Вы находитесь на странице: 1из 11

ELSEVIER

Journal of Pragmatics 31 (1999) 327-337

Associative anaphora: An attempt at formalisation *


D e n i s Mi6ville
S~minaire de Logique et Centre de Recherches S~miologiques, Espace Louis-Agassiz 1, CH-2000 Neuch6tel, Switzerland

Abstract
In this article I have tried to formalize some properties of the 'associative anaphora'. For this purpose I make use of Lesniewskian logic which is, in some respects, developmental and creative. Also I have pointed out the limitations of such an enterprise. 1999 Elsevier Science B.V. All rights reserved.

1. Prologue
It might seem unusual for a logician to take an interest in anaphora. However, the study of anaphoric facts is relevant to logic insofar as some of these facts concern entire representational networks and are closely connected to issues of reference. This interest is even easier to understand if the logician is also involved in reseaerch on the analysis of the logico-discursive operations which a speaker establishes in order to schematize, for a given audience, a sens object. From the point of view of what is called 'natural logic' (Borel et al., 1983), a distinction is made between the sense that a speaker constructs, develops, specifies (i.e. the logical theme or the discourse object), and the field of exploration (the rhematic possibilities associated with it. Also, all discourse, in its own ontology, manages objects. Their textual markers are generally nominal expressions. These objects are introduced and modified as the text develops by various operations (which do not, however, concern us in this paper). To identify the discourse theme, nominal expressions are read just as they are, in their role as indicators of object forms by which things are represented. The micro-universe schematized by discourse is postulated as containing objects which are designated by nominal constructions. In addition, the introduction of a discourse object and its development mobilize preconstructed knowledge representations. It is within this framework that this study of associative anaphora has

* Translated by Jim Kenkel and James Gasser 0378-2166/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved. PII: S0378-2166(98)00071-X

328

D. Mi~ville / Journal of Pragmatics 31 (1999) 327-337

been undertaken. We recognize in particular that associative anaphora play a basic role in the specific development of a discourse object, which itself is constructed progressively and apprehended as an object class. The associative anaphor is thus perceived as the sign of a specific construction. It is the indicator of a specifying function of a fragment of a discourse object associated with the source element of the anaphor. It is also the sign of an informative function, which can be weak if the associative pair has a strong presence in our representation; it can be strong if the discourse constructs a new associative relation. But our intention is not to add here, yet another voice to the numerous investigations of associative anaphora. We will be content essentially to propose some ideas on the relation of parts to the whole, a relation frequently mentioned by many researchers interested in this type of anaphora. However, before doing so, it will be helpful to recall some facts relevant to associative anaphora.

2. Some remarks concerning associative anaphora


It seems natural to start with its definition: "In general, the expression 'associative anaphora' designates noun phrases which simultaneously present the following two characteristics: on the one hand, a certain interpretive dependence relative to a referent previously (possibly subsequently) introduced or designated; on the other hand, the absence of coreference with the expression previously (possibly subsequently) designating this referent. An associative anaphor thus presents its referent as already known, or as identifiable, although the referent has not yet been explicitly mentioned, and moreover, does not indicate its relation to other referents or to other explicitly formulated information" (Apoth61oz, 1994: 22; translated). The canonical example of this type of anaphor is given below: (1) We arrived in a village. The church was closed. In this example the definite noun phrase (NP) 'the church' is an associative anaphor since its referent (i.e., which church?) cannot be identified solely through recourse to a referent introduced earlier by the indefinite NP 'a village' (i.e. the church of the village). Various explanations of the anaphoric mechanism have been proposed, some of which are based on quite different epistemological foundations. It is acknowledged however that the mechanism of associative anaphora rests on preconstructed knowledge representations, which, it is supposed, are shared by interlocutors (Hawkins, 1978). It is also recognized that "associative provides access to an extremely vast range of relations. Each link between facts, events, and objects, and each form of referential contiguity increases the likelihood of associative anaphora in discourse. Moreover, associative anaphora constitutes an essential process of amplification. The coherence of this amplification is more or less guaranteed by the existence in our representations of stereotyped associations, some of which are, moreover,

D. Midville / Journal of Pragmatics 31 (1999) 327-337

329

recorded in the lexicon. But discourse also generates (...) its own associative framework by presenting totally new temporary relations, which have only occasional validity. Their recognition, however, is nonetheless imposed on the hearer by virtue of the presumption of coherence wich presides over its interpretation" (Charolles, 1990:132-133; translated). These relations can be represented by the 'ingredient' relation (Berendonner, 1994) or the relation of parts to the whole (Kleiber, 1998; Charolles, 1998). However, this relation is doubly limited: 1. The relation implemented by an associative anaphor is restricted solely to the necessary or stereotypical ingredients of the entity denoted by the source element of the anaphor. 2. This relation is not transitive. The following two counterexamples confirm these limitations: (2) *We arrived in a village. The candelabras of the altar were remarkable only for being absent. The relation between 'the candelabras of the altar' and 'a village' is not acceptable. The anaphoric relation in the following example is also unacceptable. (3) *We came into a village. The stonework, of Roman origin, supported an entire structure. This limitation, tied solely to necessary or stereotypical ingredients, can be overridden through the introduction of intermediary NP's denoting fragments of the referential object, which can then support a stratified transitivity. (4) We arrived in a village. The church was open. The candelabras of the altar were reflected in the gold of the cathedral. (5) We came into a village. The walls revealed a rich past. The stonework, of Roman origin, supported an entire structure. On the basis of these few remarks, some proposals relative to the relation of parts to the whole as exploited by associative anaphora will be developed.
3. Some reflections on the relation of parts to the whole

As mentioned before, various writers refer to a relation of parts to the whole to explain the relation between the entity denoted by the associative anaphor and the entity designated by the source element of this anaphor. The clearest way to explore this relation is to propose axiomatically a basic meaning that underlies it. And, inasmuch as researchers interested in associative anaphora frequently cite Lesniewski (1992), it is from this writer that the basic meaning of this relation will be borrowed:

330

D. Midville / Journal of Pragmatics 31 (1999) 327-337

AXIOM I: If P is a part of object Q, then Q is not a part of object P. [ (VP)(VQ)(P e pt (Q) D -(Q E pt (P))) ] AXIOM II: If P is a part of object Q, and Q is a part of object R, then P is a part of object R (Lesniewski, 1992: 230). I (VP)(VQ)(VR)((P e pt (Q) A Q e pt (R)) ~ P e pt (R)) ] The study of this axiomatic base immediately reveals that the relation 'part of' possesses the properties of irreflexivity (an object is never part of itself), of asymmetry and of transitivity. The property of transitivity, however, is not appropriate for explaining the particular relation which interests us. To do so, another meaning than 'part of' must be found. The relation 'ingredient of' appears to be a good candidate inasmuch as it is mentioned by various experts on associative anaphora. My conception of this relation, like that of 'part of', also derives from Lesniewski, and for the same reasons as his. It is based on the primitive relation 'part of'. DEFINITION I: P is an ingredient of object Q when and only when, P is the same object as Q or is a part of object Q (Lesniewski, 1992: 230). I (VP)(VQ)(P e ing(Q) = (P = Q v P e pt (Q))) This relation is characterized by the following properties. It is reflexive, antisymmetrical (if P is an ingredient of an object Q and Q is an ingredient of an object P, then P and Q denote the same object), and transitive. Once again, however, these properties are not sufficient. For reflexivity leads automatically to the possibility of recognizing in the definition of associative anaphora the presence of coreference with the expression which has previously (possibly subsequently) introduced or designated the referent. (cf. Apotheloz' definition given above). As for the property of transitivity, it is exactly what should be avoided. These two relations borrowed from Lesniewski, therefore, are not appropriate to represent the relations specific to associative anaphora. It is necessary, therefore, to proceed in another way. First, those properties which seem relevant to our project need to be grouped together. We can assume that it will suffice to define a relation which possesses the properties of irreflexibility, asymmetry, and non-transitivity. But such a definition would fundamentally contradict any conception of the crucial notion of the relation of parts to the whole in that it moves away from intuitions related to the notions of fragment, agglomerates, and aggregates. Such a definition would no longer be related to the notion of collections. It is necessary then to adopt another strategy: while retaining the specific definition of a transitive relation of parts (or ingredients) to the whole, we propose the definition of a second relation of the type 'a privileged part of'. This non-transitive relation applies only to those necessary or stereotypical ingredients of the entity denoted by the source element of the associative anaphor.

D. Mi~ville / Journal of Pragmatics 31 (1999) 327-337

331

However, this option immediately raises four problems. The first has to do with the syntactic definition of such a relation. The second forces us to consider the problem of selecting what are said to be necessary or stereotypical ingredients. The third problem makes clear the necessity of defining the structure of an object which is continuously broken down to its parts. Finally, any approach which takes on these ontological problems is directly concerned with philosophical problems. Consider first the syntactic dimension by proposing the following definition of a possible relation of 'privileged parts to the whole': DEFINITION II: P is a part* of object Q when and only when, P is part of object Q and there is no object R such that P is part of object R and R is part of object Q. f (VP)(VQ)(P ~ pt* (Q) -- ~ (3R) (P ~ pt (R)A R ~ pt (Q)) ] This definition of the relation of privileged parts to the whole (part* of) is neither reflexive nor symmetrical, Moreover, it is not transitive. This means that if P is 'part* of' object Q and Q is 'part* of' object R, then P is not 'part* of' object R. It allows us to establish a relation with a fragment of an object while accounting for a certain immediacy. It can thus be interpreted as a definition which covers the idea of placing an object in relation with its necessary ingredients. Thus, in surveying an object collectively, two relations would be available: 'part of', which has properties indispensable to any partitive approach, and 'privileged part of', which is intransitive, and helps, syntactically speaking, to place a stereotyped fragment in relation to the whole. This double choice thus permits us to approach and understand any complex object in collective terms, on the one hand, and, on the other hand, to establish relations between privileged stereotyped fragments and the object itself. The semantic dimension remains to be studied, but will require a specific inquiry into the organization of the lexicon as well as psycholinguistic research into the nature of representations. This approach is indispensable in order to make it possible to attribute the privileged-stereotyped qualifier to a collective fragment of an object. The philosophical problems underlying this collective approach, will not be discussed here; instead the structure of the object, certain parts of which are crucial, will be considered. 4. Objects and the collective class So far, much has been said of partitive (or collective) relations, but little of the way the structure of the object to which these relations apply is apprehended. Traditionally, two systems of reference are mentioned in connection with the formal study of the relation of parts to the whole: Leonard and Goodman's calculus of individuals (1940) and Lesniewski's mereology (1993 [1916]). In what follows, only Lesniewski's work will be considered. Within this perspective, each object investigated in terms of parts and whole is considered as a collective class, Le. one which can be defined on the basis of the meanings of parts or ingredients previously men-

332

D. Mi~ville / Journal of Pragmatics 31 (1999) 327-337

tioned. These meanings are established in terms of a specific definition of class, on the one hand, and, on the other hand, by two new axioms which complete the axiomatic basis of Lesniewski's theory of parts to the whole. DEFINITION III: P is the class of objects a, when and only when the following conditions are fulfilled: (a) P is an object; (b) every a is an ingredient of object P; (c) for any Q, if Q is an ingredient of object P, then some ingredient of object Q is an ingredient of some a (Lesniewski, 1992: 230-231). (Va)(P e cl (a) = (3Q) (Q e a) (VR)(R ~ a D R e ing(P)) (VM) (M e ing(P) D (3D) (3E) (D e a A E e ing (D) A E e ing (B)))) AXIOM III: If P is the class of objects a, and Q is the class of objects a, then P is Q.

I (VP) (VQ) (Va) ((P e cl (a) A Q ~ cl (a)) D P e Q) I


AXIOM IV. If some object is a, then some object is the class of object a (Lesniewski, 1992:232).

[ (VP) (Va) (P e a D (3Q) (Q ~ cl (a))) I


This definition is sufficiently detailed for us to stop and consider it a moment. Before presenting the properties that characterize this collective class, it is necessary to give precise instructions for the singular propositions of the form 'A is b' appearing in the preceding definitions. Such a proposition is true if A designates exactly one individual and is one of the one or more individuals designated by b. One may read this copula as 'is the' or 'is a', 'is one of' or 'is among'. These readings are only an approximation to the formal meaning. If something is a collective class, it is considered as an individual entity. This class is accessible via elements which allow us to explore it collectively. Thus if P is the class of objects a, that means that the object P, conceived as a collective class, is explorable on the basis of the objects a. The objects a are considered as ingredients of the object P, but contrary to the distributive point of view they are not the only objects that make up P. Although they are indeed ingredients of P, they allow access to any other ingredient without thereby accepting anything whatsoever. The following example will clarify this interpretation of collective class. The four triangles, ADE, AEB, BEC, and CEF are ingredients of the class P, but they also contribute to the determination of other ingredients; for example, the parallelepipeds ADEB and BEFC and the trapezium ADEC are ingredients of the object P (therefore, P is considered to be an ingredient of itself). Moreover, the point C, the line segment EF, the point A taken together with the segment CF, and other

D. Mi~ville / Journal of Pragmatics 31 (1999) 327-337 A B C

333

E Fig.l.

things as well are ingredients of the class of triangles given in Fig. 1. It is not possible, however, to include just anything in this class. The point X does not belong in it. Thus, the elements (or element) in terms of which the collective class can be explored can be interpreted as the respect in which the speaker has chosen to present an object conceived in the collective sense. With respect to the triangles of Fig. 1, we take into account ingredients of a different sort than that of the object in which they are found. In addition, from a collective point of view, we can speak of the same object, and therefore, of these same ingredients, while treating something altogether different. It is entirely possible to account for the same ingredients of Fig. 1 in the following way: let P be the class of parallelepipeds of Fig. 1, or even the class of the parallelepiped of Fig. 1. We can therefore collectively comprehend the same object in several different ways. What follows is a summary of notions that could contribute to a formal representation of the referential organization tied to an associative anaphor.

5. An attempt at formalization
The notion of collective class gathers together as ingredients everything having to do with the aggregate, the agglomerate, the fragment, or the individualized element. The definition of ingredient [ing] is constructed on the primitive meaning of 'part of' [pt]. The definition of 'privileged part' [pt*] explains a certain immediacy between a fragment of an object and the object itself. It can be interpreted as that which includes the idea of establishing a relation between an object and its necessary ingredients. The accessibility of the ingredients of a collective class, and therefore, to the collective class itself, is always presented by means of specified ingredients. These ingredients specify the respect in which the object is collectively conceived. In what follows, the various notions developed here will be used to represent what I consider to be the referential organization linked to an associative anaphor, i.e. the relation defined by the referent of the associative anaphor and the referent of its source

334

D. Midville / Journal of Pragmatics 31 (1999) 327-337

element. I will postulate that the referent of the associative anaphor is a privileged, stereotypical, and necessary ingredient of the entity denoted by the source element of the associative anaphor. In addition, any referent of the associative anaphor, whatever it be, is subsumed under a particular aspect. Under these conditions and with the abbreviations given below, the referential organization linked to an associative anaphor can be defined as follows. P = df. the referent of the source element of the associative anaphor Q = df. the referent of the associative anaphor a = df. the name of the aspect under which the referent of the associative anaphor is subsumed r = df. the relation between P and Q (VP) (VQ) (r(PQ) (3a) (P ~ cl (a) A Ql~a A Q I~ pt* (P))) The relation between P and Q is anaphorically based, if and only if There is an aspect under which P is collectively understood and Q is one of the a's and Q is a privileged part of P Now apply this definition to the canonical example (1): the relation r between the referent of the source element of the associative anaphor, i.e., the village, and the referent of the associative anaphor, i.e. the church, is anaphorically based if and only if: There exists an aspect in relation to which this village is collectively conceived (for example, the class of its architectural areas), The church is one of these areas, The church is a privileged (stereotypical, and necessary) part of this village. This manner of perceiving things also permits the definition of the relation linked to what I called a stratified transitivity when I was considering examples of the type such as (4) above. While it is true that relative to the referential dimension, in our culture, the church is recognized as a privileged part of a village and that the collective entity - the candelabras of the altar - is a privileged part of the church, it cannot be accepted that the candelabras of the altar, while collectively considered as part of a village, are also a privileged part of a village. The relation 'part of' is tran-

D. Mi~ville / Journal of Pragmatics 31 (1999) 327-337

335

sitive, while the relation 'privileged part of' is not. The definition of the ternary relation t, given below, allows us to respect these conditions. The relation between P, Q, and R is anaphorically based if and only if: There is an aspect a under which P is collectively subsumed and Q is one of the a's and Q is a privileged part of P; and There is an aspect b under which Q is collectively subsumed. and R is one of the b's and R is a privileged part of Q. (VP) (VQ) (VR) (t(PQR) -= (9a) (P ~ cl (a) A Qea A (3b) (Q ~ cl (b) A Reb A

R E pt* (Q)))) 6. Epilogue


This attempt at formalization must be considered with some precautions. Each relation and each operation defined in the framework of a formal language derive from two levels of analysis. First of all, there is the structural level on which the relations and operations are situated, independently of the nature of the objects to which they can be applied. They have meaning only through the properties which caracterize them. But every formal language is conceived and fashioned according to its intended use. Its syntax and organization vary according to the universe to be represented and reasoning related to it. Therefore, it is necessary to interpret the language by associating it with a model. This model is not 'reality', but rather a representation of it, a representation of the investigated universe, necessarily incomplete and biased. This paper is concerned with the second level of anlaysis, a level which investigates, on the one hand, how a representation is attributed to the formal objects of the language, and, on the other, on the relative conformity existing between the constructed meaning and that which derives from an analysis of the investigated universe. The collective or mereological approach which

336

D. Mi~ville/Journal of Pragmatics 31 (1999) 327-337

has been advocated throughout this article even though it has been represented within a pluri-extensional perspective. This strategy makes it necessary to reflect on the existence of these collective individuals made up of sometimes discrete entities. A delicate choice has been made in defining the relation 'privileged part of', in that it highlights a salient part in relation to a whole. It would be illusory to think that in reality there does not exist a part which is intermediate between this salient part and the whole itself. Therefore, it is necessary to attach a functional dimension to the aforementioned relation, i.e. the relation which accounts, very locally, for a salience effect of the entity denoted by the associative anaphor relative to the entity designated by its source element, this source element appearing as the background which leads inferentially to what the anaphor denotes. The relation thus envisioned participates in the particularization function of a fragment of an object, a particularization which is both local and temporary. The extensional approach in no way considers the qualitative and intensional character of a relation. It would be interesting to introduce this dimension into the study of the underlying relations of the referential organization of an associative analysis. To encourage a discussion, I propose a rather relative and non-exhaustive typology of relations between a referential entity and its fragments or other objects, relations which I place arbitrarily in the referential mold of an associative anaphor (a-, the-). A first category of relations derives from the set of aspects to which we a referential entity can be attributed. 1. A referential entity in relation with its most internal fragments, i.e. those which participate in its identity, but which are beyond a distributive conception (A forest... Undergrowth ...) 2. A referential entity in relation with a change in its state, a process which concerns the notion of the referential entity of which it is the basis, a process which does not require the existence of an exterior agent (A forest... The growth of the forest...) 3. A referential entity in relation with an entity which underlies the existence of one or more subjects in relation to which the object is situated of qualified (A forest .... The importance of photosynthesis...) 4. Distributive notions in relation, notions which concern quantification (A forest... Spruce trees...) A second category of relations derives from the domain which the referential entity can share with other entities. 5. Relations with a possible contiguous object (A forest ... Bacteriological research ...) 6. A relation with a process which requires an exterior agent, marked or not marked (A forest... The exploitation of timber...) 7. A metaphorical relation (A forest ... Green thoughts ...) 8. A qualitative type relation relative to a certain point of view (A forest ... The trees which made Rousseau dream ....

D. Midville / Journal of Pragmatics 31 (1999) 327-337

337

These examples are somewhat provocate, but the fact remains that a study of the types of possible relations linked to an associative anaphor needs to be pursued systematically. A n d it would also be interesting to study the effects, the functions, and even the deviations that an associative anaphor establishes with the discourse in which it appears.

References
Apothrloz, Denis, 1994. R61e et fonctionnement de l'anaphore dans la dynamique textuelle. Genrve: Librairie Droz. Berrendonner, Alain, 1994. Anaphores confuses et objets indiscrets. In: C. Schnedecker, M. Charolles, G. Kleiber and L. Ujma, eds., 209-232. Borel, Marie-Jeanne,Jean-Blaise Grize and Denis Mi~ville, 1983. Essai de logique naturelle. Berne: Lang. Charolles, Michel, 1990. L'anaphore associative: Probl~mes de drlimitation. Verbum 13(3): 119-148 Charolles, Michel, 1998. Associative anaphora and its interpretation. Journal of Pragmatics 31 : 311-327. Hawkins, John A., 1978. Definiteness and indefiniteness: A study in reference and grammatical prediction. London: Croom Helm. Kleiber, Georges, 1994. Anaphore associative, ant~crdent et d~finitude.In: C. Schnedecker, M. Charolles, G. Kleiber and L. Ujma, eds., 153-172. Kleiber, Georges, 1998. Associative anaphora and part-whole relationship: The condition of alienation and the principle of ontological congruence.Joumal of Pragmatics 31: 339-362. Leonard and Goodman, 1940. The calculus of individuals and its uses. Journal of Symbolic Logic 5: 45-55. Lesniewski, Stanislaw, 1992. Collected works, Vol. 1. Dordrecht: Nijhof. Mirville, Denis, 1984. Un drveloppement des syst~mes logiques de Stanislaw Lesniewski: Protothrtique-ontologie-mrrrologie. Berne: Lang. Schnedecker, C., M. Charolles, G. Kleiber and L. Ujma, eds., 1994. L'anaphore associative: Aspects linguistiques, psycholinguistiques et automatiques. Paris: Klincksieck.

Professor Denis Mirville lectures on formal logic at the University of Neuch~tel, Switzerland. He is also director of the 'Centre de Recherches Semiologiques' of the Faculty of Arts. His work belongs to the field of Lesniewskian logic and natural logic.

Вам также может понравиться