Tense and Aspects in Discourse 9783110902617, 9783110138139

Table of contents :
Introduction
Temporal structure and discourse structure
Tense use and temporal orientation: The passé simple and the imparfait of French
Future tense and discourse representation
On so-called “tense simplification” in English
Perfect and pluperfect in German
Preterit and perfect in Dutch
Perfect, Prospective and Perspectivity
The English progressive tenses and the layered representation of Functional Grammar
Free indirect speech in French
Aspect and the semantics of noun phrases
Tense and aspect in second language learning: The Dutch interlanguage of a native speaker of English
Index

Citation preview

Tense and Aspect in Discourse

W G DE

Trends in Linguistics Studies and Monographs 75

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Tense and Aspect in Discourse

edited by

Co Vet Carl Vetters

Mouton de Gruyter Berlin · New York

1994

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter & Co., Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

Library of Congress

Cataloging-in-Publication-Data

Tense and aspect in discourse / edited by Co Vet, Carl Vetters, p. cm. - (Trends in linguistics. Studies and monographs ; 75). Includes bibliographical references and index. ISBN 3-11-013813-1 (cloth : acid-free paper) 1. Grammar, Comparative and general — Tense, 2. Grammar, Comparative and general-Aspect. 3. Discourse analysis. I. Vet, Co. II. Vetters, Carl, 1964. III. Series. P281.T375 1994 415-dc20 94-12275 CIP

Die Deutsche Bibliothek -

Cataloging-in-Publication-Data

Tense and aspect in discourse / ed. by Co Vet ; Carl Vetters. — Berlin ; New York : Mouton de Gruyter, 1994 (Trends in linguistics : Studies and monographs ; 75) ISBN 3-11-013813-1 NE: Vet, Co [Hrsg.]; Trends in linguistics / Studies and monographs

© Copyright 1994 by Walter de Gruyter & Co., D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Disk conversion and printing: Arthur Collignon GmbH, Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

Contents

Introduction Co Vet — Carl Vetters

1

Temporal structure and discourse structure Mimo Caenepeel — Marc Moens

5

Tense use and temporal orientation: The passe simple and the imparfait of French Arie Molendijk

21

Future tense and discourse representation Co Vet

49

On so-called "tense simplification" in English Renaat Declerck

77

Perfect and pluperfect in German Rolf Thieroff

99

Preterit and perfect in Dutch Theo A. J. Μ. Janssen

115

Perfect, Prospective and Perspectivity Tim van Baar

147

The English progressive tenses and the layered representation of Functional Grammar Louis Goossens

161

Free indirect speech in French Carl Vetters

179

Aspect and the semantics of noun phrases Hana Filip

227

vi

Contents

Tense and aspect in second language learning: The Dutch interlanguage of a native speaker of English

257

Alex Housen

Index

293

Introduction Co Vet - Carl Vetters

Tense and aspect are no doubt some of the most intriguing phenomena in natural language. They reflect the different ways time is conceptualized by a speech community. It remains unclear, however, why there exists such an amazing variety of ways to express these concepts and why tense and aspect distinctions generally constitute the most difficult part of the language system for non-native language learners, even if the target language is genetically very close to the native one. For the linguist tense and aspect distinctions constitute an important field of research, both from a descriptive and a theoretical point of view. First, it turns out to be very difficult to coin adequate instruments that are refined enough to state the idiosyncrasies of the particular languages in general terms. A second problem is that a considerable part of the meanings of tense and aspect forms strongly depends on contextual factors and probably on the type of text as well, so that tense and aspect cannot be properly studied if their contribution to text cohesion is not taken into account. This book contains a selection of papers that were first presented at the Conference on Tense and Aspect organized by the Belgian Society of Linguistics in Louvain-la Neuve (December 1990) and that were all thoroughly revised afterwards. They all deal with two closely related themes which constitute considerable problems for discourse interpretation rules. First, one and the same tense form often has to be interpreted in very different ways (e.g., the English Progressive can signal that the eventuality is ongoing or that it lies in the near-future). Second, many languages have more than one tense form for the same basic function: anteriority can, for example, be expressed by Simple Past and Present Perfect in Dutch and German, and by two different simple past tenses in French; many languages have two or more tenses to refer to future events, and sometimes verb forms and adverbs seem to have the same function (prospective aspect and not yet, for example). Most often the semantic differences between such pairs cannot be stated in truth-functional terms and have to be attributed to such factors as temporal orientation, text type, degree of evidence at the moment of utterance, and so forth. Some of the authors explicitly derive the different meanings of a tense or aspect

2

Co Vet - Carl Vetters

form from its prototypical or basic meaning. Other papers are more implicit on this point and concentrate on the mechanisms underlying the interpretation process, while at the same time trying to grasp the basic concepts which guide the interpretation. We begin this volume with three papers that focus on interpretation problems encountered by Discourse Representation Theory. Caenepeel and Moens' paper addresses the matter of the so-called reverse order in discourse (as in Max fell. John pushed him). They argue, against recent proposals, that inference rules based on world knowledge and reasoning are not capable of explaining an important class of exceptions. They show that it is not sufficient to take into account the relations between eventualities, but that interpretation rules also have to take into account the relations between the segments of a discourse. These discourse relations are, at least partially, determined by the discourse type (narrative versus non-narrative, for example). Molendijk argues that the difference between the two past tenses of French, the imparfait and the passe simple, can be best stated in terms of "global simultaneity" versus "absence of global simultaneity". The main outcome of his analysis is that the uses and interpretations of these tenses depend mainly on the identification of the fact or time interval which serves as the orientation point from which the eventuality is viewed. It is also shown that a class of apparent counter-examples can be handled if non-overtly expressed (i. e., presupposed or implied) facts is taken into account as well. Vet's paper presents a description of the meaning and uses of the Simple Future, the Periphrastic Future and the Futurate Present in French. It is argued that the meaning of all these tense forms have an attitudinal component, but that they differ with respect to the kind of evidence for the future eventuality the speaker has at his/her disposal at the moment of utterance. The attitudinal component of the future tenses explains why, in past contexts, they behave exactly like verbs of belief. Finally, Vet shows that Aktionsart has an important role to play in the interpretation rules for discourse containing future tense(s). Starting from more varied theoretical backgrounds, the second set of papers (by Declerck, Thieroff, Janssen, van Baar and Goossens) tackles traditionally problematic areas of the description of the tense systems of three different languages (English, German and Dutch). Declerck provides a discussion of the so-called "tense simplification" by which many traditional English grammars try to explain, for example, the seemingly special use of the Present in conditional clauses and temporal clauses referring to the future and the use of the Simple Past where one would expect a Pluperfect. He demonstrates that the idea of tense simplification can be abandoned if it is accepted that

Introduction

3

the tense system of English consists of four absolute time spheres and that other time spheres are temporally subordinated to them. The special uses of some tenses can then be explained by the fact that they can be used to refer to absolute and/or relative time spheres. Thieroff's contribution is a confrontation between the traditional view according to which the Perfects of German are aspects and his own proposal in which they are regarded as tenses. In the "aspectual" view both forms express "termination", and exclude progress in time. In traditional grammar Present, Past and Future Perfects are described as having the same time reference as the corresponding simple tenses (Present, Simple Past and Simple Future). According to Thieroff, this view is wrong, because it cannot explain why the Present Perfect in German can replace the Simple Past. Thieroff also contests Wunderlich's claim that the Present Perfect has two meanings: a "perfectic" and a "non-perfectic" one. He shows that the perfectic and nonperfectic meanings of this form depend mainly on contextual factors. The outcome of Thieroff's analysis is that the Perfects of German can be defined by the fact that the eventuality is anterior to the reference point and that the latter is not (entirely) before the moment of utterance. Janssen claims that the meaning of the Present Perfect is built up in a compositional manner from the auxiliary in the Present and the past participle of the main verb. This leads to the assumption that, as far as tense is concerned, there is no difference between Present and Present Perfect nor between Preterit and Past Perfect. Present and past tenses present eventualities respectively as being of the speaker's focal concern and of his/her disfocal concern. Disfocal can be understood temporally (i. e., the eventuality is outside the speech situation); it can also be used for the expression of potentiality, politeness, and so on. The contrast between focal and disfocal concern also explains the differences between Simple Past and Present Perfect. Van Baar's paper deals with the notions of prospectivity, perfect, and backward and forward perspectivity (expressed in English by the particles already and still respectively). Van Baar argues, against recent proposals in which the difference between aspect and perspectivity is blurred, that prospectivity and perfect are aspectual categories: they presuppose the existence of explicit transition points between situations, whereas perspectivity presupposes the existence of implicit transition points. Crucial to the difference is that aspect and perspectivity behave quite differently with respect to the scope of negation. The last paper in this group, by Louis Goossens, proposes an analysis of the English Progressive in terms of the Functional Grammar model of the layered structure of the clause. Within this framework he discusses two op-

4

Co Vet - Carl Vetters

tions: Progressive is either a realization of the category "progressive" (ongoing), which develops in the direction of a more abstract category "imperfective", or a restricted realization of the category "imperfective", whose prototypical realization is the Progressive. Goossens chooses the latter option and argues that the interpretation of the prototypical Progressive depends on the feature "dynamicity" of the predication and that the basic prototypical meaning is extended metaphorically to express near-future meaning. The last three papers of the volume address more varied descriptive and theoretical issues. Vetters provides a broad discussion of free indirect speech in French. The main difficulty for the interpretation of this kind of discourse is its ambiguity: it can be understood as reporting the thoughts of one of the protagonists or as the author's comment on previously reported events or facts. The analysis of a large number of fragments leads to the conclusion that the relations between direct, indirect, and free indirect speech are not straightforward and that the latter resists any description in syntactic terms. The outcome of Vetters' analysis is that the only clues for the interpretation of free indirect speech are provided by contextual and world knowledge. Filip explores the influence of Czech aspect on the interpretation of noun phrases, which have no article in Czech. She shows that the perfective aspect is responsible for the bounded and specific reading of mass and plural noun phrases in direct object position. She proposes a revised version of Krifka's recent theory in which the aspectual prefixes are regarded as quantifiers which operate on the domain of quantification provided by the direct object of the sentence. Filip claims that the aspectual contour of sentences is the result of the unification of the features "bounded/unbounded" and "quantized/ cumulative", which originate from the sentence's Aktionsart, aspect and Incremental Theme NP. Alex Housen is concerned with the question whether the acquisition of the temporal system of Dutch by a native speaker of English is guided by the principles of the learner's native language and/or by perceptual and semantico-cognitive universals. According to the latter view, foreign language learners develop aspectual notions (responsible for the contrast between foreground and background) before tense distinctions (the "defective tense hypothesis"). He compares the temporality systems that were operative at two stages of the acquisition process. This material reveals, quite unexpectedly, the dominant role of the Dutch verb be as a primary marker of tense at both stages of the acquisition and the slightly more frequent use of the past tense with dynamic verbs. On the whole, however, his material does not support the defective tense hypothesis: at both stages the temporal systems of the learner turn out to be tense-prominent rather than aspect-dominant.

Temporal structure and discourse structure* Mimo Caenepeel — Marc Moens

1. Temporal ordering in discourse 1.1. The role of tense and aspect It is by now generally accepted that the truth-conditions for a discourse consisting of more than one sentence cannot simply be formulated as the sum of the truth-conditions of each of the sentences in isolation. For instance, to specify the truth-conditions of a sequence such as (1)

John came in. He sat down. He poured himself a cup of coffee.

it does not suffice to state that each of the states of affairs occurred at some time in the past, as in a traditional Priorian analysis. The fact that the state of affairs described in the second sentence occurs after the one described in the first (the order of the sentences in the discourse reflects the temporal ordering of the states of affairs) forms part of the semantics of the discourse. Recent discourse-oriented work in the area of natural language interpretation has formulated proposals setting out how features of a sentence's context can be taken into account in its interpretation. This was one of the novel aspects of, for example, Discourse Representation Theory, which drew attention to the discourse role of tense: "... the significance of the tenses lies primarily in the temporal relations which they establish between the sentences in which they occur and the sentences which precede those in the texts or discourses in which those sentences figure." (Kamp-Rohrer 1983: 250) The Discourse Representation Theory approach to tense views tense as an anaphoric device. 1 The anaphoric nature of tense lies in the fact that a tensed sentence has to be interpreted with respect to a particular, previously established reference time which functions as its antecedent. If the sentence contains a definite temporal adverbial, that adverbial will be taken as identifying the reference time, and the tensed expression will be interpreted as co-referential with it. In the absence of such an adverbial the temporal antecedent is to be recovered from the sentence's context.

6

Mimo Caenepeel — Marc Moens

In such an approach, aspectual information serves as input for the discourse construction rules which specify how temporally unmodified sentences are to be ordered temporally. Sentences are classified as belonging to one of three aspectual types, namely events, processes, or states, and, depending on the aspectual type of the sentences concerned, different discourse construction rules will apply. More specifically, depending on the aspectual types of the sentences, the reference times that their tenses refer to will be manipulated in a particular way. Thus, an event sentence will introduce a new reference time "just after" the current reference time (which is usually provided by preceding discourse); and the state of affairs described by the sentence is interpreted as temporally included in this reference time. This procedure is invoked to account for the fact that successive event sentences create the impression of narrative movement. Consider the following example: (2)

John got up. He poured himself a cup of coffee.

The first sentence in this discourse is an event sentence. The state of affairs it describes (soa x \ John got up) is interpreted as occurring inside some contextually given reference time rx. The next sentence is also an event sentence. It introduces a new reference time r2 after r b and the state of affairs it describes (soa2: He poured himself a cup of coffee) is interpreted as temporally included in r2, resulting in the interpretation that soa2 occurred after soax. Process or state sentences, in contrast, are interpreted as describing a state of affairs which surrounds the current reference time, without introducing a new reference time. As a result they do not create the impression of temporal progression. The following example illustrates this: (3)

John got up. He was in a bad mood.

The event sentence introduces a new reference time r and the state of affairs of getting up is interpreted as temporally included in r. The next sentence is stative; no new reference time is introduced, and the state of affairs described by the stative sentence is interpreted as surrounding r, thus resulting in the interpretation that John's being in a bad mood temporally surrounds the event of John's getting up. Combining the notion of tense-as-anaphor with an aspectual analysis thus allows Discourse Representation Theory to account for temporal ordering phenomena in discourses such as the ones in examples (1) and (3). While the approach makes it possible to account for a great many cases, however, there

Temporal structure and discourse structure

7

are some instances where the discourse construction rules proposed by Discourse Representation Theory will yield incorrect or imprecise results. For instance, the second of two consecutive events may be interpreted as forming part of the first (as in example 4) or as simultaneous with it (as in example 5): (4) (5)

We went to London yesterday. We took the 10:30 train. The man ordered her to give him her money. He said the words in a slow, deliberate manner.

Partee (1984) acknowledges this problem, and proposes that a solution to it might lie in the introduction (by discourse-construction rules) of a free context-dependent variable over temporal relations, and deferring until later the decision whether the reference time of the clauses has moved forward or not. Potentially more problematical for Discourse Representation Theory, however, are discourses such as the following: (6) (7) (8) (9)

Frank caught the early train back to London. / gave him a lift to the station. Jane left me. She fell in love with somebody else. Annie broke her leg. She fell off her bicycle. I left the party early. My babysitter phoned up in a bit of a panic.

In all these example discourses, the second event is most plausibly interpreted as coming before, rather than after, the first one. If the standard Discourse Representation Theory rules for the processing of narrative are applied to these discourses they will receive the wrong temporal interpretation. To remedy this, we need to be able to specify why a reverse order interpretation offers itself in the interpretation of discourses such as the ones in (6)—(9) above.

1.2. The role of world knowledge

At first blush, there is an obvious answer to this question: two events will be interpreted in reverse order if world knowledge tells us that such a temporal interpretation is most plausible. This is the approach adopted by a number of

8

Mimo Caenepeel — Marc Moens

Artificial Intelligence-based accounts. For example, Dahlgren et al. (1989) argue that in interpreting discourses such as: (10)

Levine was found guilty. He broke the law.

the principle of the updating of reference times is overridden by scenariobased knowledge about court cases. This knowledge tells us that breaking the law comes before being tried, resulting in the interpretation that the second event in (10) must have occurred before the first (Dahlgren et al. 1989: 166). A similar idea is embodied in the episodic logic of Schubert—Hwang (1989): in their semantics, episodic variables are introduced into the representation of narrative clauses so that implicit, context-dependent relationships between events can be made explicit. The successive description of two events is taken to imply that the first event caused the second — an assumption which is treated as defeasible by episodically organized world knowledge. If the causal link is not defeated, then a temporal progression between the time of occurrence of the two events is assumed (Schubert—Hwang 1989: 452). Lascarides (1990) adopts a similar point of view, but integrates it into a belief- based non-monotonic logic in which knowledge about causality plays a central role. As Lascarides proposes a fairly elaborate and detailed account of the reverse-order phenomenon it is worth summarizing its central claims. Essentially, she proposes an account based on defeasible narrative and causal rules. In some cases, such as example (11), the narrative rule applies, but the causal one does not. Hence we will infer unproblematically that the events happened in sequential order: (11)

Max turned round. John hit him.

In contrast, in a discourse like example (12): (12)

Max fell. John pushed him.

both the narrative rule and a causal rule may apply. The causal rule, in this case, would be based on a causal law which stipulates that if one believes a pushing and a falling occurred2 then one may assume that the pushing caused the falling. This assumption interacts with the defeasible narrative rule which stipulates that the events happened in the order in which they are described. Neither of these assumptions is favored, Lascarides claims, so both conclusions about the temporal order between these events are possible, resulting

Temporal structure and discourse structure

9

in a weakly disjunctive interpretation for the temporal relations expressed in (12). Finally, in example (13) (13)

Max died. John poisoned him.

both the narrative rule and a causal rule may again apply. But this case differs from example (12) in that the narrative rule is overruled by indefeasible world knowledge: it is impossible to poison someone who is already dead. As a result, in the case of example (13) we can draw the conclusion that the second event happened before the first one.

1.3. Problems There can be no doubt that world knowledge plays an important role in the temporal interpretation of discourse: without the support of pragmatic plausibility the reverse temporal interpretation of consecutive event sentences cannot be sustained. Attempting to specify and formalize this role therefore constitutes an important task. Nevertheless, accounts of temporal ordering in discourse which concentrate primarily on reasoning and world knowledge leave a number of linguistic problems unaddressed. One such problem concerns some striking differences in intuitions about reverse- order discourses. For example, in the case of examples (6)—(9) Lascarides would draw weakly disjunctive conclusions; but an informal survey of native speakers' intuitions reveals a strong preference for the reverse order interpretation in such cases. Moreover, many people do not agree with Lascarides' assumption that in example (14) a reverse- order interpretation is not available. (14)

Max turned round. John hit him.

Since language users presumably share the same world knowledge about events (or certainly the type of knowledge Lascarides' account would incorporate into an event ontology) it is not clear why this difference in intuitions should arise. Moreover, if we assume that world knowledge alone licenses a reverse order interpretation for simple past sequences it is not clear why the following sequences sound odd:

10 (15) (16)

Mimo Caenepeel — Marc Moens ? Everyone laughed. Fred told a joke. ? The committee applauded. Nigel announced his promotion.

There is quite a strong causal or scenario-based link between someone telling a joke and people laughing, or between someone's promotion being announced at a meeting and people applauding, but that does not make these discourses acceptable. These observations suggest that the acceptability of reverse order event discourses is not merely a matter of world knowledge. In the rest of this paper we will try to show that a language user (reader/hearer) brings another important type of knowledge to the temporal interpretation of discourse, namely knowledge about discourse structure and discourse type. We will argue that making inferences about these constitutes an inherent part of the interpretation of reverse-order discourses; indeed, that without it, the acceptability of such discourses cannot be assessed. We will also give some examples of clues which, in decontextualized discourses, aid the reader or hearer in making such inferences.

2. Structuring the discourse Let us start by addressing the notion of discourse structure in a very simple way. In an approach which draws upon preferred causal relationships between events in an event ontology, it is not clear what the difference is, if there is one, between narrative-(or sequential-) order and reverse-order sequences. What, for instance, distinguishes the (a) discourse in the following example from the (b) discourse, given that the semantic content of both states of affairs, and the relationship (of causality) between them, are the same? (17) a. Annie fell off her bicycle. She broke her leg. b. Annie broke her leg. She fell off her bicycle. The most manifest difference between the two discourses concerns the staging of the information, or the way in which ... the linear organisation [of the discourse] can be manipulated to bring some items and events into greater prominence than others.

Temporal structure and discourse structure

11

... The notion of "relative prominence" arising from processes of thematisation and "staging" devices has led many researchers, particularly in psycholinguistics, to consider staging as a crucial factor in discourse structure because, they believe, the way a piece of information is staged must have a significant effect on the process of subsequent recall. (Brown-Yule 1983: 134) Thus, a language user would opt for the (b) discourse if the state of affairs Annie broke her leg is to be given greater prominence, for example if it constitutes the discourse topic.3 The second state of affairs is added to provide further information the hearer might subsequently require. The (b) discourse might thus occur in a context such as the following: (18)

A: Is Annie not coming to the meeting today? B: No. She broke her leg. She fell off her bicycle.

Now consider the following discourse: (19)

Annie raced home eagerly, to tell her mother the good news. When she reached the corner, however, she took too sharp a turn, and she fell off her bicycle. She broke her leg, and cried out with pain.

In this discourse, none of the events achieves the prominent status of a discourse topic followed by clauses providing subsidiary information. Instead, the discourse is propelled by the implicit question and then? which, with each new clause, leads the hearer/reader to the next event. Thus, while the semantic/referential relationship between the clauses She broke her leg and She fell off her bicycle is the same in examples (18) and (19), the structure of the discourse is crucially different. This difference becomes most clear when we look at the structure of discourse at a more abstract level — for example by invoking the tree-like structures proposed in Scha and Polanyi's Dynamic Discourse Model. This model represents the structure of a discourse as a tree, in which discourse constituent units are related to each other through a relationship of either coordination or subordination (Scha—Polanyi 1988). Each relationship is embodied by different kinds of discourse structure.4 The relationship in (17a) is narrative — a (binary, in this case) relationship of coordination; the one in (17b), on the other hand, is one of explanation — a binary relationship of subordination. In such an approach, the logical form of a sentence remains the same, irres-

12

Mimo Caenepeel — Marc Moens

pective of the type of discourse it appears in. What does differ are the manipulations the discourse- grammar rules perform over these logical forms. 5 It follows from this that we need to distinguish eventuality relations between states of affairs in a discourse, and discourse relations between the discourse segments. Temporal structure and discourse structure should be studied as separate but interrelated aspects of discourse.

3. Discourse type and context 3.1. Contextualizing discourse In the previous section we discussed the profile of reverse-order discourses in terms of staging and the relative prominence or subsidiary nature of the information conveyed by the sentences concerned. But now another problem arises: if a language user wants to express that an event in the past occurred before another already mentioned event in the past, he has a specific and unambiguous means at his disposal, namely the use of the past perfect. Consider the following examples: (20) a. Jane left me. She fell in love with somebody else. b. Jane left me. She had fallen in love with someone else. As in the case of (17) above, the semantic content of the states of affairs in the two discourses is the same; moreover, the two discourses exhibit the same eventuality relationship (cause) and the same discourse relationship (explanation). Does this mean they are identical and interchangeable? The difference between both discourses is hard to assess. The reason for this lies, we believe, in the fact that they are offered in a decontextualized manner. While a reverse-order simple past sequence will be acceptable (and processed easily) in one type of context, in another type of context the same discourse will be perceived as puzzling or odd. To see this, both example discourses in (20) need to be assessed in context. While context is a complex notion which has many parameters, the distinction relevant for our purposes can be captured in terms of a contrast between narrative and non-narrative contexts.6 We propose a definition of this contrast in terms of the relationship between the utterance and the situation of speech.

Temporal structure and discourse structure

13

If the utterance is deictically related to the actual situation of speech, so that the situational features of the latter contribute directly to the understanding of the utterance, the discourse is embedded in a non-narrative context.7 In a non-narrative context, two events can be described in reverse order.8 Thus, the use of the two simple pasts in example (20a) is appropriate in a context like the following: (21)

context: A runs into B, an old friend he has not seen for a long time. He asks Β how he's doing. Β replies: Not great. Jane left me. She fell in love with someone else.

If a simple past reverse-order discourse is offered in isolation — i. e., without information about the context it is embedded in — it will be easier to process if it contains linguistic elements which cannot be interpreted except with reference to the situatedness and the perspective of the speaker, such as deictic, expressive, and/or communicative elements. If the discourse contains such elements the reader will spontaneously construct the type of non-narrative context required for the interpretation of a simple-past reverse-order discourse, since the elements are to be related to the situational features of the speech point. This is illustrated by example (22), where the relevant situational elements have been romanized. Note that in the type of context evoked here the use of the past perfect seems inappropriate: (22)

Jane left me. Just imagine, she fell in love llhadfallen

with that stupid in love biologist

Although in principle every utterance originates in a situation of speech, a discourse may be presented as distanced from the actual time/space coordinate at which it is produced. This can be achieved by construing the material which makes up the discourse as a narrative, or a story.9 In a narrative, states of affairs are presented in a self-contained temporal continuum in which they are anaphorically related to each other, instead of deictically to the actual situation of speech. In example (23), the discourse in (20b) has been incorporated into a narrative context: 10 (23)

Three months after that, Jane left me. She had fallen in love with someone else. I was terribly upset at first, but eventually got over the shock, and started to go out with other people again. That was when I met Annie.

14

Mi mo Caenepeel — Marc Moens

This type of context requires the use of the past perfect for the expression of an explanatory relation of temporal precedence: it appears that, in the absence of the support of a direct link with the situation of speech, a relationship of temporal precedence has to be marked explicitly.11 Indeed, if the past perfect is substituted with a simple past in such a context, we will infer narrative movement: 12 (24)

Three months after that Jane left me. She fell in love with someone else, and moved to the States. I never saw her again.

It follows from this that the ease with which decontextualized simple-past reverse- order discourses are processed depends on the ease with which the appropriate context can be constructed. As already pointed out earlier, the occurrence of expressive or communicative elements facilitates this construction. Another factor which affects comprehension concerns the medium: often reverse-order discourses which seem odd or unacceptable in written form are more readily understood and accepted when they are offered as spoken discourse. This is in line with our hypothesis: the spoken medium contributes to suggesting the appropriate discourse type, and prosody constitutes an enabling, or supporting, device for simple-past reverse-order discourses. Thus, informants tend to frown at the written version of the following reverse order sequences: (25) (26)

? John went into the flower shop. He promised his wife some roses. ? Joe was discharged from hospital. He recovered completely.

If spoken, or if sufficient contextual information is added, however, the reverse- order interpretation becomes unproblematically acceptable: (27)

context: A meets Β in the street. He asks where their mutual friend C is. Β answers: He went into that flower shop over there. He promised Mary some roses.

(28)

context: Someone is at the hospital to visit her nephew, and asks a nurse for more information. The nurse answers: Your nephew was discharged from hospital earlier this morning, Mrs Jones. He recovered completely.

In this light it becomes clear why example (14) (repeated here as 29a) meets with mixed intuitions: its acceptability depends on the context in which

Temporal structure and discourse structure

15

it occurs. In a non- narrative context the reverse-order interpretation is easily accessible; example (29b) illustrates this. But in a narrative context the second simple-past sentence will be interpreted as introducing a temporal update, as in (29c). To convey explanatory temporal precedence in such a context a past perfect has to be used (29d): (29) a. Max turned around. John hit him. b. non-narrative context: A: How come John saw what happened behind them? B: He turned around. Max hit him. c. narrative context: At that point John stopped, and turned around. Max hit him. John fell, and... d. narrative context: At that point John stopped, and turned around. Max had hit him. The distinction between narrative and non-narrative contexts also enables us to explain why the examples in (15) and (16) are odd, despite the obvious causal relation between the events: it is very difficult to construct the nonnarrative context required for their interpretation. But the following discourses show it is not impossible: (30)

context: A is waiting outside a seminar room. Suddenly a lot of noise comes from inside the room. Two seconds later Β steps out of the room. The following conversation ensues: A: Good heavens! What was all that noise? B: The audience applauded. Keith announced his promotion. or: B: Everybody laughed. Fred told his parrot joke.

3.2. Discourse phenomena and discourse types Our discussion in the previous section shows that while supporting world knowledge constitutes a necessary condition for two simple-past sentences to be interpreted in reverse order, it does not constitute a sufficient one. A reverse-order interpretation of two simple-past events also requires the support of a non-narrative discourse context.

16

Mimo Caenepeel — Marc Moens

More generally, our analysis suggests that different discourse types have different phenomena associated with them. Because of this it is important to clearly separate out, on the one hand, rules for the semantic analysis of sentences, and, on the other, interpretation principles for different types of discourse. If rules for the processing of narrative discourse are applied to nonnarrative discourse, for instance, important distinctions in the use of tense and aspect will be missed. Thus in the following two discourses, which belong to the non-narrative type, the aspectual profile of the second state of affairs (which is a state in the (a) example, and an event in the (b) example) has little impact on the structure of the discourse — the discourse relation is, in both instances, one of explanation: (31) a. Jane left me. She was dissatisfied with our relationship. b. Jane left me. She fell in love with someone else. In the following discourses, however, the difference between the choice of a stative expression in the (a) example and an event in the (b) example, has a more pervasive effect on the structure of the discourse: in the (a) case the discourse relation is one of elaboration (or background); in (b) it is one of narration: (32) a. Two hours later Jane left the house. She was in good spirits. b. Two hours later Jane left the house. She hailed a taxi, and... A similar point is made in Reichman's (1984) discussion of an instructive discourse: (33)

There's an automatic control device — you can do it by hand. In fact, initially when you light off it is — when you turn on the steam plant — it s done by hand. But after you get going there is an automatic system that is sensing the flow of steam out of the steam generator. It's sensing the level in the steam generator, and it's also sensing the flow of the water through the feed reg valve. Okay?... That's a separate system... And I didn't mean to tell you about that now, but as long as you asked about it. (Reichman 1984: 356)

In this discourse the expert briefly embarks on a subdiscourse (closed off with and I didn't mean to tell you about that now). He signals this to the listener by switching to progressives in the subdiscourse. Clearly, if one were

Temporal structure and discourse structure

17

to apply the rules for the processing of narrative discourse to such a discourse, one would miss this important function of the progressive, by postulating a non-existent temporal relation between the sentences in the progressive and the events described in the preceding discourse. Clearly, this is a different type of discourse, and different processing rules are in order from the ones used to process narrative discourse. 13 A number of people working in the area of discourse, like Reichman, explicitly opt for making a principled distinction between different discourse rules for different discourse types. Dowty (1986), for instance, proposes a discourse interpretation principle specifically for narrative discourse. In other work, the distinction is adopted more casually: thus Partee (1984) acknowledges in passing that her theoretical observations apply only to narrative discourse, but she does not discuss what motivates this claim, nor what its implications are. One of the implications for Discourse Representation Theory is that, as long as the approach restricts itself exclusively to narrative discourse, it need not deal with simple-past reverse-order sequences, since they do not constitute a phenomenon that belongs to narrative discourse.

4. Conclusions

We have tried to show in this article that world knowledge about relationships between events by itself is not sufficient to explain when simple-past reverseorder discourses are acceptable; knowledge about discourse structure and discourse types also plays an important role. Event structure and discourse structure constitute separate (albeit interrelated) aspects of discourse, and the interaction between the two deserves further exploration. A full account of temporal relations in discourse should observe the division of labor between sentence semantics and discourse rules, and it is important to determine what work is done at what level. Different types of discourse require different processing rules. Narrative discourse is only one such type, albeit the one that within the Discourse Representation Theory literature has received most attention. It would be a mistake to assume that rules for narrative processing automatically apply to other types of discourse as well, or to complicate the same set of rules further so as to make them fit non-narrative discourse as well.

18

Mimo Caenepeel — Marc Moens

Notes * This work was supported by Economic and Social Research Council grant number R000231617 to the first author, and by ESPRIT Basic Research Working Group BR3351 (DANDI). The article was finished in April 1991. After delays in the book's production process, only minor stylistic changes were carried out to this paper in March 1993. 1. This idea was first suggested in McCawley (1971) and Partee (1973), but it was only developed fully in a number of approaches couched in a Discourse Representation Theory framework (e.g., Hinrichs 1986, Partee 1984, Vet 1987). More recently a number of other accounts have been suggested in different theoretical frameworks (e.g., Dowty 1986, Webber 1988). 2. Presumably involving the same participants in the same participant roles, at roughly the same time and roughly the same place. 3. For a discussion of the notion of discourse topic see van Kuppevelt (1991). 4. Thus narratives, lists, and adjacency pairs constitute examples of co-ordination; while rhetorical subordination (generalization, explanation, comment, etc.), topic/ dominant chaining and interruption constitute examples of subordinated relationships. Only narrative embodies a principle of narrative time progression; the other discourse relations embody very different manipulations of the semantic structures associated with the sentences that make up the discourse. 5. In case more than one rule applies to a given discourse, preferences are stated over the discourse-processing rules. 6. This is probably an oversimplification; it is likely that in a fuller exploration of the effect of types of context on temporal ordering more fine-grained distinctions need to be made; but this is beyond the scope of this article. 7. Most conversational exchanges belong to this type of context, although it is possible to introduce a narrative context into a conversation. 8. Always provided this is pragmatically plausible, of course. 9. Note that while a story is by no means necessarily imaginary in character, the term tends to be suggestive of fictionality. This is due to the fact that fictionality constitutes an unambiguous distancing device. 10. As the distinctive character of a narrative context lies precisely in the fact that it does not interact directly with the actual situation of speech, there is little point in describing such a context as we have done for example (21). Instead, we have marked the narrative context through the use of the anaphoric temporal adverbial in the first sentence. A similar procedure to indicate that a decontextualized discourse belongs to a narrative context is suggested by Sandström (1990). 11. A flashback initiated by a past perfect may well be continued in the simple past. This is quite a common phenomenon, especially in American English. But even in such cases, at least the starting point is explicitly indicated. 12. Or if narrative movement is pragmatically implausible, the discourse will sound odd. 13. Apart from the actual situational features of a discourse, a discourse will usually contain other clues which will enable the reader/listener to decide what type of discourse she is dealing with. The role of syntactic information in this respect

Temporal structure and discourse structure

19

depends on the language concerned. As we have seen, the simple past tense in English has both a narrative and a non-narrative use. But in other languages, such as Dutch, French, or German, the use of a simple past tense often acts as an indicator of narrativity (cf. Weinrich 1964; Janssen 1990). In these languages it is virtually impossible for simple past sentences to be interpreted in reverse order.

References Brown, Gillian—George Yule 1983 Discourse analysis. Cambridge: Cambridge University Press. Dahlgren, Kathleen—Joyce McDowell-Edward P. Stabler 1989 "Knowledge representation for commonsense reasoning with text", Computational Linguistics 15: 149—170. Dowty, David 1986 "The effects of aspectual class on the temporal structure of discourse: semantics or pragmatics?", Linguistics and Philosophy 9: 37 — 61. Hinrichs, Erhard 1986 "Temporal anaphora in discourses of English", Linguistics and Philosophy 9: 6 3 - 8 2 . Janssen, Theo. A. J. Μ. 1990 "Preterit as definite description", in: Jadranka Gvozdanovi-Theo A. J. M. Janssen, The function of tense in texts, Amsterdam: North-Holland. Kamp, Hans-Christian Rohrer 1983 "Tense in texts", in: Rainer Bäuerle—Christoph Schwarze-Arnim von Stechow (eds.), Meaning, use and interpretation of language, Berlin: de Gruyter, 2 5 0 - 2 6 9 . Kuppeveit, Jan van 1991 Topic en comment. Expliciete en impliciete vraagstelling in discourse [Topic and comment. Explicit and implicit questions in discourse], PhD dissertation, Katholieke Universiteit Nijmegen. Lascarides, Alex 1990 Knowledge, causality and temporal representation. Research Report HCRC/RP-8, Human Communication Research Centre, University of Edinburgh. McCawley, James D. 1971 "Tense and time reference in English", in: Charles Fillmore—Terence Langendoen (eds.), Studies in Linguistic Semantics, New York: Holt, Reinhart and Winston, 96— 113. Partee, Barbara 1973 "Some structural analogies between tenses and pronouns in English", Journal of Philosophy 70: 6 0 1 - 6 0 9 . 1984 "Nominal and temporal anaphora", Linguistics and Philosophy 7: 243-286. Reichman, Rachel 1984 "Technical discourse: The present progressive tense, the deictic 'that', and pronominalization", Discourse Processes 7: 337 — 369.

20

Mimo Caenepeel — Marc Moens

Sandström, Goerel 1990 Temporal adverbial clauses in narrative discourse. Dissertation, filosofie licentiatexamen in General Linguistics, University of Umeä. Scha, Remko—Livia Polyani 1988 "An augmented context free grammar for discourse", in: Proceedings of the 12th International Conference on Computational Linguistics and the 24th Annual Meeting of the Association for Computational Linguistics, Budapest, Hungary, 2 2 - 2 7 August, 1988: 573-577. Schubert, Lenhart K.—Chung Hee Hwang 1989 "An episodic knowledge representation for narrative text", in: Ronald Brachman—Hector Levesque-Raymond Reiter (eds.), First International Conference on Principles of Knowledge Representation and Reasoning, Toronto, May 1989: 444-458. Vet, Co 1987 "Temporele relaties in teksten" [Temporal relations in texts], Interdisciplinair Tijdschrift voor Taal- en Tekstwetenschap 7: 149- 166. Webber, Bonnie Lynn 1988 "Tense as discourse anaphor", Computational Linguistics 14: 61 — 73. Weinrich, Harald 1964 Tempus. Besprochene und erzählte Welt [Tempus. Described world and narrated world]. Stuttgart: Kohlhammer.

Tense use and temporal orientation: The passi simple and the imparfait of French Arie Molendijk

Introduction1 Those whose native language does not mark in an overt way what distinguishes the passi simple and the imparfait of French have as a rule great difficulty in using these tenses properly. This is even true for speakers who have acquired a relatively high degree of perfection in French. One might think that the problem is due to the fact that the meaning of the passi simple and the imparfait is a complex one. This position is adopted by Kamp (1981), among others, who states that the passi simple and the imparfait have several semantic features, none of which is absolutely permanent. Thus, the passi simple normally drives the narrative's action forward, but this situation does not necessarily occur. Similarly, the passi simple often conveys punctuality, but there exist non-punctual uses of this tense. The imparfait describes facts that obtain while other facts occur, but this is not a necessary condition for its use, and so forth.2 It will be argued in this paper that the source of the problem concerning the use of the passe simple and the imparfait does not reside in the meaning of these tenses itself: the passe simple only differs from the imparfait in that the latter invariably conveys global simultaneity, whereas the former never does. Other characteristics of the passi simple and the imparfait are merely symptoms of this underlying difference, and do not necessarily manifest themselves. For reasons of simplicity, certain problems related to the socalled picturesque use of the imparfait are disregarded here. Why is the use of these tenses so problematic then? The answer to this question can be illustrated by an example like (l): 3 (1)

The old man lit the lamp (SI). The feeble light gave the room an appearance of sadness (S2)

Should S2 of (1) be rendered in French by a passi simple sentence (la faible lumiere donna ά la piece ...) or an imparfait sentence (la faible lumiere don-

22

Arie Molendijk

nait ά la piece ...)? The following (theoretical) considerations are relevant with respect to this question. In (1), the feeble light indicates that S2 talks about what is temporally implied (i. e., "created") by SI, the temporal implication being something like there be light. Temporal implications are implications that have a temporal dimension. X temporally implies Y if (i) X implies Y and (ii) Y denotes something that is (immediately) posterior to what is referred to by X (cf. Molendijk 1990: 8 4 - 9 1 ) . As S2 speaks about what is temporally implied by SI, it refers to a fact that is posterior to E l , i. e., E2 is posterior to E l . One might conclude from this fact that S2 of (1) "should" be read as (exclusively) expressing posteriority, which would exclude the use of donnait 'gave' (imparfait) in the French equivalent of this sentence. 4 But this conclusion does not necessarily impose itself. There is another theoretical possibility. The definite article the in the subject-NP of S2 tells us that this sentence is about an existing light, i. e., about a light that is presented as "being already there" before the appearance of S2. So S2 could be said to express simultaneity with respect to an already existing fact. This fact, then, would correspond to what is temporally implied by SI: there be light. To put it another way, (1) could be considered (almost) identical to (2), the main difference between the two examples being that (2) explicitly mentions a fact that corresponds to a "hidden sentence" in (1): (2)

The old man lit the lamp (SI). There was light (in the room) now (S2). The feeble light gave the room an appearance of sadness (S3)

This would imply that S2 of (1) "should" be read as expressing (global) simultaneity, which would exclude the use of donna 'gave' (passe simple) in the French translation of S2 of (1). It is clear that the choice between the two (theoretical) possibilities mentioned above depends on what can be legitimately considered to be the (temporal) orientation point of a sentence. Defending the idea that S2 of (1) expresses posteriority rather than simultaneity is tantamount to saying that S2 is directly oriented to the fact explicitly reported by SI rather than to there be light. But claiming that S2 of (1) expresses simultaneity rather than posteriority would presuppose that S2 is to be interpreted as oriented to there be light rather than to the event mentioned in S1. The linguistic or pragmatic motivation of the choice between the two possibilities presented above will not concern us in this section (but see 3—3.1). Whatever factors play a role here, it can be argued that the question of the proper use of the passe simple and the imparfait is not primarily about the

Tense use and temporal orientation

23

meaning of these tenses itself, but about temporal orientation. In order to know what is the appropriate tense form for a given sentence S, it is necessary to identify the exact fact or moment of time that serves as orientation point for S. This topic will receive considerable attention in this paper (3—4). We will see that there are conditions under which a sentence (preferably) takes as its orientation point a fact that is not explicitly mentioned (3 — 3.1), whereas in other cases the orientation point of the sentence is a fact that is explicitly mentioned (3.2). In some cases, both possibilities exist at the same time (3.3). We will also see that the question of the orientation point of a sentence may be dependent of its syntactic context. Temporal clauses provide interesting evidence for this claim (4). But first it will be shown that the difference between the imparfait and the pass0 simple is a purely temporal one (global simultaneity vs. absence of global simultaneity; 2—2.2), and that traditional and more recent descriptions of these tenses cannot be maintained (section 1). The claims that will be made with respect to the use and the interpretation of the passe simple and the imparfait and with respect to the regularities existing in the field of temporal orientation can be seen as additions to, and modifications of, some of the proposals put forward by Kamp—Rohrer (1983) and Smith (1978). The facts that will be discussed here will not, however, be explicitly related to the theories elaborated in those papers. Judging the claims that will be made in this paper does not require familiarity with the theories

in question.

1. Descriptions of the passe simple and the imparfait The descriptions that will be briefly examined in this section are listed in (3): 5 (3)

Passe simple

Imparfait

foreground information perfective aspect punctuality of the fact narrative non-anaphoric distantiation and dimensionalisation

background information imperfective aspect durativity of the fact descriptive anaphoric absence of distantiation and of dimensionalisation

24

Arie

Molendijk

None of the characteristics given in (3) are fundamental to the passä simple and the imparfait. For instance, if the distinction between these tenses would essentially concern the "importance" of the information contained in sentences (foregrounding vs. backgrounding), then the occurrence of passe simple sentences reporting background information and/or imparfait sentences containing foreground information should be excluded. But this is not the case, as can be shown by (4) and (5):6 (4)

La Guerre de Cent Ans — qui dura d'ailleurs 116 ans (S2) — fut surtout ameme par ... (SI) 'The Hundred Years' War - which actually lasted (PS) 116 years (S2) — was primarily provoked by ... (SI)' (dura: background information, as formally indicated by d'ailleurs 'actually')

(5)

Je vis (SI) que, brusquement, Jean sortait un revolver (S2) Ί saw (SI) that John suddenly drew (IMP) a gun (S2)' {sortait: foreground information, as indicated by brusquement 'suddenly')

Similarly, if the difference between the pass0 simple and the imparfait were fundamentally aspectual (perfectivity vs. imperfectivity), then the use of the imparfait in sentences that refer to "completed" facts should be impossible. But (6) shows that this is not what actually happens: (6)

Quand j'atteignis la foret (SI), une heure sonnait (S2) 'When I reached the woods (SI), the clock struck (IMP) one (S2)' {sonnait: perfective, "completed")

As for the alleged punctuality of the passe simple and the so-called durativity of the imparfait, consider (7): (7)

Jean prit la parole (SI). II ne nous parlait que de ses affaires (S2). II ne nous parla que de ςα (S3) 'Jean took the floor (SI). He talked (IMP) about nothing but his work (S2). He talked (PS) about nothing else (S3)'

S3 refers to a fact that is temporally more extended than the one mentioned in S2: he talked about nothing but his work (S2) and continued doing so all evening/during the whole party (S3). Uses of the passe simple and the imparfait like the ones illustrated by (7) are clearly incompatible with the hypoth-

Tense use and temporal orientation

25

esis about the punctuality of the passe simple and the durativity of the imparfait. If this hypothesis were correct, imparfait facts should "take more time" than passe simple facts in contexts like (7), in which the same fact is reported both in a sentence in the imparfait and in a subsequent sentence in the passe simple. But in these cases, the passe simple fact is as a rule more extended ^temporally) than the imparfait fact. 7 If the passe simple were fundamentally punctual, this tense form should not be possible, in (7). It can also be shown that sentences in the passe simple are not invariably used for narration, and that imparfait sentences are not necessarily descriptive. Intuitively, S2 of (5) is clearly narrative, despite the (obligatory) use of the imparfait, and S2 of (4), which only allows the use of the passe simple, is descriptive. Linguists who call the imparfait an anaphoric (temporal) element (cf. Vet 1985, Vetters 1989) relate the expression "temporal anaphora" to the notion of simultaneity. (For an interesting discussion about the applicability of the notion of anaphora to temporal elements, see Kleiber (1993); cf. also Molendijk (1993). So, according to them, the imparfait expresses simultaneity, whereas the passe simple does not. One of the problems with this analysis is that it does not distinguish between several types of simultaneity. We will see in 2—2.2 that it is not primarily the absence or presence of a relationship of simultaneity that determines the (im)possibility of using the passe simple or the imparfait, but the existence of certain types of simultaneity rather than others. Thus, in an example like (8): (8)

Pierre se promenait les conditions de vie 'Pierre and his wife to her the conditions

avec sa femme (SI). II expliquait a sa femme sur Venus (S2) et lui indiqua Mars (S3) were taking a walk (SI). He explained (IMP) of life on Venus (S2) and indicated (PS) Mars'

both expliquait (imparfait) and indiqua (passä simple) indicate simultaneity with respect to the walking. (This interpretation clearly has a pragmatic aspect: facts like E2 and E3 commonly occur while facts like El are taking place). But they do not do so in the same way. Expliquait (imparfait) qualifies the walk as a whole, whereas indiqua (pass0 simple) only refers to a specific moment of the walk (see 2 - 2 . 2 ) . Let us finally consider the claim that passe simple facts differ from imparfait facts in that the former facts, but not the latter, are to be conceived of as having clear-cut dimensions and as being (physically, psychologically, or otherwise) distant from the moment of speech (cf. Waugh-Monville-Burston

26

Arie

Molendijk

1985). As for the feature [+distant], it can easily be shown that it is not fundamental to the pass0 simple. Consider (9) and (10): (9)

II fut et reste le plus grand footballeur de notre temps 'He was (PS) and remains the best football-player of our time'

(10)

Hier soir, la troisieme chaine presenta un film intitule "L 'έΐέρΐιαηί ä la peau blanche" (SI). Dans ce film, il s'agit dun homme (S2) qui ... 'Yesterday evening, the third channel featured (PS) "The elephant with the white hide" (SI). This film is about a man (S2) who...'

In (9), il fut le plus grand footballeur de ... 'he was the best football-player of ...' refers to a fact that is not entirely situated before the moment of speech, which explains why il fut ... 'he was ..·.' can be followed by et reste 'and remains'. This use of the passe simple would not be possible if this tense form had the feature [ +distant]. This feature would also forbid the passe simple to combine with an adverbial like hier soir 'yesterday evening', because of its deictic nature. But (10) is perfectly normal. As for the alleged dimensionality of the passe simple, it should be noted that this is not an exclusive property of this tense form: S2 of (6) (imparfait) clearly presents the fact it reports as having clear-cut dimensions. The conclusion that must be drawn from what has been said in this section is that the distinction between the passe simple and the imparfait must not be sought in the notional oppositions given in (3).

2. The temporal difference between the passe simple and the imparfait It will be argued in this section that the imparfait basically differs from the passe simple in that the former conveys global simultaneity, whereas the latter never has this property. Let us first focus on the imparfait.

2.1. The temporal function of the imparfait The hypothesis concerning the idea of global simultaneity contained in the imparfait can be formulated as in (11):

Tense use and temporal orientation

(11)

27

Information contained in an imparfait form occurring in a sentence P: There is a fact or moment of time X (explicitly provided by or deducible from the text that contains P, or relevant to the (speech) situation in which Ρ is asserted) such that Ρ globally holds for Χ. (X is anterior to the moment of speech).

(11) has always been illustrated by examples like (12): (12)

Vers 6 heures, Jean rentra (SI). Sa femme sortait (S2) 'About 6 o'clock, Jean came home (SI). His wife was just leaving (IMP) (S2)'

This type of examples will not be examined here, the more interesting cases being provided by examples in which the imparfait fact does not coincide with the one mentioned in the preceding sentence. This does not automatically mean that the imparfait does not express simultaneity, as we will see. These cases will be analyzed in 2.1.1 and 2.1.2. But first something must be said about the notion of global simultaneity. This notion can be informally defined as in (13). As there seems to be sufficient consensus about what it means for two entities to be simply simultaneous, the definition concentrates on the interpretation of "global": (13)

Global simultaneity ("="): X = Y if and only if X coincides with Y and: (i) X is not a proper part of Y; (ii) if X holds for a moment of time that temporally precedes Y, then it also holds for a moment of time that temporally follows Y; (iii) if X holds for a moment of time that temporally follows Y, then it also holds for a moment of time that temporally precedes Y.

According to (13), E2 (14)

Ξ

E l , in (12). The same is true for (14):

Pierre rentra (SI). IIpleuvait (S2) 'Pierre came home (SI). It was raining (S2)'

Note that in (14), E2 precedes and follows E l , temporally (cf. (13,ii — iii)). But in (7) and (8), E3 is not globally simultaneous with E2, according to the

28

Arie Molendijk

proposed interpretation of this notion, because E3 of (7) extends to the right over E2, but not to the left (meaning that (13,iii) is not satisfied), and E3 of (8) occupies a portion only of E2—meaning that (13,i) is not satisfied. We will see in 2—2.2 how these facts can be related to the use of the passe simple and the imparfait. Let us get back now to the central topic of these sections: the temporal information provided by the imparfait.

2.1.1. Imparfait and non-simultaneity What will be said in this section and in 2.1.2—2.2 about the imparfait and the passe simple implicitly rests on certain presuppositions as to what is the fact or moment of time that serves as orientation point for a given sentence. Some regularities with respect to this question will be treated rather loosely here. For reasons of clarity, more precise rules will be kept for the sections 3 and 4). Let us consider (15) - (18): (15)

Pierre alia ä son travail dans sa nouvelle Mercedes (SI). Juste ά l'entree de Gorges, il attrapa une contravention (S2): il roulait trop vite (S3) 'Pierre went to his work in his new Mercedes (SI). Just before entering Gorges, he was fined (S2): he was driving (IMP) too fast (S2)'

(16)

Le vieil homme alluma la lampe (SI). La faible lumiere donnait ά la piece un air de tristesse (S2) 'The old man lit the lamp (SI). The feeble light gave (IMP) the room an appearance of sadness (S2)'

(17)

Jean accälera son pas (SI). Comme il atteignait la foret (S2), il entendit un bruit insolite (S3) 'Jean quickened his pace (SI). Just as he reached (IMP) the woods (S2), he heard a strange noise (S3)'

(18)

"Qu 'est-ce que c 'est que ςα (S1)?" "Oh rien, c 'etait Pierre qui fermait la porte (S2)". 'What's that (SI)? Oh, nothing, that was Pierre who closed (IMP) the door (S2).'

In (15), the sentence in the imparfait is not interpreted as establishing a relationship of simultaneity with the preceding sentence.8 Similar remarks

Tense use and temporal orientation

29

can be made about the imparfait sentences occurring in (16) - (18): none of the imparfait facts mentioned in (16) — (18) is simultaneous with the one mentioned in the preceding sentence. Yet it can be maintained that the tense forms occurring in the imparfait sentences of (15) — (18) express simultaneity. Before giving positive arguments for the claim about the temporal information provided by the imparfait, let us assume for the sake of the argument that in (15) — (18) the imparfait does not express simultaneity. Then it would seem logical to admit that in (15) and (18), for instance, the imparfait expresses anteriority, because E3 of (15) and E2 of (18) are anterior to the fact referred to by the preceding sentence of these examples. For similar reasons, the imparfait in (16) — (17) would express posteriority. But these assumptions give rise to serious problems. If one admits that roulait 'was driving', ex. (15), and fermait 'closed', in (18), express anteriority, one has to explain why this kind of temporal relationship cannot be expressed by the same forms in examples like (19) and (20):9 (19)

Pierre se coucha (SI). Le lendemain, il fut jeti en prison (S2): il (*)roulait trop vite (S3) 'Pierre went to bed (SI). The following day, he was sent to prison (S2): he was driving (IMP) too fast (S3)' (The asterisk concerns the non-habitual reading of S3)

(20)

"Pourquoi la porte η 'est-elle pas ouverte (SI)?" "Pierre la *fermait (S2)". 'Why isn't the door open (SI)? (Literally:) Pierre closed (IMP) it (S2>*

And the claim about the relationship of posteriority allegedly established by donnait 'gave', in (16), and atteignait 'reached', in (17), would force us to explain why these forms are no longer appropriate in contexts like (21) and (22): (21)

Le vieil homme alluma la lampe (SI), et une faible lumiere * donnait ä la piece un air de tristesse (S2) 'The old man lit the lamp (SI), and a feeble light gave (IMP) the room an appearance of sadness (S2)'

(22)

Jean accelera son pas (SI). Bientot, il * atteignait la foret (S2). II entendit un bruit insolite (S3) 'Jean quickened his pace (SI). Soon he reached (IMP) the woods (S2). He heard a strange noise (S3)'

30

Arie

Molendijk

Note that (21) and (22) would be normal with respectively donna (pass0 simple) and atteignit (passe simple) (see 2.1.2—2.2). These problems cannot be solved on the basis of the claims that have been hypothetically made above about the meaning ("non-simultaneousness") of the imparfait in (15) — (18). But they can, as we will see, if one admits that not only in "easy" cases like (12), but also in more complicated ones such as (15) — (18), the imparfait expresses simultaneity.

2.1.2 More about non-simultaneity and the imparfait It can be argued that the only difference between the two cases described at the end of 2.1.1 is that in (12), but not in (15) — (18), the fact or moment of time with which the imparfait establishes (global) simultaneity is explicitly mentioned in the preceding sentence. When cases like (15) — (18) occur, the temporal relationship established by the imparfait frequently concerns the imparfait fact on the one hand and a fact of type (i), (ii) or (iii) on the other: 10 (i)

(ii)

(iii)

a fact reported by a sentence that does not precede the sentence in the imparfait, or that, if it does, does not immediately precede the imparfait sentence; or a fact corresponding to a sentence that is temporally presupposed (for this notion, see below, this section) or temporally implied by a sentence preceding the imparfait sentence; or a fact that is part of the speech-situation.

As for (i), this case applies to (15). According to (11), the tense form occurring in S3 of (15) could be said to "force" the hearer to look for a fact or a moment of time with which S3 can establish the relation " = " . The recipient does not consider E2 as entering into this relationship, because world knowledge rules out this reading (cf. what has been said about (15)). But these pragmatic considerations do allow the hearer to consider E3 as going on "at" E l , meaning that S3 establishes a relationship of global simultaneity between the fact reported by S3 and the one mentioned in SI. (i) does not only apply to (15), but also to (17), the only difference being that in (17), but not in (15), " = " concerns the imparfait fact and a fact reported by a sentence (clause) that follows the imparfait sentence (clause). We will see in section 4 that this situation normally occurs when the imparfait is contained in a subordinate clause introduced by a conjunction like comme 'just as'. 11 As for (ii), this situation occurs in (16). SI temporally implies il y avoir de la lumiere 'there be light', which immediately follows (temporally) the

Tense use and temporal orientation

31

fact that is mentioned explicitly. Assuming that (11) is right, we can say that donnait 'gave' (imparfait) is used to establish global simultaneity between the imparfait fact and il y avoir de la lumiere 'there be light', temporally implied by SI of (16). So we see that one of the theoretical possibilities that were discussed in the introduction with regard to (1) is realized in French, but an explanation of this phenomenon has not yet be given (cf. section 3.1). Cases like (16) only differ from examples like (23): (23)

Pierre rentra (SI). Le soleil lui brulait les ipaules (S2) 'Pierre went into the house (SI). The sun was burning (IMP) his shoulders (S2)'

in that the imparfait fact of (23) does not coincide globally with a fact corresponding to a temporal implication, but with one that corresponds to a temporal presupposition, i. e., with Pierre etre dehors 'Pierre be outside', temporally presupposed by SI. 1 2 The situation described in (iii) above is illustrated by (18). It is common knowledge that a present tense question like qu'est-ce que c'est que ςα? 'what is that?' can be used to refer to a fact that has in fact already stopped at the moment one actually utters the question. This fact can be a noise, as in S2 of (18). This sentence is used to "describe" the noise, which implies that it expresses global simultaneity with respect to it. So it can be argued that in (18) too, the imparfait has the temporal value described in (13) (cf. also Tasmowski-De Ryck 1985). The analyses proposed for (15) — (18) make it easy now to understand why it is not possible to use the imparfait in the last sentence of (19), (20) and (22). (For (21), see below). Roulait 'was driving', in (19), is unacceptable because (19) does not provide an explicitly or implicitly mentioned fact or moment of time with which the imparfait can establish the relation " Ξ " . Similar remarks can be made about (20) and (22). For instance, substituting bientot 'soon' in (22) for comme 'just as' in (17) eliminates the possibility of temporally relating S2 to the subsequent sequence, so that there exists no fact any more with which S2 can establish a relationship of simultaneity. One of the functions of bientöt 'soon' is to express posteriority with respect to an already mentioned fact or moment of time. As for (21), the situation slightly differs from the ones discussed above. Despite the fact that SI of this example implicitly "reports" a fact (i. e., the one corresponding to il y avoir de la lumiere 'there be light') with which E2 globally coincides, donnait 'gave' (imparfait) is excluded. The reason probably is that the presence of et 'and' in S2 of (21) explicitly forces us to conceive of E2 as the immediate consequence of El (see also 3.2). In other

32

Arie Molendijk

words, we are forced to interpret S2 as oriented to le vieil homme allumer la lampe 'the old man light the lamp' rather than to il y avoir de la lumiere 'there be light'. This implies that S2 expresses posteriority (with respect to il allumer la lampe), not global simultaneity (with respect to la lampe etre allumie), which explains the unacceptability of donnaitP What has been said in this section and in 2.1-2.1.1 about the imparfait does not only account for its use in examples like (15) — (18) and (23). It also explains the use of the imparfait in cases that were shown to be problematic to the descriptions of the imparfait given in section 1. Sonnait 'struck' 0imparfait), for instance, in (6), is explained now by the fact that E2 of this example is globally simultaneous with El. Parlait 'talked' (imparfait), in (7), can be said to signal the existence of a relationship of global simultaneity between E2 and what is temporally implied by S1. Phenomena of the type discussed in this section will be treated in more detail in sections 3 and 4. But first something must be said about the passe simple.

2.2. The temporal function of the passe simple It has often been said that the passe simple expresses posteriority, which would explain the frequent use of this tense form in combination with adverbials like puis 'then = after that\finalement 'finally', bientöt 'soon', plus tard 'later on', and so forth. But it can easily be shown that the passe simple does not express posteriority only. Its temporal meaning is rather vague and can be better described in terms of what it does not express: the passe simple expresses what the imparfait cannot express (and, of course, what the other tense forms of French do not express). This seems tantamount to saying that the passe simple is used to express (i) posteriority, (ii) non-global simultaneity, or (iii) simple anteriority with respect to the moment of speech ("deictic" use of the passe simple, cf. Molendijk 1990: 185-188). In fact, this way of speaking about the temporal content of the passe simple is not correct, insofar as the passe simple itself does not have three distinct temporal values: it only has one "negative" temporal value. What makes that one of the possibilities mentioned above is realized in a given context is not, strictly speaking, the passe simple itself. Thus, what actually makes that S2 of (21), for instance, with donnait 'gave' (imparfait) replaced by donna (passe simple), situates E2 temporally after El is not the passe simple itself, but a complex of factors among which can be mentioned the property of this tense form of lending itself to be used in sentences expressing

Tense use and temporal orientation

33

certain kinds of temporal relationships (posteriority, for instance). So if in (21) (with donnait replaced by donna), the passe simple is said to express posteriority, this must be considered a way of speaking used for reasons of simplicity only. Having made these reservations, we can say that the passe simple expresses posteriority in (21) (with donna). As for the uses of the passä simple mentioned in (ii) and (iii) above (this subsection), they are illustrated by (7) — (8) (non-global simultaneity of the passe simple fact with respect to the fact of the second sentence) and by (4), (9) and (10) (deictic use of the passe simple: anteriority of the passe simple fact with respect to the moment of speech). The claim about the deictic use of the passe simple may demand some explanation. When we say that a fact Ε is presented as primarily and directly related to the moment of speech, we mean that considerations about the temporal relationship of Ε with respect to any fact E' that does not coincide with this moment are to be considered of secondary relevance. Were these considerations of primary relevance, Ε would indeed be directly oriented to E', i. e., Ε would not be directly oriented to the moment of speech. So fut 'was' (passi simple) in (9) is deictic because it presents the fact reported by the sentence containing this tense form as directly oriented to the moment of speech, not as primarily oriented to a past moment of time. II fut ... le plus grand footballeur ... 'he was the best football-player ...' does not mean something like "in that period, he was the best"—which would imply a temporal orientation to a certain period and for that reason would justify έίαίί (imparfait) instead of fut (passe simple)-but simply "he always was the best". For similar reasons, dura 'lasted' (passe simple) and presenta 'presented' (passe simple) are deictic, in (4) and in (10). Apparent exceptions to what has been said up to now about the passe simple can all be shown to be not exceptional at all, in the same way allegedly deviating uses of the imparfait were shown in 2.1.1-2.1.2 to conform to the rule. As these uses of the imparfait were treated in some detail, and as the apparently deviating uses of the passe simple can be dealt with in the way "dissident" imparfait behavior was explained, little needs to be said about uses of the passe simple that look like exceptions to the rule, but that actually are not. Consider, for instance, the following example: (24)

Jacqueline insulta I'imperatrice (SI). Puis eile fit une autre betise (S2). Elle refusa de faire la reverence (S3) [quand I'empereur entra] 'Jacqueline insulted the empress (SI). Then she made another big mistake (S2). She refused (PS) to make a curtsy (S3) [when the emperor entered (the room)]'

34

Arie

Molendijk

S3 of this example illustrates a phenomenon that is very similar to the one illustrated by the last (imparfait) sentence of (15). It was shown that the imparfait occurring in this sentence does not express anteriority, notwithstanding the anteriority of E3 of (15) with respect to E2. Similarly, in spite of the fact that E3 of (24) globally holds for E2 — an interpretation that forces itself upon us because S3 "explains" the content of S2 — the passe simple form itself occurring in S3 of (24) does not express global simultaneity. If this were indeed the case, then it would not be possible to move puis 'then' to the sentence containing refusa 'refused': (...) puis eile refusa de faire la rivärence: eile fit {done) une autre betise '(...) then she refused to make a curtsy, (so) she made another big mistake'. This seems to imply that not only S2 of (24), but also S3 of this example is in the scope of puis 'then'—which explains why refusait 'refused' (imparfait) cannot be used in (24), the ideas of posteriority (puis 'then') and global simultaneity being incompatible. It can be concluded from these facts that refusa 'refused' (passe simple) is used in (24) to express posteriority, not global simultaneity. There is still one thing to be noted about the passi simple. The temporal relationship established by this tense between facts does not necessarily concern explicitly mentioned ones (cf. what has been said about the imparfait). Thus, in an example like (25): (25)

Jean se mit en marche (SI). Quelqu'un I'appela (S2) 'Jean started walking (SI). Someone called (PS) him (S2)'

the temporal relationship established by the passe simple concerns E2 and Jean marcher 'Jean walk', temporally implied by SI (relationship of nonglobal simultaneity) rather than E2 and El (posteriority). This analysis is in agreement with the fact that E2 of (25) is quite naturally understood as occurring at a moment belonging to the time span occupied by Jean marcher 'Jean walk'. An analysis according to which E2 is just posterior to El would not account for this fact.

3. Temporal orientation We have seen that the proper use of the passe simple and the imparfait presupposes knowledge of certain principles related to temporal orientation. Some of these principles were indirectly and rather loosely treated in the preceding sections. In this section, a more direct treatment of certain phenom-

Tense use and temporal orientation

35

ena related to temporal orientation will be attempted. Their relevance for tense use in French (passe simple vs. imparfait) will be demonstrated. Among the principles that can be described in some detail, the ones given in (26) (below) can be mentioned. Strictly speaking, a sentence does not have another sentence as its orientation point, but a fact or moment of time referred to by it. For the sake of simplicity, this distinction will not always be made here. At first sight, the rules of (26) may look rather enigmatic. They will be illustrated in 3.1 - 3 . 3 . (26) (i) Given an independent sentence S with non-temporal presupposition Spresup nt , and a sentence S' preceding S with temporal implication S'impl t and/or temporal presupposition S'presup t : (a) If the (semantic) subject SUB J of Spresup nt corresponds to a definite description in S, then, if Spresup nt = S'impl t , there is a strong tendency to temporally relate S to S'impl t , i. e., S preferably takes S'impl t as its orientation point (' = ': identity). (b) if Spresup nt = S'presup t , then S'presup t serves as the orientation point for S. (c) If the (semantic) subject of Spresup nt corresponds to an indefinite description in S then, if the other condition mentioned in (a) still applies, there is a strong tendency to temporally relate S directly to S' i. e., S preferably takes S' as its orientation point. (d) If S = S'impl t then either S' or S'impl t serves as orientation point for S. But if S is modified in such a way that S is not (entirely) identical anymore with S'impl t , then there is a certain tendency to relate S to S'impl,, i. e., S preferably takes S'impl t as its orientation point. If S is modified in such a way that S is the negation of S'impl t , then S invariably has S'impl t as its orientation point. (ii) Given a complex sentence: (a) If a subordinate clause of the complex sentence is linked to the main clause by a temporal connective, then they do not both have their respective orientation points outside the complex sentence. One of them has its orientation point inside the complex sentence. (b) A member of the complex sentence that has its orientation point inside the (complex) sentence is not temporally oriented to a temporal implication or a temporal presupposition, i. e., it takes a fact explicitly mentioned in the complex sentence as its orientation point.

36

Arie Molendijk

These regularities seem to have a rather universal character (for certain language- specific rules, see 4). The question whether or not they can be explained by more general principles of interpretation will not concern us here. Let us first consider (26,i,a) and (26,i,b).

3.1. Temporal orientation to (facts corresponding to) implications or presuppositions (26,i,a) can be illustrated by an (English) example like (1), repeated here as (27): (27)

The old man lit the lamp (SI). The feeble light gave the room an appearance of sadness (S2)

S2 of (27) presupposes there be light (in the room). The semantic subject of this (non-temporal) presupposition corresponds to a definite description in S2 (the feeble light), and the presupposition itself is identical with what is temporally implied by SI. So the situation described in (26,i,a) is realized. This implies that S2 of (27) is supposed to have there be light as its orientation point rather then the old man light the lamp. In other words, S2 of (27) is supposed to express simultaneity with respect to there be light — because it is pragmatically impossible for a sentence like S2 of (27) not to establish this kind of relationship with its orientation point — rather than posteriority with respect to the old man light the lamp. This result may surprise those who would be inclined to conclude from E2's posteriority with respect to El that S2 exclusively expresses posteriority. It can be argued, however, that S2 of (27) is more directly connected with what is temporally implied by SI than to SI itself, the intended reading of (27) being closer to (28) than to (29): (28)

The old man lit the lamp (SI). The light that illuminated the room now (S2) was feeble and gave it an appearance of sadness (S3).

(29)

IThe old man lit the lamp (SI), as a consequence of which the feeble light gave the room an appearance of sadness (S2)

If S2 of (27) were directly related to SI (direct temporal relationship), then, considering that S2 speaks in this particular case of a certain result attached

Tense use and temporal orientation

37

to El, it should be possible to paraphrase (27) as in (29). But the definite article in the feeble light prevents (29) from being a correct paraphrase of (27). In fact, (29) is odd, precisely because of the definite article. Some would feel indeed that it forces us to interpret (29) as referring to a situation in which an already existing light starts to give a certain appearance to a given room. For that reason, they would prefer a feeble light to the feeble light, in (29). It will be clear now why many speakers of French prefer donnait 'gave' (,imparfait) to donna (passe simple), in the French equivalent of (27): S2 of this example expresses global simultaneity with respect to there be light (i. e., with respect to its orientation point) rather than posteriority with respect to the old man light the lamp. The validity of what has been said about the use of the imparfait in S2 of the French equivalent of (27) is not affected by the fact that transforma 'transformed' (passe simple), not transformait (imparfait), would be used in the French equivalent of (30): (30)

The old man lit the lamp (SI). The feeble light transformed the room into a place full of sadness (S2)

What distinguishes (30) from (27) is that a fact like the one reported by S2 of (30) is pragmatically unlikely to realize during the whole interval occupied by there be light, implied by SI, whereas this possibility is the one that is very likely to realize in the case of the second sentence event of (27). It should be noted here that E2 of (30) is punctual, whereas E2 of (27) is not. 14 Apart from that, (30) and (27) are not essentially different. In both (27) and (30), S2 is supposed to take there be light as its orientation point (cf. (26,i,a)). The use of transforma (passe simple), in the French equivalent of (30), is explained by the fact that the relationship of simultaneity existing between E2 of (30) and what is temporally implied by SI is not of a global nature (cf. what has been said about the passe simple). We have seen that (31) (below) mainly differs from examples such as (27) in that S2 of (27) has as its orientation point a (fact corresponding to a) temporal implication, whereas the orientation point of S2 of (31) corresponds to a temporal presupposition (cf. what has been said about (23)): (31)

Paul went into the house (SI): the sun was burning his shoulders (S2)

This is predicted by (26,i,b). S2 of (31) has he be outside as its orientation point, for he be outside is a (non-temporal) presupposition of S2 as well as

38

Arie Molendijk

a temporal presupposition of SI. The fact that S2 is supposed to be temporally oriented to he be outside explains why this sentence would be rendered in French by a sentence in the imparfait: the burning of the shoulders by the sun globally holds for the interval during which Paul is outside. We have seen that (27) and (30) are characterized by the fact that the semantic subject of a non-temporal presupposition attached to a sentence S corresponds to a definite description in S. The situation that arises when the semantic subject of a non-temporal presupposition of S corresponds to a nondefinite description in S is different. This will be the topic of the next section.

3.2. Temporal orientation to facts that are explicitly mentioned It was argued that S2 of (27) preferably takes the implication attached to SI as its orientation point. A slight modification of (27) suffices in order to change this situation: (32)

The old man lit the lamp (SI). A feeble light gave the room an appearance of sadness (S2)

It follows from (26,i,c) that in (32) S2 is temporally oriented to the lighting of the lamp rather than to what is implied by SI. This means that S2 is supposed to express posteriority rather than global simultaneity (cf. what has been said about (27)), because E2 of (32) is posterior to El, i. e., to its orientation point. This corresponds with the fact that S2 of (32) is not about an existing light (as in (27)) but about a "new one", because of the indefinite article. In other words, rather than describing an already existing situation, S2 introduces a (chronologically) new fact. It refers to a change, to a transition. This explains why many speakers feel inclined to insert and between SI and S2 of (32).15 For those speakers, the use of a feeble light instead of the feeble light is only justifiable if and is inserted, or at least "implicitly present". This also explains why (32) can be paraphrased as in (33): (33)

The old man lit the lamp (SI), as a consequence of which a feeble light gave the room an appearance of sadness (S2) (cf. (29), which is unacceptable).

What has been said about (32) is clearly in accordance with the fact that many speakers of French would render S2 of (32) by a sentence in the passe

Tense use and temporal orientation

39

simple, because this tense form lends itself to be used in sentences expressing posteriority. In spite of all this, a reading according to which S2 of this example expresses global simultaneity (with respect to there be light) must not be entirely excluded (cf. (26,i,c)). It should be noted indeed that it is possible to add "static now" to S2 of (32). So S2 could be interpreted as containing a "hidden" now referring to the "result" of S1: now a feeble light gave the room an appearance of sadness ( = 'now that he had lit the lamp', 'now that there was light'). This is in accordance with the fact that there are speakers of French who do not reject donnait 'gave' (imparfait) in the French equivalent of (32). This section and the preceding one were about sentences having non-temporal presuppositions that were identical with certain temporal implications or presuppositions attached to preceding sentences. In the next section, it will be shown that what has been said about the orientation point of these sentences does not entirely apply to examples in which a sentence itself is (quasi-) identical with certain temporal implications.

3.3. Sentences that are (quasi-) identical with temporal implications The cases that will be examined here are illustrations of the rule given in (26,i,d). Consider (34): (34)

He said yes (before the altar) (SI). He was a married man (S2)

In this example it is S2 itself, not what it presupposes, that is identical with what is temporally implied by SI. S2 is indeed (trivially) identical with he be (a) married (man), which is temporally implied by SI. In other words, (34) is not of the same type as (27) and (32). According to (26,i,d), S2 of (34) admits a reading implying that S2 is directly oriented to he say yes (relationship of posteriority). This possibility explains why as a consequence of which can be inserted between the two sentences of this example: (35)

He said yes (before the altar) (SI), as a consequence of which he was a married man (S2)

But (26,i,d) also states that S2 can be read as oriented to what is temporally implied by SI (relationship of simultaneity) rather than to he say yes. In other

40

Arie

Molendijk

words, (26,i,d) also predicts an interpretation according to which S2 of (34) explicitly "repeats" what is already temporally implied by SI. This interpretation is intuitively correct: S2 may indeed be read as static, as containing a "hidden" now. now he was a married man, as globally relating the fact of being married to an "already existing" moment of time. This moment, then, would be the one for which the temporal implication of SI is true. Note that the notions of "statism" and simultaneity are intimately related. A static reading of S2 of (34) would be perfectly normal, for instance, in a context which presents E2 as the thoughts of the protagonist. If S2 of (34) is replaced by a sentence adding details to what is temporally implied by SI, as in (36) (below), then a static reading of the sentence that is substituted for S2 of (34) tends to be the more "natural" one. In other words, the sentence preferably takes the temporal implication of SI as its orientation point: (36)

He said yes (before the altar) (SI). He was married to a beautiful girl with twinkling eyes and long legs (S2).

This result should not surprise us. As the details mentioned in (36) {to a beautiful girl with...) are attached to the "inevitable consequence" of SI, they are globally simultaneous with it. So S2 of (36) is quite naturally interpreted as static. What has been said about (34) and (36) provides us with an explanation for the fact that in the case of the French equivalent of an example like (34) both fut 'was' (passe simple) and etait (imparfait) can be used, έίαϊί being the more "normal" tense form in the case of (36). Note that in the French equivalent of an example like (37) (below), fut would be positively excluded: (37)

His friends all left him (SI), but he was not alone (S2)

This is predicted by (26,i,d). According to this part of (26), S2 of (37) cannot take SI as its orientation point, because S2 is the negation of what is temporally implied by SI. Put differently, it follows from (26,i,d) that S2 of (37) expresses global simultaneity, because it takes the temporal implication attached to SI as its orientation point and denies its truth. That S2 is not supposed to express posteriority is intuitively correct; it is in accordance with the fact that it is impossible to insert an expression like as a consequence of which between SI and S2 of (37).

Tense use and temporal orientation

41

4. Temporal orientation and complex sentences Which orientation point serves as temporal reference for a given sentence (clause) S partly depends on its syntactic context. The question whether or not S is an independent sentence or a member of a complex sentence plays an important role indeed in determining its orientation point. Most interesting regularities can be observed here, some of which have been mentioned in (26,ii). According to (26,ii,a), SI and S2 of (38) (below) do not both have their orientation point outside the complex sentence: (38)

When he finally drew the circle (SI), it was an odd one (S2)

Combining this fact with the content of (26,ii,b) yields the following prediction: either SI takes as its orientation point (the fact explicitly reported by) S2 (i), or S2 takes as its orientation point (the fact explicitly reported by) SI (ii). Now, it can be shown that (i) does not apply to cases like (38) (see below, this section). So S2 of (38) takes as its orientation point (the fact explicitly reported by) S1. What is important here is that S2 of (38) is not supposed to orient to there be a {specific) circle, temporally implied by S1. This is an interesting point, because according to (26,i,a), S2 would be oriented to this temporal implication if when were suppressed, as in (39): (39)

He finally drew the circle (SI). It was an odd one (S2)

In other words, S2 of (39) is supposed to express global simultaneity—because its orientation point is there be a (specific) circle—whereas S2 of (38) is not supposed to express this temporal idea—because its orientation point is he draw the circle.16 According to what has been said about (38) and (39), our intuitions should "tell" us to interpret S2 of (39) as purely static, whereas S2 of (38) should be interpreted as more or less "dynamic": when he finally drew the circle, it turned out to be an odd one. In other words, our intuitions should tell us that S2 of (38) is about "something new", whereas S2 of (39) is about something that already exists. It is perhaps not obvious that this is indeed what seems to distinguish (38) and (39). The following observations may be relevant here. That the presence of when in a complex sentence such as (38) forces us to interpret the main

42

Arie

Molendijk

clause as talking about something new is in accordance with the following. 17 If when is added to SI of an example like (1) above, the definite article in the subject NP of S2 of this example has to be replaced by the indefinite article: (40)

When the old man lit the lamp (SI), a feeble light gave the room an appearance of sadness (S2)

In other words, adding when to SI of (1) forces us to conceive of the light as a "new" one (because otherwise the definite article should be possible). In the same way, the presence of when in (38) forces us to interpret S2 of this example as referring to something new. As we have seen, this can be (indirectly) accounted for by the rule given in (26,ii,b). We will see in a moment that all this is crucial with respect to the question which tenses should be used in the French equivalents of (38) and (39). (26,ii,a — b) does not contain information about what exactly is the sentence (fact) that functions as orientation point for the other one in complex sentences like (38). So the fact that in (38), S2 is oriented to SI, not the other way round, is not predicted by (26,ii,a — b). It seems difficult to give general rules with respect to this question. Something can be said about it, though. It does not seem hazardous to suppose that the orientation point of a given member of a complex sentence functions as its temporal theme. As Ducrot (1979) points out, if it is possible to relate a sentence containing a temporal expression X to a question also containing X, then X functions as temporal theme of the sentence. If this possibility does not present itself, X is not the temporal theme of the sentence. So in (41): (41)

Last year I bought a car

the adverbial functions as the temporal theme of the sentence, because (41) can be considered an answer to (42): (42)

What did you do last year?

But in (43) (with the adverbial not separated from the rest of the sentence by a comma indicating a pause): (43)

I bought the car last year

Tense use and temporal orientation

43

last year does not function as the temporal theme, because (43) is not an appropriate answer to a question like (44), but to (45), for instance: (44) (45)

What did you do last year? When did you buy the carl

Note also that the temporal theme of a sentence is often (but not always) at the beginning of that sentence (cf. (41) and (43); see Ducrot 1979). According to what has been said about temporal themes and about their orientation point function, and assuming that what applies to non-sentential adverbials also applies to sentential ones (as far as theme assignment is concerned), we can say that in (38) above the when-clause functions as the orientation point for the main clause, not the other way round. The w/zefl-clause is indeed the temporal theme of (38), as shown by (46): 18 (46)

What was the circle like (SI) when he finally drew it (S2)? When he finally drew the circle (S3), it was an odd one (S4)

For similar reasons, S2 of (47) functions as the orientation point for SI, cf. (48). (47) (48)

When he reached the woods (SI), he heard a strange noise (S2) What happened (SI) when he reached the woods (S2)? When he reached the woods (S3), he heard a strange noise (S4)

But in (49) (below), the clause introduced by just as does not function as the orientation point of the main clause, because (50) is not quite natural: (49) (50)

Just as he reached the woods (SI), he heard a strange noise (S2) ?What happened (SI) just as he reached the woods (S2)? Just as he reached the woods (S3), he heard a strange noise (S4)

Rather than constituting an appropriate answer to SI - S2 of (50), (49) is related to a question like (51): (51)

What happened then?

The analyses proposed for (38), (39), (47) and (49) provide some interesting facts about the use of the passe simple and the imparfait in the French

44

Arie Molendijk

equivalents of these examples. We have an explanation for the fact that S2 of (38) would be rendered by a sentence containing a passä simple, whereas S2 of (39) would be translated into French by a sentence in the imparfait.19 We have seen indeed that (the fact explicitly reported by) SI of (38) functions as the orientation point for S2. As E2 does not globally coincide with its orientation point, i. e., with E l , S2 has to be rendered in French by a sentence in the passe simple. But S2 of (39) would contain έίαίί 'was' (imparfait), because according to (26,i,a), S2 of (39) has there be a {specific) circle as its orientation point, the relationship involved being a relationship of global simultaneity. As for (47), this case is comparable to (38), which explains entendit 'heard' (passe simple). Note that E2 of (47) does not globally coincide with El, the most natural reading being that E2 is immediately posterior to El (or perhaps contained in El). Finally, what has been said about (49) explains that SI of this example would be rendered in French by a sentence in the imparfait. As this sentence does not function as the orientation point for S2, the latter necessarily functions as the orientation point for the former (cf. (26,ii,a)). And as El globally coincides with E2, i. e., with its orientation point, SI must be translated by an imparfait sentence. 20 It should be observed that in the French equivalent of (47), SI would not be an imparfait sentence (cf. (49)). We have seen that SI of (47) functions as the orientation point for S2. In other words, SI does not take S2 as its orientation point. So it must take its orientation point in the part of the text preceding SI. Now, the event reported by a French non-habitual quand-sentence normally is not globally simultaneous with this orientation point. As a rule, this type of sentence expresses posteriority with respect to it. This is in accordance with the fact that the. event it reports is normally rendered in French by a sentence in the passe simple. We have seen that the syntactic context of a sentence (clause) may determine the "choice" of its orientation point. The rules given in (26,ii) illustrate this. Other rules could be given. It can be shown, for instance, that a subordinate clause S that is dependent on a simple past main clause S' containing a verb of perception invariably takes (the fact reported by) S' as its orientation point. Thus, S2 of (5) (section 1) takes (the fact reported by) SI as its orientation point. This explains why the subordinate clause in (French) examples like (5) never contains a passe simple form. 21 An elaborate treatment of this type of phenomena will be kept for another occasion.

Tense use and temporal orientation

45

5. Conclusion It has been argued that the difference between the imparfait and the passe simple of French can be described in simple temporal terms (global simultaneity vs. absence of global simultaneity), and that certain traditional and more recent descriptions of these tenses cannot be maintained. The difficulties that non-native speakers of French feel in using the passe simple and the imparfait properly were shown not to be caused by the allegedly "complex" semantics of these tenses. It was demonstrated that the source of the problem is intimately related to the notion of temporal orientation. In order to know which tense should be used in a given sentence, it is necessary to identify exactly the fact or moment of time that serves as the orientation point for this sentence. From a theoretical point of view, this means that the rules describing the temporal functions of the passe simple and the imparfait are required to interact with general rules describing regularities in the field of temporal orientation. Some of these regularities were discussed in detail.

Notes 1. I would like to thank Georges Kleiber, Rita Landeweerd, Henriette de Swart and Co Vet for useful comments on an earlier version of this paper. 2. "Fact" is used as a general term referring to both events and states. 3. SI: first sentence; S2: second sentence; E l : fact reported by SI; E2: fact reported by S2, etc. 4. It is assumed here that the passe simple and the imparfait differ in the way described above. 5. The notions of foregrounding, backgrounding, perfectivity, imperfectivity, punctuality, durativity, narration and description are often used by school grammars of French to explain the difference between the passe simple and the imparfait. The opposition non-anaphoric vs. anaphoric is used, among other linguists, by Vet (1985) and Vetters (1989). As for the notions of distantiation and dimensionalisation, see Waugh—Monville-Burston (1985). 6. The relevant French tense forms are in roman letters; PS: passe simple·, IMP: imparfait. 7. Note also that (6) does not allow sonna 'struck' (passi simple). 8. Automobilists are not normally handed police tickets while driving, but afterwards; for the sake of simplicity, I will not consider the situation in which automobilists are fined "by radar". 9. (19) and (20) would be perfectly acceptable with, respectively, the plus-que-parfait avait route 'had driven' and the passe compose a ferme 'has closed'.

46

Arie Molendijk

10. This implies that the sentence in the imparfait is oriented to one of the facts given in (i) - (iii). 11. Here, the syntactic dependency of S2 of (17) with respect to S3 has as its semantic counterpart the semantic (temporal) dependency of E2 with respect to E3: E2 is presented as temporally related to E3 (relationship of global simultaneity) rather than as temporally related to El (relationship of posteriority). 12. This reading is pragmatically the most natural one. Temporal presuppositions differ from temporal implications in that the former type refers to (not explicitly mentioned) facts that are anterior to explicitly mentioned ones, whereas temporal implications correspond to facts that are posterior to explicitly mentioned ones (cf. Molendijk 1990: 84—91). Thus, Pierre etre dehors 'Pierre be outside', temporally presupposed by SI of (23), is anterior to the fact that is explicitly reported by this sentence. 13. The fact that S2 of (21) has an inchoative reading is perhaps a consequence of what has been said about this example. 14. This explains the unnatural character of transformer + durative adverbial. 15. And is often used to indicate that there is a transition. 16. Note that it be an odd circle and he draw the circle are not globally simultaneous. 17. I will only consider here cases in which the w/zen-clause precedes the main clause. 18. Note the repetition of when he finally drew the circle. 19. I assume that French quand 'when' and English when are similar, semantically. 20. This sentence would be introduced by (temporal) comme, whose semantic value is comparable to that of English just as. 21. The event reported by a simple past subordinate clause S that is dependent on a simple past main clause S' containing a verb of perception necessarily coincides (globally) with the main clause event.

References Ducrot, Oswald 1979 "L'imparfait en franpais", Linguistische Berichte 60: 1—23. Kamp, Hans 1981 "Evenements, representations discursives et reference temporelle", Langages 64: 39—64. Kamp, Hans—Christian Rohrer 1983 "Tense in texts", in: Rainer Bäuerle-Christoph Schwarze—Arnim von Stechow (eds.), Meaning, Use and Interpretation of Language. Berlin: Mouton de Gruyter, 250-269. Kleiber, Georges 1993 "Lorsque l'anaphore se lie aux temps verbaux", in: Carl Vetters (ed.), 117-166. Molendijk, Arie 1990 Lepasse simple et l'imparfait: une approche reichenbachienne. Amsterdam-Atlanta: Rodopi. 1993 "Presuppositions, implications, structure temporelle", in: Carl Vetters (ed.), 167-192.

Tense use and temporal orientation

47

Smith, Carlota 1978 "The syntax and interpretation of temporal expressions in English", Linguistics and Philosophy 2: 43 — 99. Tasmowski-De Ryck, Liliane 1985 "L'imparfait avec et sans rupture", Langue Franqaise 67: 5 9 - 7 7 . Vet, Co 1985 "Univers de discours et univers d'enonciation: les temps du passe et du fiitur", Langue Frangaise 67: 38—58. Vetters, Carl 1989 "Grammaire generative et textuelle des temps verbaux", Recherches linguistiques de Vincennes 18: 101 — 145. Vetters, Carl (ed.) 1993 Le temps: de la phrase au texte. Lille: Presses Universitaires de Lille. Waugh, Linda-Monique Monville-Burston 1985 "Aspect and discourse function: the French simple past in newspaper usage", Language 62: 846—877.

Future tense and discourse representation Co Vet

Introduction In the last decade Discourse Representation Theory, developed by Kamp (1981a, 1981b), has given a new impetus to the study of tense, especially of past tense in narrative discourse. This article will focus on the meaning and discourse functions of the two future tenses of French, the Simple Future (SF) and the Periphrastic Future (PF). For example: (1)

Fido mangera le bifteck. (SF) 'Fido will eat the steak.'

(2)

Fido va manger le bifteck. (PF) Fido goes eat the steak. 'Fido is going to eat the steak.'

In the literature the opinions diverge as to the meaning and use of these forms. The following, partially incompatible views have been defended (see Fleischman [1982: 18-19]; Lorenz [1990: 7 - 8 ] for more detailed overviews): a. the Simple Future and the Periphrastic Future are mere stylistic variants; b. the two futures express a semantic difference: the Periphrastic Future is used to refer to eventualities in the near future, the Simple Future to denote eventualities in the remote future. c. The Periphrastic Future is not a future tense but the Present of the verb aller 'go' followed by the infinitive. d. The Simple Future of French is not a tense, but a mood. It will be shown that most of these points are not entirely wrong; however, a different analysis will be put forward, according to which the future tenses of French (including the Futurate Present) have in common that (part of) their

50

Co Vet

meaning should be seen as a propositional attitude. The meaning difference resides exclusively in the type of evidence which the speaker has at his/her disposal at the speech time. The second aim of this paper is to formulate rules capable of predicting the temporal relations in fragments consisting of sentences in the Simple Future and/or Periphrastic Future. The structure of this paper is as follows. In section 1 I will give a brief outline of Discourse Representation Theory. In section 2 it will be shown that point (a) cannot be maintained because the Simple Future cannot replace the Periphrastic Future (or vice versa) in all contexts. In section 3 the notion "association" will be used to account for the complex meaning of the Periphrastic Future. In section 4 I will discuss point (d) (the Simple Future is a mood). Section 5 will be devoted to the Futurate Present. In section 6 it will be shown that the attitudes I postulated as (part of) the meaning of Periphrastic Future, Simple Future and the Futurate Present are capable of explaining the different meanings of the past forms of these tenses as well and that they follow the same pattern as attitudinal verbs like croire 'believe'. Finally I will examine, in the last section, whether and under what conditions a Simple Future or a Periphrastic Future sentence introduces an event or a state in the Discourse Representation. This is important for the interpretation of the temporal relations between the eventualities a discourse is about.

1. Discourse Representation Theory Discourse Representation Theory was developed by Kamp (1981a, 1981b) in order to deal with nominal anaphora (pronouns) in discourse and with the relations between the eventualities introduced into the discourse domain by the successive sentences of a text. In both cases the main idea is that sentences containing a nominal or a temporal anaphoric expression depend for their interpretation on elements (discourse referents) previously introduced into the Discourse Representation. Kamp (1981a: 40) gives for instance the following example of the way pronouns can be dealt with: (3)

Discourse: Pedro possede un äne. II le bat. 'Pedro has a donkey. He beats it.'

The Discourse Representation of Figure 1 can be thought of as the mental image the hearer/reader of the discourse (3) constructs when he/she interprets

Future tense and discourse representation

51

u ν [Pedro possede un äne] u - Pedro äne (ν) u possede ν [// le bat] il le bat u le bat u bat ν Figure 1. Discourse representation of (3)

it. Discourse Representations consist of a list of discourse referents (u and ν in our example) and a list of conditions on these discourse referents. The pronouns il 'he' and le (here 'it') are replaced by discourse referents which are already present in the Discourse Representation. The discourse (3) is true and the Discourse Representation of Figure 1 is correct if Figure 1 can be extended into a model M, that is if there is a function which assigns specific members of the domain of Μ to u and ν and if the atomic formulas (u possede ν and u bat v) are true in Μ. I will not go into this matter here (see Kamp (1981a, 1981b) for further details on the modeltheoretic interpretation of Discourse Representations). In the Discourse Representation of Figure 1 the discourse markers u and ν stand for the individuals the discourse of (3) is about. Discourse Representations that account for the temporal relations between the eventualities of a discourse need discourse referents for eventualities. Kamp distinguishes two kinds of eventuality: events (e) and states (s). According to Kamp, in French narrative discourse states are introduced into the Discourse Representation by sentences in the imparfait (Imperfect Past, IP) and events by sentences in the passe simple (Simple Past, SP). States behave as anaphoric elements in that they have to be related to a temporal antecedent previously introduced into the Discourse Representation. As an example Kamp (1981a: 47) gives a Discourse Representation for the following discourse: (4)

Pedro entra (SP) dans la cuisine. Marie faisait (IP) la vaiselle. II passa (SP) au salon et alluma (SP) sa pipe. 'Pedro entered the kitchen. Marie was washing the dishes. He went to the drawing-room and lit his pipe.'

52

Co Vet

η ei u ν Sj w e2 x e3 y [Pedro entra dans la cuisine] u = Pedro ν = la cuisine ei: u entrer-dans ν

e! e

1


e

2


e

3

Figure 18. Temporal interpretation of (34)

It may be concluded that, from an aspectual point of view the Simple Future is neutral, that is, it does not change a state into an event (as does the Simple Past in French) or an event into a state (as does the Imperfective Past in French). The event-like or state-like character of the eventualities is entirely determined by the Aktionsart of the sentence. The Discourse Representation construction rules for Simple Future sentences can be formulated as follows: (41) a. If a sentence S, has the structure Simple Future[predication -TR] (i. e., if the predication has non-transitional Aktionsart)·. (i) introduce the discourse markers pj, i (and the discourse markers for the individuals involved in the predication) and the conditions POS (i, pj) and pj: K'(where K' is a Discourse Representation); (ii) Introduce into K' the discourse marker s k and the conditions s k : predication, and s k Ο Ε (where Ε is an antecedent previously introduced into the Discourse Representation and n