clearly, all this will need to be firmed up to make progress on the broader question. In one respect, that can be done immediately. There is behind the whole thing a kind of equivocation that needs to be corrected. It’s between two ways of characterizing a phrase. There is a difference between labels for the internal characteristics of phrases and for the external role they play in the constructions they enter into. Terms like ‘subject’, ‘complement’, ‘adjunct’, ‘resultative’, or ‘purpose clause’ all unambiguously characterize constituents by the role they play as part of larger ones, their external role. Terms like ‘noun’ unambiguously name lexical categories, and no one is inclined to use them to mean, say, ‘complement to a verb’ (setting aside sloppy talk of ‘acting as a noun’ in first-semester undergrad assignments and prescriptivist harangues). They’re characterizations of an internal property of a word and of the phrases it heads, not of their relation to larger expressions. The term ‘modifier’ is uncomfortably perched astride this fence. It characterizes both a family of (internal) lexical semantic characteristics, and a family of (external) distributional ones. That, I think, may account for some of the conceptual muddle. The internal sense of ‘modifier’, then, to a very crude first approximation, may amount to just this: you’re a modifier if you’re an adjective or an adverb. That probably makes you pretty good at gradability. The external sense of ‘modifier’ has to do with crosscategorial parallels in the role an expression plays. You’re a modifier if you’re adjoined to something that you’re not a semantic argument to. You very well might have a semantics that can be expressed with and : a red dinosaur is red and a dinosaur. Obviously, the distinction doesn’t instantly cut through the haze. But it is useful because, for the external sense, it’s possible to provide a straight- forward and rigorous (if imperfect) definition of modification in terms of semantic type. As we’ll see in subsequent chapters, on one classical way of thinking, a modifier is any expression that maps a type to the same type: that is, anything whose denotation is type h ⌧ , ⌧ i , where ⌧ is a type. When ⌧ is a predicate type, anything with this kind of meaning is called a PREDICATE MODIFIER. An example:

WHAT, IF ANYTHING , IS MODIFICATION? 11

(12) a. π red dinosaur ∫ = x. π red ∫ ( x ) ^ π dinosaur ∫ ( x ) b. π mutter quietly^ ∫ = e. π quietly^ ∫ ( e ) ^ π mutter^ ∫ ( e )

The result is the conjunction-based semantics we need. McNally (to appear), who wrestles with the same conceptual problem we face, adopts a working definition of ‘modifier’ based on a feature the two approaches share: a modifier is ‘an expression which combines with another expression to produce a result with the same semantic type’. We will explore all these ideas in much greater depth in subsequent chapters. The important point for the moment is that they establish that providing a semantic definition of ‘modifier’ on the basis of certain external properties of an expression is possible.^2 They certainly don’t establish that there is an interesting natural class here beyond the types, though. Whatever the answers to these questions might be, something about this purely type-based characterization doesn’t seem fully satisfying. Shouldn’t there be more to modification than mere combinatorics? Here, we might need to turn to the other, internal sense of modification. And here, it’s not so clear whether types alone will help. They may, as we’ll see in chapter 3, but it’s not obvious. Syntactic category helps, too—‘adjectives and adverbs’ seems straightforward enough—but again, it doesn’t get at an essentially semantic internal notion of modification. So this will have to remain a bit murky. Still, having the question hanging in the air as we proceed will be useful, if only because it frames the discussion. We’ve now taken a few tentative, incremental steps toward a clearer way of thinking about it. But the question remains: What, if anything, is modification? In light of all this big-picture rumination, the question suggests some topics we will need to confront. We will need to talk about compositional principles. That’s crucial to the external sense of modification. But we’ll also need to talk about syntactic category, adjectives and adverbs in particular, and something deeper: what kind of lexical semantics is especially asso- ciated with them. In the combinatorial principles briefly glimpsed above, there was a perfect symmetry between the adjectival and the adverbial case. That’s distinctively modifier-like. If the notion of modification is to be useful, it may well be in characterizing regularity in cross-categorial behavior—a crucial element of being a modifier may be behaving consis- tently irrespective of the syntactic neighborhood you’re in. As we proceed, I’ll occasionally draw attention to such crosscategorial connections, and I’ll address some remaining ones in chapter 6.

(^2) This definition isn’t bullet-proof, as we’ll see when we return to the issue in the final chapter.

12 PRELIMINARIES

1.5 Roadmap

Chapter 2 is concerned with what adjectives mean, both as a matter of their lexical semantics and in the context of particular constructions. It also examines the occasionally quirky way they combine semantically with their neighbors. Chapter 3 turns to the semantics of vagueness, an age-old philosophical puzzle, and its grammatical cousin, gradability. It considers two major analytical approaches to the problem, and then deploys one of them to probe into how scalar notions are represented in the semantics of gradable predicates and what variation there is in this respect. Com- paratives and other degree constructions (equatives, superlatives, etc.), have long been one the major topics of semantic inquiry, and the principal proving ground for theories of gradability. They are the focus of chapter 4. Chapter 5 takes up a less well-understood topic, the semantics of adverbs. One of the great challenges in that area, which adjectives don’t present in the same way, is providing an account of the different interpretations they receive in different syntactic positions. By chapter 6, the essential background will be in place to confront some remaining phenomena that cut across more than one category, or indeed across virtually all of them. Finally, chapter 7 concludes with a brief return to some of the questions raised in this chapter.