9h-9h30 Welcome and coffee (hall)

9h30-9h45 Presentation of the European Science Foundation, by Ms Mariane Yagoubi (ESF)

9h45-10h Opening of  the conference, by Marc Jeannerod, director of ISC

10h-10h45 Henk Verkuyl (University of Utrecht, NL)
 Last and cost ; for and in

10h45-11h15 Coffee Break

11h15-12h30  Public lecture
 Renaat Declerck (University of Leuven, B)
A model of the English tense system (hand- out) summary

12h30-14h Lunch (ISC)

14h-14h45 Ilse Depraetere (K.U. Brussel, B)
The use of tense in relative clauses, and a note about the interaction between (a)telicity, multiple numerical NPs, (un)boundedness and the progressive

14h45-15h30 Ronny Boogaart (Free University, Amsterdam, NL)
 Aspect and temporal ordering : A contrastive analyse of Dutch and English

15h30-16h Coffee Break

16h-16h45 Jacques Moeschler (University of Geneva, CH)
 How to infer time direction in discourse ?

16h45-17h30 Myriam Bras (IRIT, University Paul Sabatier, Toulouse, F) et Anne le Draoulec (University of Toulouse 2, F)
 French adverbial puis between temporal structure and discourse structure

17h30-18h Discussion of the day contributions, chaired by Sheila Glasbey (University of Dundee, Scotland)

Thursday 24 february 2000

9h15-10h Anastasia Giannakidou (University of Groningen, NL)
 Temporal connectives in a crosslinguistic perspective : Greek “mexri” and English “until"   abstract”

10h-10h45 Ian Roberts (University of Stuttgart, D)
 The history of the future

10h45-11h15 Coffee Break

11h15-12h30  Public lecture
Alessandra Giorgi (University of Venice, I)
 On the morphosyntax of sequence of tense in Italian   abstract

12h30-14h Lunch

14h-14h45 Laurence Danlos (University of Paris 7, INRIA, F)
 On event coreference

14h45-15h30 Hans Kamp (University of Stuttgart, D)
An Algorithm for the representation of tense and aspect in German sentences

15h30-16h Coffee Break

16h-16h45 Henriette de Swart (University of Utrecht, NL), Arie Molendijk (University of Groningen, NL)
 The Passé Composé in French : aspectual and discourse values

16h45-17h30 Louis de Saussure (University of Geneva, CH)
 Modeling time in procedural pragmatics

17h30-18h Discussion of the day contributions, chaired by Sheila Glasbey (University of Dundee, Scotland)

Friday 25 february 2000

9h15-10h Juergen Bohnemeyer (Max Planck Institute for Psycholinguistics, Nijmegen, NL)
 Inferring from what ? The impact of grammaticalization of aspectual operators on the derivation of temporal inferences in discourse

10h-10h45 Alice ter Meulen (University of Groningen, NL)
 Reasoning with Dynamic Aspect Trees   abstract

10h45-11h15 Coffee Break

11h15-12h30  Public lecture
 Mark Steedman (University of Edinburgh, Scotland)
 Temporal semantics and temporal reasoning
postscript full text

12h30-14h Lunch

14h-14h45 Hans Smessaert (University of Leuven, B)
 Temporal reasoning with aspectual adverbs

14h45-15h30 Round Table I : discussion of Bohnemeyer’s video (J. Bohnemeyer, A.ter Meulen, M. Steedman, R. Boogaart)

15h30-16h Coffee Break

16h-18h Round Table II : Temporal reasoning,linguistic variation and cognitive structure (A. ter Meulen, S. Glasbey, V. Déprez, A. Reboul, H. Kamp, J. Moeschler, M. Steedman)

last and cost; for and in
Henk Verkuyl

In this talk I will first discuss some differences between the Dutch verbs "duren" (last) and "kosten" (cost). It seems as if the Dutch and the English verbs more or less show a similar behaviour, but given the subtleties of the differences between the verbs I will start with the Dutch data. The differences between the two verbs are reflected in the differences between the temporal adverbials "gedurende een uur" (for an hour) and "in een uur" (in an hour), the preposition "gedurende" being formally related to the participle "durende" (lasting). The verb "duren" turns out to pertain to a monotonous measure function, whereas "cost" is not monotonous. This transfers to the prepositional phrases, which means that there is a severe problem for Krifka's 1987-analysis of "in an hour", based on entailment. It turns out that "in an hour" should not be taken as `within an hour', as some people appear to think. Given the important status of the two PPs in the well-known aspectual tests, it seems important to establish a well-founded analysis of their meaning.

Given their aspectual relevance, some predictions can be made with respect to Slavic languages and Romance languages. And indeed, one may predict that the differences between the two verbs shows up in differences in interpretation given the presence or absence of aspectual prefixes in Slavic languages. I will compare the Dutch data with English, Polish and Spanish and possibly other languages (French?) in order to see what sort of measure functions are involved in the use of temporal adverbials. One of the obvious results of this comparison is that it can be proved now that a perfective prefix can have or undergo a long distance effect.

up

A model of the English tense system (hand out)
Renaat Declerck (University of Leuven (K.U.L.))

Summary:

In his lecture R. Declerck will present his model of the English tense system which he has developed over the last twelve years. This model not only deals with the use of tenses in independent clauses but also (and especially) with the use of tenses in subordinate clauses and in discourse. The concepts and hypotheses forming the building stones of the model will be systematically introduced and defined. It will be argued that the semantics of the various tenses in English are the temporal structures which they realize. These temporal structures will be visualized by the use of figures, with further figures added to illustrate the tense structures of complex sentences.
  Perhaps the most important (and, for some, most controversial) claim underlying the model is that English has relative as well as absolute tenses. Absolute tenses relate the time of the situation referred to directly to the 'zero-time' (which is usually the time of speech) and in doing so create a 'temporal domain' (i.e. a set of times related to each other by tenses) in one of the 'absolute sectors' of the time line, viz. the past, the pre-present, the present or the post-present. Relative tenses express a temporal relation — simultaneity, anteriority or posteriority — in such a temporal domain. In I felt that he was lying, there is an absolute past tense form (felt) and a relative past tense form (was lying) expressing simultaneity. The assumption that there are two different past tenses in English is not uncontroversial, but it is confirmed by a number of data. For example, the hypthesis that English has a relative past tense (representing the time of the situation referred to as simultaneous with some other time in a past domain) is the only way of explaining the use of had in Yesterday John promised he would do it as soon as he had time, probably next week. The situation of having time is interpreted here as following (rather than preceding) the zero-time. The past tense form had can therefore not be analysed as locating its situation before the zero-time, as felt does. Was feeling must therefore be treated as a relative tense form expressing simultaneity in the past domain established by the absolute tense form felt.

  Declerck will explain in detail which principles and rules determine the tense(s) to be used when the speaker incorporates the time of a new situation into a past, pre-present, present or post-present domain. He will also discuss various moot points of the literature on tense, such as: Is there a future tense in English? How exactly do we define 'the time of a situation', 'simultaneity', 'absolute-relative tenses', 'time of orientation', etc? Does English have a 'consecutio temporum' (sequence of tenses) rule? How come we have to use the present tense in I'll be there before you {are / *will be}? How can we explain the choice of tenses we observe in When they {first saw / had first seen} the house it had looked rather dreary, or in I talked to them in French because they {didn't / don't} understand English? How come we say If John comes tomorrow, there will be trouble rather than If John has come tomorrow, there will be trouble, although John's coming will actually precede (and cause) the trouble? And so on.

up

The Passé Composé in French: aspectual and discourse values
Arie Molendijk   &  Henriëtte de Swart
University of Groningen  Utrecht University
 

According to Vet (1980), the French tense system implies two axes. The first axis is set up around the deictic centre, the speech point S (R=S).  The three relations which an event E can have with respect to R give rise to definitions of  the Présent (simultaneity), the Passé Composé (anteriority) and the Futur (posteriority). The second axis is set up around a moment in the past S’ (R=S’ < S). We can use the three relations which an event E can have with respect to R to define a second set of three tenses, namely the Imparfait (simultaneity), the Pluperfect (anteriority) and the Futur du Passé (posteriority). Assuming with Vet (1980) and others that the Passé Simple is the perfective variant of the Imparfait, we observe that the Passé Composé and the Passé Simple are defined on two different axes. This is rather surprising in view of the observation that, in the spoken language at least, the Passé Composé is normally used instead of the Passé Simple. The three-way distinction we find in written language (Passé Composé-Passé Simple-Imparfait) is reduced to the two-way distinction between Passé Composé and Imparfait in spoken French. The question arises how we can reconcile the insights of Vet (1980) with the observations about the role of the Passé Composé in spoken French.
 We will approach this question from two angles. First, we will study a number of cases where the Passé Composé and the Passé Simple are *not* interchangeable. It is well known that the Passé Composé can be used in imperfective contexts, which exclude the Passé Simple, such as the construction depuis + Passé Composé (Martin 1971, Molendijk 1990). It has also been observed that the Passé Simple has difficulty combining with a deictic adverb such as hier, while the combination of a Passé Composé and a deictic adverb is perfectly acceptable (Vet 1980, Molendijk 1990, etc.). Finally, we observe that causal inversion is hardly possible in sequences of two affirmative sentences in the Passé Simple, whereas we can easily reverse the descriptive order of two events with a sequence Passé Composé + Passé Composé or a sequence of two negative sentences in the Passé Simple (compare Saussure 1996, Vet 1996, Molendijk & de Swart 1999). The combinatory differences between Passé Simple and Passé Composé lead us to posit interpretational differences between the two past tenses, which have repercussions for the discourse level.
 The second angle is the comparative study of perfect tenses in French, English and Dutch. Boogaard (1996) observes that in English and Dutch, we cannot use the present perfect to tell a story, but obviously, we preferably use the Passé Composé in spoken French to take over the narrative function of the Passé Simple. Presumably, these contrasts reflect different stages in the grammaticalization process which leads from resultative to perfective interpretations (Boogaard 1996). In this respect, it is interesting to observe that the use of the Passé Composé does not necessarily imply that the result state is relevant at the speech point. In a construction like Quand Pierre est rentré, il s’est couché tout de suite, there is no implication that at the present moment, Charles is at home/in bed, while the presence of such a result is the normal condition for the use of the present perfect, both in English and in Dutch.

Selected references

Boogaart, R. (1999). Aspect and temporal ordering. Amsterdam: VU (thesis).
Martin, R. (1971). Temps et aspect. Essai sur l’emploi de temps narratifs en moyen français, Paris.
Molendijk, A. & H. de Swart (1999). L’ordre discursif inverse en français, Travaux de Linguistique 39.
Molendijk, A. (1990). Le Passé Simple et l’Imparfait: une approche reichenbachienne, Amsterdam: Rodopi.
De Saussure, L. (1996). Encapsulation et référence temporelle d’énoncés négatifs au passé composé et au passé simple, Cahiers de linguistique française 18.
Vet, C. (1980). Temps, aspects et adverbes de temps en français contemporain. Essai de sémantique formelle, Genève: Droz.
Vet, C. (1996). Anaphore et deixis dans le domaine temporal. Anaphores temporelles et (in)cohérence, Cahiers Chronos I, Amsterdam: Rodopi.
 

up
 

A test case for Declerck (1991): the use of tense in relative clauses and a note about the interaction between (un)boundedness, (a)telicity, multiple numerical NPs and the progressive
Ilse Depraetere, K.U.Brussel
ilse.depraetere@pandora.be)

In the first part of this presentation, I would like to report on a study (Depraetere 1996) in which Declerck’s (1991) model of tense was applied to relative clauses. The question I set out to answer was whether the syntactic differences between a RRC and a NRRC are also reflected in different temporal systems. Relative clauses that indicate future time will serve as a starting-point for the discussion. On a more general level, I will discuss the merits and possible shortcomings of Declerck’s system.

Secondly, I would like to show that the constraints on the use of the progressive in combination with multiple numerical NPs can be explained if we take into account the role the NP has in establishing not only telicity but also boundedness (Declerck 1991, Depraetere 1995, 1996). This is particularly obvious in the case of the past progressive:
(1) a. He killed five chickens. (telic and bounded)
 b. Because we had so many guests, he was killing five chickens. (Depraetere & Reed, to appear)
(2) a. He killed five chickens. (atelic and bounded)
 b. *He lost control of the car and skidded across the farmyard. He was killing five chickens. (Depraetere & Reed, to appear)
I will argue that a multiple numerical NP does not necessarily have the effect of making a sentence telic or even bounded. If it refers to an inherent (i.c. intended) endpoint (i.e. if the non-progressive sentence is telic and bounded (cf. (1a)), the progressive marker can be added. If the NP only serves to bound the situation, the progressive cannot be used (cf. (2b)). If the NP does not represent the situation as bounded or telic, the progressive is again acceptable:
(3) Little Sarah was sipping two drinks. (at the same time) (unbounded, atelic)
If these observations are correct, they prove that ‘intention’ has to be taken into account when defining telicity.
The approach outlined does not appear to explain all possible uses of the indefinite present perfect progressive (PPP) in sentences with multiple numerical NPs. Although the examples in (4) and (5) refer to bounded situations, the progressive is all the same acceptable:
(4) I have been making three cakes. That’s why I’m covered in flour.
(5) What has she been doing? She has been performing three swallow-dives.
The sentences in (4) and (5) probably prove that the multiple numerical NP plays still another role in these contexts, an issue which I hope to discuss with the participants in this workshop.
 

Declerck, Renaat. 1991. Tense in English. Its structure and use in discourse. Routledge.
Depraetere, Ilse. 1995. On the necessity of distinguishing (un)boundedness and (a)telicity, Linguistics and Philosophy 18: 1-19.
Depraetere, Ilse. 1996. The tense system in English relative clauses. Mouton de Gruyter.
Depraetere, Ilse & Susan Reed. (to appear). The present perfect progressive: constraints on its use with numerical NPs, English Language and Linguistics.
 

up
 

Puis entre structure temporelle et structure discursive
French Adverbial Puis between Temporal Structure and Discourse Structure
Myriam Bras, Anne Le Draoulec, Laure Vieu
IRIT-ERSS, Toulouse

Abstract

In this paper, we'll tackle again an old issue in discourse studies, namely the interaction between temporal structure (i.e. the set of the eventualities described by a text and the temporal relations between them) and discourse structure (i.e.  the set of the discourse constituants or segments and the rhetorical relations between them). Many studies have focused on the role of tenses in this interaction. We propose to examine the lexical item puis in French in order to bring new elements to this discussion.

Previous work on puis have brought out the role of puis in discourse as a marker of (i) argumentation, (ii) enumeration as discourse structuration and (iii) temporal ordering.  In this work, we will be concerned by the temporal role of puis, that is (iii) and some cases of (ii) when puis is implied both in discourse and temporal structuration. We leave the study of Et puis apart.We will consider the role of puis in combination with tenses and also with discourse structure. We restrict the kind of text to be studied to narratives, so we will analyse texts conjugated with the famous Passé Simple / Imparfait pair. We will not only consider the cases when puis connects one sentence with the next one in the textual order but also the cases when puis links two discourse segments within a more complex discourse structure.

We will study the contribution of puis to the temporal structure and to the discourse structure. We will show that puis affects both structures but also that its interpretation varies in context. This analysis will be done within the framework of Segmented Discourse Representation Theory (SDRT) which has proved to be a good one to give a formal account of this kind of phenomena.

Considering that puis is not an adverbial of temporal location as it does not introduce any temporal referent but that it is rather a relation marker, we will have to determine whether this relation is a simple temporal relation of succession or if it has an effect on the discourse structure beyond this simple succession. SDRT describes puis as indicating a Narration relation. This fits with the usual observation that puis only links events, in our case, described by Passé Simple sentences, as in :
Et à midi ils s'arrêtèrent pour manger, puis ils se remirent en marche et nous les suivions toujours.
We will see however that puis can be involved in other schemes, in particular using Imparfait as in:
Le petit continuait de trembler, et il ne répondait pas aux questions de sa mère ; puis il éclata en sanglots.
We will also see that puis behaves in a special way as far as causality relations are concerned. Even if Narration and Result are two compatible discourse relations, puis is not compatible with Result, as in :
L'acide tomba dans le mélange. # Puis une explosion s'ensuivit.

We will take these facts into account in formalizing the semantics of puis in SDRT and its role at the semantics-pragmatics interface. This will lead us to refine the standard axiomatics of SDRT.
 

up
 

Temporal connectives from a crosslinguistic perspective: until and Greek mexri
Anastasia Giannakidou, KNAW/University of Groningen

Some temporal connectives are known to exhibit selective patterns regarding the aspect of their arguments. It has been observed, for instance, that connectives like ??until ??require their ? argument to denote a durative eventuality (a state or a process). Because the ??argument may also be a negative sentence (Karttunen’s 1974 negative polarity (NPI) until), it has been argued that negative sentences should be treated as durative too (see especially de Swart 1996). In this paper, I consider the Greek counterpart of ??until ???mexri, in constructions where the ??is non-clausal. Although like until, mexri combines only with statives and processes(1-3), unlike until, mexri is incompatible with negative sentences (4).

(1) I prigipisa  egrafe   ena grama mexri ta    mesanixta.  (process)
the princess  wrote.imperf.  a    letter   until    the midnight
‘The princess was writing a letter until midnight.’
(2) I prigipisa  kimotane   mexri ta    mesanixta.  (state)
the princess  slept.imperf.   until    the midnight
‘The princess was sleeping until midnight.’
(3) * I prigipisa  egrapse  ena grama mexri ta    mesanixta.  (event)
the princess  wrote.perf  a    letter   until    the midnight
‘The princess was writing a letter until midnight.’
(4) * I prigipisa  dhen kimithike  mexri ta    mesanixta.  (negative event)
the princess   not  slept.perf.  until    the midnight
‘The princess did not sleep until midnight.’
= It was only at midnight that the princess fell asleep.

The intended meaning of (4) is supplied with prin apo ta mesanixta ‘before midnight’ and para mono ta mesanixta ‘but only at midnight’. The latter is an NPI since para mono is unacceptable without negation. Note, additionally, that Greek marks durativity with imperfective aspect. Crucially, negative imperfectives are indeed compatible with mexri as we see in (5); but in this case we have only the reading in (5’) which is reminiscent of Mittwoch’s (1977) reading where negation takes scope over the whole proposition:

(5) I prigipisa  dhen kimotane   mexri ta    mesanixta.
the princess   not  slept.imperf.  until    the midnight
‘It is not true that the princess was asleep until midnight.
(5’) ? ?s (sleep (the princess, s) ? until-midnight (s))

(5) is neutral wrt when the princess woke up, and we cannot positively infer that the princess woke up earlier; continuations like ‘in fact she was still asleep after 2 am’ are possible, thus supporting Mittwoch’s proposal that such an inference is a conversational implicature.
The facts presented above have a number of interesting consequences. First, they seem to confirm Karttunen’s position that there is an NPI-use of the connective in (4), but they also show that the until-relation under negation is not an NPI crosslinguistically. Second, they seem to question the uniform analysis of negative eventualities and statives, and support Karttunen’s idea that the NPI-use of until is punctual. Third, they cast doubt on Karttunen’s thesis that under negation, until and before are equivalent, since unlike mexri the Greek counterpart of before, prin, is indeed compatible with negative ? arguments. Last but not least, they suggest that the NPI-analysis (Karttunen) and the wide scope negation analysis (Mittwoch) are not semantically and pragmatically equivalent, as argued in de Swart (1996). In Greek, mexri is not an NPI; but still mexri can be interpreted in a proposition inside the scope of negation as long as negation does not apply directly to it. This reading is clearly distinct from the NPI-reading of negation applying locally to mexri.
 

up
 

The History of the Future
Ian Roberts
University of Stuttgart

This paper compares the historical development of the neutral expression of futurity in two fairly well-known cases: the development of the future tenses in most of the Romance languages from the combination of infinitive + habere ("have") in Latin and the development of the English auxiliaries
will and shall. What we observe is that the two cases are strikingly similar. The comparison suggests natural pathways of syntactic and semantic change.
 English: the development the English future forms is a subcase of the well-known development of the English modal auxiliaries. Lightfoot (1979), Roberts (1985, 1993) and Warner (1993) have shown that  this change involved a categorial reanalysis of a subclass of verbs as I-elements, and that this change took place early in the 16th century. However, if we take into account a wider range of data, the picture becomes more complex.
First, Warner (1983, 1993:147) formulates the following generalisation for ME (his (3)): (1) Preterit-present verbs subcategorized for the plain infinitive which denote necessity, obligation and related notions of futurity are finite only. The generalisation covers the core deontic modals of ME (including shall). We account for Warner's generalisation in terms of a subpart of Cinque's (1999) proposals for  the functional structure of the clause (see Benincà & Poletto (1997)). The relevant substructure of Cinque's split-I system is the following: (2) ModEpistemic T(Past) T(Future) MoodIrrealis ModNecessity ModPossibility … ModRoot … Aspect VP Suppose that the relevant OE and ME modals were, in the relevant interpretations (i.e. as in (1)) able to be merged directly into the relevant functional heads (i.e. the modal field below T). From this position they moved higher, at least as far as T, just as all finite verbs did at these periods (see Roberts (1985)). Merging these modals directly rather "high" in the functional structure meant that they could not be licensed in non-finite forms, as this involves heads which are situated lower in the structure. Second, Warner shows that epistemic interpretations of premodals emerge in ME (see also Lightfoot (1979, Chapter One), Roberts (1985)). We can interpret this as a further reanalysis of (some) premodals as being merged in the MoodEpistemic position in (2). This idea has two consequences.  Third, there is evidence that certain modals may have been restructuring verbs in the sense of Rizzi (1982). Cinque (2000) proposes that restructuring verbs are functional heads. If so, then the evidence that premodals were restructuring verbs supports the proposal that they were merged directly into the functional system. Romance: the future and conditional tenses of most of the modern Romance languages originate in a periphrastic construction in Latin formed by an infinitive followed by habere "to have". The full lexical Latin verb habere was reanalysed as the future/conditional ending in the modern Romance languages in at least two stages (there is evidence for a third stage, but we will leave that aside here). First, habere was reanalyzed as a future auxiliary comparable to will or shall in Modern English. Second, the auxiliary habere, an autonomous word, was reanalyzed as a syntactic affix. The first change arguably took place in the 3rd century, according to Benveniste (1968) (see below). The second change may be a direct reflex of the first (Fleischmann (1982)).  The first change is directly comparable to the ME modals of "leakage" of verbs with certain interpretations into the functional system. In fact, the parallels are closer. Although habere had the full range of Latin tenses, voices and moods, only the present and imperfect (Gallo- and Ibero-Romance) or perfect (Italian) indicative active forms were reanalysed as futures and conditionals respectively. Thus we do not find future participles based on a Latin infinitive plus non-finite form of habere (and this despite the fact that Classical Latin had future participles, which, like the Latin future tenses, are entirely lost in Romance). The absence of future participles and the like in Romance suggests that only a relatively small number of finite forms of habere were reanalysed. In other words, reanalysed habere was morphologically defective in a way which is directly comparable to the ME premodals.  A further parallel with the ME premodals arises from observations made by Benveniste (1968). Benveniste clarifies a number of aspects of the developments in Late Latin. He points out that the periphrastic construction infinitive + habere originates early in the 3rd century AD. The "overwhelming majority of examples", according to Benveniste, indicate that the periphrasis involved a passive infinitive. The periphrasis "acts as the equivalent of the future passive participle" and "served to indicate the predestination of an object to follow a certain course of events."  It seems clear that habere here has a modal interpretation that essentially involves the notion of futurity. We can thus tentatively apply Warner's generalisation for the ME premodals, given in (1), to habere, as follows: (1') 2nd conjugation stative verbs subcategorised for an (passive) infinitive which denote  necessity, obligation and related notions of futurity are finite only. Since habere was the only 2nd-conjugation stative to have the relevant semantic property, (1') singles out just this verb (debeo ("owe") may have been another candidate, and in fact become a future auxiliary in one variety of Sardinian). We can account for (1') exactly as we accounted for (1) in the previous section. If habere is inserted directly into a modal head high in the functional structure, it is unable to be licensed as non-finite by aspectual or infinitival heads lower in the structure. Benveniste shows that by the end of the Imperial period the periphrasis clearly had a straightforward future meaning. By this time, then, habere was able to be merged in T(Future).  Similarities: (i) Evidence for incremental reanalysis "upwards" through the functional system, first to non-epistemic Mood, then to T (then to epistemic Mood; we will give evidence of Modern Romance varieties where chanterai forms are restricted to epistemic interpretations). This can accounted for in terms of a theory of change which treats the loss of movement as the central mechanism in categorial reanalysis; since movement is always local and upward, categorial reanalysis is local and upward. (ii) Morphological marking defines elements as exceptional: preterit-present morphology in English, 2nd-conjugation in Latin. All reanalysed verbs are stative, and arguably therefore more prone to an "athematic", functional interpretation. Differences: the obvious difference in the two developments is that habere became an affix, while the English modals remain morphologically free elements. This requires us to take closer account of Latin word order. Although Latin word order was rather free, there is a general consensus that the unmarked order was OV. This in turn implies that auxiliaries followed main verbs, following standard typological generalisations. Following Kayne (1994:141, n.15) and Zwart (1993:334f.)), we might propose a structure like the following for a Late Latin future form with a direct
object:
(3)  [ [VP Obj tV  ] ..  [T  V  [T habet ]]]
We can now account for the different developments in Latin/Romance and English in terms of the simple fact that Latin was OV, while 16th-century English was not. Hence habere was host to V-movement, as shown in (3), while English modals never were. Since ENE was VO, the modals did not develop into affixes.

Ian Roberts
Scienze della communicazione
Facolta' di lettere
University of Siena
via del Giglio 14
Siena
Italy

up
 

Aspect and temporal ordering: A contrastive analysis of Dutch and English
Ronny Boogaart
Free University, Amsterdam, NL

It is well known that grammatical aspect, i.e. the distinction between perfective and imperfective aspect, has a role to play in the process of inferring temporal relations in discourse: aspect imposes semantic and, therefore, indefeasible constraints on temporal reasoning. To explain the temporal interpretation of Dutch and English sentences containing (simple) past tense forms, this finding may not seem particularly useful given the fact that both the Dutch and the English past tense is unmarked for grammatical aspect. However, there are many instances in which the unmarked tense is unambiguously either perfective or imperfective and, more interestingly, these instances are not always the same in Dutch and English.

In this talk, I will show that the (occassionally) different aspectual interpretation of the aspectually unmarked past tense in Dutch and English is predictable as long as the grammaticalization of aspectually marked forms in both languages is taken into account. More specifically, perfect and progressive verb forms have grammaticalized to different extents in Dutch and English and this has systematic consequences for the interpretation of the unmarked form in both languages and, accordingly, for the way in which the unmarked forms constrain temporal reasoning in discourse. I will argue that the contrastive approach sheds new light on two hotly debated issues within the domain of temporal ordering in discourse, namely the interpretation of tense and aspect in indirect speech (sequence of tenses) and the so-called reverse-order phenomenon.
 

up
 

How to infer time direction in discourse ?
Jacques Moeschler
Department of linguistics
University of Geneva

Discourse semantics, as Kamp’s DRT, Asher’s SDRT, or ter Meulen’s DAT, have as main issues the question of relations between eventualities in discourse. In SDRT, discourse relations (more specifically « rhetorical » relations) as Narration, Explanation, Elaboration, Background and Results are conditions of discourse coherence and are inferred on the basis of default inference rules (Narration) and constraints on specificity (the more specific relation wins). In this talk, two types of discourse relations will be focused on, i.e. Narration and Explanation. The main issue I would like to answer is the following : what are the constraints (linguistic and non linguistic) which are responsible for the inference of the correct discourse relation ?
The answers I will propose belong to a radical pragmatic device, devoted to uttereance interpretation and not to discourse relations determination. This model (the Directional Model, cf. Moeschler et al. 1998) is a specification of Relevance Theory for temporal reference assignment, and emphasises the role of contextual information in utterance interpretation. DM is based on two basic assumptions :
1. Time directions in discourse (forward or backward inferences) are results of computation on directional features, either encoded by linguistic information or inferred from implicated (contextual) premisses. Thus, each utterance receives a provisory time direction which will combine with next utterance time direction (at this discourse level, the directional model is partially compositional).
2. Time direction in discourse is the result of principles of strenght on directional features : three constraints seems to be generalized (on the basis of possible combinations of tenses, connectives and predicats in French) :
A. Contextual information (that is, implicated premisses) is stronger than linguistic information.
B. Procedural information (that is directional information encoded in non lexical categories as connectives and tenses) is stronger than conceptual information (encoded in lexical categories).
C. Phrasal procedural information (connectives, adverbials) is stronger than morphologically incorporated procedural information (tenses).
Thus, for any utterance, time direction is computed via these contraints, and the inferred time direction for the first utterance is combined with the time direction computed for the second one. Finally a scale of optimality in discourse formation will be presented, based on the following principle : the higher the constraints is violated, the odder the discourse will be. As an empirical conclusion, the hierarchy of constraints on time direction proposed is not only a way of describing how we infer time direction in discourse : it should provide an interesting explanation of why some sequences of utterances are better than others when one wants to convey the same directional information.
Nevertheless, the question of discourse composition (how to compose time direction from more than two utterances ?) should be addressed. The second part of the paper is devoted to a short discussion of examples drawn from Crichton’s novel Airframe and concerns the composition of time inferences in paragraphs. I will use, as a tool for representing events relations in discourse, the theory of mental representation (TMR), which is a formalism based on Relevance Theory devoted to the representation of reference in discourse (cf. Reboul et al. 1997). The main issue will be the following : how are the criteria on time direction composed to allow or block event composition in discourse ? The answer I will submit for discussion is based on the principle of relevance : compose as far as information conveyed by composed eventualities is worth to, and stop composition as soon as your expectations are satisfied.

References
Moeschler, J. et al. (1998), Le temps des événements. Pragmatique de la référence temporelle, Paris, Kimé.
Reboul, A. (1997), Le projet CERVICAL : Représentations mentales, référence aux objets et aux événements, Nancy, LORIA, ms.
 

Representing time in procedural pragmatics
Louis de Saussure
University of Geneva

Abstract

Recent works in radical pragmatics of time have pointed out that temporal representations of eventualities are ordered by a conflit-resolution engine based on the strengh of various identified factors : contextual assumptions, connectives, tenses and conceptual rules (cf. Moeschler & alii 1998).
The aim of this talk is to provide a fully procedural approach for pragmatics of time as an example of how a globally procedural pragmatic theory can handle tough discourse linguistic problems. Following the reasearch on procedural expressions in Relevance Theory and extending the notion of procedure to the  whole utterrances, I will describe temporal relations by a cognitive context-building procedure, i.e. a set of organised cognitive operations. The temporal procedure takes one utterrance as input, the current utterrance, whereas other approaches, including Moeschler’s MID and Asher’s SDRT, take two utterrances. The current utterrance is interpreted relatively to the hearer’s cognitive environment, which contains information about already processed utterrances. The information we focus on is a set of mental representations (cf. Reboul & Moeschler 1998) of eventualities denoted by former utterrances, each of them containing temporal coordinates as a specific type of data, known as Reichenbach’s E and R (cf. Reichenbach 1947).
In this talk we postulate that when interpreting an utterrance that denotes an eventuality, the hearer :
- accesses the variables E and R if any in his cognitive environment ;
- recovers linguistic and non-linguistic temporal informations : tense, connective, conceptual rule, the latter beeing extracted from his world knowledge or inferred. All these informations formate as interpretative instructions or constraints on the coordinates ;
- evaluates potential conflicts between these instructions and solves it by a relevance-researching engine based on the strengh of the different conflicting factors ;
- applies the relevant temporal instruction on the coordinates E and R, the latter beeing considered here as a non-trivial referring coordinate in all cases (of utterrances in past tenses at least).
This procedural model accounts for discourse relations between both close and distant utterrances, and gives an unexpected contribution to the research on narrative discourse by explaining the highly efficient combination of utterrances in some specific past tenses.
This approach has been developped following numerous linguistic observa-tions that have lead to the procedural modeling of french tenses and connectives made by the Research Group on Temporal Reference directed by Moeschler in Geneva. Reason for which it will mainly be examplified with french data ; however, it will give rise to some comparison between french and english tenses, such as passé simple and simple past.

References

Asher N. (1993), Reference to Abstract Objects in Discourse, Dordrecht, Reidel.
Lascarides A. & Asher N., « Temporal Interpretation, Discourse Relations and Commonsense Entailment » in Linguistics and Philosophy 16, 437-493.
Moeschler J. (1998), Les relations entre événements et l’interprétation des énoncés, in Moeschler J., Jayez J., Kozlowska M., Luscher J.-M., de Saussure L. & Sthioul B., Le temps des événements. Pragmatique de la référence temporelle, Paris, Kimé.
Reboul A. & Moeschler J. (1998), Pragmatique du discours, Paris, Armand Colin.
Reichenbach H. (1947), Elements of symbolic Logic, New-York, Free Press.
Saussure L. de (to appear), Pragmatique temporelle de la négation. Dire et ne pas dire le temps, thesis, University of Geneva.
Sperber D. & Wilson D. (1986, 2e éd. 1995), Relevance. Communication and Cognition, Oxford, Blackwell.
Ter Meulen A. (1995), Representing time in natural language : the dynamic interpretation of tense and aspect, London - Cambridge Mass., The MIT press.
 

up
 

On the morphosyntax of sequence of tense
Alessandra Giorgi
(University of Venice, I)
(joint work with F. Pianesi)

 In this talk we address some questions concerning the temporal interpretation of embedded clauses, on the basis of empirical evidence from Italian. We will show that the language-specific properties of the morphosyntactic structure interact with very general mechanisms at the syntax/semantics interface, giving rise to a wide range of empirical patterns.
 The general overview of our proposal is the following: languages have a different inventory of morphosyntactic verbal forms -- for instance Italian is richer than English with respect to the variety of tenses and moods available in the various contexts. In spite of this, however, all languages establish the same kind of temporal relations between superordinate and subordinate events/states -- for instance simultaneity or precedence. As a consequence of this observation, one is forced to draw the conclusion that very general mechanisms at the syntax/semantics interface mediate between morphosyntax and meaning.
 The goal of this work is to clarify the nature and the properties of such mechanisms and to shed light on their functioning in a language-specific domain. In particular, we will argue that the patterns of Sequence of Tense in Italian can be accounted for by means of the interplay of the following three „parameters‰:
? The nature of the complementizer projections: C vs. MOOD.
? The morphosyntactic properties of the embedded verbal form.
? The presence vs. absence of a propositional attittude predicate in the superordinate clause.
 

up
 

On Event Coreference
Laurence DANLOS
TALANA, Université Paris 7, & LORIA

Works on temporal relation between two eventualities e1 and e2 examine exclusively whether one eventuality precedes, includes or overlaps with the other one (Moens & Steedman 1988, Asher 1993, Pustejovsky 1995). All these temporal relations suppose that e1 ? e2. I will concentrate on the cases where e1 = e2, i.e. on event(uality) coreference.

Event coreference has rarely been studied in detail, except for (pro)nominal phrases referring to an event (Webber 1988, Asher 1993). However, event coreference can be observed between two sentences, (1) and (2), and between an abstract description not syntactically realized (e.g. description of the causing sub-event implied by a causative verb) and a sentence (3) and (4).

(1)a Fred treated a tree. He pruned an oak.
     b Fred slapped Sue. He hit her today.

(2)a Fred pruned an oak. Therefore he treated a tree.
     b Fred slapped Sue. Therefore he hit a woman.

(3) Fred cracked the carafe. He hit it against the sink.
(4) Fred hit the carafe against the sink. He cracked it.

To study event coreference, it is necessary to distinguish two types of coreference relations: particularization (PART) and generalization (GEN). Let D1 and D2 be two successive descriptions of the same entity x (event or object). D2 = PART(D1) if D2 conveys some new information about x when compared to the information known from D1. D2 = GEN(D1) if D2 does not bring any new information about x.

In (1), the event coreference between the two sentences is of type PART: the two sentences refer to the same event, the second one bringing new information on this event. In (2), GEN is involved. I will emphasize the unusual coreference relations between NPs in these discourses (e.g. coreference between two indefinite NPs).

Examples (3) and (4) are causal discourses in which the result is expressed by means of a causative verb, here crack. The cause sentence describes an event which is interpreted as coreferent with the causing sub-event implied by the causative verb, i.e. an underspecified act by Fred on the carafe. I will show that the event coreference relation should be of type PART when the discourse relation is Explication as in (3) and of type GEN when it is Result as in (4).

The data will be presented in English, but they are the same across several languages (e.g. French, Italian, Arabic, Korean).

References
Asher, N., 1993, Reference to Abstract Objects in Discourse, Kluwer, Dordrecht.
Danlos, L., 1999, "Event Coreference Between Two Sentences", in Proceedings of the Third International Workshop on Computational Semantics (IWCS'99), Tilburg.
Danlos, L., 2000, "Event Coreference in Causal Discourses", in P. Bouillon et F. Busa, (eds), Meaning of Word, Cambridge University Press.
Moens, M., Steedman, M., 1988, "Temporal Ontology and Temporal Reference", Computational Linguistics, vol. 14, 15-28.
Pustejovsky, J., 1995, The generative Lexicon, The MIT Press.
Webber, B., 1988, Discourse deixis: Reference to discourse segments, in Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics (ACL'88), Buffalo, NY, 113--123.

Interpretation of German Tenses and Temporal Adverbs.Principles for the construction of Underspecified Discourse Representation Structures.
Hans Kamp, Kai-Uwe Karstensen, Uwe Reyle, Antje Rossdeutscher, Jasmin Saric
& Judith Tonhauser.

A major problem for a semantics of German tensed discourse is the high degree to which the German tense forms are ambiguous (when compared, say, to those of English, French and other Romance languages). Often these ambiguities can be resolved, but only on the strength of information located elsewhere in the sentence or discourse.For an approach towards sentence and discourse interpretation within a general underspecification framework such as Reyle's UDRT, the ambiguities inherent in the German tenses pose a special challenge, since they often interact with other kinds of ambiguity or underspecification, such as e.g. quantifier (more generally: operator) scope ambiguities and underspecification of the collectivbe-distributive distinction..

The talk wil present some of the particular problems that the German tense system presents, as well as the general architecture and principles of a UDRT-based interpretation theory for German tenses and temporal adverbs.

up
 

INFERRING FROM WHAT?
The impact of the grammaticalization of aspectual operators on the derivation of temporal inferences in discourse
J. Bohnemeyer, MPI Nijmegen

Abstract

In previous research, I have shown that defeasible temporal inferences can be derived from the use of aspectual operators in discourse according to a set of Gricean ‘generalized conversational implicatures’. I have proposed a theory of such inferences that presupposes a principled inventory of six ‘notional’ aspectual operators. These are ‘notional’ in the sense that they abstract away from the actual grammatical realization of aspectual operators in any particular language. It is widely known, however, that languages differ drastically in this domain. That is, they differ in the number of aspectual operators they grammaticalize, in the degree of grammaticalization of particular aspectual operators, in the conflation of different aspectual operators in one morpheme or construction, and in the conflation of aspectual operators with non-aspectual operators (such as tenses and modalities) in particular morphemes or constructions.

The aim of the present paper is to investigate in a case study how language-particular differences in the grammatical realization of aspectual operators affect the derivation of temporal inferences from the use of these operators in discourse. The case study is focussed on the contrast of simple tenses and progressive constructions in three closely related languages: Dutch, English, and German. While the lexical and grammatical expression of temporality is in broad outline compatible across these languages, they differ significantly in some details, including in the degree of grammaticalization of their progressive constructions. It can be shown that the progressive constructions of the three languages form a cline from least strongly grammaticalized in German to most strongly grammaticalized in English, with Dutch as an intermediary case. In order to examine the consequences of these differences for the derivation of temporal inferences, data was collected with a video stimulus from native speakers of the three languages. A referential communication task was designed around this stimulus so as to make sure that the subjects focussed on communicating the same event orders represented in the videos. In discussing the results, I will in particular pursue the question as to what extent the aspectual interpretation of simple tense forms, and the temporal inferences they trigger in discourse, depend on the degree of grammaticalization of a progressive construction.
 

up
 

Reasoning with Dynamic Aspect Trees
Alice ter Meulen (University of Groningen, NL)

Temporal reasoning is modelled as situated inference in Dynamic Aspect Trees (DAT, ter Meulen (1995)) and characteristic inferences are formulated as natural deduction rules. Constraints on permutation, monotonicity and cut are based on the fundamental distinction between static information, which lives on the DAT structure and is portable within subtrees,  and dynamic information, which drives the architecture of the DAT.
The data are limited to the English tense and aspectual morphology only, but applications to other languages are invited.
 

up
 
 

Temporality in Plans and Natural Language Semantics.
postscript full text
Mark Steedman
University of Edinburgh (Scotland)

The paper discusses some resources from logic and artificial intelligence research that can be used to build knowledge representations that support common sense reasoning about events and times, so as to support a semantics for verbs, temporal adverbials and other natural language categories like tense, mood, and aspect.  While these categories are commonly talked of as "temporal", the paper argues that they are primarily causal and teleological, and can be captured in a dynamic logic-based variant of the McCarthy/Hayes/Kowalski situation calculi called the Dynamic Event Calculus.  It discuss various criticisms of such calculi that have been mounted in the AI literature, based on supposed inability to handle various versions the AI Frame Problem, including the Ramification and Qualification Problem.  It shows that on the contrary, when properly drawn up in the way that the natural language problem demands, such systems provide natural solutions to such problems.
 

up
 

Temporal reasoning with aspectual adverbials
Hans Smessaert (FWO-Vlaanderen & K.U.Leuven)
Alice G.B. ter Meulen (Universiteit Groningen)

In this talk two groups of Dutch adverbial expressions are assigned a semantic representation, namely the adverbials of aspectual continuity, such as nog (still) and nog niet (not yet) and those of aspectual focus, such as nog altijd niet (still not) or al niet meer (?already no longer). In both cases a bit-string formalism is proposed which generalizes the basic distinction between positive and negative polarity. The continuity adverbials receive a three-dimensional bit-string representation (ABC), wheras the focus adverbials require a five-dimensional bit-string representation (ABCDE) in order to capture the idea of alternative expectations.

   A: actual polarity    1 = positive polarity:   0 = negative polarity
   B: event-internal polarity transition  1 = beginning   0 = finishing
   C: speaker’s perspective on the event 1 = retrospective   0 = prospective
   D: evaluative focus w.r.t. expectations 1 = positive focus (faster) 0 = negative focus (slower)
   E: counterfactuality of the alternative 1 = primary focus(opposite) 0 = secondary focus (identical)

These bit-string representations are then first of all used to account for static temporal inferences, which concern one single information state. Two elementary valid patterns are distinguished: from focus with ABCDE in (1a) to continuity with ABC in (1b), and on to elementary polarity with A in (1c):

 (1a)  Jan slaapt nog altijd niet om t1 (John is still not asleep at t1)  0 1 0 0 1 (ABCDE)
 (1b)  Jan slaapt nog niet om t1  (John is not yet asleep at t1)  0 1 0  (ABC)
 (1c)  Jan slaapt niet om t1  (John is not asleep at t1)   0  (A)

Both with continuity in (2) and focus in (3) the relationship of internal negation, which reverses the AB-values, gives rise to logical equivalences when the AB-reversal on the adverbial is neutralized by the substitution of the contradictory verbal predicate. The unidirectional inferences in (1) and the equivalences in (2-3) can furthermore be combined in a lattice structure of valid static temporal inferences:

 (2a)  Jan slaapt nog niet om t1  (John is not yet asleep at t1)  0 1 0  (ABC)
 (2b)  Jan is nog wakker om t1  (John is still awake at t1)   1 0 0  (ABC)
 (3a)  Jan slaapt nog altijd niet om t1 (John is still not asleep at t1)  0 1 0 0 1 (ABCDE)
 (3b)  Jan is nog altijd wakker om t1 (John is still (always) awake at t1) 1 0 0 0 1 (ABCDE)

The dynamic perspective on temporal reasoning then tries to account for the fact that when new information is added, certain components are updated, whereas others remain constant across information states. More in particular, the minor premiss of the syllogistic argumentation pattern in (4b) triggers a polarity transition which switches the C-value from 0 in the major premiss to 1 in the conclusion. As a consequence, the values of both the A- and the E-parameter in the major in (4a) have to be reversed in the conclusion (4c):

 (4a)  Jan sliep om t1 nog altijd niet. (John was still not asleep at t1.) 0   1   0   0  1
(4b)  Jan viel om t2 in slaap.  (John fell asleep at t2.)          (switch ACE)
 (4c)  Jan sliep om t3 eindelijk.  (John was finally asleep at t3.)  1   1   1   0   0

In other words, valid reasoning depends on preserving the B-value of the polarity transition (beginning or ending) as well as the D-value of the subjective focus evaluation (‘faster or slower than expected’). In a final step it is demonstrated how the static and the dynamic perspectives on temporal reasoning may interact to yield complex but valid inference patterns. Along the way there will be ample occasion to comment on the considerable differences in lexicalisation potential for aspectual focus adverbs between Dutch, French and English.

up