Representing events in language and discourse
MOESCHLER, Jacques. Representing events in language and discourse. In: Coene, M. ; De Mulder, W. ; Dendale, P. & D'Hulst, Y. Traiani augusti vestigia pressa sequamur. Studia lingvistica in honorem Liliane Tasmowski. Padova : Unipress, 2000. p. 461-479
Disclaimer: layout of this document may differ from the published version.
1 / 1
Representing events in language and discourse
Department of Linguistics University of Geneva
Among the domains of linguistics in which Liliane Tasmowski-De Ryck’s works have been very important, the description of verbal tenses and, more generally, temporal reference are foremost. I will only refer here to one of her contributions (Tasmowski-De Ryck 1985), because this paper has had an enormous impact on my own work about temporal reference in recent years.
Why? Quite simply, because Liliane successful showed how classic and modern descriptions failed in their description of narrative imperfective shift (imparfait narratif de rupture). Since that paper, a precise set of properties, examples and counter-examples are to be found in litterature about time in language.
What I propose to do in the present paper is to illustrate a problematic linked with temporal reference and verbal tenses, that of discourse relations, which is directly linked with (temporal and non-temporal) uses of verbal tenses. More precisely, I will talk about an ontological category, events, which has to do with language and language use. But, in order to show the relations between events and language, I will not adopt the starting point of ontology. I will rather deal with linguistic categories linked with time and aspect, and also with relations between events in discourse. I will thus speak of linguistics (mainly linguistic categories), pragmatics (language use) and also of discourse, i.e. of the relations which ontological categories manifest in discourse and communication.
The position on language which I will defend is this: languages, and more specifically French, have as their main property to express relations with the world through their use. They allow reference to individuals, to objects, to events, that is, they allow speakers to represent the world. These representations have a very precise cognitive function: having a representation of the world is only relevant and interesting for an individual if the representation is true. It would indeed not be relevant and would quite costly
cognitively to entertain false representations of the world. It may well be that some of the representations which we have of the world are not correct, but one of the aims of cognition is the ability to adapt the representations of the world which we have to reality: when a representation which was entertained as true is proved to be false, i.e. it does not accord with reality, the interest of the individual is to change its evaluation and to consider it false. If he or she does not, he or she may encounter highly disagreable situations which might even threaten his or her life. Entertaining as correct as possible representations of the world is thus one of the most urgent tasks of human cognition, and understanding how representations are built through language is one of the tasks of contemporary linguistics. Explaining how these representations are accessible through communication is the main task of pragmatics1.
In this paper, I will only deal with one specific type of representations: event representations. We are indeed able to represent events in time, to see them as punctual or durative, as bounded or unbounded. The daily reading of newspapers allows us to learn what happened today or yesterday; reading novels and fictions allows us to represent events which did not happen and even to deal with them meaningfully. We will see how events are represented and evaluated. Finally, we will show how, in discourse, event representations are ordered relative to one another.
But before introducing ontological caegories, let us take a look at linguistic categories and how they are linked to the world.
2. The conceptual content of lexical categories
I would like to present a classical conception of linguistic categories which I will then reinterpret in cognitive and pragmatic terms. Contemporary linguistic theories agree on a two-part division of types of linguistic categories: they can be either lexical or grammatical.
Lexical categories for French include Nouns, Verbs and Adjectives. They have an important common characteristic: the expressions which come under them are defined by their lexical meaning. If cat and dog are different nouns,
1 See Reboul & Moeschler (1998a) for a more precise account.
this is because they designate different classes of individuals. If runing and reaching the summit have different meanings, it is because they designate different types of events. Each type of things designated by an expression belonging to one of the lexical categories is defined by very precise conditions, which correspond to its meaning. If we can distinguish between dog and cat in language, this is because there is a difference between two types of mammal and that these types are differenciated in French. In the same way, if we can distinguish between an activity like runing and an event like reaching the summit, it is because these eventualities have different meanings. One can say that Paul has reached the summit only if Paul went from the bottom of the mountain, climbed the mountain and reached the top of the mountain.
Lexical categories have a property which is directly linked to their lexical meaning: they can encode concepts. What do I mean by “encoding concepts”?
A more simple way of speaking is to say that a noun, a verb, an adjective have conceptual contents. If they have a meaning, it is because they express a concept. Let us come back to the cat and dog example. If we, French speakers, can designate an object in the world as un chat (a cat) and not as un chien (a dog), it is because we are able to recognize certain properties which may be or not be necessary, which define the concept CHAT. It happens that it is the properties of individual cats which determine the concept CHAT and not the word chat which defines what it is is to be a cat.
If we can have representations of the world, it is because we can have representations of what dogs are, of what are cats and the same goes for many other things. But we can as well have representations of cats and dogs independently of language: I can right now represent to myself my dog, Ego, or have a general representation of Newfoundland dogs, I could draw it for you, etc. But if I can have linguistic representations, it is because language allows me to refer to it: either to one specific member of the class (for instance through a possessive NP, my dog, my cat, to refer respectively to Ego, a 7 years old Newfoundland, and to Perceval, a 5 years old siamese), or to the entire class, cats, dogs, or to a subset of the class (my cats, my dogs).
Reference is thus done in language use through referential expressions:
definite descriptions, indefinite descriptions, possessive or demonstrative NPs,
pronouns, proper names, etc. Each time a speaker uses a referential expression, he or she aims to designate, either referentially or attributively, a thing in the world (see Donnellan 1979). He or she then allows his or her hearer to access a representation of that thing. Suppose you are in my kitchen. You get more and more surprised as you see a cat on a cupboard, a cat on the fridge and a cat purring near the cooker. If I now say: “My cat is near the fridge. He is called Perceval”, you will not only identify an object in the world, you will also be able to build a representation of my cat. It is male, siamese, likes to lie near the cooker and, as you just walked up to him and pet him, he is not easily frightened and he purrs easily. Henceforward, we will have no problem when speaking of Perceval. You will know that we are speaking of my cat and we will be able to exchange informations about him, about cats, about cat-lovers, etc.
Now, you could object that this does not tell us anything about language: I have only spoken about some uses of language and, notably, I have alluded to the fact that proper names and definite descriptions refer to things in the world.
But in my utterance, “My cat is near the cooker. He is called Perceval”, I used a term which defines a concept. By using it, I may have taught you, if you do not speak French very well, that animals who mew are called chats in French, that cats like heat, that they sleep a lot, etc. Apparently, I have not said anything about the word chat, though I have said a few things about the concept CHAT.
The question which comes to mind now is what is the interest of having a whole repertory of expressions belonging to lexical categories if their meanings are not linked to objects in the world or, at least, to the mental representations we have of these objects. I cannot see any clear answer to that question and the conclusion which must be drawn from this fact is that the interest of concepts, which have lexical counterparts in the lexicon of natural languages, is precisely to allow us to access true representations of the world.
3. The procedural content of grammatical categories
I would like now to introduce an important distinction which will modify the representation we have of language in an important way. There are expressions, notably grammatical morphemes which have neither lexical meaning nor conceptual content. And yet, it would not be correct to consider
that they are not linked to concepts. These expressions, personal pronouns, deictics, cunjunctions, some adverbs, verbal tense have meaning, but they have procedural meaning. This means that there are concepts which only exist through words in language.
Considering that grammatical morphemes have meaning is nothing new.
What is new, though, is considering that they have procedural meaning.
Indeed, the whole lexicographic tradition, just as grammatical tradition did, considered that the meaning of grammatical morphemes is conceptual. Most lexicographic descriptions of connectives and grammatical descriptions of berbal tenses link these morphemes to a set of meanings, all of which depend on a general meaning, or in modern terms on a “basic semantism” or “common meaning”, etc. Contemporary descriptions of connectives and verbal tenses, whatever their theoretical framework, have however insisted on the difficulties in describing the conceptual meaning of grammatical morphemes. Which concept should we associate with et (and), si (if) and parce que (because)? The concepts of logical cunjunction, of implication, of causality are indeed not precise enough to account for the temporal, causal, counterfactual uses of et, for the performative, biconditional and counterfactual uses of si or for the causal enunciative or inferential uses of parce que, as respectively illustrated through examples (1) to (3):
(1) a. Max est rentré et le téléphone a sonné.
(Max came in and the telephone rang) b. Marie a crié et le bébé a pleuré.
(Mary shouted and the baby cried)
c. Pierre: Hier soir, j’ai découvert la théorie de la relativité.
Marie: Et moi, je suis le pape!
(Peter: Yesterday evening I discovered relativity theory. Mary: And I’m the pope!)
(2) a. Si tu as soif, il y a de la bière dans le frigo.
(If you’re thirsty, there’s beer in the fridge) b. Si tu es sage, nous irons au cinéma.
(If you’re nice, we’ll go to the movies)
c. Si j’étais riche, ma 2CV serait une Rolls.
(If I were rich, my 2CV would be a Rolls).
(3) a. Max est malade, parce qu’il a trop mangé.
(Max is sick because he ate too much)
b. Est-ce que Max est malade? Parce que je ne l’ai pas vu depuis trois jours.
(Is Max sick? Because I haven’t seen him for three days)
c. Max est tombé à mobylette, parce qu’il a le bras dans le platre.
(Max fell from his motorcycle because his arm is in plaster).
In the same way, attributing conceptual content to a verbal tense such as imperfective (Imparfait) such as unbounded past state does not account for such simple uses as counterfactuals, represented speech and thought ou temporal or narrative uses of the imperfective:
(4) a. Si Marie était là, Max serait heureux.
(If Mary were here, Max would be happy)
b. J’ai rencontré Jean. Il partait pour l’Espagne et avait rendez-vous avec des amis d’enfance.
(I met John. He was leaving for Spain and was meeting with childhood friends).
c. Max alluma une cigarette. Le tabac avait un coup de fiel.
(Max lit a cigarette. The tobacco tasted foul.)
It is easy to see why a content or a conceptual meaning cannot be attributed to grammatical morphemes: their value varies relative to contexts and to utterances in which they occur. In other words, they are context-sensitive expressions.
Until a few years ago, when relevance theory was devised by Dan Sperber and Deirdre Wilson (see Sperber & Wilson 1995, Wilson & Sperber 1993a), there was no satisfactory account of the pragmatic and semantic functioning of these morphemes. Linguistic descriptions, though very rich and detailed, did not amount to anything like a theory. For instance, some accounts would use notions like discourse or coherence: anaphoric prounouns, verbal tenses, connectives, etc. would be cohesion marks signalling discourse coherence. But
these accounts unfortunately relied on circular definitions: discourse was defined through coherence and coherence through discourse. What is more, it has been shown (see Reboul 1997) that cohesion marks (connectives, verbal tenses, referential expressions) are neither necessary nor sufficient to coherence.
The approach which we advocate and which is precisely developped in Reboul and Moeschler (1998a) annd (1998b) gives an account of linguistic expressions with procedural content. This account has two aspects: a linguistic aspect and a cognitive aspect.
4. The cognitive explanation of procedural expressions
Let us come back to concepts. Some concepts have a concepual content: they allow us to access a representation of the world, i.e. to represent situations, events, things (individuals, objects) and to evaluate them as true or false. Beside concepts with a conceptual content, we distinguished concepts with a procedural content. Which content should be associated to these concepts? The concepts which have a procedural content give instructions on the way representations built from concepts with conceptual content should be treated or on the way their referents should be accessed.
Let us suppose that we must attribute a content to the concept JE (I). It seems, a priori, legitimate to associate to the linguistic expression je a concept and this, for linguistic rather than cognitive reasons. The most ordinary way to designate oneself is not to use a definite description: thus, to refer to myself, I will not ordinarily use the expression Jacques Moeschler nor any definite description such as le pragmaticien radical de Genève (the radical pragmatician in Geneva). I will designate myself through the use of a pronoun such as je (I), moi (I), me (me) or, if we do not understand one another, I will designate myself through a demonstration (by pointing to my chest). Now, how should one explain that the content of this concept should be procedural? It could be thought that the content of the corresponding concept is “the speaker of this sentence”. But, unfortunately, there is a devastating argument which shows
that this is not correct. If it was the right analysis, then sentence (5) would always be false, which is not the case2:
(5) Je n’existe pas.
(I do not exist.)
Substitute to je the synonymous expression, i.e. “the speaker of this sentence”:
(6) Le locuteur de cette phrase n’existe pas.
(The speaker of this sentence does not exist).
But (5), when uttered by myself, merely means:
(7) Jacques Moeschler n’existe pas.
(Jacques Moeschler does not exist.)
(5) and (7) are false, but not necessarily false, whereas (6) is necessarily false.
I might indeed not have existed, if, for instance, my parents had not met, or my older sister had been a boy. The only content which can be attributed to je is thus a procedural content which could spelled explicitely as:
(8) Look for the speaker.
If je has a procedural meaning, it is then impossible to give an incorrect analysis of (5): it will be recognised that (5) is false, but not that it is necessarily false.
The cognitive account of the description of procedural concepts is thus as follows: to access mental representations, whether through language or through any other means, et to evaluate these representations, it is necessary to have informations on the way to treat representations and to identify their referents.
5. The linguistic explanation of procedural concepts
There is also a linguistic account of procedural concepts. The linguistic argument relies on the way utterances are interpreted. Interpreting a utterance
2 See, for more detailed ccounts, Moeschler & Reboul (1994, 364-365), Reboul &
Moeschler (1998, chapter 7) and Reboul (1992), based on the original account by Kaplan (1989).
implies applying linguistic rules which yield a partial interpretation (the logical form of the utterance), but also implies completing, i.e. enriching the logical form to obtain a completly propositional form, which can be evaluated. A simple way to go from logical form to propositional form is through the use of informations drawn from the situation, which are more precisely called the context. The context gives access to one or more other propositions which allow the individual to draw (contextual) implications which correspond to the speaker’s intention. If, for instance, in the small dialogue under (9), Marie can access (10) (the context) she will be able to draw conclusion (11) and to understand what Pierre meant:
(9) Marie: Quelle heure est-il?
Pierre: Le facteur vient de passer.
(Mary: What time is it? Peter: The postman just came) (10) The postman comes at nine.
(11) It is a few minutes past nine.
There is, however, another way of accessing contextual information, i.e.
information which is needed to determinate the propositional form of the utterance: the speaker can use procedural expressions and the hearer can apply the procedure linked to them. We have already seen the example of the procedure linked to a first person pronoun. Let us now turn to a more complex situation:
(12) Anne m’a dit: “j’ai lu ton article pour Liliane”
(Ann told me: “I’ve read your paper for Liliane”)
How should we apply the procedures linked to procedural expressions? First person marks are characterized by the following procedure:
(13) Look for the speaker and identify the referent of the first person marks with him or her.
In the same way, second person marks are characterised by procedure:
(14) Look for the hearer and identify the referent of the second person marks with him or her.
In the situation described in (12), there are two speakers, Jacques Moeschler and Anne Reboul and the hearer of Anne Reboul’s discourse is Jacques Moeschler. The propositional form of (12) is just complete if each referent has been correctly attributed and if it agrees, in the situation described, with (15):
(15) Anne Reboul a dit à Jacques Moeschler: “Anne Reboul a lu l’article pour Liliane de Jacques Moeschler”
(Anne Reboul told Jacques Moeschler: “I’ve read Jacques Moeschler’s paper for Liliane”).
6. The procedural meaning of past perfect (passé composé)
Let us turn to a last example of procedural meaning, that of a verbal tense such as past perfect. There is a non controversial fact about utterances with past perfect: they represent events, situations accomplished in the past which generally have a relation with the present. This relation can be characterised through a procedural content:
(16) Identify the temporal interval between the moment at which the event occurred and the speech moment3.
The problem is thus whether the speaker and hearer will proceed to identify the interval. Let us consider two situations represented in (17) and (18) (borrowed from Wilson & Sperber 1993b):
(17) Votre collègue vient vous chercher dans votre bureau pour vous emmener déjeuner. Vous lui répondez: “Désolé, mais j’ai déjà déjeuné”.
(Your colleague comes to have lunch with you. You answer: “Sorry, but I already had lunch”)
(18) Vous discutez entre amis de vos plus importants voyages. Vous déclarez: “Moi, je suis allé au Tibet”.
(You are discussing of your farthest journeys with friends. You say: “I have gone to Tibet”).
3 Things are more complex indeed as some implications of the past perfect are bounded and others are not. See Sthioul (1998), Luscher & Sthioul (1996) and Luscher (1998).
How does your hearer manage to understand in (17) that the interval between the time at which you had lunch and the time of your speech is of some minutes rather than hours, days or years? How do your hearers in (18) understand that the relevant interval is your life, or at least your conscious life?
The explanation is not linguistic. The meaning of the past perfect has nothing to do with these indications. Rather, the explanation is pragmatic. It does not lie in the meaning of the past perfect, but in the nature of the events represented.
Here it is:
A. In the first situation, we know that the event described is cyclic: unless under peculiar circumstances, in our western society, we have lunch once a day. Now we know that if a speaker says that he or she already had lunch, we do not interpret his or her utterance as meaning that he had lunch but as communicating that this event occurred in the very recent past. If, for instance, our companion answers to the question “what did you do today?” by “J’ai déjeuné” (I had lunch), we will consider that in order to be the great event of his or her day, that lunch must have had a special importance which luncheons generally do not have. It is thus not difficult to describe the procedure for the past perfect:
(19) Look for the shortest possible interval given the circumstances.
B. In the second situation, it is common knowledge that an ordinary european citizen does not travel to Tibet very frequently. Interpreting the utterance “Je suis allé au Tibet” (I went to Tibet) thus implies that the event be situated in the interval of the speaker’s whole life. But once again the procedure indicated under (19) can be applied. It is indeed more relevant for the speaker to communicate a recent event than a very ancient one.
Suppose for instance that you lived for the first six months of your life in Tibet. In this context, the utterance “Je suis allé au Tibet” (I went to Tibet) is true, but not relevant: you probably do not remember anything of that time and mentioning an event which actually occurred but which you cannot remember is not truly relevant in communication. If you went to Tibet when you were 10 years old, your utterance will be more relevant because your hearers will be able to draw some effects from it: you remember that journey, it was noteworthy, etc. Make the interval shorter
and you will find out that the utterance gains relevance through additional effects. In the case of (18), which does seem rather extreme, the procedure of the past perfect applies.
This shows the interest of having procedural expressions to represent time.
Procedural concepts are necessary to access representations of events and of participants in these events: the conceptual informations given in the utterance are not sufficient to access a complete propositional form. Procedural concepts have precisely as their function to allow such an identification. Indeed, if interpreting a utterance is having a representation of the event it represents, one must be able to pin point this event in time; if the event has participants, they must be identified. This is precisely what procedural expressions allow.
7. A minimal ontology of events
I began this paper by reminding my readers of the fact that language function is to give access to representations of the world and that it is relevant, for individuals, to entertain true representations of the world. We will now take a closer look at event representations.
We can choose between two different methods. We can either work with classical linguistic methodology, trying to give representations of events on the basis of what language offers us. Or we can, as philosophers do, give ourselves an ontology of events and determine its relation to language.
The first strategy, traditionnally called a semasiological approach (i.e. from signs to concepts), has been very often used and has, notably, led to aspectual theories or to Aktionsarten: the kinds of things which sentences represent are a function of the semantic properties of the linguistic units combined in sentences. But, apart from the lack of generalisations due to the specificities of particular languages, this approach also fails on ontological clearness. How can it be explained, indeed, that we can represent events, that these representations are true of the world, if the categories offered are linked to language and are thus subject to time and space variations?
The second strategy takes ground in a simple constatation: either ontological categories are universal or they are not. As ontological categories are supposed
to be universal in this second strategy, it requires a minimal definition of events. We will relie on the following remarks4:
I. Events are processes which occur in a specific period of space-time: they have a begining and an end, which implies that they are defined through an initial boundary and a final boundary. The time interval defined by these two boundaries defines the time of the event.
II. Events are defined relative to a zone which precedes them (the pre-state) and a zone which follows them (the post-state). An event is thus what succeeds a state in which the event has not yet occurred and a state in which the event does not occur anymore. The post-state is often called the resulting state; it is generally defined by the implications of events.
III. States are characterized by their unboundedness: they have a duration, but not initial or final boundary. They are, what is more, homogeneous, as there is no change occurring between the initial and terminal boundaries.
IV.Some events, or processes, are characterized by the fact that they have an intrinsic end whereas others do not. For instance, building a house is a process defined by such an ending (a culmination), while pushing a cart is not (the ending of the process is not the supermarket cart or the car or anything else). It is said that building a house is telic, while pushing a cart is atelic.
V.Finally, the last distinction is that some processes take time while others directly describe their intrinsic ending: for instance, reaching the summit described the intrinsic ending of the process, without any interval preceeding it; on the other hand, though building a house defines an intrinsic ending, it includes a preceeding time, necessary for its accomplishment.
To come back to Vendler’s by now classical ontology and categories, we will make the following ontological distinctions:
4 See, notably, Vendler (1867), Mourelatos (1981), Asher (1993) and (1997) and Kozlowska (1988) for a synthetic overview.
A. We will first distinguish between states and processes: states are unbounded, atelic, do not imply change of any kind, i.e. they are homogeneous; processes are intrinsically or extrinsically bounded, telic or atelic, homogeneous or heterogeneous.
B. Among processes, we will distinguish between activities and events:
activities are atelic, not intrinsically bounded, whereas events are telic and bounded.
C.Finally, we will distinguish among telic and bounded processes, that is among events, between accomplishments and achievements: achievements are punctual, they do not have duration, while accomplishments do.
We can sum up this ontological taxonomy by the following figure:
This taxonomy, though it is classic, will allow us to understand how we can access representations. Representations of eventualities are based on the fact that we can distinguish between states of affairs on the basis of criteria of:
(a) progression/non-progression: states do not progress, processes do;
(b) homogeneity/non-homogeneity: activities are homogeneous, events are not
(c) punctuality/non-punctuality: achievements are punctual, accomplishments are not.
These criteria are ontological. They group with other criteria, such as telicity and boundedness. To see how they come in the representation of events, we must now examine how conceptual and procedural informations combine in the building of eventuality representations.
8. The combination of procedural and conceptual information in the interpretation of events
We now know what types of things utterances allow us to represent: states, activities, accomplishments and achievements. These different eventualities
correspond to different aspectual classes. They define the aspect in the world of the things we are talking about.
We will now ask how aspectual classes are built in utterance interpretation and how utterances combine in discourse.
The first question has to do with how we can represent things in the world as states, activities, accomplishments and achievements. Following the principles given above, conscious access to these kinds of things can only be combination of conceptual and procedural information.
Conceptual information allows us not only to give a name to the thing, but also to have a semantic representation of it. Suppose that I want to allow you to represent a specific event, the destruction of Carthago for instance. There are several things which I can do to linguistically express that event:
(20) a. La destruction de Carthage.
(The destruction of Carthago)
b. La destruction de Carthage par Scipion (The destruction of Carthago by Scipio) (21) a. Carthage a été détruite.
(Carthago was destroyed) b. Scipion a détruit Carthage.
(Scipio has destroyed Carthago)
In (20), I used an action noun (destruction) which is respectively associated with one or two arguments, the patient (Carthago) and the agent (Scipio). In (21), the predicate is not nominal anymore but verbal, but the diathesis is identical: in (21a), the predicate only has one argument, while in (21b) it has two arguments. The conceptual representations, given here as logical forms, are respectively as follows5:
(22) a. λx[destruction’(Carthago)(x)]
5 In the standard use of λ-calculus (see Dowty, Wall & Peters 1981, Moeschler &
(23) a. λx[destroy’(Carthago)(x)]
These two sets of logical forms are not identical: predicates do not have the same aspectual consequences: destruction is a process, while détruire (destroy) can be seen as either an accomplishment or an achievement, notably in its passive form.
In the expressions I used, as well as in the logical representations I gave of them, informations are only conceptual: concepts DESTRUCTION and DESTROY, associated to the words destruction and détruire (destroy), yield encyclopaedic informations as well as aspectual information on what a destruction or the destroying of something, a town for instance, is. But the representation given by conceptual contents is very abstract: we know what type of eventuality is implicated, we know who the protagonists are, but we cannot yet actualise this representation under a propositional form. This is where procedural informations come in, that is the procedural content, notably of verbal tenses.
The past perfect activates two types of information: a conceptual information (the event is past) and a procedural information (“look for the shortest interval in the circumstances which makes the utterance relevant”). For the information to be relevant, it must either correspond to a true proposition or produce implicatures which correspond to true propositions. It is thus necessary to access a complete representation which can be evaluated for truth-value.
Suppose now that my son, Alexander, who has just begun learning latin, announces during lunch: “Scipion a détruit Carthage” (Scipio destroyed Carthago). His younger brother, Nathanaël, who only know roman civilization through Gosciny and Uderzo’s works, replies: “Tu as lu ça dans le Journal des enfants?” (Did you read that in the Children’s magazine). Self-evidently, the event reported by his brother is not correctly localised in time. Nathanaël knows that it occurred in the past, but he thinks that it was in the recent past.
He does not know who Scipio is or what Carthago is. He can think that it is a town, a character in a comic strip, etc. But, we, reasonably well-informed adults, have informations on concepts SCIPIO, CARTHAGO, stocked in their
encyclopaedic entries. We know that Scipio’s destruction of Carthago occurred in 146 before Christ, that Scipio was a Roman general and Carthago a town in North-Africa, near Tunis. The representation under propositional form which we can access:
(25) λP[P(Scipio’)], where P = Roman general responsible for Carthago’s destruction
(26) λP[P(Carthago’)], where P = Punic capital destroyed by Scipio in 146 before Christ.
(27) λaλb [detroy’ (b)(a)], where destroy = make b disappear
(28) Scipion a détruit Carthage: ∃e∃t (destroy (e) ∧ agent (Scipio, e) ∧ patient (Carthago, e) ∧ occur at (e, t) ∧ t = 146 before Christ
The representation with propositional form can be interpreted as true, because the informations it contains correspond to what occurred. Now, it can very well be that I am mistaken on the date of the destruction, or, if I must treat the sentence “Carthage a été détruite” (Carthago was destroyed), believe that Caesar destroyed Carthago. I can thus, from informations given in the utterance, but also from encyclopaedic informations given by conceptual contents to which my interpretive device has access, produce false representations. I can even keep these false representations in memory and they only will be supressed if they are confronted to stronger contradictory informations. For instance, before producing this example, I consulted an encyclopaedia: some of my representations of Romans and Punics were false and I suppressed them.
What does this example show? Very simply, it shows that to build an event representation, here Scipio’s destruction of Carthago, I must complete the logical representation given by the linguistic informations in the sentence with two types of informations: conceptual informations given by lexical categories and procedural informations given by grammatical categories. The development of logical form into propositional form leads to the propositional form (that is to a complete propositional representation), which can then be evaluated as true or false.
What is of interest in accessing representations under propositional form which can be evaluated? I already have given a first answer, at the begining of this paper, and I would now like to complete it. The true representations allow access to a correct representation of the world and thus, allow the hearer to enrich the informations he has about the world. If he enriches his knowledge of the world with false representations, he will, as we have seen, meet with difficulties. It is thus cognitively relevant to entertain true representations of the world.
Now, these representations, when they are built through communicative acts, notably linguistic acts, are not only interesting for the individual through their truth. They are also interesting through their relevance. This simply means that they produce a minimum of cognitive effects. What are these effects? In Relevance theory, three types of effects are defined. A representation is relevant in as much as it
I. produces new information: for instance, the individual did not entertain that representation as true before;
II.reinforces an old information: for instance, the degree of certainty with which it is entertained may either grow or diminish;
III.annulates an old representation, because it is inconsistant with it and because it is stronger than it.
A way of being relevant in communication is thus for the speaker to allow access to representations which produce at least one cognitive effect. In a parallel way, gaining relevant information is tantamount for the hearer to get at least one cognitive effect in the interpretation process of the utterance.
9. A pragmatic treatment of discourses
Let us sum up the interpretation process for utterances, such as we described it previously:
A. The utterance is translated in a logical form, which is propositionally incomplete.
B. The logical form, which is a structured sequence of concepts, is enriched and transformed into a propositional form: the complete propositional
form is the result of the activation of both conceptual contents and procedural contents of the concepts in the logical form of the utterance.
C.The propositional form together with contextual informations yields cognitive (contextual) effects. These contextual effects, when they are the result of the cunjunction of contextual and linguistic informations are implicatures of the utterance. The propositional form, as well as the propositional attitude of the speaker and the illocutionary force of the utterance, are the explicatures of the utterance, that is the informations which are a development of linguistic informations without being the result of inferential processes; on the other hand, the implicatures of the utterance are the result of non demonstrative inference, of a deductive nature.
Determining aspectual class belongs, among other things, to the explicatures of the utterance. Thus, understanding an utterance means being able to determine
(a) its explicatures, notably its aspectual class;
(b) its implicatures.
Now, what are the factors which contribute to explicature and implicature determination in utterances? To answer this question, we will compare the two short dicourses below:
(29) Max a poussé Jean. Il est tombé.
(Max pushed John. He fell) (30) Jean est tombé. Max l’a poussé.
(John fell. Max pushed him)
The correct interpretations are respectively:
(31) e1(PAST(PUSH(Max, John)) TEMPORALLY PRECEEDES e2(PAST(FALL(John)) & e1 CAUSES e2.
(32) e1(PAST(Fall(John)) TEMPORALLY FOLLOWS e2(PAST(PUSH(Max, John)) & e2 CAUSES e1.
In other words, the interpretations are
1. the forward temporal (or narrative) reading, in which the first event described in discourse temporally preceedes and causes the second event;
2. the backward causal (or explicative) reading, in which the second event described in discourse temporally preceedes and causes the first event.
These two reading are symetrical. Yet, it must be explained why the forward temporal reading is not possible in (30) and why the backward causal reading is not possible in (29). I will propose a solution which gives an important role to conceptual, procedural and contextual informations. Let us examine each of these two readings:
A.The forward temporal reading of (29) is the easiest to access because it contains conceptual informations and procedural informations which converge toward the same solution. Indeed, concepts associated with predicates tomber (to fall) and pousser (to push) are linked by a causal rule:
(x pushes y) CAUSES (y falls). Then, according to the principle which says that causes temporally preceede effects, the temporal reading is obtained through conceptual informations only. This reading is confirmed by procedural information as well. We begin by preferentially infering, from the order in which events are described, the temporal reading, then, the past perfect triggers, by default, a temporal reading. These two procedural informations perfectly combine with the conceptual information associated with pousser (to push) and tomber (to fall): the forward temporal reading is consistant and activated. Thus the representation accessed by the hearer is the one in which the two events have the relationships of temporal order and causality.
B. The backward causal reading of (30) is the most immediately accessible one. But here, the informations given by conceptual contents contradict those given by procedural contents: for the backward causal reading to be the case, we must go back against the direction of discourse. This means that conceptual informations are stronger than procedural informations and that when there is conflict they dominate.
But it only needs a little attention to see that these two readings, which we will call the preferential readings, are not alone: (30) can have a forward
temporal reading though, admittedly, it is less accessible than its backward reading, and even (29) can have a backward causal reading. How can we explain that?
C.the forward temporal reading of (30) implies that conceptual informations are nullified: in other words, the causal rule push-fall does not apply. If it does not, then procedural informations given by discourse and verbal tenses do. But how can the causal rule be blocked? Let us consider example (30’) which makes explicit another temporal and causal relationship:
(30’) Jean est tombé. Ensuite, Max l’a poussé jusqu’à un banc.
(John fell. Then, Max pushed him toward a garden seat)
If (30’) describes a situation in which John fell because he stumbled on a root and then Max pushed him to allow him to rest and be out of the way while he was going for help, then the temporal reading will dominate.
This has an important implication: contextual implications are what validates conceptual or procedural informations. In (30), accessible contextual informations validate conceptual informations, in (30’) they validate procedural informations.
D. What about the backward reading of (29)? Is it possible? I would say yes, if and only if (29) describes the situation explicitly described in (29’):
(29) Max a poussé Jean jusqu’au chemin. Il est tombé en se prenant les pieds dans une racine.
(Max pushed John toward the path. He fell by stumbling on a root)
Clearly the second event explains the first one. This reading is possible here because, though procedural informations are contradicted, informations given in the utterance block the causal rule push-fall. But it should be agreed that given that all the informations, conceptual and procedural, are blocked, contextual informations must be extremely accessible for this interpretation to emerge and for the utterance to explicitely describe them, because (29) is a very bad way of representing backward causality. It should noted that (30) is not a usual way of representing temporal order either.
Thus it can be seen how conceptual and procedural informations combine:
they play different parts relative to backward or forward discourse relations and the complete interpretation of discourse depends on their combination and on their compatibility with accessible contextual informations.
I have only presented here the main lines of an integrated approach to conceptual, procedural and contextual aspects of utterances in discourse. More detailed studies, implying among other things other verbal tenses and connectives (see Moeschler 1998a and 1998b), have shown that the method is descriptively adequate: it describes the facts correctly and yields interesting predictions. Yet, and independantly of the descriptively or explicatively adequate property of the model outlined here, its main virtue lies in that it can account for discourse facts without using notions drawn from discourse analysis. We have expressed our misgivings relative to those notions elsewhere (see Reboul & Moeschler 1995, 1996 and 1998b): here, the treatment of discourse is based only on the hypotheses of a theory of utterance interpretation. Here is a path which pragmatics can follow to account for the linguistic complexity of discursive phenomena.