«INFORMATION STRUCTURE, GRAMMAR & STRATEGY IN DISCOURSE Jon Stevens A DISSERTATION in Linguistics Presented to the Faculties of the University of ...»
INFORMATION STRUCTURE, GRAMMAR & STRATEGY IN DISCOURSE
Presented to the Faculties of the University of Pennsylvania
Partial Fulfillment of the Requirements for the
Degree of Doctor of Philosophy
Supervisor of Dissertation
Associate Professor, Linguistics
Graduate Group Chairperson _______________________
Robin Clark, Associate Professor of Linguistics Dissertation Committee Robin Clark, Associate Professor of Linguistics Anthony Kroch, Professor of Linguistics Florian Schwarz, Assistant Professor of Linguistics !
INFORMATION STRUCTURE, GRAMMAR & STRATEGY IN DISCOURSE
COPYRIGHT2013 Jon Scott Stevens The truth of the thoughts here communicated seems to me unassailable and deﬁnitive.
—Wittgenstein, 1918 I have been forced to recognize grave mistakes in what I wrote in that ﬁrst book.
—Wittgenstein, 1945 iii Acknowledgements I am grateful to my family, my friends, my wife Julie, and everybody close to me who has supported me in this endeavor, and who has made my life outside of linguistics enjoy- able and fulﬁlling. I remain indebted to my advisor Robin Clark, without whose guidance, open-mindedness, criticism and occasional indulgence of my ﬂights of fancy this disser- tation could not have been written. I am also indebted to my committee members Florian Schwarz, whose feedback has been more helpful than I ever could have asked for, and Tony Kroch, whose encouragement and instruction ﬁrst sparked my interest in issues of infor- mation structure. This dissertation also would not have been possible without the helpful feedback of Julie Legate and Beatrice Santorini, as well as the methodological guidance of Georgia Zellou. Charles Yang has been immensely inﬂuential for me as a linguist and a scientist more generally—without him I never would have learned computational linguis- tics, for which I am grateful. I thank all of my fellow Penn graduate students who have made my time at Penn all the more exciting. I am particularly grateful to Chris Ahern for reading (often underdeveloped) drafts of papers which ultimately led to this research, to Caitlin Light for helping me sharpen my thoughts about information structure, and to Yong-Cheol Lee for our ongoing discussions about the nature of information structure and intonation. I also gratefully acknowledge the efforts of Claudia Consolati, Norman Rusin, Marina Johnston, Johannes Eichstaedt and Stefan Heim for help with translating and obtaining judgments from languages that I do not speak. Finally, a huge thanks to everyone who I have forgotten to include here. There are likely many of you.
This dissertation examines two information-structural phenomena, Givenness and Focus, from the perspective of both syntax and pragmatics. Evidence from English, German and other languages suggests a “split” analysis of information structure—the notions of Focus and Givenness, often thought to be closely related, exist independently at two different levels of linguistic representation. Givenness is encoded as a syntactic feature which presupposes salience in prior discourse and either (1) prevents prosodic prominence (in languages like English and German), or (2) drives syntactic movement (in languages like Italian). On the other hand, Focus, which introduces strong prosodic prominence and a contrastive interpretation, exhibits none of the expected properties of a syntactic feature, and is therefore analyzed quite differently. I argue that Focus is the result of purely pragmatic principles which determine utterance choice in the face of grammatical optionality. The syntactic and phonological systems often generate multiple possible formulations of an utterance, and communicative principles can be invoked to explain the correspondences between certain kinds of discourse contexts and certain patterns of linguistic form. The application of communicative principles to problems of utterance choice is modeled mathematically using the tools of game-theoretic pragmatics. From this perspective, utterances are taken to be strategically chosen in order to maximize communicative effectiveness. Ultimately, the strong differences between Focus and Givenness emphasize a methodological point: both syntactic and pragmatic perspectives are necessary to fully determine the space of possibilities in natural language. Neither perspective should be ignored.
1.1 Prior Questions Chomsky (2011), challenging Enﬁeld (2010), asks if language is like “today’s weather”, an accidental collection of distinct phenomena which conspire toward a single effect, or if there is indeed a singular mechanism, unique to the human species, that is specialized for language. The Minimalist conception of language (Chomsky, 1995, 2005) leans strongly toward the latter: there is posited to be a faculty of language (FL) which is a specialized cognitive computational system responsible for the fact that humans—and no other animals—acquire an unconscious rule-based system of grammatical competence. That this grammatical competence is used for communication is seen as secondary from this perspective. But even taking it to be true that the core
system underlying the use of grammar was not selected or “designed” by evolution for any particular communicative purpose, this does not rule out a strong role for communicative principles in a complete theory of language.
At ﬁrst glance this may seem to be either a contradiction or else a conﬂation of two distinct senses of the word “language”. The word is ambiguous between: (1) the computational combinatoric system FL as it is narrowly deﬁned within mainstream Minimalist linguistic theory, and (2) the more colloquial sense of the word, which encompasses the externalization and communicative/social uses of FL. Under deﬁnition (1), assuming Chomsky’s thesis to be correct, it would be a contradiction in terms to claim that communicative principles play any role in determining natural language possibilities. However, under deﬁnition (2) it is a perfectly reasonable, even obvious, thesis—the space of possible utterances is determined by both grammar and pragmatics. But is the broader second definition of “language” useful for studying FL? The reﬂex of many theoreticians may be to
answer in the negative, adopting the following view from Chomsky (2011, p.275):
Sometimes externalization is employed for communication—by no means always, at least if we invest the term “communication” with some signiﬁcance.
Hence, communication, a fortiori, is a still more ancillary property of language, contrary to much conventional doctrine—and of course language use is only one of many forms of communication.
In general this seems to be true—nothing about communication can explain why “colorless green ideas sleep furiously” is a sentence of English but “sleep ideas green colorless furiously” is not. But there is one area of linguistic research where principles of pragmatics and communication collide with principles of FL: the study of information structure.
By deﬁnition, information structure consists in those areas where discourse context partially determines linguistic form. In this dissertation, I analyze the information-structural phenomena of Givenness and Focus which leads to a particular view on the relationship between syntax and pragmatics. Namely, we must allow a role for discourse context and communicative efﬁcacy in determining judgments about what is acceptable or possible in a language.
This view is arrived at by ﬁrst showing that one aspect of information structure—the marking of Givenness—respects rules of syntactic structure formation. These rules are well-established in syntactic theory and are, as Chomsky predicts they would be, semantically and pragmatically arbitrary. That is to say, there is no obvious communicative reason why the following contrast should exist—the explanation is purely structural.
(1) a. Which professor did you have for biology?
b. *Which did you have professor for biology?
The generalization here is that the entire wh-phrase which professor must move to the front of the sentence to form a question—partial movement is not allowed. Chapters 3 and 4 are dedicated to showing that there is evidence for similar structural constraints on the marking of Givenness (i.e. salience in prior discourse). This gives us a template for what a syntactically encoded information-structural feature should look like. Crucially, when we examine the distribution of Focus (i.e. prominence due to contrast, in a particular sense), the template is violated. This suggests a different view of Focus: there is no feature in the syntactic derivation of a sentence, as is often assumed, which is responsible for Focus.
2 Rather, the phenomenon arises from communicative principles which regulate how utterances are chosen when multiple grammatical alternatives exist. FL generates all possible Focus structures, independent of context, and pragmatics ﬁlters them.
This is a somewhat different conception of linguistic judgments than what is sometimes assumed by syntacticians in that strong negative intuitions about sentences are not necessarily due to ungrammaticality. Rather, a strong negative intuition about a sentence can arise either due to failure of FL and/or the parameters of the language in question to generate that sentence, or else due to a pragmatic mismatch between the sentence and the context in which it is situated. Nonsensical sentences like “colorless green ideas sleep furiously” can exist in a contextual vacuum, yielding the intuition that this is an acceptable sentence of English. But other sentences, e.g. “JOHN kissed Mary” (with strong prosodic prominence on the subject) become contextualized—one cannot help but evaluate them with respect to a discourse context—and the sentence is only acceptable if the context that is given (either explicitly or perhaps implicitly) is appropriate. This does not mean that the sentence is ungrammatical when the context does not allow it; it is merely infelicitous.
This is not a new idea; it is exactly the hypothesis utilized by followers of the approach of Vallduv´ (1990), who posits an autonomous “module” of grammar, i.e. of FL, called ı Information Structure which imposes correspondences between syntax and “information packaging”, i.e. instructions on how to functionally interpret an utterance. Eilam (2011, p.6), following this approach, nicely distills the proposed difference between grammaticality and felicity in context.
(T)he lack of attention to context and Information Structure leads to the misinterpretation and misanalysis of data; in particular, the unacceptability associated with phenomena which are driven by Information Structure considerations is mistaken for ungrammaticality, and these phenomena are erroneously analyzed as syntactic and/or semantic in nature.
I take this to be correct in spirit. However, I provide an analysis which does not require any modiﬁcation of the accepted architecture of FL, an architecture which typically does not include an Information Structure module. I hold that there is no autonomous domainspeciﬁc mechanism responsible for phenomena such as Focus. Rather, I aim to provide a more parsimonious account whereby Focus is epiphenomenal, resulting from the interaction between general communicative principles and optionality in grammar.
This account has two advantages. First, it is more in line with Minimalist principles, in that FL is strictly limited to the set of computational operations required to mediate sound 3 (or sign) and semantics, with no additional constraints introduced by information structure. Second, it offers a deeper explanation of why Focus exists (to facilitate communication). Finally, the current work illustrates that in fact, some of what is called informationstructure, i.e. phenomena related to the notion of Givenness, really are due to principles of FL. Therefore, what is called “information structure” is indeed like “today’s weather”—it is an accidental combination of a syntactic feature whose interpretation yields particular discourse effects, and the (likely unconscious) application of general principles of communicative optimality to the problem of optionality in grammar.
1.2 Roadmap Chapter 2 begins with a survey of previous literature on information structure, providing a basic deﬁnition of the concept and analyzing some previous claims about the informationstructural notions of Focus, Topichood, Contrast and Givenness. The terminology surrounding these ideas is muddy, confusing and often contradictory. I attempt to ﬁlter the useful from the superﬂuous, making some assumptions along the way, and ultimately arriving at a particular set of terminology to be used in the remainder of the volume. In this chapter I also outline some syntactic assumptions and introduce the style of notation to be used throughout. Finally, I give a brief primer on the ﬁeld of game-theoretic pragmatics, the tool which I use to model the communicative theory of Focus which I eventually develop.
Chapters 3 and 4 are devoted to the phenomenon of Givenness. A recent previous analysis is considered and then refuted, which leads to a syntactic generalization regarding certain interesting data. These data are introduced in Chapter 3, where I use both intuitions and results from two audio-based judgment tasks to argue for a particular analysis of “de-accenting” in English. I show that de-accenting due to Givenness (i.e. salience in prior discourse) is in fact limited by syntactic principles. For example, a partial adjunction structure cannot be marked as Given (and subsequently de-accented) without marking the entire XP as Given. This is analyzed in terms of a syntactic feature. Chapter 4 applies this analysis to German, showing that the same principles hold, and also extending the analysis to scrambling constructions in German.