Tense, Aspect Temporal Reference and Marc Moens Ph. D. University Edinburgh of 1987 Eß1, Jý1V ýýGý 0 CM x JJ ' Declaration I declare has been by that this thesis composed myself and that the research reported therein has been by indicated. conducted myself unless otherwise Edinburgh, September 15,1987 Acknowledgements Before I begin I like have to tell to my story, would mention a number of people who played important it dissertation. an part over the years took me to complete this Mark Steedrnan, I forced by who suggested to me the topic of tense and aspect. see myself him for There is idea in dissertation the to thank this. this rules of etiquette probably no which I did discuss Mark lunch, dinner, drinks during in his the not with over or many sessions - And be he fixed times to office. when got really rough and nothing seemed going well, my car itself His friendship radio and said that everything else would sort out as well. support and have beyond for. thank gone what one can someone Barry Richards, kept holes in I him to who as a supervisor punching whatever presented as If is in follows, it is due in argument. there some coherence what certainly no small respect to his debate issues. continuing willingness to the My friends Centre, for the and colleagues at remarks and comments at workshops or seminars. I hope dissertation finally jokes like "Are feeling that submission of the will put a stop to you " "Isn't it ". tense? time talk? and about you're giving a Mijn die bleven doorheen de jaren, hebben ouders, me steunen ook al ze zich waarschijnlijk het lang duurde. toch vaak afgevraagd waarom zo And Mimo. (words fail) 4 Abstract English exhibits a rich apparatus of tense, aspect, time adverbials and other expressions that locate in can be used to order states of affairs with respect to each other, or to them at a point Ideally for these time with respect to the moment of speech. one would want a semantics between demonstrate expressions to that an orderly relationship exists any one expression and it Yet linguistic formal leave the meanings conveys. most existing and semantic accounts something to be desired in this respect, describing natural language temporal categories as descrip- being full indeterminacies, of ambiguities and apparently escaping a uniform semantic tion. from It will be argued that this anomaly stems the assumption that the semantics of these is directly linear familiar from logic expressions related to the conception of time temporal or be physics an assumption which can seen to underly most of the current work on tense and - involved in aspect. According to these theories, the cognitive work the processing of temporal discourse intervals line consists of the ordering of events as points or on a time or a set of time lines. There however, for is this time the are, good reasons wondering whether concept really one linguistic directly it be that to; that our categories are most related will argued a semantics of in defining temporally referring expressions and a theory of their use the temporal relations of different the events require a and more complex structure underlying meaning representations is A be developed, based than the that commonly assumed. semantics will on assumption like tense, to categories aspect, aspectual adverbials and propositions refer a mental representa- is tion of events that structured on other than purely temporal principles, and to which the notion of a nucleus or consequentially related sequence of preparatory process, goal event and is consequent state central. It be identification is logical that the the to the will argued of correct ontology a preliminary formal being in design the choice of any particular representation scheme, as well as essential language front-ends for databases. It be how temporal the of natural will shown ontology developed here be implemented in database information that time-related can a contains about is be by language that to events and queried means of natural utterances. Contents Abstract 4 ............................................................................................................................. 1. The temporal 7 procedural study of expressions ............................................................. 2. The tense 14 referential nature of ...................................................................................... 2.1 Tense 14 as anaphor ................................................................................. 2.2 Referential 21 temporal accounts of expressions ....................................... 2.2.1 Reichenbach's 22 times reference .................................. 2.2.2 Computational 24 approaches ......................................... 2.2.3 Formal 27 approaches .................................................... 2.3 Non-referential 32 approaches .................................................................. 2.4 Problems 34 the with referential account ................................................. 2.5 Conclusion 36 ........................................................................................... 3. Temporal intervals 38 ontologies: points, and events ........................................................ 3.1 Introduction 38 ......................................................................................... 3.2 Points, intervals, 39 events ....................................................................... 3.3 Conclusion 49 ........................................................................................... 4. A typology temporal descriptions 51 of ............................................................................ 4.1 Introduction 51 ......................................................................................... 4.2 Tense, Aktionsarten 51 aspect and ........................................................... 4.3 The 55 taxonomy ...................................................................................... 4.4 Conclusion 66 ........................................................................................... 5. An temporal 67 analysis of expressions ............................................................................ 5.1 The 67 aspectual nature of some temporal expressions ............................ 5.1.1 For 67 and until .............................................................. 5.1.2 In-adverbials 71 .............................................................. 5.1.3 Negation 74 .................................................................... 5.2 Progressives Imperfective Paradox 75 the and ......................................... 5.2.1 Progressive 75 states ...................................................... 5.2.2 A functional 79 account ................................................. 5.2.3 The imperfective 86 paradox .......................................... 5.3 Perfects 94 ................................................................................................ 5.3.1 Some descriptions 94 semantic ...................................... 5.3.2 Consequent 96 states ...................................................... 5.3.3 McCawley's Comrie's 4 types 101 and of perfect ........... 5.4 The Perfect Progressive 103 ....................................................................... 6 Contents 5.5 Conclusion 104 ........................................................................................... 6. The discourse temporal 106 structure of ............................................................................. 6.1 Introduction 106 ......................................................................................... 6.2 Temporal Connection 107 .......................................................................... 6.2.1 When 107 ......................................................................... 6.2.2 Other temporal 116 connectives ....................................... 6.2.3 Presuppositions descriptions 118 and event ..................... 6.3 The discourse 120 temporal structure of .................................................... 6.4 Conclusion 129 ........................................................................................... 7. The descriptions 130 aspectual classification of event ........................................................ 7.1 Introduction 130 ......................................................................................... 7.2 Aspectual 130 tenselessness categories and ............................................... 7.3 Diagnostic tests for the 134 categories ....................................................... 7.3.1 Stative 136 expressions .................................................... 7.3.2 Event 140 expressions ...................................................... 7.4 Aspect 145 and compositionality ................................................................ 7.4.1 Mass terms, bare 150 plurals and aspectual class ............ 7.5 Conclusion 154 ........................................................................................... 8. Temporal information for language 156 natural processing ................................................ 8.1 Introduction 156 ......................................................................................... 8.2 Extracting information from temporal text 157 .......................................... 8.3 Temporal 163 representation models .......................................................... 8.3.1 The 163 situation calculus ................................................ 8.3.2 Ontologies 165 temporal with characteristics ................... 8.3.3 Event-based 171 ontologies .............................................. 8.3.4 Conclusion 175 ................................................................. 8.4 An implementation 175 the of episodic representation scheme .................. 8.4.1 Introduction 175 ............................................................... 8.4.2 The database 179 .............................................................. 8.4.3 Aspect 184 and aspectual adverbials ................................ 8.4.4 Anaphoric 190 tense and when-clauses ............................ 8.4.5 Program 193 and sample session ..................................... 8.5 Conclusion 198 ........................................................................................... References 201 ......................................................................................................................... Chapter 1 The temporal procedural study of expressions Our language bias in its Relations date ordinary shows a tiresome treatment of time. of are ex- This bias is itself alted grammatically as relations of position, weight and color are not. of an inelegance, breach Moreover, form it or of theoretical simplicity. the that takes that of re- - form is quiring that every verb show a tense peculiarly productive of needless complications, - it demands lip is farthest from (Quine since service to time even when time our thoughts. 1960: 170) It has been degree though the often remarked, not always with same of exaggeration or is in Quine's that the temporal exasperation as exhibited words, expression of relations plays important in language, an role and that as an obvious correlate the processing of temporal - - information important language constitutes an part of natural understanding. It is have been develop logical therefore that to not surprising many attempts made systems function in language. Ideally, that try to the temporal give an account of of expressions sys- like be in tems these the they should abstractions of processes people engage as produce and English is, logic process sentences; that a of temporal expressions should explain what takes in information in language the temporal place process of relaying and extracting natural discourse, from then this that the and abstract account simple, general rules govern usage of temporal expressions. In information be described terms, the temporal very general processing of could as consisting (for the the of construction of a serial representation of events, processes and situations all of I "states in discourse. Constructing the term talked the which will use general of affairs") about involves the to the such a serial representation anchoring states of affairs our conception of dimension, i. line, lines in time e. ordering them along the real time or along expected time the future, imaginary lines These be or along alternative time now. states of affairs can ordered instance with respect to each other, or with respect to some pre-established on the temporal dimension The usually, though not necessarily, the moment of speech. cognitive work - involved in language temporal understanding natural expressions consists of creating and (cf. Miller & Johnson-Laird the these updating particular contents of serial representations 1976: 413). It important be seems, then, that an task to carried out when constructing a temporal semantics Chapter 18 Temporal expressions for a natural language is to determine the nature of the underlying time concept that is involved in this. all As by Bach (1981), try to this kind by was pointed out one should not answer of question ask- ing looks like. The they think the time native speakers what underlying concept correct be, from language approach seems to rather, to work a grammatical analysis of temporal its basis. towards an account of conceptual Yet, for has be formidable To begin this turned to task. various reasons, always out a with, as Miller & Johnson-Laird (1976: 411) information be by temporal point out, can conveyed words in (begin, (year, all the major syntactic categories, such as verbs end, precipitate, recur), nouns (recurrent, later), (eventu- season, precedent, minute), adjectives successive, present, adverbs (at, during, in) (before, ally, often, shortly, subsequently), prepositions on, and conjunctions just An language involve temporal as, until, as soon as). analysis of would a study of virtu- language, be ally all categories of the and only then would one able to say something about the language the natural temporal conceptual apparatus underlying system. Moreover, information that temporal merely collecting all expressions convey and studying import interact in their these semantic will not suffice, since many of expressions complicated inexplicable in Thus, it patterns, revealing unexpected and often seemingly changes meaning. like describe to the would seem a straightforward and attainable goal semantics of a simple for describe present tense; such an account could, example, the present tensed state of affairs holding, in (la): as true at the moment of speech or as currently as (1) (a) John loves Mary Yet, drastically in the this tense to meaning of simple morpheme seems change rather combina- different (as in lb), in tion with a verb of a type and yet again combination with an adverbial in (c) (cf. Miller & Johnson-Laird 1976: 412): as (1) (b) John kisses Mary (c) John kisses Mary today The in (b) describes habitual expression an state of affairs, or a running commentary on a it be As specific state of affairs. a commentary, may reporting a past state of affairs using the is happen. The in (c) historic present, a currently ongoing one, or one that yet to expression has happen. describes that to a state of affairs yet unambiguously Since in (1) be the ambiguity exhibited example seems to related to the type of the main verb, Chapter 19 Temporal expressions in be define the next step the research could to the temporal nature of verbs and classify them to their temporal A like love, for be according profile. verb example, could classified as stative, different interpretation in giving rise to a semantic when used the simple present than an event like kiss. And kiss describes verb whereas an event that can take some time, an event verb like discover be described being illustrated in (2a). But could as punctual or non-extended, as influence is illustrated in these temporal profiles can change under the of other expressions, as (2b) (c) discover has iterative (cf. L. Carlson 1981): and where an meaning (2) (a) I discovered hidden in an old staircase my cellar. (b) I discovered beach in Corfu last many a nice summer. (c) Many discovered beach the people white sandy near the this olive yard year. Hurtig (1974) intractable notes an even stranger and apparently more related phenomenon, illustrated in (3): (3) (a) For 12 Mary her than more years, shared all worries and joys dead with the man. (b) For 12 Mary her than more years, shared all worries and joys dead with a man. The information be temporal wide range of grammatical categories with which can expressed, interaction between fact the complicated patterns of them, and the that one and the same diverse has it temporal expression can convey a set of seemingly unrelated meanings made difficult discern language. The to the temporal any systematic principles underlying grammar of language discover basis to the then these strategy of analyzing underlying conceptual and use build logic for language doomed failure. concepts to a temporal natural expressions seems to Because insurmountable began building formal of these apparently problems, researchers their from direction, frameworks bor- accounts the opposite taking as their starting point conceptual from logical language. the time, these to the rowed study of and applying study of At first blush, have lot it. After logi- this strategy also seems to a to commend all, temporal have cians studied extensively the nature of time and can provide us with a conceptual All is it is apparatus the properties of which are well understood. that needed, so seems, an how language account of natural expressions map states of affairs onto this conceptual struc- ture. And is in One its this yet, strategy very misleading a number of ways. of misconceptions, the deep-rooted one, has to do with the very structure of time. Most temporal logicians most will Chapter 1 10 Temporal expressions agree that time consists of past, present and future, and that the future is just like the past - from fact it follows apart the that the present rather than preceding it. However, linguists and descriptive language data semanticists concerned with a proper account of natural noted early basic distinction in language like English is on that the temporal a twofold, not threefold: English has device for future from grammar no whatever cutting time away the time of speak- ing. (Joos 1964: 135) On the basis of the tenses, it is possible only to distinguish the past from the non-past. Thus, (4a) is like (4b) is clearly past, combining most naturally with an adverbial yesterday; clearly non-past, not allowing combination with an adverbial referring to a past time: (4) (a) John played squash (b) John plays squash As discussed before, tense in (4b) describes habitual was a present as used probably an or future like In state of affairs, combining most naturally with an adverb often or tomorrow. indicates does its the tense other words, present only non-pastness and not on own allow a further disambiguation between future To in English time. that present or communicate some- is "at be thing the time", the really present or going on present progressive can used: (4) (c) John is playing squash But in (c) denotes logicians' is to say that the expression the present time complicating matters in (c), describes rather than resolving them: the present progressive again, only a non-past state location described in (4c) be both in the the the of affairs; actual of state of affairs could be for future. the present, or planned The found in logics, future is logically temporal that the other conceptual notion often viz. image is in language. As Lyons speaking the exact mirror of the past, also not reflected (1977: 677), futurity in language is it temporal remarks never a purely concept; necessarily includes There is an element of prediction or some related modal notion. no a priori reason languages future why could not treat time predictions as grammatically parallel with statements but in don't. the they about past, general framework It be from this brief discussion that taking a pre-established conceptual should clear lead to to study the grammar of temporal relations can very easily unnecessary complications in the leave things that need explaining. grammar, or will unexplained Linguists formal line be to this of research will then and semanticists who still want pursue forced logical framework from. This, in defeats to adapt the they started off a sense, the pur- for from pose of their enterprise: one of the reasons starting off the theories and concepts