TEL AVIV UNIVERSITY THE LESTER AND SALLY ENTIN FACULTY OF HUMANITIES THE SHIRLEY AND LESLEY PORTER SCHOOL OF CULTURE DEPARTEMENT OF COGNITIVE STUDIES OF THE LANGUGAE AND ITS USAGE DYNAMIC SITUATIONS: ACCOUNTING FOR DOWTY’S INERTIA NOTION USING DYNAMIC SEMANTICS Master’s Thesis By Ido Ben-Zvi [email protected] Under the Supervision of Professor Fred Landman March 2005 ii ACKNOWLEDGMENTS This thesis was long in the making. And now, within but a brief late night hour or so, it will be sealed. It is time to thank those who had been in the know, as well as those who were not but whom I love in any case. First, my deepest thanks go to Prof. Fred Landman, who had provided constant guidance and support throughout. Especially for managing to convince me that this progressive feat will someday reach past simple. I thank Galit Sassoon, who took the time to read this stuff carefully and to discuss it, and also others who braved this manuscript, or who sympathized from aside. Good friends in Tel-Aviv, Tullamore Dew at the Jajo, and of course Mishmish, who seems to think that laptops were invented by god to warm up a sleepy cat’s belly. Finally, I want to thank my parents Yael and Amos Ben-Zvi. My truest support comes from them. ii iii TABLE OF CONTENTS 1 INTRODUCTION, OR: WHY BOTHER READING ON.........................................6 The Problem.................................................................................................................6 The Theory Propounded..............................................................................................7 What Follows Next......................................................................................................8 2 DOWTY’S MODAL ACCOUNT OF THE PROGRESSIVE..................................10 Dowty’s Imperfective Paradox..................................................................................10 Combining Modality and Temporality......................................................................11 Initial Formalization of the Progressive Operator.....................................................13 Later Formalization of the Progressive Operator......................................................14 3 THE INERTIA CONSTRAINT................................................................................16 Inertia as a Compound...............................................................................................16 Conflicts between the Two Constituents of Inertia...................................................17 Further Support for the Epistemic Modal Base.........................................................19 An Intuitive Comparison of the Ontological and Epistemic Bases...........................21 Normality Revisited...................................................................................................22 4 DYNAMIC SEMANTICS.........................................................................................26 Dynamic Semantics in Context.................................................................................26 Structures : Worlds, Referent Systems, Possibilities, Information States.................27 Information Growth...................................................................................................30 Object Language Semantics......................................................................................32 5 DYNAMIC SITUATIONS........................................................................................37 Why We Need Situations...........................................................................................37 Preliminary Definitions.............................................................................................40 Syntactic Properties of Sentences and Syntactic Sets........................................40 Notational Conventions......................................................................................41 The Linguistic Context..............................................................................................42 Non-Discernibility.....................................................................................................45 iii iv Discernibility by Abstracting on Information States..........................................45 Discernibility Based on Possibility Truth Values..............................................46 New Formal Structures..............................................................................................49 Situations............................................................................................................49 Information States are Redefined.......................................................................52 Situation Extension.............................................................................................54 Dynamics of Linguistic Context Updates...........................................................55 Object Language Semantics......................................................................................57 Variable assignment...........................................................................................57 The Object Language.........................................................................................58 Compatibility with Dynamic Semantics....................................................................60 Discourse Based Information States...................................................................60 Transition Function Between DS States and Situation Based States.................62 6 THE PROGRESSIVE IN A DYNAMIC SEMANTICS ENVIRONMENT............64 Further Assumptions and Articulations.....................................................................64 Basic Assumptions on Event Temporal Relations.............................................64 Articulations regarding the compositional structure of Dowty’s aspectual classes65 Knowledge Up to a Point...........................................................................................65 Normal Course of Events...........................................................................................67 Partiality of Being......................................................................................................69 Minimal Eventiveness...............................................................................................71 Dynamic Semantics of the Progressive.....................................................................74 Some Concrete Examples..........................................................................................75 7 CASE STUDIES, COMPARISON AND CONCLUSION.......................................80 Case studies...............................................................................................................80 Treatment of activities........................................................................................80 Problem of creation verbs...................................................................................80 Problem of interruption......................................................................................84 Problem of non-interruption...............................................................................84 Problems of continuation I.................................................................................86 Problems of continuation II................................................................................86 Problems of perspective.....................................................................................89 A note on the issue of multiple viewpoints........................................................89 Comparison with circumscription..............................................................................91 Conclusion.................................................................................................................93 A DISCERNIBILITY BASED ON POSSIBILITY TRUTH VALUES.......................96 B PROOFS OF THEOREMS IN CHAPTER 5..........................................................101 iv v Theorem 1................................................................................................................101 Theorem 2................................................................................................................103 Theorem 3................................................................................................................103 Theorem 4................................................................................................................103 Theorem 5................................................................................................................105 Theorem 6................................................................................................................106 Lemma 1..................................................................................................................108 Lemma 2..................................................................................................................109 Lemma 3..................................................................................................................110 Theorem 7................................................................................................................110 BIBLIOGRAPHY...........................................................................................................116 v CHAPTER 1 INTRODUCTION, OR: WHY BOTHER READING ON The Problem Imagine this: a zebra is pasturing on the endless plains of the Serengetti savanna. It is finishing off a patch of greenery. According to a semantic analysis of the progressive offered back in 1979 by the linguist David Dowty (Dowty 1979), this more or less means that there are still some sprouts to eat but if all goes on normally as it did up until now they will soon be gobbled up by the zebra. Take another look at the zebra’s surroundings. Note the feline silhouette crouching but a few meters away. A hungry lioness is in the final stages of hunting down the zebra. What this means, according to that same linguist, is that the zebra is still alive right now but if all goes on normally as it did up until now it will shortly be served as lunch to the lioness’ cubs. Intuitively it seems reasonable enough to interpret a progressive sentence like (1) as (2), and to interpret (3) as (4). (1) The zebra is finishing off the patch of greenery (2) There are still some sprouts but if all goes on normally as it did until now then the zebra will shortly eat every one of them. (3) The lioness is hunting down the zebra (4) The zebra is still alive but if all goes on normally as it did until now then the lioness will shortly kill it. Unfortunately, while (1) and (3) may both be true of the same scenario, (2) and (4) are very hard to reconcile. Moreover, usage of intuitively appealing phrases like ‘all goes on normally as it did up until now’, or to be more precise ‘..the future course of events 6 7 develops in ways most compatible with the past course of events’1 still leaves a lot to be desired from an analytic theory of meaning. The problem examined in this thesis then is not Dowty’s renowned ‘imperfective paradox’ of the progressive aspect in general. Accepting Dowty’s initial and in my mind very intuitive reasoning about the progressive, I will be concerned here with the inconsistencies that it brings about, and with trying to formalize the primitive notion of normality that remains as a rather large chunk of unanalyzed intuition right at the core of Dowty’s theory. Both issues will be resolved through highlighting the crucial part played by partial information in the semantics of the progressive. The Theory Propounded The progressive aspect has served as troubled waters testing the craftsmanship of many a theory faring the semantic seas. This has been particularly true in the wake of Dowty’s proposed analysis of it. It is fascinating to observe the multitude of morbid examples that are associated with these theories: people get run over by trucks, they get hit by lighting, they drown, they are eaten by bears, they walk into minefields, etc, etc. Apart from telling us something about the theorists involved, this also points out something about the progressive itself. Death, as we all know, comes ever uninvited, unexpected. The progressive aspect, I maintain, is very sensitive to the relation between what is expected and what is unexpected. The theory I advocate is three fold. First, while trying to follow closely in the footsteps of Dowty’s intuitively appealing concept of inertia (the idea of ‘things going on in a normal fashion’), I hold that the modal basis for this concept is epistemic and not ontological. This may seem to be in line with Dowty’s own theory, at least with that fuzzy part about things going on normally. But I will show that Dowty’s modality is either completely ontological, in which case it does not provide the required results, or else is an inconsistent mix up of an ontological and an epistemic base. Second, I hold that the notion of partiality plays a critical role in the semantics of the progressive. I think that at the intuitive level this too is an enticing conviction. The 1 (Dowty 1979) page 148 8 progressive appears to be a kind of commonsensical projection of what we know on to the parts of reality of which we do not know. Thus the zebra may truly be said to be finishing off the greenery if its (or our) partial knowledge does not include data about the approaching feline death. In trying to analytically bite off a chunk from the vague notion of normality I will take partiality a step further and use it to formally explain what it means for nothing unexpected or out of the ordinary to happen. This is a particularly difficult notion to catch formally because of the double use of negation: not only are we after those ‘things’ which are un-expected, but also are we interested in those cases where they don’t happen. This leads us to the third pillar on which this thesis rests. Partiality will give us an explanation of what the unexpected happenings are, and my third point is that built into the progressive operator is a kind of minimality constraint. Being interested only in those cases where nothing unexpected happens means throwing away all those cases where something superfluous does happen if we can also imagine a similar case where it does not. Once again, my aim is to crystallize this intuition in a formal way. So the theory brought forth here is a formal theory of partial information, with a proposed application to the progressive aspect. In theories of knowledge it is often the case that what is known is structured using a discourse context or common ground. The novelty in my approach to partial information is that it also offers us a chance to structure what is only speculated but not asserted. This domain of the doxastic stands midway between the common ground, to which we are fully committed, and the completely uncontemplated, which is irrelevant for our reasoning processes. What Follows Next Chapter 2 closely investigates the modality involved in Dowty’s proposed semantics for the progressive. The chronological evolvement of the theory hints at the problem of inconsistency associated with it, because the demand for normality was only added at a later stage of the theory. This inconsistency is brought out more clearly in chapter 3, which also proposes the epistemic modal base as a possible solution to it. Turning to an epistemic modality also beckons us to try and further analyze normality itself, and an initial analysis is indeed offered. Having proposed a different approach, but 9 merely by waving of hands and other forms of intimidation, we turn to formalize a framework which will support the required modality and offer a chance to define normality. Chapter 4 gives a summary of dynamic semantics, an epistemic framework which is particularly friendly to notions of partial knowledge. This same framework is adopted, and further adapted in chapter 5. There the notion of a situation, a partially defined possible scenario, is defined. Situations are the basic discernibles of my information state, and so their existence and distinctness from one another depends on the fine grainedness of the linguistic (or conceptual) context in which I operate. Chapter 5 is where the bulk of formal development takes place in this work. Chapter 6 is where it all comes together. The progressive operator is given a dynamic semantics analysis which crucially depends on the concept of situation defined earlier in order to formalize the notion of minimality that is required. Finally, chapter 7 takes on a bunch of formidable puzzles of the progressive and pits the theory advocated in this thesis against them. Many of the puzzles are handled by the theory, and this without need to resort to a primitive notion of normality. The one challenge that it does not rise to meet is when a sentence seems to simultaneously incorporate multiple contrasting viewpoints. It does however point the way toward a proper treatment of even such sentences through multiple agent epistemic systems. 10 CHAPTER 2 DOWTY’S MODAL ACCOUNT OF THE PROGRESSIVE This chapter provides a brief summary of Dowty’s theory of the progressive as outlined in (Dowty 1979) and (Dowty 1977). The impefective paradox is described, and in examining Dowty’s ways out of it, special emphasis is laid on the ontological nature of the modality involved. Tracing the chronological evolvement of the theory hints at a possible inconsistency, which will be brought out fully in chapter 3. Dowty’s Imperfective Paradox What is the imperfective paradox? Even after it has been identified and christened by Dowty, the exact phrasing of the paradox is not at all clear. Both (Landman 1992) and (Vlach 1981) phrase the paradox as the problem of explaining the difference in truth conditions for the progressive between accomplishments and activities. Hence the paradox is exemplified by asking how can it be that (5) entails (6), but (7) does not entail (8). (5) John was pushing a cart. (6) John pushed a cart. (7) John was drawing a circle. (8) John drew a circle. By itself this comparison does not seem very paradoxical. But the paradox does not just apply to this comparison, rather it takes into account also Dowty’s semantic modeling of the distinction between the verb classes, his ‘Aspect Calculus’. This calculus reduces the various verb classes, as set out by the seminal (Vendler 1967), to different combinations of stative predicates and special aspectual operators such as DO, BECOME, AT and CAUSE. In this calculus accomplishments are analyzed as a structure containing process and result, connected by causality. Only now, having accepted this