274 106 2MB
English Pages 235 [243] Year 2009
Bruno Langlet • Jean-Maurice Monnoyer (Eds.) Gustav Bergmann Phenomenological Realism and Dialectical Ontology
Philosophische Analyse Philosophical Analysis Herausgegeben von / Edited by Herbert Hochberg • Rafael Hüntelmann • Christian Kanzian Richard Schantz • Erwin Tegtmeier Band 29 / Volume 29
Bruno Langlet • Jean-Maurice Monnoyer (Eds.)
Gustav Bergmann Phenomenological Realism and Dialectical Ontology
Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 [email protected] United Kingdom, Ireland, Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]
Livraison pour la France et la Belgique: Librairie Philosophique J.Vrin 6, place de la Sorbonne; F-75005 PARIS Tel. +33 (0)1 43 54 03 47; Fax +33 (0)1 43 54 48 18 www.vrin.fr
2009 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 978-3-86838-35-4 2009 No part of this book may be reproduced, stored in retrieval systems or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use of the purchaser of the work Printed on acid-free paper FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany by buch bücher dd ag
G. BERGMANN : PHENOMENOLOGICAL REALISM AND DIALECTICAL ONTOLOGY Table of Contents INTRODUCTION
1 - ERWIN TEIGTMEIER Bergmann on Brentano
7
2 - FRED WILSON Bergmann’s Hidden Aristotelianism
17
3 - GUIDO BONINO Space, time, concrete, abstract
69
4 - JEAN-BAPTISTE RAUZY Sellars et Bergmann lecteurs de Leibniz. La querelle des particuliers
87
5 - L. NATHAN OAKLANDER A critique of Degree Presentism
103
6 - JIRI BENOVSKY La théorie des faisceaux et la théorie des substrats
117
7 - LUCA ANGELONE & GIULIANO TORRENGO Bare particulars and persistence in Bergmann
139
8 - FREDERIC NEF Bergmann et l’ontologie de la connexion
157
9 - E. JONATHAN LOWE Modes of exemplification
173
10 - LUC SCHNEIDER On Ties and Copulae within the Ontological Square
193
11 - BRUNO LANGLET & JEAN-MAURICE MONNOYER Gustav Bergmann et les complexions meinongiennes
209
Introduction
The essays collected in this volume were read at the Gustav Bergmann (1906-1987) International colloquium, held in Aix-en-provence, from December 9th to December 11th 2006, for the centennial anniversary of Bergmann’s birth. The conference was organized within the framework of the Séminaire de Métaphysique (SEMa), which is hosted by the IHP (Institut d’Histoire de la Philosophie). The (non-official) purpose was to promote some liberal exchanges and debates between some Bergmann’s interpreters (American and European) focusing on the “actuality” of his thought ; the official one was to greet the first meeting devoted to this philosopher in France, where he is still widely little-known. We publish some of these contributions in French, with the hope of a better understanding of the great influence his philosophy should exert in the next future. Bergmann’s situation within the history of philosophy is somehow difficult to synthesize, as there is always a risk to create a damaging twist of his evolutive research. This major philosopher bears witness of many ruptures spreading out through the last century, with “phenomenalism”, “physicalism” or with “representationalism” (still very much alive today). He never ceased to reflect on logic and ontology, and to think about their correct relationship, complicating this task with the hypothetical necessity of linking the assay of entities with honest phenomenology. Thus, he came to present a kind of methodology of the revelation of “uncomplete ontologies” with their own pressing problems. Here lies the significance of the so-called truncated world denounced in Realism : a strange world really without minds or without intentional relations, a world in bits, which would be carved in fictitious unities by the resort of our anthropomorphic projections. Nowadays, it is in substance this very world his innovating thinking tried to exorcize, at the level of sociology of knowledge, as well as at the level of epistemology and philosophy of language. From Bergmann we often keep in mind the revival of the doctrine of universals, and
his characterization of bare particulars. Yet, by contrast, he will remain in reality the man of complexities that cannot be ultimately reduced — these of “facts”, of “circumstances” and “classes” — and also most likely as the thought-upsetting last hero of some drastic approach of fundamental ontology (although this one is theoretically and paradoxically concerned at first with the simplest entities). Gustav Bergmann has never ceased to place himself sidelong and to confront directly the critical core of the American philosophy, which he identified and recognized very well under the canonical form it was to take from the end of the Fifties. He was also the intractable interlocutor of Carnap and the cutting critic of Goodman and Strawson. It would be unfair to say that his own difficult, sometimes delphian, and always inventing way of writing may explain he was often misunderstood. Someone once claimed clarity is not enough : the word fits him as a glove. Bergmann is not only the best chronicler of the « linguistic turn », but above all the sworn enemy of that « linguistic idealism », leading to the semantic nominalism by the way of a kind of subreption (his term) of syntax into semantics which Bergmann used to condemn and, what’s more surprising, which he rejected quite early in his philosophical career. To put it somewhat bluntly, Bergmann sees « mean » as a predicate, in the classical sense. This rules a « pseudo-relation » (a logic relation). Ontology doesn’t have to contract itself in the concealed form of an intensional discourse, or as it is sometimes asked for today, in a « metaontological » form, nor to stand up in a haughty stance without any consideration of facts. In short, Bergmann’s philosophical program is thoroughly “analytical”, and it gave a beneficial updating of the scientific philosophy of the XXth century, by limiting its abusive pretentions. Without tempting to reconstruct the critical and “rational” history of his exceptional record, which began with the first important articles published between 1941 and 1947, we choose to scrutinize the posterity of some of Bergmann’s intuitions. They renew the topics of a “positive” metaphysics and modify the false understanding of the Viennese Legacy, from which metaphysics is righting itself, thanks precisely to Bergmann, and his extensive reading of Brentano and Meinong, and even though it seemed to be definitely condemned by Carnap himself.
2
A doctor in mathematics (1928), and a latecomer to philosophy, G. Bergmann was from his school-years a close friend to Gödel, with him a member of the Vienna circle, and more later in charge of teaching after emigration (1938), as the Kurt Lewin’s assistant in psychology, in IowaCity, but he progressively changed his mind in a profound way, with the beginning of his work in the USA. Even though he might be judged today as a renegade or an apostate, Bergmann belongs without any doubt to the history of philosophy of the XXth century, and he produced also a very interesting historical contribution in itself. He studied, for instance, the raising of individual accidents in Occam’s work, what we call nowadays “tropes”. His first strong recommendation was the reading of Russell’s Principia Mathematica : he never let go his conception of an “Ideal Language” though, for him, the decisive point would be to know how finite classes are to be represented in this very same language, and which psychology has to escort our definite descriptions when we have lost acquaintance with the world. The second step, according to him, is marked by the Quine’s Mathematical logic, which he would regard as fixing some kind of orthodoxy, and its too strong and enduring influence on so many philosophers. For this reason, from 1967, as against the prevalence of a diffuse nominalism, Bergmann intended to come back to a “realistic” position, finding part of his inspiration — and at the same time relieving himself — from the legacy of Brentano and Meinong. In his unfinished last work, published in 1992 : New Foundations of Ontology, clearly presented in this book by F. Nef, Bergmann goes however ahead, as he endeavours to radicalize a part of the ideas of his Viennese masters. In retrospection, we may be allowed to think that he was the unwittingly precursor of so called “meinongian logics”. But here again stands another kind of two-sided key to assay the richness of Bergmann’s work : on the one hand, Bergmann’s intuitions justify to go back to the insights of Brentano and Meinong ; in the other sense, these latter can be seen as providing useful entrances into the issues he was to work, and by which, as a happy consequence, Austrian classical themes were to be rediscovered, and debated anew. Previously, the publication of The Metaphysics of Logical Positivism in 1953 disturbed the academic consent. Then, his numerous papers kept challenging the best rivals he could be offered in the American scene, like
3
Goodman and Sellars. His masterwork, Realism – A critique of Brentano and Meinong (1967), was received with suspicion. Rehabilitation came first from Herbert Hochberg, with a true verve, and indirectly from the diversified and suggestive works of some of its students, such as Reinhardt Grossmann, Fred Wilson and Nathan Oaklander - and then, somewhat later, from the young Italian school of metaphysics in the early Nineties, the undisputable talent of which should be acknowledged here. The issue of the Rivista di Estetica (n° 25, anno XLIV) which came out in 2004 (“Il realismo ontologico di Gustav Bergmann”, Rosenberg & Sellier, Torino) gave a first echo of his reception in Europe and a useful bibliography of his works. The essays in this volume are very diverse in nature, and cover a wide open range of topics in relation with bergmannian themes, problems and influence, both external and internal to his work. They are still only drawing the active flux which is impulsing the real and posthumous devenir of the Bergmann’s philosophy : around metaphysics of time, of existence, of concreteness and abstraction ; or about very old and technical ontological problems of substrata and particulars ; others essays focused on “connexion” and “exemplification” puzzles, but also in an appropriately related way around Sellars, Aristotle, Brentano and Meinong, as a mark of the freshness of a new wind blowing on the truncated world. Erwin Tegtmeier, who gave one of the best and briefest accounts on the relationship between Meinong and Bergmann and who directed the edition of the Collected works, was the instigator of this colloquium. He deserves our special gratitude. Two volumes of proceedings concerning Bergmann’s philosophy have been already published by Ontos Verlag : Ontology and Analysis (2006, Edited by L. Addis, G. Jesson and E. Teigtmeier) and Fostering the ontological turn (2008, Edited by R. Egidi and G. Bonino). This third volume provides a complement to these noteworthy, substantial and well-directed studies. The Aix-en-Provence conference happily ended with the founding of the Gustav Bergmann Society (Presidence : F. Wilson ; Secretary : L. Schneider), which will very soon have opportunity to reunite the members in order to draw up the balance sheet of prospective inquiries that Bergmann’s thinking leaves for now on
4
borrowed time, in several domains (set theory, cognitive phenomenology, logic of perception). We wish to thank the IHP1 (Institut d’histoire de la philosophie, EA 3276) and his director Alonso Tordesillas, and the Mission Recherche of the Université de Provence, for their financial support and assistance. We are greatly indebted to Pr. R. Hüntelmann for welcoming these proceedings in this collection. B. Langlet and J.-M. Monnoyer
1
We thank also Lynda Gaudemard, whose help was decisive for the organisation before and during the colloquium (including the translation of numbers of drafts and papers for the french assistance), and Diego Covu, who translated the pre-paper of E. Tegtmeier.
5
Bergmann on Brentano ERWIN TEGTMEIER
Bergmann's major work Realism bears the subtitle “A Critique of Brentano and Meinong”. That suggests that the whole book is oriented to that aim, even the first two parts titled General Ontology and Representationalism. The third part is especially about Brentano. Bergmann singles out Brentano and Meinong for systematic and detailed criticism because he appreciates both philosophers highly and shares with them the primacy of ontology and the tendency to epistemological realism.
In the first part of realism Bergmann develops a division of all ontologies into reist or thing ontologies, and fact or complex ontologies. Reist ontologies have to reject in addition to facts also external relations and universals. In the second part, he shows that reist ontologies because they reject universals, external relations and facts lead to representationalism and ultimatly to idealism. Part III of Realism reconstructs Brentano's ontology and epistemology mainly with the question: did Brentano succeed in overcoming representationalism and in arriving at epistemological realism? The answer can be easily anticipated since Bergmann characterises Brentano as an extreme reist. Nevertheless, the exact reasons for the negative answer are very illuminating just because Brentano was, also in Bergmann's judgment, a great philosopher, a first-class ontologist. There is no need to justify the classification of Brentano as a reist. It is obvious. The term "reism" has been coined for Brentano's philosophy and Bergmann extended it to the whole Aristotelian tradition, even to all traditional ontology. Bergmann implies that Brentano saw the gist of the traditional ontology more than anyone else. He was convinced that it is an ontology of simple objects. Bergmann emphasizes to what incredible degree Brentano succeeds in avoiding complex entities. However, as was mentioned, Bergmann calls Brentano an extreme reist. He does not consider him a typical reist. The typical reist is reconstructed in Realism as a gamma-ontologist and it turns out that Brentano is not a proper gamma on-
tologist. Meinong is more of it. Pivotal in Brentano's ontology is not gamma (a function mistaken for a nexus, as Bergmann characterises it) but rather unilateral separability. From any physical object a space time cell, as Bergmann calls it, is separable since the object could move and leave the cell. Brentano was an absolutist with respect to space and time and accepted empty places and time points. Now, the space-time cell of a physical object is its individuator and contains its substantial determinations. The latter shows, as Bergmann emphasizes, that Brentano is not an Aristotelian with respect to substance and essence. That unilateral separability plays are role similar to gamma means that Brentano does explain by it the property ownership and the standing in relations of the objects. These two phenomena usually create complexity even if only one relation or property is taken into account. But in Brentano's ontology they do not. That is because there is a unilateral but no bilateral separability. If the space-time cell is separated nothing is left. There is no second part of the object which can go together with it to form a complex. That is Brentano's ingenious move which makes his strict reism possible. In part I Bergmann points out that a reist ontology is patently inadequate. He argues that an ontology without facts cannot explain ontologically what is explained in the epsilon ontologies (the alternative to the gamma ontologies) by facts. And he maintains that ontologies without facts usually have at least near-facts. Bergmann thinks that unilateral separability and the relation derived from it, namely modal inclusion, form Brentano's near-facts. They can be described only as near facts because there is no category of facts in Brentano's ontology (the later Brentano vehemently rejected facts) and modal inclusion is assumed not to create complexes consisting of space-time cells and objects. In view of the role of space-time cells as individuators and substantial determinations Bergmann notes that Brentano was a great Aristotle scholar but not an Aristotelian ontologist. He also refers to Brentano's rejection of Aristotle's form-matter doctrine and of Aquinas' composition of objects of essence and existence. The role played by space-time cells is similar to that of a traditional substance and Brentano even calls it “substance” but it is not an Aristotelian substance at all since it is momentary and not an essence. The phenomena with respect to which Brentano continues the Aristo-
8
telian view is that of relations. His relations are internal to their relata. Bergmann distinguishes two versions of the theory of internal relations, the moderate and the extremist version. According to the former a two-place relation is based on both relata and according to the latter it is based on one of the relata, namely on the first relatum, which was termed “fundamentum” in the tradition. Bergmann describes the development of Brentano's theory of relations which leads to an exclusion of Aristotle's subclass of relations of comparison from the category of relations. The later Brentano claims that relations of comparison are no genuine relations and he accepts as subcategories of the category of relations only modal inclusion, wholepart relations, intentional relations, cause-effect relations and boundary relations. The latter are pivotal in Brentano's theory of space and time. Bergmann is particularly critical of Brentano's theory of relations. He himself advocates external relations and thus a position diametrically opposed to Brentano's. He argues that Brentano does not really remove the difficulties of the Aristotelian theory which Aristotle had already discovered but had left in abeyance. However, Bergmann is not only interested in scrutinizing Brentano's ontology for its own sake but also with respect to its epistemological implications. The first section is titled “The Truncated World” which means that it deals only with Brentano's ontology of the physical world. The world is in Bergmann's terminology truncated insofar as it is a world without mind. Bergmann proceeds in this way in order to apply what he calls his ground plan test (p.140). The aim of this test is to find out whether a philosopher needs additional categories to assay mind ontologically. If that is the case Bergmann concludes that the ontology has a tendency towards idealism. It is structurally idealist, as Bergmann puts it. The ground plan test demands, of course, to look next at Brentano's ontological assay of the mental. According to Brentano only what has intentionality is mental and only mental acts have intentionality. Mental acts are members of the same category as bodies. Temporal determinations are also substantial for them though not spatial determinations since the mental is not localized spatially according Brentano. There is another ontological difference between the mental and the physical in Brentano which causes Bergmann to doubt whether Brentano passes the ground plan test without reservations. The selves to which the mental acts belong are continuants,
9
i.e., they persist. In the physical worlds there are only momentary objects. Brentano needs persistent selves to solve the problem of the individuation of mental acts. Their temporal determinations are not sufficient to solve it since there are mental acts with the same temporal determinations and since they lack spatial determinations. Nevertheless, Bergmann concludes that there is a certain tendency towards idealism in Brentano's ontology. This diagnosis is supported by the hidden representationalism which Bergmann finds in Brentano's ontology of minds. Representationalism holds that we know the world by ideas or concepts. Now, Brentano explicitly rejects representationalism. He denies that there are ideas and concepts. But Bergmann insists that Brentano still continues in representationalist ways of thinking. The distinction between conceiving (that is how Bergmann translates Brentano's traditional term “Vorstellung”) and judgment, for instance, is according to Bergmann, representationalist. The German term “Vorstellung” is, of course, the synonym of the French “idée” and the English “idea”. But that is not Bergmann's point. Rather, the critical point is the interaction between conceiving and judging. The conceiving is meant to represent and present the object and the task of the judgment is to acknowledge or reject what is conceived. Bergmann points out that what is involved is the conceiving of an idea and the believing that the idea represents (or does not represent) something. This interpretation jibes well with Brentano's view that all judgments are basically existential, either positive or negative. The interpretation is also supported by Brentano's theory of evidence and its pivotal role in his theory of knowledge. Bergmann has shown in his reconstruction of representationalism that it produces the need of a criterion. Bergmann wonders why Brentano arrived at an explicit rejection of representationalism although he in many respect continues to think in its way. He offers four reasons which he attributes to Brentano. Bergmann values the first reason as good and very impressive in contrast to the three others. He connects the latter with the three predicaments he earlier made out as the historical sources of representationalism. Brentano's first reason was the insight that even if there were ideas what one is thinking when one is thinking of a horse, for example, is something physical and not something mental such as an idea. After representationalism had blurred the distinction between the mental and the physical for centuries the Early Bren-
10
tano made it one of his main task to clarify and sharpen the distinction again. Bergmann emphasizes that this insight is phenomenological and thus dependent on Brentano's new methodological approach. The first predicament is the scientific predicament which arises from the view that physics entails the subjectivity of all the qualities we perceive. Brentano realizes that those qualities cannot be ideas in the mind since they are exemplified by physical things. The problem of what to make of perceptual qualities disappears if ideas are rejected. Then perceptual qualities can be taken to be nonexistent. That allows Brentano to assume that there is only on the one side of intentionality the mental act and the other side of it the scientific object. Brentano holds that the physical objects we perceive simply do not exist. The second predicament is the skeptical one. The assumption of ideas in representationalist epistemologies was designed to make sure that there is always something we know even in cases of error or imagination. But skepticism had a new starting point by asking how we know what the ideas represent and whether they do represent anything if we know only by means of ideas. Brentano's view that the intentional relation needs not necessarily have a second relatum offers a solution of the problem of error and imagination which avoids the assumption of ideas. Hence, if ideas are rejected it seems to be the only solution. Therefore, the rejection of idea supports Brentano's view of the intentional relation. The third is the nominalistic predicament. By this Bergmann means the difficulty or even impossibility of getting along with an ontology without universals. He diagnoses ideas to be surrogates of universals and characterizes ideas as “universals in exile from reality”. Now, Brentano vehemently rejects universal and can strengthen that rejection by also rejecting ideas. The irony with Bergmann's rational explanation of Brentano's explicit anti-representationalism is that the reasons Bergmann attributes to Brentano for rejecting ideas are basically the same as those which led historically to their introduction in the first place. However, that is not unrealistic. The same complex of thought may play different and even opposing roles in different contexts. Concerning Brentano's ontological assay of the mental act Bergmann criticizes two aspects: First, that the intentionality of a mental is not guar-
11
anteed. Due to Brentano's view of relations which makes them attributes of one of the relata there is no guarantee that there is an object of the act (i.e., an existent terminus of the relation) and a connection between act and object. Bergmann contrasts his own assay with it and emphasizes that his intentional nexus is a genuine connector and there is always a fact to which the nexus connects the act, be it a fact in the mode of potentiality. Second, like the representationalists Brentano takes for granted that there is a priori or conceptual knowledge. He cannot ground this knowledge on connections among ideas as he does not acknowledge ideas. Rather, he holds that all a priori knowledge is negative, more precisely a rejective judgment which has evidence. Conceiving something both square and red causes a rejective judgment and that causal connection is also known that is what furnishes evidence to the rejection. That accounts for a priori knowledge in terms of reflexive knowledge. After bringing out Brentano's hidden representationalism Bergmann examines two issues which seem to him particularly instructive about Brentano's ontology of mind. The issue of self-knowledge and the assay of judgments on positive non-relational facts. Brentano follows a long tradition of philosophy which teaches that there is a privileged kind of knowledge. This kind of knowledge is assumed to be infallible and is considered to be the only knowledge in the strict sense. Brentano hold that all knowledge in the strict sense is self-knowledge and that only self-knowledge is evident. That is why a priori knowledge has to be reflexive. Brentano adopts Aristotle's view that each mental act in addition to whatever else it may know also knows itself. Each act has itself as a secondary intention. Bergmann has serious doubts about such multiple intentionality which continue in has discussion of double judgments in section 15. As far as selfknowledge is concerned he hold there has to always another mental to know a given mental. That implies that not all acts can be known. Brentano maintains the traditional view that a knowing subject should be aware of all its mental acts and that there must be no unconscious knowledge. Bergmann reminds the reader of the difficulties of the representationalist assay of the knowledge of facts which he had earlier in Realism pointed out. Those difficulties arose in the attempt to explain complex representatives (mental contents). Brentano admits only simple mental contents. He was able to do so because of his problematic reduction of all judgmental
12
mental acts to simple existential judgments. That was the price he had to pay for it, Bergmann notes. At the end of the section on Brentano's ontology of mind Bergmann introduces a “conversion rule”. He arrived at it by comparing numerous historic and contemporaneous ontologies which move towards idealism and it says that the poorer the ontology of the physical the richer the ontology of the mental. If, e.g., the ontology of the physical has no categories of facts and of universals the ontology of the mental will have a substitute. That suggests that certain categories are contributions of mind to the knowledge of the non-mental. In view of the conversion rule Brentano does relatively well in spite of his very frugal one-category ontology. However, when it comes to the knowledge of relations Brentano falls in with the conversion rule and moves off from epistemological realism. With respect to relations Brentano assumes what Bergmann calls double conception and double judgment. As was mentioned already, Brentano sticks to the Aristotelian view of relations which takes them to be properties of one thing with respect to another thing. How are relations and relational facts known? Brentano holds that all knowledge (i.e., all judging) is based on conceiving. If an object is to be known it has first to be conceived, then it is judged. Judging means either acknowledging or rejecting what is conceived. Now, in case of relational facts there are two objects to be conceived. Brentano is forced by his view of relations to assume that there are two conceptions involved in the knowledge of one relational fact. The two objects are conceived in different modes though, the fundamentum is conceived modo recto and the terminus is conceived modo obliquo. Bergmann considers Brentano's double conceptions and the double judgments to be incompatible with the correspondence view of truth and he calls that view the common sense core of epistemological realism. He explicates it by the condition that every knowing must have one constituent which corresponds to the fact which is known. He calls it the core of the knowing in contrast to the fringe which makes the knowing, e.g., a conceiving or a rejective judgment. That condition is fulfilled, Bergmann thinks, by Brentano's analysis of non-relational judgments as existential judgments but not by his analysis of relational judgments. There is according to Brentano's analysis not one judgment with one core but one judgment in which another inheres and therefore two cores. The non-correspondence between the cores of
13
knowings and the known is particularly obvious where the terminus object is supposed not to exist and therefore cannot be acknowledged. To support his diagnosis concerning the anti-realistic implications of Brentano's ontology of knowledge Bergmann can refer to Brentano's explicit repudiation of the correspondence theory of truth. Originally, he had adopted it from Aristotle. It must be mentioned that Brentano soon gave up the correspondence view because he understood it, like Aristotle, in terms of separation and connection in the sentence and in reality. Brentano argued that the correspondence view does not fit existential judgments where there is no connection or separation involved. Now, the correspondence view can be rendered in such a way that it accommodates existential judgments. Bergmann's explication of it does and the correspondence view surely is a core of epistemological realism. It represents the idea of the independence of the known not merely from the individual knower but also from knowing in general. Brentano himself offered his evidence theory of truth as an alternative to the correspondence view though he did not take it to be in conflict with epistemological realism. Evidence is in Brentano's epistemology a quality of certain mental acts, namely of judgments. He claims that we are presented with this quality in self-knowledge but not in other kinds of knowledge, especially not in perception. The evidence quality is not a some certainty but rather something like reasonableness and correctness. Nevertheless, as Bergmann emphasizes, the evidence theory of truth does not derive the truth of a judgment from the known facts. Brentano paved the way to this theory, as Bergmann explains, by arguing that the existence and non-existence of objects is literally nothing. The arguments are connected with Brentano's reduction of all judgments to singular existence judgments. Bergmann finds here a “poignant example of the conversion rule” (Realism, p.303). Existential facts in the physical world are substituted by existential judgments in the mental world and their truth is based on a quality of a mental entity. Taking into account how Brentano defines evidence one cannot accuse him of making truth dependent on the individual mind. However, as Bergmann points out, Brentano substitutes objectivity (correspondence with facts) by inter-subjectivity (conformity with inter-subjective rules of checking). Therefore, Bergmann also disagrees with those who praise Brentano for having established an objective ethics. Like the truth of judgments the cor-
14
rectness of moral approvings and disapprovings is based according to Brentano on the subjective quality he calls evidence. Bergmann describes Brentano's evidence theory of truth with respect to his aim of developing a realist epistemology as the second major breakdown. The first major breakdown according to Bergmann's diagnosis was the doctrine of double judgments. The third and final major breakdown in Brentano's attempt to arrive at epistemological realism, described by Bergmann, consists in his explication of existence. It was mentioned already that Brentano explicates existence basically in terms of acknowledgment, i.e., in terms of mental act. That makes existence mind-dependent which is hardly compatible with epistemological realism. The epistemological realist mainly holds that the object exists mind-independently. However, Brentano is a critical realist. He holds that the objects we perceive do not exist, but that perception is caused by objects external to mind we do not perceive. These objects are the micro-physical objects which Brentano wants to assume to exist independently of mind. Bergmann draws attention to the inconsistency in that due to Brentano's ontological analysis of existence. He also points out that Brentano cannot refer to the physical objects we perceive to make clear the notion of mind-independent existence because he takes them to be nonexistent. Bergmann's case of the third major breakdown does not seem to me as clear as he thinks. One has to take into account that the problem of existence is very subtle. Bergmann agrees with Brentano that existence is not a property. Bergmann assays existence as an entity in each existent. It is categorized as a mere subsistent and it cannot be represented in Bergmann’s ontological symbolic language. The existence of a thing is not a fact in Bergmann. Moreover, Bergmann advocates the entity existence only in his book Realism. Later he dropped it without explanation. Bergmann's later and earlier position as well as Brentano's could be expressed by saying that existence is not an entity at all. Nevertheless, both have to face the circumstance that we have the notion of existence and that it is crucial for ontology. Hence, both have to find a way of fixing and explicating that notion. Brentano's way is to refer to positive affirmative judgment in order to indicate what existence means. This seems adequate since he has only the category of objects in his ontology and since he assumes that a
15
judgment is always directed to an object. The earlier Bergmann similarly claimed that we are presented with an entity's existence if we are presented with it although he did not acknowledge existence as an entity in all entities. Thus, Bergmann's conclusion that Brentano means that existence is mental does not seem to me sound.
LITERATURE G. Bergmann: Realism. A Critique of Brentano and Meinong. Part III, in: G. Bergmann: Collected Works III. Frankfurt/Lancaster, 2004.
16
Bergmann’s Hidden Aristotelianism FRED WILSON The ontology of Gustav Bergmann contains properties and particulars and, composed of them, facts. It has facts in which a particular exemplifies a property, and facts in which several particulars exemplify a relation. These are the atomic facts. But it contains some other facts besides these. In particular, it contains general facts: besides the facts that this A is B, that A is B, and so on, and facts like this other thing which is not A is B or maybe it is not B – besides individual facts like these, there is the general fact that for everything, if it is A then it is B. That is, besides the facts Fa, Fb, Fc, ... there is also the general fact Every individual is F or, in symbols, the general fact (x)(Fx)1 1. Gustav Bergmann, “Generality and Existence,” in his Logic and Reality (Madison, WI: University of Wisconsin Press, 1964), pp. 64-84. Compare his Realism (Madison, WI.: University of Wisconsin Press, 1967): “Generality and existence are the two subsistents represented by the universal and existential operators, respectively.” (p. 94n). But the arguments for these two entities are in “Generality and Existence.” In his posthumous New Foundations of Ontology, ed. W. Heald (Madison, WI: University of Wisconsin Press, 1992), Bergmann argues that there could not be any laws of nature if generality were not in the world: “If [generality] were not in the world, there could not be any laws of nature.” He goes on to remark that “Ever since Hume, philosophers have been so absorbed by the question of how, if there are such laws, we may come to know them, that they have become oblivious of this necessary, although to be sure not sufficient, ontological ground for there being any.” (p. 173) But again, the arguments are essentially those of “Generality and Existence,” so those are what we shall concentrate on. It should be noted, however, that the analysis of generality as an entity changes to a very novel account in New Foundations. It is perhaps worth noting that Hume is not concerned just with the issue of how we know generalities to be true; as we shall see, below, he is also interested in the ontological issue. In fact, as we shall see, he is in effect arguing that generality in Bergmann’s sense does not exist. In that sense, Bergmann is right: for Hume, generality as such does not exist and there are therefore no laws of nature, there are only regularities, patterns among the particular facts that make up the world. It should perhaps be noted that for Bergmann the entities he says to be subsistents all exist: subsistence is a mode of existence. Or rather, this is so by the time of Realism. In “Generality and Existence” subsistents are there, in the world, but they are
Bergmann is here disagreeing with Wittgenstein, who proposed in the Tractatus that the general statement, Every individual is F, is to be understood as an indefinitely long conjunction of statements of individual fact. In holding this position, Wittgenstein was disagreeing with Russell, who argued for general facts in his “Lectures on Logical Atomism.”2 Bergmann is on Russell’s side. Indeed, he repeats Russell’s argument in justifying his own view. And he offers a second argument of his own. Bergmann in an early essay on “Logical Positivism” made one of the defining characteristics of that philosophical position is that all who adopt that position “hold Humean views on causality and induction...”.3 Bergmann is among these positivists.4 What I propose to argue that when Bergmann introduces generality he introduces an entity that is contrary to that Humean position on causality, and moreover in fact violates the Principle of Acquaintance which, as he argues, provides the basic argument for the Humean position on causality. The inconsistency can be removed, I shall argue, by abandoning generality as an existent. The case that Bergmann makes for generality consists of two arguments. I propose to examine this case, both his arguments. But before getting on to that I want to trace a tradition in philosophy, one in which Bergmann’s not fully existents. The point is relevant for many aspects of Bergmann’s thought, but not for what we are about: we may take it that, for all intents and purposes, generality is there, in the world, and in that sense does exist. It should be remarked that Heald’s Introduction to New Foundations is important for anyone interested in Bergmann’s later thought. See also P. Butchvarov, “Bergmann and Wittgenstein on Generality,” Metaphysica, 7 (2006), pp. 123-145; and “Metaphysical Realism and Logical Nonrealism,” in Richard Gale, ed., Essays in Metaphysics (Oxford: Blackwell, 2002), pp. 282-382. 2. B. Russell, “The Philosophy of Logical Atomism,” in his Logic and Knowledge, ed. R. C. Marsh (London: Allen and Unwin, 1956), pp. 175-282. 3. Gustav Bergmann, “Logical Positivism,” in his Metaphysics of Logical Positivism (New York: Longmans, Green & Co., 1954), pp.1-16, at p. 2. 4. For a systematic and historical treatment of Bergmann’s thought, see the very insightful study by Herbert Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism (Amsterdam and Atlanta: Rodopi, 2001). See also the long essay, F. Wilson, “Placing Bergmann,” in L. Addis, G. Jesson, and E. Tegtmeier, eds., Ontology and Analysis: Essays and Recollections about Gustav Bergmann (Frankfurt: Ontos Verlag, 2007), pp. 185-276.
18
views can in fact be situated, though I doubt that Bergmann would see things in quite this way. In any case, locating him in this tradition will lay the groundwork for my reasons for thinking his view on general facts is mistaken. We have the following syllogism: Water when heated boils This water is being heated so, this is (will be) boiling In symbols this is: (x)(Fx e Gx) Fa so, Ga The pattern is familiar. We have all seen explanations and predictions of this form. We have the event Fa and it falls under the pattern recorded in the major premise. If the generality which is the major premise is true then, given that the event Fa occurs then we know that the further event Ga will occur. That is, given the generalization, and supposing it to be true, the one event is succeeded by the other event: if the generalization is true, then the two events are connected. It is the nature of this connection that we have to discuss. Is this a generalization that is true just in case all its instances are true, saying no more than that all those instances are true? Or is this generalization not only true but also a fact that is there in addition to its instances and irreducible to them? The early Bergmann answered the first of these questions affirmatively and the second negatively, but, accepting various arguments and feeling certain pressures, he eventually changed him mind and answered the first question negatively and the second affirmatively: generality, he came to insist, has ontological status – it is a subsistent which exists alongside other existents. It is this later view that we wish to examine. Bergmann was, of course, not the first to hold that some at least among the general facts have a special ontological status. In fact, such a view appeared early in the history of philosophy. Here is some of this background. We can in fact begin at no better place than at the beginning, with Socrates. The scene is set in the neatest of Plato’s dialogues, the Phaedo.5 5. Plato, Phaedo, in Plato, Five Dialogues, trans, G. M. A.Grube (Indianapolis, IN: Hackett, 1981).
19
Socrates is sitting in prison. According to the earlier dialogue, the Crito, his friends offered to arrange things so he could escape to Thebes. But Socrates refused: following the demands of justice he determined that it would be wrong to escape his native city and run off to Thebes. He is, instead, determined to remain in his cell and drink the hemlock. So, we have one event, Socrates sitting in his cell. There are two events that could follow that event, namely, running off to Thebes and remaining in Athens and drinking the hemlock. There is nothing in the first event, Socrates sitting there, that indicates or determines that it will be followed by one rather than the other. But in fact it is the drinking of the hemlock that follows. Why? Socrates introduces into the discussion the philosophy of Anaxagoras.6 Anaxagoras had a sort of naturalism, in terms of which he hoped to explain what happens in the world.7 The world begins as a mixture of everything with everything, every part contains within itself an infinity of parts. These parts are of many kinds, bright and dark, for example; whatever else one can say, it is clear that the parts have properties of sensible things: explanation proceeds in terms of these parts, and it is naturalistic in the sense that things in this world are all of sensible sorts.8 There are no kinds in Anaxagoras that transcend and lie outside the natural world but in terms of which we are expected to explain causally what happens in the world of ordinary experience. Of course, some of the parts, no doubt most of them, are too small for the eye to notice, but they are like the things we sense and are therefore parts of the world we sensibly experience. In answer to 6. Phaedo, 97c1 ff. 7. See G. S. Kirk and J. E. Raven, The Presocratic Philosophers: A Critical History with a Selection of Texts (Cambridge: Cambridge University Press, 1957), Chapter XV, pp. 362-294. 8. See Kirk and Raven: Anaxagoras writes (according to the fragments) that “But before these things were separated off, while all things were together, there was not even any colour plain; for the mixture of all things prevented it, of the moist and the dry, the hot and the cold, the bright and the dark, and of much earth in the mixture...”. (p. 368, fragment 496). Or again, “... this [initial] rotation caused the separating off. And the dense is separated from the rare, the hot from the cold, the bright from the dark and the dry from the moist.” (p. 373, fragment 503) And so we can get flesh and bones, but the presence of the “dark” allows for shadows and the “bright” allows for rainbows: the world is as much phenomenal as it is corporeal – though this more recent terminology is no doubt misleading.
20
Parmenides, the world consists of a real diversity of parts each of which is in itself unchanging. What changes are their relations, once set in motion these parts begin to accumulate to form the bodies as we know them, including our own bodies. In fact, everything contains infinitely many parts, and among this infinity are parts gathered together as bits bone, and an infinitely of such bits. There are, too, infinitely many groups which are bits of flesh. Moreover, every bit of flesh contains not just bits of flesh but also bits of bones and bits of everything else. After an initial starting rotation, the parts started to come together. The initial movement was caused by mind or nous. Mind was itself a sort of substance. To be sure, it is a special sort of substance: it is fiery. But it is of a sort of which there is no question that we can become aware of it. It too is part of the natural world of things we can know in our ordinary experience of things. Now, ordinary things have parts of the initial mixture concentrated or gathered together in various special ways: ordinary things arise out of the original mixture through the motions of its parts. Originally at rest, the mixture is put into motion through a push from mind or nous which gives rise to an initial rotary motion. Mind or nous is concentrated before its initial push,9 but then divides itself among the individual things that began to emerge.10 After the initial push, the world emerges through changes that seem to have been more or less mechanical, their patterns determined by the natures of the bits and of the initial jolt, with mind playing only a little role, now organizing and directing individual things.11 But mind itself is, as we said, a natural thing.12 The person seems, on this account of Anaxagoras, to have 9. See Kirk and Raven: “And the things that are mingled and separated and divided off, are all known by Mind. And all things that were to be, all things that were but are not now, all things that are now or shall be, Mind arranged them all...”. (p. 373, fragment 503) 10.Thus, Anaxagoras writes (so we are told) that “...Mind, which ever is, is assuredly even now where everything else is too, in the surrounding mass and in the things that have been either aggregated or separated.” (Kirk and Raven, p. 374, fragment 506) 11. This is how Kirk and Raven (p. 375) read this fragment: “And when Mind initiated motion, from all that was moved separation began and as much as Mind moved was all divided off; and as things moved and were divided off, the rotation greatly increased the process of dividing.” (p. 373, fragment 504) 12. It is perhaps misleading to suggest, as do Kirk and Raven, The Presocratic Philosophers, that “mind, like matter, is corporeal...” (p. 375) The term ‘corporeal’ suggests that mind is something more bodily than simply consciousness. Their earlier
21
been considered a body, consisting of skin and bones, structured and moved by mind, which organizes things so as they survive as wholes for as long as they can, but in all this persons, it seems, are not distinctively different from dogs, and perhaps not even from oysters. Socrates objects.13 If Anaxagoras were correct, then, like a dog seeking only bodily well-being and survival, he would have run off to Thebes. But he is not doing that, so Anaxagoras must be wrong. The difficulty is the role of mind or nous. Socrates has already established that mind or nous, the soul, moves the body. Thus far he agrees with Anaxagoras. But what is needed to explain Socrates’ sitting in jail being followed by his drinking the hemlock is the fact that he is moved by a vision of the just. This leads him into his alternative explanation. Earlier he has established two things. One is that things in the world of sensible experience, one’s body, in particular, is moved by the soul and that this active entity is simple, an indivisible pure activity, and therefore unlike the things of the world of sense which are divisible and separable into parts, as, e.g., Socrates’ sitting there in his cell is separable from the events that will succeed it. The second thing that he has established is the existence of the forms as perfect exemplars of the things we observe in sense. Having secured agreement that all equals in our ordinary experience are imperfectly equal, he then argues on the basis of the principle that in order to judge something to be not F one needs to have the concept of F – in order to know that something is not red one must know what it is to be red. So, to judge something to be imperfectly equal we must have the concept of perfect equality. The form of perfect equality must therefore exist, not in the world of ordinary experience, but in another world. What holds for equality holds for other things, perfectly straight lines, perfect triangles, and, of course, perfect moral virtue or justice. He uses these metaphysical principles to give an explanation of the sequence of events that Anaxagoras is unable to give. Socrates sitting in his cell is followed by his drinking the remark that “Anaxagoras ... is striving... to imagine and describe a truly incorporeal entity” (p. 374) seems to be more just. The point is that these things which are incorporeal (“fiery”) can causally affect and causally interact with the things of flesh and bones that are more fairly described as material and are in that respect like things corporeal. 13. For an elaboration of the Socratic position, see F. Wilson, Socrates, Lucretius, Camus – Two Philosophical Traditions on Death (Lewiston NY and Queenston ON: Edwin Mellen Press, 2001), Chapter Three.
22
hemlock because the soul, his soul, is striving to imitate in his outward sensible appearances the form of perfect human justice, the ideal good that moves his soul in one direction and not another, the form the vision of which moves him to do what is right or just and not what the doggie would do.14 There is a problem:15 why does the soul of Socrates strive after the form of human justice rather than the form of doggie justice or oyster justice? Separation is to be explained in terms of a connection being effected by a soul imitating a form. So the soul striving after human justice is explained by the soul striving after the super-form of justice – from which it is separate so its striving after that needs another form to strive after, that of the super-duper-form of justice – and so on without end. And since there is no end, nothing ever gets explained. This is the Third Man Argument of Aristotle. To solve the problem in a way that saves the basic form of explanation Aristotle ends the separation of the forms from the souls. Each soul, as a Aristotle comes to speak, each substance has intrinsic and inseparable from itself a form or essence. Then patterns in the world of ordinary experience which are separate from one another – apparently separable, so far as our senses know – because they are not really separate: rather they are connected by the activity of the substance in a direction determined by its form or essence. The sequences of the separable events of sense, of ordinary experience, are explained through their being a tie between them such that one event necessitates the succeeding event by virtue of the activity of the soul in which those events occur where the observable pattern reflects the necessary structure or form or essence of the active substance. The sequences of ordinary events in the world of sense are explained in terms of entities outside this world, the sequence is really one in which that parts are not after all separable but connected by the striving of the active substance qua having a certain form that determines the direction of the striving. Moreover, the explanation is in terms of best: the form of human is the form of human justice, and the way things are is 14. Here we have what has been called, not unjustly, the “natural philosophy” of the Phaedo. For more on this, see Wilson, Socrates, Lucretius, Camus, Chapter Three. For the basic argument that what we have here is the pattern developed for the explanation of ordinary events, see G. Vlastos, “Reasons and Causes in the Phaedo, Philosophical Review, 78 (1960), pp. 291-325; and R. G. Turnbull, “Aristotle’s Debt to the ‘Natural Philosophy’ of the Phaedo,” Philosophical Quarterly, 8 (1963), pp. 131-143. 15. See F. Wilson, Socrates, Lucretius, Camus, p. 121ff.
23
explained by the soul aiming at the best. What is is explained in terms of what ought to be.16 Contrast this to the world of Anaxagoras. In that world, there is no unity, only separable parts. There are sequences of events, and even regular patterns: after all, stuffs are stuffs and each kind behaves in its own ways and combines and separates in its own ways. So there is regularity and patterns. But these patterns are contingent, separable and without any real or necessary connection to one another. Even mind or nous, while of a unique sort of substance, and not mixed with the other stuffs, is not itself a unity; it too consists of parts. What is not a unity cannot unify and therefore cannot explain. Mind, moreover, is of this world, the world that exists in space and time, the world we know by sense. Mind is not something that is outside this world and certainly not something that grasps the eternal forms and therefore partakes of that eternal being. There is no explaining of what is in terms of the best, no teleology that determines the way things are in terms of striving for what ought to be. So we have ordinary events explained by being instances of patterns: All human (H) is animal (A) Socrates is a human (H) Hence, Socrates is animal (A) On this, everyone agrees, Anaxagoras, Socrates, Plato, and Aristotle, and everyone since then. If we ignore the initial shove, this is all there is to explanation on Anaxagoras scheme. But for the Platonist and the Aristotelian, this is not enough: there is not yet any real unity. For the Platonist and the Aristotelian, the major premise, recording the pattern All human is animal if there is to be an explanation must do more than merely record a pattern observed in ordinary experience: it must of course represent that, but further must represent a real connection, something necessary and not merely contingent: it must represent a necessary structure that so binds the characters that H in its very being is inseparable from A in its very being. H and A must be so intimately connected that neither the contrary, stating their incompatibility, No man is animal nor the contradictory 16. Cf. F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies (Toronto: University of Toronto Press, 1999), Study One; and “Science and Religion: No Irenics Here,” Metaphysica, 7 (2006), pp. 159-170.
24
Some man is not animal is conceivable. The connection of H and A must be a timeless structure that rules these out as possibilities. As Aristotle worked these things out,17 the required structure is that of a species as part of a genus – the species man within the genus animal, with the connection between effected by the specific difference, in this case the specific difference of being rational. This necessary structure is displayed in what Aristotle referred to as a scientific syllogism: All rational is animal All human is animal All human is rational This syllogism displays the logical and ontological structure of the species human. This form is given equivalently in the real definition of the species human: human is rational animal This real definition is a necessary truth, a timeless truth about the logical and ontological structures of the forms of substances. More generally, and putting it in terms of a more recent logic, we have characteristics F and G, and the observable pattern All F are G or (x)(Fx e Gx) We have this connecting the two events Fa and Ga: (x)(Fx e Gx) Fa so, Ga According to the experience of sense, being F and being G are separable events. The pattern represented by “All F are G” is merely a regularity, a pattern amongst events that is merely contingent. There is no deeper or stronger connection, and therefore no genuine connection. In order that there be a genuine explanation, there must be a necessary connection. This is in effect Socrates’ argument against Anaxagoras, or part of that argument anyway. The very same argument has been put more recently by Fred Dretske. In his essay on “Laws of Nature,”18 he argues that mere regularities 17. See F. Wilson, The Logic and Methodology of Science in Early Modern Thought: Seven Studies, Studies One and Two. 18. F. Dretske, “Laws of Nature,” Philosophy of Science, 44 (1977), pp. 248-68.
25
are simply unable to explain anything. To say that a law is a universal truth having explanatory power is like saying that a chair is a breath of air used to seat people. You cannot make a silk purse out of a sow’s ear, not even a very good sow’s ear; and you cannot make a generalization, not even a purely universal generalization, explain its instances. The fact that every F is G fails to explain why any F is G, and it fails to explain it, not because its explanatory efforts are too feeble to have attracted our attention, but because the explanatory attempt is never even made. The fact that all men are mortal does not explain why you and I are mortal; it says (in the sense of implies) that we are mortal, but it does not suggest why this might be so.... Subsuming an instance under a universal generalization has exactly the same explanatory power as deriving Q from P & Q. None.19
The important point is that laws, that is, generalizations that genuinely explain, “tell us what (in some sense) must happen, not merely what has and will happen (given certain initial conditions).”20 Mere matter of fact regularities do not explain: this claim is the substance of Dretske’s argument (as it was the substance of Socrates’ argument against Anaxagoras). Statements of such generalizations amount logically to nothing more than conjunctions. And conjunctions do not explain their conjuncts. If a generalization is to explain, then it must show how the events in question were necessitated, not merely record their occurrence. In some sense, causal laws must be necessary; they must be such that an event that violates them is impossible, they must be such that their contraries must be inconceivable. This argument is of a piece with that of Socrates against Anaxagoras. Thus, the regularity represented by “All F are G” will not by itself be explanatory unless it is the reflection in sense of a necessary connection among forms which are the ontological reality behind the world of sensible appearances. That is, we must have a relation among forms (^) R(F, G) where R is a second order relation amongst first order characteristics F and G. Since this is a relation amongst forms, it is a relation which holds amongst them as outside the world of sense and of space and time. It holds not merely everywhere and everywhen, not merely omnipresently and 19. Dretske, “Laws of Nature,” p. 262. 20. Ibid., p. 263.
26
omnitemporally but eternally, that is, outside the world of space and time, existing not in this world we know by sense, but existing as beings in the calm of a timeless and unchanging order. Since the order is timeless and therefore changeless, it is necessary. This necessity secures as necessary the connection, only apparent to sense, amongst the characteristics F and G, so that the pattern represented by “All F are G” is not merely a contingent regularity but a connection which is necessary and therefore genuinely explanatory. And, since the fact that R(F, G) holds timelessly, it can be argued that it is known a priori, as Plato and Plato’s Socrates argued, or at least in a non-perceptual intuition, as Aristotle argued. However, there are problems. It is worth noting, first, that, while there was a certain neatness to the necessary connection in the real definitions of the Platonic-Aristotelian metaphysics, once one moves away from this traditional ontology, things cease to be so neat. Causal regularities come to take on more complicated forms. The second-order relation that constitutes the tie that secures causal relatedness should be seen as a somewhat odd relation. We not only have the above two-place instance of it, but also three- and many-place instances of it. Thus, we could also have the regularity (x)[(Fx & Gx) e Hx] If this is to have the necessity of a genuine causal collection, if it is to be a law and not a merely contingent regularity, then we will need the secondorder relational statement R(F, G; H) where this will have to be distinguished from R’(F; G, H) which guarantees that (x)[Fx e (Gx & Hx)] is a law, if it be a law. We also have laws like For every shark there is a pilot fish close by which would go into the symbolism as (x)[Sx e (›y)(Py & Cxy)] which, one supposes, is guaranteed by some second order fact like R”(S; P-; C-,-) or perhaps something like R”’[Sx, (›y)(Py & Cxy)] or whatever – it is in fact quite difficult to say exactly what it would look like. But one gets the idea. The point here is that the necessary tie among 27
properties or characteristics of things that is intended to secure as causally necessary what otherwise would be contingent regularities among separable facts. Dretske’s causal connection is not just one connection but a whole family of relations. We are owed a story about how the members of this family are to be understood ontologically. Dretske does not give us that story. Be this last as it may, the Dretske suggestion for securing the metaphysical or ontological necessity those regularities that are to count as causal has of late been taken up by several other metaphysicians, among them D. M. Armstrong,21 M. Tooley,22 and J. Brown.23 Unfortunately, there is a fatal flaw in the position that these philosophers have advocated. The crucial point is that in order for the fact that R(F, G) to guarantee the truth of the regularity, it must also be true, and not merely true but necessarily true, that (@) R(F, G) e (x)(Fx e Gx) The difficulty is that in terms of the standard notion of entailment given by the usual rules of formal logic, the sentence (@) is not a necessary truth: the logical form of (@) is not that of a logical truth, since the antecedent is an atomic statement (of the second order) and the formal relations of standard logic imply that no atomic statement ever entails a general truth. But (@) is precisely the sort of statement that must be necessary if structural relations among properties are to provide a metaphysical ground for regularities among the individuals that exemplify those properties. Moreover, the relational statement (^) R(F, G) so far as formal logic is concerned is an atomic statement and, again, therefore not a necessary truth: the logical form of (^) is that of an atomic statement and our standard logic implies that all such statements are contingent, not necessary. The problem is that (^) and (@) are not what they are required to be: necessary. Their problem is to show the necessity of the relational fact (^) and the principle of inference (@). As John Earman has put it, 21. D. M. Armstrong, What Is a Law of Nature? (Cambridge: Cambridge University Press, 1983). 22. M. Tooley, “The Nature of Laws,” Canadian Journal of Philosophy, 7 (1977), pp. 667-98. 23. J. Brown, The Laboratory of the Mind (London: Routledge, 1991).
28
What remains to be worked out [on the view that causal relations are relations among universals] is the formal semantics of the entailment relation [that holds between the statement about universals and the matter of fact regularity]; whether this can be done consistently ... remains to be seen.24
Earman puts the matter rather too cautiously: there is no reason to suppose that it can be done, no reason to think that one can find in the formal logic of the empiricist grounds to make plausible the claim that a second order atomic statement about properties, “R(F, G)”, should entail a first order generalization, “(x)(Fx e Gx)”. Earman is making the point that was in effect made earlier by Russell in his Inquiry into Meaning and Truth.25 Considering the generality All humans are mortal26 he takes up what he calls, not inappropriately, the “Aristotelian” suggestion, that there is an a priori connection among the predicates: “We might be tempted to say that one of these predicates implies the other, and use this as an analysis of what is ‘expressed’ by ‘all humans are mortal’.” But, he goes on to argue, This Aristotelian interpretation ... overlooks the fact that the connection is not between the predicates as such, but only between the predicates as predicated of one subject. “A is human”27 involves “A is mortal,” but not “B is mortal.”28
The objection was clearly and explicitly developed by Bergmann, in his essay “On Non-Perceptual Intuition,”29 and it has been developed by others 24. John Earman, “Laws of Nature: The Empiricist Challenge,” in R. J. Bodan, ed., D. M. Armstrong (Dordrecht, The Netherlands, 1984), p. 221n21. 25. B. Russell, Inquiry into Meaning and Truth (New York: W. H. Norton, 1940). 26. I have substituted ‘human’ where Russell has ‘man’. 27. Russell at this point uses ‘human’ instead of the earlier term ‘man’. 28. Russell, Inquiry into Meaning and Truth, pp. 116-117. 29. G. Bergmann, “On Non-Perceptual Intuition,” in his Metaphysics of Logical Positivism (New York: Longmans, Green & Co., 1954), pp. 228-231. Earman seems to have overlooked this important essay, which makes the point he is making; at least, it is not in the bibliography to Earman’s essay. Bergmann develops the same point in a rather different context in his essay on
29
also.30 It turns out that Dretske’s proposal is hardly original. Bergmann gave the criticism originally in response to C. D. Broad, who had argued Dretske’s thesis in his commentary on McTaggart,31 who had also defended a version of Dretske’s thesis. Unfortunately, Dretske seems not to know either the history of his proposal nor the criticisms which have been raised against it. The tradition of Plato and Aristotle had in fact a way out of this problem. That was the active soul. Recall from the logic text tradition that Aristotle argued with regard to syllogisms that the first figure M is P S is M S is P was the perfect figure, which could justify syllogisms in the other figures by deriving or “reducing” those of the other figures from those of the first figure, but the syllogisms of which could not itself be so justified by reduction. However, Aristotle also argued, the syllogisms which occur in this first figure could be justified by what came to be called the dictum de omni et nullo. This is the maxim that “whatever can be affirmed (or denied) of a class can be affirmed (or denied) of everything included in the class.” In the traditional metaphysics this did indeed have a place. According to that tradition, the proposition (a) S is P is about substantial forms. But it is true about individuals that (b) All S are P in the sense that (b’) All individuals which are S are also P If no substantial forms exist, as Anaxagoras would have it, this latter is of “The Ontology of Edmund Husserl,” in his Logic and Reality (Madison,, WI: University of Wisconsin Press, 1964), pp. 193-225. Earman misses this discussion too. 30. I have previously developed the following criticism of these philosophers in F. Wilson, Laws and Other Worlds (Dordrecht, The Netherlands: D. Reidel, 1986). See also F. Wilson, Logic and the Methodology of Science in Early Modern Thought, Study Two. 31. See C. D. Broad, Examination of McTaggart’s Philosophy (Cambridge: Cambridge University Press, 1933), vol. I; and J. M. E. McTaggart, The Nature of Existence (Cambridge: Cambridge University Press, 1921), vol. I.
30
course the only sense that could be given to the asserted generality: it merely records a pattern amongst things. The problem facing McTaggart, Broad, Dretske, et al., is to account for why (a) implies (b) = (b’). The traditional ontology deriving from Aristotle and from Socrates’ critique of Anaxagoras has a solution to this problem. On this ontology, the forms as the forms of active entities, souls or substances, cause the individuals of which they are the forms to be in a way that guarantees the truth of (b) = (b’): what is predicable of the universal or form is predicable of the individual substances subordinate to it. S is P This is S so, This is P The metaphysics of active substances thus guarantees that (b) = (b’) is true if (a) is true. But furthermore, this ontology of active substances ensures that (b) = (b’) is not just true but necessarily true. For, the active substance, the This, has as its form the species S, and has that form necessarily. So, of its own ontological necessity, since it is S, it of necessity acts to make itself P: it brings it about that it conforms in its outward appearances to the necessary structure of its inner form or essence. And this will be true of every thing, that is, every substance which has that same form or essence. The form thus guarantees that every object with that form will have the same pattern in its outwards appearances. (b) = (b’) will therefore not only be true but be necessarily true. And, given the structure of the metaphysics, since (a) is about forms and timelessly true, and therefore knowable a priori, so also the truth of (b) = (b’) be knowable a priori. On the substance metaphysics deriving form Socrates’ encounter with Anaxagoras, Earman’s problem is solved, and Bergmann’s objection is met. Thus, on the traditional view, the dictum de omni et nullo expresses a fundamental principle or law about the ontological structure of the universe; as John Stuart Mill was later to put it, the principle stated that “the entire nature and properties of the substantia secunda formed part of the nature and properties of each of the individual substances called by the same name...”32 32. John Stuart Mill, System of Logic, Bk. II, Chapter ii, § 2; volumes 7 and 8 in his Collected Works, ed. J. Robson (Toronto: University of Toronto Press, 1974), vol. 7, p. 174. For discussion of Mill’s point, see F. Wilson, The Logic and Methodology of Science in Early Modern Thought (Toronto: University of Toronto Press, 1999), pp. 231-240.
31
Now, as Broad presents the suggestion that Dretske later takes up, he quite rightly notes, appropriately as a commentator, that the proposal is present already in McTaggart, who had argued that observed instances of things being both F and G, no matter how numerous, have no tendency to support the conclusion that ‘All instances whatever of F are instances of G’.33 Hence, Unless some further premise, which is known a priori, be added to the empirical premise, the latter can never give us any ground to suspect the presence of the relation of conveyance [McTaggart’s term for the relation of causal necessitation].34
But at the same time, McTaggart also holds, as Broad puts it, that In the case of laws of nature, human beings cannot see by direct inspection that the relation of conveyance holds between the terms.35
This problem is not even acknowledged by Dretske, but it is significant. For, this problem, which Broad calls epistemological,36 implies that this account of causation that is being proposed, as a second-order relation between properties, even if we waive Bergmann’s objection, is of no use in solving the problem of induction. Hume was the first to make the point clearly. He argued that there is no necessity to causation: Anaxagoras was right, all that is involved in explaining the processes of this world, this world in which we carry out our ordinary lives, and which we know by means of our senses (and by means of our inner awareness), is the regularities that obtain among the ontologically separable events that we find in this world and have imposed upon us. If we examine carefully our knowledge of this world we find neither within it nor within ourselves as part of this world, any entities that could transform these regularities into objective necessary connections. Of course, these patterns are regularities, generalities which are timeless patterns, but nothing about them or the entities in the world that makes it a matter of 33. Broad, p. 223; McTaggart, p. 275. 34. Broad, pp. 223-4. 35. Broad, p. 227; cf. McTaggart, p. 275. 36. Dretske, “Laws of Nature,” p. 227.
32
ontological necessity that the world unfolds according to these patterns rather than those. Hume proposes,37 or, rather, argues that objectively causation simply is regularity: to ascribe a causal relation is to propose a general pattern describing the way things actually are. The justification for this claim is based on the failure of those who defend objective necessities to direct our attention to these objective necessities. After all, we make causal judgments all the time – my flicking the switch turns on the light, the bark of the tree is rough to the touch, always rough, my willing my arm to rise is inevitably (though not necessarily) followed by my arm going up (most of the time, at least) – and since making causal judgments occurs all the time, if necessary connections were required for us to make such a judgment, then we should constantly be aware of them, they should be a constant part of or experience of the world and of our being in it. But they aren’t, and so the reasonable person would exclude them from his or her ontology. As Hume puts it, The small success, which has been met with in all the attempts to fix this power, has at last obliged philosophers to conclude, that the ultimate force and efficacy of nature is unknown to us, and that ’tis in vain we search for it in all the known qualities of matter.38
And so we have his definition of “cause” as nothing more than objective regularity among entities that ontologically are separable: “an object precedent and contiguous to another, and where all the objects resembling the former are placed in a like relation of priority and contiguity to those objects, that resemble the latter...”.39 Wittgenstein made the same point in his usual oracular way.40 “The world is everything that is the case.” (Tractatus, pr. 1) These facts are 37. David Hume, Treatise of Human Nature, ed. L. A. Selby-Bigge, second edition, ed. P. Nidditch (Oxford: Oxford University Press, 1978). 38. Hume, Treatise, p. 159. 39. Hume, Treatise, p. 172. 40. L. Wittgenstein, Tractatus Logico-Philosophicus, trans. D. F. Pears and B. McGuiness, with Intro. by Bertrand Russell (London: Routledge and Kegan Paul, 1961). References are by numbered propositions. See also P. Butchvarov, “Bergmann and Wittgenstein on Generalitty,” for references to, and comments on, this and other parts of the Wittgenstein corpus.
33
represented by elementary propositions and “One elementary proposition cannot be deduced from another” (Tract., pr. 5.134); he then comments to emphasize the point, “There is no possible way of making an inference from the existence of one situation to the existence of another, entirely different situation” (Tract., pr. 5.135). The facts of this world as it is given to us in experience, and the objects, particulars and characteristics, are logically and ontologically self-contained: one does not imply any other, each is what it is and neither is another entity nor is tied in its being to another entity. Hume made the relevant point: with respect to inferring one basic fact from another, “There is no causal nexus to justify such an inference” (Tract., pr. 5.136). And again, to emphasize and make the point more dramatically, he goes on to state that We cannot infer the events of the future from those of the present. Belief in the causal nexus is superstition. (Tract., pr. 5.1361)41
Bergmann was later to propose that “as long as one sticks to cautious generalities” then it is safe to say that “all Logical Positivists,” including of course himself, among other things “agree that they ... hold Humean views on causality and induction....”42 This is based, like Wittgenstein’s similar view, on the logical and ontological independence of basic facts, that is, as Bergmann also puts it, upon an “insistence on the self-containedness of the given...”.43 This latter means that factual generalities are all contingent, including such statements as those of colour incompatibilities: Whatever is red is not green or in symbols (x)(Rx e ~Gx)44 To know an entity, a particular, a characteristic or a fact, we do not need to know any other entity. To know what redness is it is not necessary to know greenness, and to know that something is red it is not necessary to know 41. The emphases throughout are Wittgenstein’s. 42. Bergmann, “Logical Positivism,” p. 2. 43. Bergmann, “Logical Positivism,” p. 14. 44. Cf. Bergmann, “Logical Positivism, Language, and the Reconstruction of Metaphysics,” in his Metaphysics of Logical Positivism, pp. 30-77, at p. 50; and “The Revolt against Logical Atomism,” in his Meaning and Existence (Madison, WI: University of Wisconsin Press, 1960), pp. 39-72, at p. 57.
34
that it is not green. As Bergmann also put the point “There are several things with which we become acquainted if they are once presented to us. If one such thing is presented to us again, we recognize it.”45 This being true, the world as we know it in ordinary experience is “atomistic,”46 the parts as we know them in their givenness are logically and ontologically separable: there are no objective necessary connections. McTaggart attempts to get around this point. He argues that he can establish a priori that there are such connections among properties: “...it is possible to know a priori that some such laws are valid if anything does exist.”47 But, as Broad indicates, we cannot know specifically which relations hold in the actual universe.48 For, we are not acquainted with any such relation: it is not given to us in our experience of the sensible world, nor is the activity or soul that makes the connection between the world of forms and the world ordinary experience given to us in our ordinary experience. In that experience all can and do record is mere regularity. But if this is so, that is, if we assume as given the epistemological problem raised by Hume, then there is an insurmountable difficulty with McTaggart’s claim upon which one must insist. The point is that we must know the sort of thing that is claimed to be there if that claim is to be intelligible. If we claim that there are unicorns, then we must know what sort of thing a unicorn is if that claim is to be intelligible. If we claim that there are boojums, then we must know what sort of thing a boojum is if that claim is to be intelligible. We know the sort dog because we are acquainted in ordinary experience with Fido, Spot, etc., that is, instances of the sort dog. We know the sort unicorn not because we are acquainted with instances of that sort – after all, there are no unicorns – , but because we can define the concept “unicorn” in terms of other sorts – horse and horn – which are such that we have been acquainted with instances of those sorts in ordinary experience. But we do not know the sort boojum because we can neither define it nor are we acquainted with instances of that sort. Boojums are snarks, and we have never seen a snark either, so we don’t even know the sort of this sort. These words, ‘boojum’ and ‘snark’ are just 45. Bergmann, “The Revolt against Logical Atomism,” p. 44. 46. Bergmann, “The Revolt against Logical Atomism,” p. 45. 47. McTaggart, p. 274. 48. McTaggart, p. 239.
35
sounds.49 So the claim that there are boojums is unintelligible where the claims that there are dogs and that there are unicorns are both intelligible. Similarly, we can claim that there are objective necessary connections of the sort defended by McTaggart, etc., only if either we are acquainted with instances of the kind or we can define them in terms of kinds instances of which we are acquainted with. But such relations are held to be unanalysable. And then there is the epistemological problem: we have not encountered any of these entities in ordinary experience. But if such relations are unanalyzable and we are not acquainted with them in experience, then the concept of such a relation is like the concept of a boojum: unintelligible. This is a problem for the recent advocates of this position such as Dretske, Tooley, Brown or Armstrong, and also a problem for its earlier advocates Plato and Aristotle. To be sure, the recent advocates do make a reply of sorts: they suggest that we can treat the required relations among universals as “theoretical entities.” Thus, Brown introduces the objective causal or nomological relation as a relation among forms as a “theoretical entity posited for theoretical reasons.”50 Armstrong makes much the same point: “The postulation of a connection between universals can provide an explanation of an observed regularity in a way that postulating a Humean uniformity cannot.”51 However, first off, one does not “postulate” a Humean uniformity: such a uniformity is simply an observed regularity. The use of such language is nothing other than a cheap attempt to make it seem that the two positions are somehow parallel, in that they both “postulate” something – but affirming an observed regularity is very different from “postulating” an unobserved tie of ontological necessity. Moreover, “postulating” a tie as a “theoretical entity” is also unlike the “postulation” of the “theoretical entities” in physics, the affirmation, usually on good inductive grounds, of entities, atoms or quarks or whatnot, that are unob-
49. They do receive a definition of sorts in Lewis Carroll’s poem on “The Hunting of the Snark,” but it is only what one calls an “implicit definition,” which is no definition at all – we learn some syntax for these terms but no semantics: they remain without any reference in terms of what we are acquainted with, and are therefore really and simply meaningless. 50. Brown, Laboratory of the Mind, p. 83. 51. Armstrong, What Is a Law of Nature? p. 104.
36
servable by sense.52 The “postulation” of a causal nexus or a tie of ontological necessity is totally different from the postulation of small parts that we cannot see – for example, the postulation of a speck of dust to explain why my watch won’t work.53 In this latter sort of postulation, the entities postulated share certain features with the entities of which we are aware in ordinary experience. Although we may not experience the entities themselves, and, indeed, even though they may be too small or too far away for us ever to experience them, nonetheless, they are of kinds with which we are acquainted.54 But this does not hold for the nomological connection: by hypothesis this is something totally different in kind from anything with which we are acquainted. This raises problems, surely, for anyone – problems that should not be brushed under the carpet by a blithe and unexamined use of the term ‘postulation’.55 Broad in fact attempts to avoid this problem. He argues that we are in fact aware by a sort of “non-perceptual intuition” of instances of the crucial relation, “conveyance” as he calls it, following McTaggart. Broad indicates56 that we have in the example of everything that is shaped has extension: “we recognise by mere inspection and reflexion that nothing
52. For more on this notion, see G. Bergmann, “Outline of an Empiricist Philosophy of Physics,” American Journal of Physics, 11 (1943), pp. 248-258, pp335-342. (This was re-printed in H. Feigl and M. Brodbeck, eds., Readings in the Philosophy of Science [New York: Appleton-Century-Crofts, 1953].) See also F. Wilson, “Empiricism and the Epistemology of Instruments,” The Monist, 78 (1995), pp. 207-229. 53. This example, interestingly enough, is from Hume; see his Treatise, Bk. I, part iii, sec. 12, p. 132ff. It should make interpreters of Hume hesitate before they read him as a phenomenalist or a subjectivist or a sceptic. For discussion of these points in detail, see F. Wilson, The External World and Our Knowledge of It: Hume’s Critical Realism: An Exposition and a Defence (Toronto: University of Toronto Press, forthcoming). 54. For a discussion of these inferences in the context of empiricism, see F. Wilson, Hume’s Defence of Causal Inference, Ch. 1. 55. Those criticized by Wilson in his Explanation, Causation and Deduction, sec. 3.6, were equally unforthcoming with regard to our knowledge of the nomological connective which, in spite of that lack of knowledge, they nonetheless insist must exist. Credo ut intelligam. 56. Broad, Examination of McTaggart’s Philosophy, vol. I, p. 225.
37
could possibly have shape and lack extension”.57 But this will not do. It may well be that the proposition that everything that has shape is extended is a synthetic a priori truth. But Broad has not solved the problem of giving us an instance of the relation of conveyance. For, if conveyance is a relation among properties that guarantees the truth of the generalization in question, then it must entail that regularity. That is, we must have (@) as a necessary truth, as Bergmann argued. And Broad does not provide us with an account of entailment that will do the job. So Broad has after all not provided us with an instance of the sort of relation that he and McTaggart need in order give conceptual content to the claim that there are second order relations among properties constituting ties of causal necessity. Actually, Broad elsewhere gives another example that he thinks is a case where we experience a tie of causal necessity. This is the case of the will. “It is perfectly plain,” he tell us, “that, in the case of volition and voluntary movement, there is a connection between the cause and the effect which is not present in other cases of causation and which does not make it plausible to hold that in this one case the nature of the effect can be foreseen by merely reflecting on the nature of the case.”58 Hume argued that this case is in fact no different from other cases of causation: cause and effect are logically and ontologically separable.59 But, Broad goes on, “The peculiarity of a volition as a cause-factor is that it involves as an essential part of it the idea of the effect.”60 But surely Broad is confusing the causal connection with the intentional connection. In the case of volition, the volition which is the cause contains an idea which has as its intention the event which is the effect. It does not follow, as Broad suggests it does, that cause and effect are not logically and ontologically separable: it is easy to conceive the cause occurring but not the effect. So I think Hume is correct 57. Broad, Examination of McTaggart’s Philosophy, vol. I, p. 225. 58. C. D. Broad, Mind and Its Place in Nature (New York: Harcourt, Brace, 1925), p. 102. 59. See Hume, Treatise, Book II, part ii, sec. 1. Also Hume, Enquiry concerning the Human Understanding, in D. Hume, Enquiries concerning Human Understanding and concerning the Principles of Morals, ed. L. A. Selby-Bigge, 3rd edition, revised, P. H. Nidditch (Oxford : Oxford University Press, 1975), Sec. VII, part i. 60. Broad, Mind and Its Place in Nature, p. 102.
38
here also and Broad is wrong, we no more experience a causal tie in the case of volition than we do in the case of billiard balls. It is no doubt true that in willing the effect I do not predict the effect. But there is nothing in the idea that causation is regularity that requires that in willing my arm to go up and its so going up I must think of the regularity that must – and does – obtain with respect to volitions of that sort and movements of that sort if the former are to cause the latter. So again Broad has not provided us with an instance of the sort of relation that he needs in order give conceptual content to the claim that there are second order relations among properties constituting ties of causal necessity. This is true in general. The defenders of the thesis that laws are relations among universals provide no clear account of entailment such that a statement of such a relationship entails a generalization about the particulars that exemplify those universals. Thus, though Dretske, Armstrong, Brown and Tooley, following Plato and Aristotle, all assert that the connection (@) is an entailment relation they make no effort to indicate the logical form that provides the ground of the necessitation. The antecedent of (@) is a relation among properties, and therefore in some sense a necessary truth, at least in the sense of being timeless. However, on any standard account of entailment, there is no reason to expect a non-general relational fact about properties to entail a generalization about the particulars which have those properties. Whatever else a law is, it is a regularity. The concern is whether there is anything more to laws. Dretske et al., insouciantly following Broad, argue that there is something more, to wit, the relation among universals. But if that is to make sense, then the relation among universals does indeed have to entail the matter-of-fact regularity, as they quite rightly see. But, alas! they provide no account of that entailment relationship. To assert that one exists is hardly to explain it. What precisely is the logical form that is the ground of the necessity alleged to hold between the antecedent and the consequent, between the relational fact about universals and the generalization about the particulars falling under those universals? In the absence of any reasonable account of logical form that would show the necessity of (@), the notion that we can construe laws as relations among universals is simply a non-starter. Plato and Aristotle of course do provide an answer where Dretske et al. do not: they do ensure the necessary truth of (@). They do this by appeal to the activity of a soul or substance to guarantee that the connection among the forms is exemplified in their outward appearance as observable things. This is not an entailment relation, but it is an ontological guarantee. However, just as 39
the Humean and Wittgenstein and Bergmann reject objective necessary connections, so they also reject the active soul or substance: we are acquainted with neither, and therefore neither can be admitted into our ontology. Indeed, Dretske et al. also reject this Platonic or Aristotelian solution of an active substance – at the least, they invoke no such entity – but they thereby deprive themselves of a solution to the problem of guaranteeing the necessity of the observed regularities among ordinary things. Here is where we are. What the defender of objective necessary connections needs is something like this. We have F caused G (or: F causes G) which can be construed as This event, which is F-like, causes this other event, which is Glike or more briefly as (This being F-like) causes (this being G-like) or still more briefly as (c) Fa causes Fb We can see here the logical features that a causal nexus or necessary tie must have.61 First, the tie causes connects two facts. It is a connective of sorts, connecting atomic sentences into complex sentences. Second, as (c) stands, the connective causes is not a truth-functional connective: the truth-values of the two atomic sentences in (c) do not by themselves determine the truth-value of the complex sentence. The truthvalue of the complex sentence will depend on other things, too, e.g., F and G cannot be any old kinds, they must be kinds that are causally connected. Any connection of the atomic facts that purports to be the causal tie must be truth-functional, on the one hand, and yet also, on the other hand, depend on the kinds that are said to be the causally relevant kinds. This is in effect the third point that is to be emphasized: the statement (c) will be true only if the atomic sentences mention kinds that are causally connected. The kinds must be what are often called “natural kinds,” that is, kinds that are in themselves so bound together that they determine that a fact of the one kind necessitates causally the existence of a 61. For these points, see my earlier discussion in F. Wilson, Explanation, Causation and Deduction (Dordecht, The Netherlands: D. Reidel, 1985), pp. 293ff.
40
fact of the other kind. They must as kinds themselves be bound together in the causal nexus. In this respect they must in effect be Aristotelian forms or essences. Fourth, the causal tie must vary from context to context. Sometimes the “natural kinds” will determine a regularity of one logical form, sometimes a regularity of another logical form, sometimes involving relations at other times not. Fifth, it must such that the kinds F and G are “natural kinds” in the sense that when (c) obtains then, necessarily, the regularity (r) (x)(Fx e Gx) also obtains. These are the five crucial features of any connective that can do what the causal nexus or tie is supposed to do, according to those who propose it. The proposals of Dretske et al. all fail to satisfy these crucial features. There are other issues. The proposed nomological connective causes is on the face of it non-truth-functional. If we make it truth-functional by making it a second order relation among characteristics as in R(F, G), thereby making it an atomic fact among atomic facts, separable from and logically unconnected to other atomic facts, then it does not do the required job; this was the point made by Bergmann against Broad (and McTaggart), the point which was later taken up by Earman. But if it is non-truthfunctional, then it comes into conflict with the standard explication of the analytic-synthetic distinction which presupposes that the only primitive connectives are truth-functional. Call this the first epistemological problem. Further, if the causal tie is taken as primitive then there is the problem of how we know when it obtains and, more deeply, how we even know what it is. This is the difficulty Hume raised. It was referred to by Broad as the “epistemological problem” facing one who proposed anything like McTaggart’s nexus of “conveyance.” Let us refer to it as the second epistemological problem. With all this in mind, we can now turn to Bergmann’s general facts, the general facts that exist over and above particular atomic facts. We have Fa & Ga Fb & Gb Fc & Gc ........ and so on but in addition to these we have as an additional fact the general fact 41
(d) (x)(Fx e Gx) This, I am going to suggest, has the five crucial features needed by a “connective” if it is to count as giving us the tie or nexus of causal necessity. It is even possible to give it a form that simulates the form ‘R(F, G)’ that is proposed by Dretske & Co. to account for causal necessity. Let us define ‘C(f, g)’ is short for ‘(x)(fx e gx)’ where ‘f’ and ‘g’ are free variables. Then, given (d), we have (dd) C(F, G) as a (defined) relation among the characteristics F and G understood as universals.62 It of course immediately follows that (de) C(F, G) e (x)(Fx e Gx) is analytically true, ex vi terminorum, thus meeting Earman’s objection. Now look at the five crucial features which anything that purports to play the role of a causal connective or tie as in (c) Fa causes Ga I am proposing that in Bergmann’s ontology the relation ‘C(F, G)' as given in (dd) satisfies the five crucial features and therefore functions as a nomological tie or nexus. First: The tie or connective must connect any two atomic facts representing the cause and the effect. But if Fa obtains and so does the general fact (d) then the fact Ga must exist. The existence of the general fact guarantees that, if the facts which is the cause obtains then it is required that the fact which is the effect also obtains. Since the general fact exists in its own right, this guarantee is ontological and objective, there in the world, linking cause and effect. Second: The general fact (d) connects the atomic facts as (c) requires in a way that is in effect truth-functional: there is nothing about the logical 62. That is, characteristics considered as universals but not necessarily as (Platonic) forms – though there are some defenders of the idea that nomological necessity derives from a relation among characteristics who take those characteristics to be universals in the sense of (Platonic) forms. Thus, Brown, Laboratory of the Mind, apparently holds that causal necessity derives from a relation among properties or characteristics which he takes to be Platonic forms; and Brown follows Broad who made the same proposal in his Examination of McTaggart’s Philosophy. The difference is of course simple. One can, like Bergmann (and like G. E. Moore), take characteristics to be universals and also hold that such universals are given to one in sense experience; but if they are taken to be Platonic forms then they are taken to be entities that are not given in sense, but only in some sort of rational or (as Broad put it) non-perceptual intuition.
42
form of (d) that conflicts with the standard truth-functional logic deriving from Russell’s (and Whitehead’s) Principia Mathematica by way of Wittgenstein’s Tractatus. At the same time, the general fact (d) connects the atomic facts in a way that clearly depends on the kinds occurring in those atomic facts. Third: The kinds are straight forwardly “natural kinds” simply because they do exist as kinds which occur in the general fact (d), or, if you wish, in (dd) understood as a relational fact connecting characteristics of things understood as universals . Fourth: The causal tie must vary from context to context. But whatever the form of the causal connection, the general fact of the required form will exist. Fifth: If the general fact (d) constitutes the causal tie (c), then automatically the regularity (r) obtains. As Bergmann would have it, (r) simply is the general fact (d) that ties together the causally connected atomic facts: so, of necessity the fact (d) that ties together the atomic facts in (c) implies regularity (r). To put the same point in another way, one that makes clear that Earman’s objection is met, the fact (d), in the form of (dd), guarantees that (de) is analytic and that the regularity (r), following from (dd) and (de), holds. Thus, Bergmann’s irreducible general facts have the five crucial features that any connection must have if it is to constitute the causal nexus that is supposed to tie together individual facts with an objective necessity. I conclude that the general fact that Bergmann insists exists over and above its instances constitutes what is in effect a causal tie or nexus that structures the kinds that occur in it in such a way that it is guaranteed if something falls under the one kind (or kinds) it will also be of the other kind (or kinds). Notice, first, that it links the kinds, as defenders of causal ties from Aristotle to Broad to Dretske have required; but notice, second, that it so binds these that the relevant pattern obtains amongst the individuals exemplifying those kinds, as the tie proposed by Broad or Dretske does not do, i.e., it meets the Bergmann-Earman objection to causal ties; and, notice, third and finally, that it secures as its reflection among observed particulars the pattern required by the connection amongst the characteristics without introducing the active soul or substance by which that correspondence was secured by Plato and Aristotle. What I am arguing, of course, is that Bergmann’s irreducible general
43
facts constitute a sort of Aristotelianism – a hidden Aristotelianism.63 But of course we have also listed two epistemological problems. How does Bergmann’s tie, his irreducible general facts, fare with these? The first epistemological problem has to do with the analyticsynthetic distinction. Any proposed causal tie or connective must satisfy the condition of being truth-functional; otherwise the traditional explication of the analytic-synthetic distinction is compromised. But there is nothing in the logical form of the general fact (d) that is in conflict with the standard 63. Bergmann, in his posthumous New Foundations of Ontology, offers a revised account of general facts. Here Bergmann wishes to rid his ideal language of variables, since they cannot, he suggests, reasonably be held to represent anything existent, even by the ontologizer who aims to create the most luxuriant of worlds. Now, he proposes, the fact that all F’s are G’s is “built” by the function V “not just from one argument but, indifferently, from an indefinite number of alternative arguments ...from , from , and , and so on” (p. 235, I have modified Bergmann’s notation slightly). The function V thus effects a many-one mapping from the pairs or 2-tuples , etc., onto the general fact. The ontological analysis of the general fact is not “(x)(Fx e Gx)” but rather V[a, Fa e Ga] or any of its “variants” such as V[b, Fb e Gb]. Bergmann suggests that “Each [of these variants] is an alternative assay of the one and same general fact.” (p. 232) The mind, in having before it, the general fact also has before it the sentence ‘(x)(Fx e Gx)’ as its text. In the text there is no hint, or perhaps only a hint, of the multiplicity of assays. In thinking the general fact one also thinks the text. As he puts it, “One cannot believe, or doubt, or remember, and so on, any generality without perceiving the appropriate words” (p. 204, his italics). This part of the more general thesis that “all awarenesses, except primary Perceivings and Imaginings (and undoubtedly some ‘Feelings’), are inseparable from their texts.” (p. 234) All this is important for what Bergmann is about in New Foundations. One of these things is to get rid of the variables. For what is conventionally represented by ‘(x)(Fx)’, the singular fact Fa won’t do rather than the 2-tuple because “V Fa” does not make it explicit which constituent of the atomic fact it is that the quantifier operates upon (which is the job customarily done by the quantified variable). The quantifier must both retain the atomic fact and also recognize as it were the constituent being generalized over. So the generality operator V must operate on both, that is, on the 2-tuple (p. 168). But all this makes clear that the general fact does, still, satisfy what we have called the five crucial features that something must fulfill if it is to count as an irreducible causal tie. Indeed, the new assay makes it even more clear than before that the general fact satisfies these five crucial features. So the changes in Bergmann’s assay of general facts does not affect what we are about.
44
explication of the analytic-synthetic distinction. So this first epistemological problem is solved. The second epistemological problem involves our knowledge of the proposed tie: the objection to the tie as proposed by Plato and Aristotle or as proposed more recently by Broad and McTaggart and even more recently by Dretske, Tooley and others is not given to us in our ordinary experience of the world. This is a serious objection to Bergmann’s view – it is essentially the Humean objection to any sort of tie meant to secure objective causal necessities. He attempts to meet this objection head-on: he asserts that he is acquainted with some cases of the general facts which he takes to be facts over and above the individual facts that are presented. It might be argued that Bergmann claims the general fact is something that he is not really acquainted with, and that he has argued himself into thinking that he really is presented with an irreducible general fact. Certainly, he does offer an argument that there must be general facts, an argument which appears originally in Russell. I think there is a point to such a comment. In any case, there are two aspects to Bergmann’s case for general facts, the argument deriving from Russell, and the claim that he (Bergmann) is acquainted with such facts. To be sure, there are only a few general facts with which he is acquainted. So most of our causal judgments are simply guesses as it were that the regularities of the world are this way rather than that. For most of our causal judgments we judge that a regularity obtains but we are not acquainted with the general fact we would have to know, have to be acquainted with, if we are to know that the regularity we are guessing at really is a genuine causal regularity – one where the separable atomic facts that have that pattern are not merely conjoined but are genuinely tied together by an irreducible general fact. So most of our causal judgments are simply as the Humean says they are. But at least we do know some irreducible general facts, and therefore know the entities that in these facts constitute the causal nexus. This means that Bergmann is at least one step better than the Platonist or the Aristotelian or Broad or Dretske. Unlike these latter defenders of the causal nexus or nomological tie, Bergmann can at least claim to know the sort of entity that the tie is: since he knows in his experience of the world some instances of the tie, he can claim to know the sort of entity it is. Where these others claim to know what the causal tie is but can in fact give no instances in our experience of that tie, they are confined in practice to holding that causal 45
judgments are judgments of regularity only: in their theories they may be anti-Humean but in their practice they are after all Humeans. Bergmann, too, is in practice like them, a Humean – “merely” a Humean, if you wish – , but unlike them he can reasonably claim to know what it is that a regularity must be like if it is to be genuinely causal and not merely a “mere” regularity. That is, he can claim reasonably so to know, provided that the case he makes for the existence of generality is sound. That case we must now examine. There is, in the first place, the argument, deriving from Russell, for the existence of general facts. And there is, second, he claim to be acquainted with generality. We can deal with them in order. The argument deriving from Russell is this.64 The ontologist aims to give a complete description of the world. Let us suppose we have a small world of four objects, two particulars a and b, and two characters F and G, where these form the two facts (i) Fa, Gb It would seem that the sentences (i) give a complete description of the facts of this world. But Russell argues that the description is complete only if we include in our knowledge claims about the world that these are all the facts in this world, only if we know, for example, that the only particulars in this world are a and b – in symbols (ii) (x)(x = a v x = b) Our description of the model world is not complete if all we have are the descriptions (i), it is complete only if we add (ii) to the facts we know. So, besides the individual facts (i) we must include in our inventory of facts the general fact (ii). The argument can be put in another way. Suppose our world consists of a and b and that these are both H. That means that the atomic facts are given by the list (iii) Ha, Hb Instead of saying that the world is described by the list (iii), we might say that it is described by the conjunction (iv) Ha & Hb But the statement that (v) (x)(Hx) 64. Bergmann, “Generality and Existence,” pp. 69-70; Russell, “The Philosophy of Logical Atomism,” p. 236.
46
is stronger than (iv): according to the usual rules of logic, (v) entails (iv) but (iv) does not entail (v). So, in order to make our description logically complete we must include (v) among the list of facts that exist. Generality is indispensable. Here is how Russell puts it: It is perfectly clear, I think, that when you have enumerated all the atomic facts in the world, it is a further fact about the world that those are all the atomic facts there are about the world, and that is just as much an objective fact about the world as any of them are. It is clear, I think, that you must admit general facts distinct from and over and above particular facts.65
Or so the argument goes. But it seems to me not to be sound. Consider the world (i). Surely, if in this world a is F and b is G, then (i) does give a complete picture of how things stand in that world. If I am asked to draw a picture of, let us say, our cat, then I draw the face, the mouth, the nose, the ears, the whiskers, and so on, through all the features of the cat. I then give you the picture. I do not then have to add that those are all the features of the cat. It is built into the idea of giving you a fair picture of the cat that everything is included in the drawing that I hand you. It is true that I have given all the features in what I have drawn, but it is not an additional feature that just those are all the features. Similarly, for the world of (i), this list provides a complete picture of what obtains in that world: there is no additional fact that just this is everything that is in the world. Similarly for the world (iii) where everything is H: it is in the idea of a complete picture of that world that the list (iii) includes everything that is there to be pictured, and that that picture shows that everything is H. It is not necessary to include as an additional part of the picture the universal statement (v) representing as part of the picture the somehow additional the general fact (v). So, the general statement is not indispensable to a description of the world. So the Bergmann-Russell argument for the existence of general facts fails. This argument against general facts is, I think, essentially Wittgenstein’s. In the Tractatus he tells us that the world is a world of (atomic) 65. Russell, “The Philosophy of Logical Atomism,” p. 236.
47
facts, “everything that is the case.” These facts are pictured by elementary propositions. It is part of the very idea of giving a complete picture of the world that every (atomic) fact is pictured, and the names of objects are such that every name occurs in at least one picture. If the objects [in the world] are given, then at the same time we are given all objects. If elementary propositions are given, then at the same time all elementary propositions are given. (Tractatus, pr. 5.524)
A universal proposition is a conjunction of elementary propositions. If η has as its values all the values of a function fx for all values of x, then N(ή) = ~(›x)(fx). (Tractatus, pr. 5.52)
In the notation of the Tractatus, ‘η’ is a variable that takes propositions as its values, and ‘N(ή)’ represents their joint denial.66 Thus, if we have ‘fa’ and ‘fb’ as values of η, then ‘N(ή)’ is ~fa & ~fb which is equivalent to ~(fa v fb) which, Wittgenstein is saying, is equivalent to ~(›x)(fx) And this is equivalent to (x)(~fx) The point is that here we see Wittgenstein affirming that the existentially quantified statement is equivalent to a disjunction of atomic propositions, which means in turn that a universally quantified statement is equivalent to a conjunction of atomic statements. So, if the list fa, fb gives a complete picture of the world (this is a really small world, but that does not affect the point), then the generality Everything is f , that is, the statement that (x)(fx) amounts simply to the statement that 66. Actually, the Tractatus uses the Greek letter ‘ξ’ and then the same letter with a bar over it. I couldn’t duplicate the latter with my word processor, so I made a substitution which does the required job.
48
fa & fb This what Wittgenstein meant when he said that Any proposition is a truth-function of elementary propositions. (Tractatus, pr. 5)67 Any general proposition describing the world is simply a conjunction of atomic propositions. There is no general fact over and above the individual or particular atomic facts. Russell was later to make much the same point in his Inquiry into Meaning and Truth: Theoretically, given sufficient capacity, we could express in the objectlanguage every non-linguistic occurrence.68
that is, every non-linguistic occurrence in the world that we are acquainted with in our experience of things, and where, as he speaks, “The words ... [which are] characteristic of logic”, including ‘true,’ ‘false,’ ‘not,’ ‘some,’ and (what is important for us) ‘all’, are “words which occur in the secondary language and in all higher languages, but not in the objectlanguage,”69 that is, not in the language that records the facts of the world as given to us in sensible experience. As he went on to put the point, “‘Had we but enough time,’ we could dispense with general propositions. Instead of ‘all humans are mortal’, we could say ‘Socrates is mortal,’ ‘Plato is mortal,’ [‘Hypatia is mortal,’] and so on.”70 We do need general propositions, but not in our picturing of the world as we experience it, and therefore not as part of our ontology: In fact, however, this [enumerating ‘Socrates is mortal’ etc.] would take too 67. I have slightly altered the translation. Pears and McGuinness have the translation, “A proposition is a truth-function of elementary propositions.” This English translation is slightly misleading. The German is “Der Satz ...” which has the meaning of “the proposition ...”, that is, “any proposition...” and not simply “a proposition...”, as the translators would have it. Compare the phrase “The whale” in “The whale is a mammal”: the latter is understood as “any whale is a mammal” or “all whales are mammals”. 68. Russell, Inquiry, p. 94. 69. Russell, Inquiry, p. 95. 70. Russell, Inquiry, p. 320.
49
long, and our vocabulary of names is insufficient. We must therefore use general propositions. But the subjective character of logical words appears in this, that the state of the world which makes a general proposition true can only be indicated by means of a general proposition.71
Russell is here arguing that his former commitment to the existence of general facts alongside atomic facts is mistaken. It is indeed the case, as he argued in the “Logical Atomism” lectures, and as Bergmann also argued, that the general proposition is indispensable, and says something we cannot say with atomic propositions alone. In the logic of our language, generality is indispensable. But in the language that is used to picture the world (what Russell here calls the “object-language”), there are only statements of atomic fact. As we go about describing the world, statements of general fact are indispensable, but the world that we are describing does not contain anything but atomic facts.72 So, once again we may conclude that there is no general fact over and above the individual or particular atomic facts. Herbert Hochberg has criticized this argument against the existence of irreducible general facts.73 He does not suggest that every regularity is an irreducible general fact – some regularities are just regularities – , but he does argue that there are some generalities that must be taken a irreducible in one’s ontology. Consider a three-membered universe: {a, b, c} Hochberg argues that it is a fact and, indeed, a irreducible fact that this is a three-membered universe. This fact, he argues, not merely shows itself, as we have argued, following Wittgenstein, but must be recognized in the truth of the generality (x)(x = a v x = b v x = c) which gives the structure of the universe as three-membered. If we have, say, the property F in this universe, and have moreover that Fa 71. Ibid. 72. Russell, Inquiry, pp. 319-320. 73. H. Hochberg, “Negation and Generality,” in his Logic, Ontology and Language: Essays on Truth and Reality (Munich: Philosophia Verlag, 1984), pp. 296-312, at p. 303ff.
50
then Hochberg’s point is that the fact that Fa is irreducible in this universe and that the generality describing the universe as being three-membered is equally a fact and equally a fact that is irreducible. One can, however, make the same point without acknowledging that there exists a general fact to that effect, a fact that is there but is irreducible to any non-general fact about the universe. At least, so I would suggest. Hochberg’s point requires the special predicate x=avx=bvx=c which specifies the universe by enumerating its members. Let us abbreviate this predicate by the expression Ex Call such a predicate an “enumeration predicate.” Clearly, such a predicate can be defined for any universe (so long as it is finite – but that’s another story, to consider the infinite universe would take us far afield). Now, the predicate ‘Ex’ that we have just defined holds of or is true of each and every member of the universe, there is no member for which it is not true. If we now look at a four member universe {a, b, c, d} then our just-defined ‘Ex’ is true of three of its members but not of the fourth. And if we now look at a two member universe {a, b} then if we apply ‘Ex’ then its first two disjuncts are true but the third disjunct is meaningless since the term ‘c’ does not refer to any entity in this universe (assuming the rule of unum nomen unum nominatum). Evidently, for each universe there is an enumeration predicate, and only one enumeration predicate such that (1) it applies to each individual in the universe, (2) it applies to every individual in the universe, and (3) it contains no disjunct that is meaningless because the constant which appears in that disjunct does not refer. Call such a predicate the enumeration predicate characteristic of that universe. So ‘Ex’ is the enumeration predicate characteristic of our little three-membered universe. Hochberg holds that we cannot capture the fact that our threemembered universe is of a certain size, that is, is three-membered, by holding that its three-memberedness shows itself in any picture of that world (as Wittgenstein held and as we have argued). Rather, we can capture that fact of three-memberedness only if we hold the generalization (x)(Ex) to be true (which it is) and to be irreducible to any statement or combination of statements about the atomic facts of that universe. But surely all 51
that we need to recognize is that the conjunction Ea & Eb & Ec is true, that there is no conjunct of ‘Ex’ which is meaningless, and that there is no statement of the form Ex for example Ed which is false. Surely recognizing that the enumeration predicate characteristic of an n-membered universe is true of each and every one of the members of the universe and that it has no meaningless disjuncts, captures as it were the truth the universe is n-membered: it is not necessary to go beyond this and add And, ah yes, besides all this, it is also the case that there is an additional fact represented by ‘(x)(Ex)’ where ‘Ex’ is the enumeration predicate characteristic of the universe. So Hochberg is wrong: there is no need to include among the list of irreducible facts in the universe besides the atomic facts also the general fact represented by that universal generalization which states that the enumeration predicate characteristic of the universe holds of every individual in the universe. I therefore reject Hochberg’s argument for the existence of at least a few irreducible general facts. The criticism that we have developed following Wittgenstein and Russell of the Russell-Bergmann74 argument for the existence of general facts based on the logical indispensability of generality seems to be a sound criticism. Return now to Bergmann, to the other argument that he gives for the existence of generality, the argument based on the claim that he is acquainted in his experience of the world with generality as part of the logical structure of that world. Bergmann has argued consistently that ontology should be done in conformity with the Principle of Acquaintance – the rule that nothing is to be admitted into one’s ontology unless one is acquainted with that entity or with entities of that sort. In his case of the ontologization of generality, his making it into an entity in his ontology, is justified by the Principle of Acquaintance. That is, he claims that generality is presented to him in his experience of the world. 74. This latter Russell is of course Russell at an earlier stage.
52
I merely claim that (a) generality is presented on some such occasions [that is, occasions when “one uses or could use, for the best of reasons, a general sentence”], and (b) if it were not, we would not on any occasion know what the quantifier meant.75
“I take (b) to be obvious,” he comments.76 But in reply one might perhaps suggest that he is not so much acquainted with generality as convinced that he is acquainted with it. In any case, he gives the following example.77 We have two individuals a and b. Individual a has the property C of being a circle, while b has the property S of being a square. Besides the characteristics C and S, there is Rxy the relational characteristic of x being inside y. The situation is this: a is inside b, and is the only square inside b. Bergmann supposes himself to say (!) This square is the only one inside this circle The transcription of (!) is not (+) Rab that is, not this alone, but rather the conjunction of this together with (++) (x)(x = a v ~Rxb) or, more fully, taking into account the characteristics square and circle, (x)[(Sx & Rxb) e x = a] Bergmann now notes that (++) contains the generality operator. He then comments that I conclude that generality is on this occasion presented to me in connection with the fact [Rab].
and adds that Instead of saying that it is (on this occasion) presented in connection with [Rab], one might as well say that it is (on this occasion) presented 78 in addition to [Rab].
75. Bergmann, “Generality and Existence,” p. 70. 76. Ibid. 77. Bergmann, “Generality and Existence,” p. 71. 78. Ibid.
53
What can one say? Take the situation that Bergmann has described. It is there, in the world, given to us in our sense experience of that world and of those facts in the world. If one is asked to give a linguistic picture of the situation, then one would give: (s) Sa , Cb, Rab That is how one would describe the situation, picture it in language. The point is that (s) is the complete picture. To be sure, not everything that we know about the situation is there in the linguistic picture we make of it. Thus, from the picture we can see that the square is inside the circle and that it is the only square that is inside the circle. That this is true, is something we discover in looking at the picture. It is not something that is said in the linguistic picture (s), as a fact alongside as it were the facts that are pictured, that is, the “other facts.” It is not there as part of the picture, but still, it is a fact that is shown to us by the picture we make of the situation when we made a complete picture of the situation. In that picture, it is shown that b is the only square inside the circle a: if there were another circle there then there would be another atomic fact of the form Sx and another relational fact of the form Rxb. Given that the picture is complete, given that it is a complete picture of this little world, then there is no need to add that in addition to the facts pictured, there is beside them the further fact that b is the only square inside the circle a. Bergmann’s situation is given in sense experience, and we know about that situation what Bergmann records in the quantified statement. But that knowledge is not given to us in our sense experience of the world. That general fact is before the mind, but it is not a fact with which we are acquainted, that is, the general fact which is before our mind is not one with which we are acquainted in our ordinary sense experience of the world. It derives from our thinking about that world, but is not in the world as we sensibly experience it. That is why, as Russell put it in the Inquiry, the description of the world as we experience it appears in our object-language but generality as a logical concept appears only in our secondary language which we use as we reflect in thought about the world described in the object-language. I conclude that Bergmann’s claim that he is acquainted with generality is not sound. However, it must also be said that he general fact is certainly before his mind, it is in the intention of an act of thinking about the situation – indeed, one cannot doubt that that act, that mode of thinking, can be called perceptual: the general fact is not in the world as we know it in our sense experience, but nonetheless it is something about the world that we per54
ceive. Indeed, Bergmann later came to the conclusion that generality is in fact not given in our sensible experience of the world. In his New Foundations of Ontology, our sensible experience of the world consists of acts which he refers to as “primary Perceivings and Imaginings (and undoubtedly some ‘Feelings’).”79 But general facts are, ho now acknowledges, entities that we cannot perceive or imagine. The awareness of a general fact must be a believing or an entertaining, not a perceiving or an imagining.80 Bergmann is therefore here granting Russell’s point and ours that general facts are not really among the facts of the world as it is given to us in sensible experience. They are therefore excluded by the Principle of Acquaintance. But Bergmann gives up that Principle, at least in its original sense, where ‘acquaintance’ means ‘sensible awareness.’ He now expands ‘acquaintance’ to mean any act of awareness, thus including believings and entertainings as being ways of being acquainted with facts. He now holds that “whatever is thinkable exists.”81 Thus, the intention of any act exists. And we can think of or about general facts. So they exist. But why accept that principle? Surely it gets us into insurmountable problems. Thus, I can think that Toronto is east of Brockville For, that is something I disbelieve, and in disbelieving it, I am thinking it, and so it exists. So Bergmann’s new Principle of Acquaintance commits us to the existence of false facts – surely something intolerable, even if these false facts are pervaded by the mode of potentiality rather than actuality. Put in these terms, one can begin to see how Bergmann could come to hold that, since a general fact is the object or intention of thought, he is acquainted with a fact existing alongside other facts, other intentional objects of thought. We therefore have as the intention of a thought a fact, namely, the general fact, that we are not presented with in our sensible experience of the world. What, then, do we say about intentional objects? Bergmann says they all exist. Is that so? That seems intolerable as a conclusion, no matter what Bergmann says about such things as the mode of potentiality. We must ask, what gets one into this problem? 79. Bergmann, New Foundations of Ontology, p. 234. 80. Bergmann, New Foundations, p. 219. 81. Bergmann, New Foundations, p. 61.
55
Not every intentional object exists. Yet Bergmann insists that every such object does exist. There seems to be something that has gone wrong with how he is thinking about intentionality. So, let us recall Bergmann’s formula for representing thoughts: ┌p┐Mp which says that the thought that p means (the fact) that p. ‘p’ as a sentence is the text of the thought. This text is perhaps actually said or, as most often occurs, there is a short-term disposition to utter that text. ‘┌ p ┐’ is the name of a simple character, the thought. So at least Bergmann proposes. Actually, it would seem better to think of ‘┌ p ┐’ as a definite description: it is a definite description to this effect, “the thought of which ‘p’ is the text.”82 The important feature in Bergmann’s account of intentionality is ‘M’. When Bergmann first introduced this to represent the intentionality of thought, he thought of ‘M’ as linguistic, and that a statement to the effect that “┌ p ┐ M p” is true by virtue of its linguistic form alone – which makes it a sort of logical truth, since, like all the positivists, he took logical truths to be truths which hold by virtue of their linguistic form alone.83 But he gradually came to the conclusion that, like all of logic, this formal fact too had to be ontologized: the linguistic ‘M’ had to be construed as representing something existing in the world.84 This last meant treating M as a sort of a relation, that is, a relation similar to an ordinary relation like kicks. But if it is true that Beverley kicks Georgie then both the relata, Beverley and Georgie, must exist. So, if M is a relation, then its relata must exist. Hence, if we have ┌p┐Mp then both relata, and in particular the relatum p must exist. Hence, if we are 82. See F. Wilson, “Effability, Ontology and Method,” Philosophy Research Archives, 9 (1983), pp. 419-470. 83. See for example, Bergmann, “Professor Ayer’s Analysis of Knowing,” in his Metaphysics of Logical Positivism, pp. 214-227. 84. See G. Bergmann, “Realistic Postscript,” in his Logic and Reality, 302-340; and so “Acts,” bid., pp. 3-44. For a critical discussion of this journey that Bergmann made from logic as linguistic to logical realism, see H. Hochberg, The Positivist and the Ontologist: Bergmann, Carnap, and Logical Realism. See also W. Heald’s Introduction to Bergmann’s New Foundations of Ontology.
56
thinking of or aware of or entertaining the fact that (++) (x)(x = a v ~Rxb) then we have ┌ (x)(x = a v ~Rxb) ┐ M (x)(x = a v ~Rxb) and the second relatum must exist. Similarly, if Bergmann is presented with the fact about the sensible world that Ca then we have ┌ Ca ┐ M Ca and again the second relatum exists. So, when Bergmann is aware of the fact that Ca which exists, he similarly is aware of the general fact (++) which also exists, exists alongside the individual fact Ca. So there we have it: there are general facts, there in the world, alongside particular or atomic facts. We can think them so they must exist. There is a mistake here. It is just wrong, we have been arguing, to hold that the general fact exists alongside the particular facts; such a generality does not exist (though the generality often enough is true). Something has clearly gone wrong. What has gone wrong, I suggest, is Bergmann’s treatment of M as a relation. More correct is his original view that this represents something linguistic, rather than being something like a real relation similar to kicks. If we think of ‘p’ means p as a linguistic statement about the meaning of the sentence ‘p’, then means, that is, M, will allow us to have the thought that p, i.e., ┌p┐, to have p as its object without there being any requirement that p exist. Thus, in Toronto is east of Brockville’ means Toronto is east of Brockville we have a true statement of meaning even though it is false that, and therefore not an existing fact that Toronto is east of Brockville This, as we have said, is intolerable: there are not in the world alongside the (positive) facts another set of entities, the false facts. Bergmann is committed to the view that these objects of thought which do not exist do after all exist.85 Surely we must, as Russell once put it, as philosophers have a robust sense of reality; false facts violate, or should violate, any85. Cf. F. Wilson, Critical Review of R. Grossmann, The Categorial Structure of the World, Canadian Journal of Philosophy, 16 (1986), pp. 163-180.
57
one’s robust sense of reality. Besides, and this is what is crucial, we are not acquainted with them as being in the world as it is given to us in our sense experience of it: they violate the Principle of Acquaintance. Thus, thinking of the M in ‘p’ M p as linguistic yields an account of intentionality that does not force one into the position in which Bergmann finds himself, that of requiring that the object of thought, the intention, whatever it may be, exist. One can accept the Principle of Acquaintance without requiring oneself to hold that every object of thought is a case where we are acquainted with something that exists. Treating ‘M’ as linguistic enables one to hold, on the one hand, that generalities are among the objects of thought, while also holding, on the other hand, that such an object of thought is not a fact alongside the facts we are acquainted with in our sensible experience of the ordinary world. The sentence ‘p’ is the text of the thought. Its meaning, in the linguistic sense, is given by the various patterns of language that govern it and the terms that occur in it – these are regularities and dispositions that constitute world-word connections (e.g., rules of reference), word-word connections (e.g., rules of syntax, rules of logic, statements of generality), and word-world connections (rules for speech acts), and involve, in complicated ways, the causal role of rule expressions (resolutives and imperatives).86 This meaning that determines what ‘p’ means, and in particular if it is atomic what it refers to and pictures, is a matter of patterns and dispositions. These by their very nature are not wholly present in consciousness when we use ‘p’ or are disposed to use ‘p’. Patterns and dispositions are known by inference, yet when we use ‘p’ we are aware non-inferentially of that meaning: There is present in consciousness a characteristic or feature that is the simple feature of that conscious state that causes us to use or be disposed to use ‘p’. ‘┌ p ┐’ is a definite description of this feature: it is the feature that causes us to utter or be disposed to utter ‘p’, the text of the conscious state; and ┌p┐Mp 86. Cf. W. Sellars, “Some Reflections on Language Games,” in his Science, Perception and Reality (London: Routledge, 1963), pp. 321-358. See also F. Wilson, “Marras on Sellars on Thought and Language,” Philosophical Studies, 28 (l975), pp. 91-102, and “Review of Tuomela, Human Action and Its Explanation,” Dialogue, 21 (1982), pp. 571-577.
58
represents the fact that the feature referred to by ‘┌ p ┐‘ causes the fact that one utters or is disposed to utter the text ‘p’. Thus, the meaning of the conscious state, what it intends, is given by the linguistic patterns governing ‘p’ and the terms that occur in it. ‘M’ gives it that the conscious state has as its meaning or intention the fact that ‘p’ means, it also represents that there is present in consciousness the thought that causes us to utter or to be disposed to utter the sentence ‘p’ with that meaning. Understanding intentionality in this way means that we are not treating the text of every act of awareness, every intentional state, as having an object that has any ontological status. And so, perceiving that the a is the only square inside the circle b does not require one to give ontological status to the general fact. However, there remains another point that Bergmann makes, again following Russell. Logically speaking, the generalization (x)(Fx) is stronger than Fa & Fb & Fc & ... no matter the length of the conjunction. So “generality” cannot mean “very long conjunction.” What, then, does it mean? Bergmann suggests that one couldn’t know the meaning of generality unless it was an entity that is presented to one. Since there are no general facts over and above the atomic facts, or so we have argued, we cannot have acquired our concept of generality from acquaintance with the entity generality as part of a general fact. We have, therefore, yet to explain the meaning of the concept and how we acquire that meaning. Wittgenstein makes the relevant point, I think. He holds, correctly we are arguing, that ontologically there is no difference between the general proposition ‘(x)(Fx)’ and the conjunction ‘Fa & Fb & Fc & ...’: recall his comment that Any proposition is a truth-function of elementary propositions. (Tractatus, pr. 5.53)
This does not mean that generality is the same thing as ‘(long) conjunction’. Thus, he tells us that What is peculiar to the generality-sign [as in ‘(x)(Fx)’] is first, that it indicates a logical prototype, and second that it gives prominence to constants. (Tractatus,
59
pr. 5.522)
What ‘(x)(Fx)’ does is give us a prototype for a fact: this prototype is ‘Fx’.87 This prototype gives prominence to, in particular, the constant ‘F’: the prototype is therefore the prototype of a fact in which the characteristic F is exemplified. Thus, what ‘(x)(Fx)’ means is more than what a conjunction means: it means that as one continues to construct pictures of the facts of the world, that is, as one continues to give the elementary propositions that picture the facts of the world, then every individual no matter what, will fit the prototype ‘Fx’: for every individual there is or will be or was, that individual is prototypically F. The generalization is not a conjunction but a declaration that pictures of the world will have in them facts that fits this prototype, an anticipation of the structure of facts in the picture of the world, whatever picture it is, however large the world becomes. This, then, is the meaning of the general quantifier. The meaning of an elementary proposition is given in terms of the state of affairs that it pictures. An elementary proposition is correct as a description of the world just in case that the state of affairs that it pictures is there in the world as it is given to us. The meaning of the general proposition lies in its giving a picture of the pictures that will be there in any complete description of the world.88 87. Russell makes more or less the same point in his, Inquiry into Meaning and Truth: what is important about a general proposition is that it gives a prototypical form for statements of individual fact. “When we judge ‘A is a man but not mortal’, we accept ‘A is a man’ but we reject ‘A is mortal.’ The various acts of this kind, putting B, C, etc. in place of A, all have something in common; what they have in common is a belief expressed by the words ‘some man is not mortal.’ When we reject this belief, we are in a state expressed by the words ‘all men are mortal’.” (Inquiry, pp. 319-320) And again, Russell remarks that, when I believe that everything is F, then I believe that Fx is true for all values of x. a may be such a value, but I may never have heard of a. Thus, “the belief that [Fa] is one belief, and the generality is part of that belief. Moreover, it is intensional in the sense that I can have the belief without knowing all [individuals] there are. As soon as I understand the [word F], [and] the subjectpredicate form, ... I have everything, except generality, that is required for understanding [(x)(Fx)].” (Inquiry, p. 317) 88. As Russell was to put it in his Inquiry into Meaning and Truth (p. 319), the elementary propositions form a sort of object language describing the world as we experience it, and the general proposition is as it were meta-linguistic, in a language that comments on that object language.
60
As for a regularity of the form (x)(Fx e Gx) this too should be understood in terms of prototypes. It says that if a fact fits the prototype Fx then it will also fit the prototype Gx while if there is a fact that does not fit the first of these prototypes then it may or may not fit the other prototype. It lays down, in other words, that future pictures will fit one of these three prototypes: Fx, Gx ~Fx, Gx ~Fx, ~Gx89 And now we are brought back to Socrates and Anaxagoras. The world as we experience it consists of separable parts, parts that are logically and ontologically separable, parts that are conjoined but merely conjoined. If Anaxagoras is correct, then there is nothing more, only these parts and that patterns that describe how parts of just these sorts are associated with parts of just these other sorts. If Socrates is correct, then this account of the world misses something: what it misses is the tie that provides the reason why these parts are associated with those parts. Why is there this pattern rather than that? Because, Socrates proposes, there is a reason for this, an ontological ground, for this pattern being more than a mere pattern, but instead one that must hold of the world as we experience it, a ground or tie which makes of this pattern more than a “mere” pattern. Recall that Socrates put his point in terms of his own situation. Here he is, Socrates, sitting in prison. According to the earlier dialogue, the Crito, his friends offered to arrange things so he could escape to Thebes. But Socrates refused: following the demands of justice he determined that it would be wrong to escape his native city and run off to Thebes. He is, instead, determined to remain in his cell and drink the hemlock. So, we have one event, Socrates sitting in his cell. There are two events that could follow that event, namely, running off to Thebes and remaining in Athens and drinking the hemlock. There is nothing in the first event, Socrates sitting there, that indicates or determines that it will be followed by one 89. Let us agree here that we will not worry here about the meaning of negation, that is, about whether there are negative facts as well as positive facts. For this issue, see F. Wilson, Review of Grossmann’s Categorial Structure of the World.
61
rather than the other. But in fact it is the drinking of the hemlock that follows. Now, Socrates complained that Anaxagoras account of things could not explain this sequence that actually occurs: the explanation must be in terms of Socrates “aiming at the best”. But Anaxagoras’ scheme has no model of the “best” for Socrates to aim at: there are no ideal forms and in particular no ideal form of human justice. Lacking an objective standard of virtue, Anaxagoras cannot explain Socrates’ action. So at least Socrates argues. But (here we are giving a helping hand to Anaxagoras), Anaxagoras need not deny that Socrates is aiming at the “best.” But this best is what Socrates values as the best. Anaxagoras can argue that Socrates is in fact moved towards this goal that he identifies as the best, but that there is nothing to this beyond that valuing, no objective standard in the ideal forms of things. At least, so he could argue: certainly, that is the way that Hume was much later to argue when he developed a similar case against the Platonists and other moral objectivists of his own day. There are no objective values but there are relative values. Anaxagoras thus has a reply to Socrates: he can explain the events that Socrates wishes to explain without any appeal to transcendent entities, either simple souls or ideal forms, entities not of this world that somehow effect a tie of necessity between events in the ordinary world. All he needs is regularity: Socrates values what he feels to be just, and we know that he regularly acts to make the world, himself in particular, be as justice, so envisioned, requires. Relative values and regularities about things in the world, including people, suffice. There is no need for objective values or forms nor is there any need for a tie or causal nexus or objective necessary connection. Socrates and after him Plato and Aristotle, and then, in our own age McTaggart and Broad, and more recently Dretske, have all sought for such a tie. Fruitlessly: their claims notwithstanding, none have passed the test established by the Principle of Acquaintance. We seem to be left with the atomism of Anaxagoras, of Hume, and of the logical positivists, and, among the last, Bergmann. Interestingly, we found that Bergmann, starting as a logical positivist, came to a position on the ontology of logic which seemed in a way to do what Socrates & Co. were trying to do, namely, give a tie that would bind the separable facts of our sensible experience into necessary unities. This came when he argued that besides the particular facts that we find in the world as we experience it, there are also general facts. But when we examined those arguments, one based on the inferences 62
of logic and one based on acquaintance, we found that neither of them withstood scrutiny. We found both these rooted in a mistake that Bergmann made about the “relation” – the apparent “relation” – of intentionality. Thinking of it as a real relation, he naturally concluded that there really were general facts that were given to him in our perceptual experience and that he had no choice but to grant them an irreducible ontological status. But we also argued that an alternative account of intentionality, in which, as Bergmann originally held, whatever its grammatical form, it is not a real relation, would eliminate that structural reason for thinking that there are general facts.90 So we are back to the metaphysics of logical positivism from which Bergmann began his ontological journey. And here we must say that there is a certain point to what Socrates was arguing against Anaxagoras. Here I do not mean the point about value. Anaxagoras’ ontology, unlike Socrates’, lacks objective value. Anaxagoras’ ontology does have souls and they do move bodies, just as do Socrates’ souls. But unlike Socrates’ souls the minds or portions of mind that move individual things on Anaxagoras’ scheme do not transcend the world of ordinary experience. They consist of separable parts and we can see Anaxagoras’ struggling with language to describe them. They are fiery, or like fire, he says.91 Their parts are in rapid motion – we might recall Hume’s comment on selves that “they are nothing but a bundle or collection of different perceptions, which succeed each other with an inconceivable rapidity, and are in a perpetual flux and movement.”92 Recall Moore’s point that mental acts are diaphanous or transparent.93 However the fiery bits go together, the separable 90. I shouldn’t be taken, when I say this, that I am saying that the alternative account of intentionality that we have proposed is exactly what Bergmann had in mind when he first defended a “positivistic metaphysics of consciousness”: Bergmann’s suggestion was different, clearly so, from our alternative.. See G. Bergmann, “A Positivistic Metaphysics of Consciousness,” Mind, n.s. 54 (1945), pp. 193-226. It must be said that Bergmann is not entirely pellucid in his account of meaning or intentionality. For a clear discussion of Bergmann’s development, see H. Hochberg, The Positivist and the Ontologist. 91. As Kirk and Raven, The Presocratic Philosophers, note, “Anaxagoras in fact is striving, as had several of his predecessors, to imagine and describe a truly incorporeal entity” (p. 374). 92. Hume, Treatise, p. 252. 93. G. E. Moore, “The Refutation of Idealism,” in his Philosophical Papers (London:
63
parts of mind or nous, they do separate themselves from other bits of nous, and direct some at least of the corporeal groupings or bodies which emerge out the primordial mixture. Like Socrates’ souls, these groupings of fiery bits give bodies direction: they are purposive. But there are no transcendent forms that constitute a set of objective values. For Anaxagoras, if one may put it this way, there are valuings but no values, that is, no objective values, but there are things that are valued, these are the goals towards which mind or nous directs the body which it moves. Or, in other words, there are in Anaxagoras no objective values, but there are relative values, the ends towards which the bodies are directed. This is the view of Hume, of the logical positivists, and of Bergmann.94 It is not this that I mean when I say that there is a point to what Socrates says. I have in mind, rather, the point about the need for a tie. Without such a tie, without objective necessities, nothing can be explained. This is the point that Plato and Aristotle, and Broad and McTaggart and Dretske are all trying to get at. Just this sort of tie is absent from Anaxagoras – at least from what we know of Anaxagoras. That at least was the reason that formed the basis of Socrates’ criticism of Anaxagoras. Such a tie of objective necessity is also absent from the Humean world of the logical positivists and of Bergmann. This tie or necessary connection that is missing from the worlds of these philosophers would seem to be just that real connection that distinguishes those patterns or regularities that are laws and those that are “mere” regularities, mere “accidental” generalities. In fact, as we have argued, these are entities with which we are not acquainted in the world of our ordinary sensible experience. They are excluded from our ontology by the Principle of Acquaintance. But at the same time we seem to need them in order to distinguish those patterns which are lawful or necessary and those which are accidental and merely contingent. We seem, after all, to be driven, like Socrates, to find a world outside the world of ordinary experience, the world that transcends our ordinary world, but the world in which we find the tie that we must grasp if we are truly to have the reasons which explain things in the ordinary world. And so Hume gives his first definition of ‘cause’ as regularity: Kegan Paul, Trench, Trubner and Co., 1922). 94. For the latter, see Bergmann’s essay, “Ideology,” in his Metaphysics of Logical Positivism, pp. 300-326.
64
ontologically there is no distinction between laws and accidental generalities. But then he immediately goes on to acknowledge, along with Broad and McTaggart and Dretske, and of course Socrates, that after all one must distinguish two sorts of regularity, those that are causal and those that are merely accidental. A violation of an accidental generality is not impossible, but an event that violates a causal law is somehow impossible. So Hume asks that question, “What is our idea of necessity, when we say that two objects are necessarily connected together?”95 There is, after all, a necessity for which we must account. Hume saw this. It could not be some sort of objective necessity: all regularities are, objectively, equally contingent. Hume concludes that therefore such necessity as there is must be a subjective necessity. Just as there are no objective values but there are relative values, so also, although there are no objective necessary connections, there are relative necessities, subjective attitudes of felt necessity that distinguish those generalities that we take to be lawful from those that we take to be merely accidental. It is clearly possible, I suggest, to have much of Bergmann’s philosophy, e.g., almost all his philosophy of science,96 even if we give up generality. Some logical positivists were prepared to abandon the Humean concept of causation as they attempted to account for the required necessity. Carnap was to compromise on this issue when he introduced the notion of predicates introduced by “reduction sentences” to make a place in the language of science for disposition terms like ‘soluble.’97 Others were to introduce unanalysed modalities, objective causal necessities,98 when faced with the issue of justifying the assertion of counterfactual conditionals such as “If I were to heat this water, it would boil” which can be 95. Hume, Treatise, p. 155. 96. See G. Bergmann, Philosophy of Science (Madison, WI: University of Wisconsin Press, 1956). 97. See R. Carnap, “Testability and Meaning,” Philosophy of Science, 3 (1936), pp. 419-471, and 4 (1937), pp. 1-40; and also C. G. Hempel, “The Concept of Cognitive Significance,” Proceedings of the American Academy of Arts and Sciences, 80 (1951), pp. 61-77. 98. Cf. A. Pap, Semantics and Necessary Truth; An Inquiry into the Foundations of Analytic Philosophy, with a foreword by Brand Blanshard (New Haven: Yale University Press, 1966).
65
justified by “water, when heated, boils”, where I am not justified in asserting “If this coin [which is a loonie] were in my pocket, it would be silver” even though it happens – that is, happens accidentally – to be the case that “All the coins in pocket are silver.”99 Bergmann saw the way out: the difference between a statement of law and a mere accidental generality is one of context, that is, psychological context, the way we use the generality in our inferences.100 It is not that the counterfactual conditional (%) If a were F then it would be G is justified by the fact that the regularity (%%) All F are G is objectively necessary, but rather that we count (%%) as necessary because we in fact use it to support the assertion of (%), where if we didn’t so use it we would count it to be a “mere” accidental generality. Hume had made this point already. He had given his first definition of ‘cause’ as regularity. This was causation or lawfulness understood objectively. And objectively there was no necessity: objectively there is no difference between laws and accidental generalities. But for all that, Hume recognized, as Bergmann was later to recognize, there is a necessity that must be accounted for. Hume provides this account in his second definition of ‘cause’ as “An object precedent and contiguous to another, and so united with it in the imagination, that the idea of the one determines the mind to form the idea of the other, and the impression of the one to form a more lively idea of the other.”101 If C’s cause E’s, then if I observe a C, 99. The connection between laws and contrary-to-fact conditionals was emphasized by R. Chisholm, “Law Statements and Counterfactual Inference,” Analysis, 15 (1955), pp. 97-105. 100. Bergmann, “Comments on Professor Hempel’s ‘The Concept of Cognitive Significance’,” in his Metaphysics of Logical Positivism, pp. 255-267, at pp. 262-263. 101. Hume, Treatise, p. 172. For discussion of Hume on these definitions, and a defence of his position against the charge of subjectivism or even scepticism, see F. Wilson, Hume’s Defence of Causal Inference (Toronto: University of Toronto Press, 1997). John Stuart Mill’s position is similar to that of Hume; see F. Wilson, “The Logic of John Stuart Mill,” in Handbook of the History of Logic. Volume 4: British Logic in the Nineteenth Century, ed. Dov Gabbay and John Woods (The Hague: Elsevier, 2008), pp.229-281, sec. 6 (“Mill’s Defence of Deductive Logic”), pp. 248250.
66
that is, in Hume’s terms, have an impression of it, then there is an association in the mind such that I am led to expect an E, that is, I come to have the idea of an E. In this case, on the basis of the regularity I am predicting that the C I observe will be E. But I may also simply suppose that something is a C: in this case I have the idea of a C. In this case, the established association leads me to the idea that this something is also an E. If this something were C then it would be E. In other words, I use the casual regularity to make contrary to fact assertions. The causal regularity thus describes not only what is actual but what must be in the actual (predicting what must be) and placing limits on what is possible (contrary to fact assertions). Thus, a violation of a causal law is not only contrary to fact but something that is not possible. What is crucial is that this necessity is not objective but rather is subjective: a causal law simply is a regularity towards which we have the psychological attitude of being prepared to use it to predict and to make contrary to fact assertions. An accidental generality is, in contrast, one that we do not use to predict or to make contrary to fact assertions. This is our idea of necessity: it is rooted in the pragmatics of law assertions. Of course, some people assert that If a black cat were to cross my path then bad luck would ensue Here they are connecting the supposed events by the generality that Whenever a black cat crosses one’s path then one will have bad luck. This is superstition. But these people are asserting a contrary-to-fact conditional on the basis of a generality. That makes the generality lawlike, just as Water, when heated, boils is lawlike because we use it to assert conditionals like If this water were to be heated, then it would boil The regularity about water is reasonable. The one about cats is superstition. We therefore do not have an objective criterion for distinguishing those generalities which are lawlike from those which are not. But the objective
67
necessary connections of Plato and Aristotle, of Broad and McTaggart, and of Dretske provide just such an objective criterion. Surely, then, we have been too hasty in our dismissal of those ontologies. Otherwise, do we not fall into a scepticism about which generalities are laws and therefore reasonable to assert and those generalities which are not laws and which are therefore unreasonable to assert? Do we not fall into a subjectivistic relativism? But there is a way to distinguish science from superstition without invoking any mysterious objective necessary connection. In point of fact, the generalization about water is well supported by empirical evidence, the generality about black cats is not. The objection depends upon confusing the notion of treating a generality as lawlike with the notion of reasonably treating a generalization as lawlike. When we can assert a generalization on the basis of evidence that conforms to the norms of good scientific practice, then we can reasonably treat it is as lawlike, and use it to predict and to assert contrary-to-fact conditionals, and therefore to explain. Whether or not a generality is lawlike is a subjective matter, but whether or not the evidence we use to justify asserting the generality is good scientific evidence or not is an objective matter. The necessity of a law is subjective, the reasonability of such necessity is objective. We do not fall into an irrational subjectivism and scepticism about laws after all. The account of causal connection may be Humean, but that should be no cause for alarm: no subjectivistic scepticism looms before us. Indeed, given that the claim that there is an objective causal tie or necessity is one that cannot be found to hold in the world of everyday experience, the only world we know, it is the defender of objective necessities who falls into scepticism. The Humean viewpoint may be second best, given what the defender of objective necessities hopes for, but since that hope cannot be fulfilled, the second best is the best that we can do: the reasonable person will simply decide that the Humean position is the only reasonable one to adopt. Bergmann began as a positivist, accepting a Humean position on laws. For a variety of reasons he came to accept the claim that over and above particular or individual atomic facts there are irreducible general facts. He even suggested that without such general facts there could be no laws. We have examined his reasons for introducing generality as an existent into one’s ontology, and have concluded that they are in fact unsound. So the Humean account of laws and causation stands: the earlier positivist metaphysics emerges as after all defensible. 68
Space, Time, Concrete, Abstract GUIDO BONINO 1. Bergmann and contemporary ontology There are many ways of doing ontology; one could say, to use Bergmann’s words, that there are different styles of ontology. Bergmann’s own style is not very widespread today, and that in part explains why, even though ontology is enjoying a remarkable renaissance, Bergmann’s work is not much read, studied or commented on. Naturally enough, the historical and theoretical causes of such a situation are diverse and complex, as always happens with this kind of things. My aim is that of pointing out one of the reasons why Bergmann’s way of doing ontology seems outmoded today to many philosophers, and why most work done in this discipline does not display significant connections with Bergmann’s views. In other words, I am trying to show some of the differences between Bergmann’s views and the mainstream of contemporary analytic ontology. I have no original ontological theses to put forth, since the aim of the paper is, in a sense, historical. In making clear some of the assumptions underlying Bergmann’s conception of ontology, I am trying to understand the reasons of the distance between such a conception and those which are more common nowadays. The starting point may be that of providing two examples showing in what sense Bergmann’s style of ontology is felt as extraneous by many contemporary ontologists. The paper aims at explaining one of the roots of this sense of extraneousness, which of course can also be looked at from Bergmann’s point of view, according to which most of contemporary ontology would probably sound as philosophically irrelevant, or at least not properly philosophical. One of the examples comes from a recent book by Peter van Inwagen, the other has to do with the notions of concrete and abstract. In the Introduction to Ontology, Identity, and Modality1, Peter van Inwagen distinguishes between two kinds of ontology: A-ontology and Bontology, where ‘B’ explicitly stands for Bergmann, the purported champion of such a conception. The A-ontology van Inwagen puts forth is 1
Van Iwagen (2001).
conceived of in a “broadly Quinean” way2, and the whole book provides an illustration of what an A-ontology is like. As to B-ontologies, van Inwagen professes not to understand either their key-terms, such as ‘trope’, ‘bare particular’, ‘bundle’, etc., or their main aim, i.e., that of investigating the structure of objects. Van Inwagen maintains that the concept of structure is basically a spatial concept, and that when one asks about the structure of, say, a chair, the people who are most entitled to give a meaningful answer are carpenters, chemists and physicists. To be sure, the sense of the term ‘structure’ can be legitimately extended beyond the original one, which is spatial; that has been successfully done, e.g., in mathematics or in linguistics. Yet the ontological extension does not seem to van Inwagen to have a clear and definite sense: “What I cannot see is how a chair could have any sort of structure but a spatial or mereological structure. And, in the matter of mereological structure, I cannot see how a chair could have any parts but smaller spatial things”3. It is not difficult to grasp from such remarks the deep difference between van Inwagen’s and Bergmann’s views concerning the nature, the purpose and the method of ontology. It is less easy to pinpoint what the roots of such a wide cleavage are. Following a cue provided by the passage quoted, one can reasonably guess that one of the basic differences has something to do with space: whereas A-structures (if such a concept makes sense at all) are primarily spatial, B-structures do not have any clear and obvious – or at least any privileged – connection with space. Probably the notion of time can be associated with that of space, so that one can think of an A-structure as a spatio-temporal structure, and of a B-structure as a structure that is not necessarily spatiotemporal. Therefore, different ways of conceiving the notions of space and time and their role in ontological discourse seem to account – at least in part – for the reciprocal remoteness of A- and B-ontologies. The second example I want to set forward concerns the notions of concreteness and abstractness. ‘Concrete’ and ‘abstract’ are certainly two keywords in contemporary ontology. The meaning of these terms is not always the same in different authors; yet they are usually associated in some way with the notions of space and time. Very roughly speaking, they are often thought to mean something like spatio-temporally localized and not spatio-temporally localized respectively. From a historical point of view, maybe the recent wide acceptance of this way of characterizing concreteness and abstractness is due in large part to the influence of Nelson 2 3
70
Van Inwagen (2001), p. 3. Van Inwagen (2001), p. 2.
Goodman and Willard Van Orman Quine’s works. Some of the traditional ontological terms have been “redefined” (with a sort of “persuasive definition”) as a consequence of the importance attached to these notions. Following a suggestion by David Armstrong – hinting at the fact that both Goodman and Quine come from Harvard –, one could speak of a “Harvard” influence and a “Harvard” terminology with reference to these “redefinitions”4, leaving the differences between Goodman’s and Quine’s positions as well as some philological niceties on one side. For instance, the term ‘realism’, traditionally referring to a doctrine according to which universals exist, has been redefined by Harvard as a doctrine according to which abstract objects exist (in fact Harvard usually speaks of Platonism rather than of realism, but that is a relatively minor point). To be sure, universals are abstract objects according to Harvard, but not all abstract objects are universals. For example, one of the most typical forms of Platonism is – in Harvard’s opinion – that which admits the existence of classes, but not every philosopher would agree that classes are universals. According to Armstrong, for instance, universals are repeatable in that they are identical in different instantiations; in this sense classes (at least classes of first order particulars) do not seem to be repeatable, and therefore they do not seem to be universals5. Furthermore, Armstrong notoriously claims that universals are not abstract in Harvard’s sense, since – contrary to a widespread belief – they are spatio-temporally localized6. The details of this debate do not concern us here; however, the very existence of the debate shows that the straightforward identification of nominalism with the rejection of abstract objects – an identification stemming from Harvard –, and the implicit identification of universals with abstract objects, is not accepted by everyone. In particular, such an identification is certainly not accepted by Bergmann. Yet a more or less “Harvardian” terminology is pervasive today, and with the terminology comes a way of thinking of ontology itself, which does not fit at all with Bergmann’s. These few remarks should suffice to show that Bergmann’s style of ontology is very far from that which is most common in contemporary ontology, or at least from the Goodman-Quine-van Inwagen style (Quine’s influence on contemporary ontology has been great indeed, and – as has already been hinted at – van Inwagen himself claims that his conception of ontology is “broadly Quinean”). More importantly, the examples suggest 4 5 6
Cf. Armstrong (1997), p. 120. Cf. Armstrong (1989), p. 10; Armstrong (1997), p. 188. At least in a sense: cf. Armstrong (1988) and Armstrong (1989), p. 99.
71
that one of the main aspects of the difference between Bergmann and a significant strand of contemporary ontology concerns in some way the conception of space and time. In other words, one of the reasons of Bergmann’s “outmodedness” has to do with his way of conceiving space and time and their role in ontological analysis. To be more precise, Bergmann’s attitude seems to be characterized by a deliberate playing down of the notions of space and time, at least in comparison with the fundamental role attributed to them by “mainstream” ontology. The predominantly historical aim of the paper lies in trying to unravel and trace down some of the theoretical reasons which account for the sense of estrangement – so clearly expressed by van Inwagen – that many contemporary ontologists feel with respect to Bergmann’s ontology. 2. An overview: Bergmann’s philosophical method and his conception of ontology The remaining part of the paper aims at showing in what the difference which has just been pointed out exactly consists. To show that, a Bergmannian perspective is adopted, since it is assumed that the other side of the opposition is more widely known and more familiar. In order to do that, a certain amount of mere exposition of Bergmann’s views will be unavoidable, to provide the required frame of reference for the discussion that will follow. In fact, Bergmann’s analysis of the notions of space and time can only be understood in the context of the whole of his ontological scheme, which must therefore be summarily presented. But before doing that, another preliminary step is in order, concerning Bergmann’s philosophical method: the ideal language method. Bergmann conceives of philosophy, along positivistic lines, as the clarification of “philosophical” uses of words. Whereas ordinary (commonsensical) uses do not require any special explication (if we cannot understand them, what can we understand at all?), some uses of some words in traditional philosophy – if taken in their literal sense – give rise to absurdities. Typical examples are theses such as ‘Physical objects do not exist’ or ‘Time does not exist’. The only way to perform adequately the clarificatory task which characterizes philosophy is by means of an ideal language. First of all, one has to construct a language in a purely formal (i.e., syntactical) way. Once interpreted, such a language, in order to be the ideal language, must satisfy three conditions:
72
(i) it must be such that it is possible, by means of it, to say everything that is said in the ordinary (commonsensical) use of language; (ii) it must be such that it is not possible, by means of it, to say any of the absurdities which are characteristic of traditional philosophy; (iii) it must be such that it is possible, by speaking commonsensically of its syntactical and semantical properties, to solve all traditional philosophical problems. If the constructed language does not satisfy the conditions (i)-(iii), then it is not the ideal language. For instance, if it does not satisfy condition (i), that means that it is not “complete”, in the sense that it leaves some aspects of our experience out of consideration. Of course one can never be certain that the language he has put forward is really the ideal language, since the taking into account of aspects of experience previously neglected or the discovery of new philosophical problems might always prove it to be inadequate. The best one can do is to propose a language and show, case by case, as they present themselves, that it is adequate. Bergmann’s method is in many ways indebted to Carnap’s distinction between a material and a formal mode of speech. What is nonsense in the material mode of speech can be translated into the formal one. The crucial difference is that according to Bergmann that does not mean dissolving the problem, but rather solving it. In Carnap’s view the problem is dissolved since it is reduced to a choice between different languages, and such a choice is devoid of any cognitive import since it is governed by pragmatic considerations. According to Bergmann, the choice is not governed by pragmatic considerations; and, more than that, it is not a choice at all. One cannot choose the ideal language; rather one can put a language to the test, and find out whether it can be the ideal language. As long as it satisfies the three conditions mentioned, it could be the ideal language. But that is – so to speak – an empirical question. To be sure, nothing of what is said in the ideal language itself can be the solution to a philosophical problem, or even a statement of it (cf. condition (ii)); and what is said in the (commonsensical) metalanguage does not refer directly to the philosophical problems themselves, but only to the syntax and the semantics of the (ideal) object language. Yet the fact that that language, rather than another, is (or could be) the ideal language tells us something not about the language, but about the world. The fact that a certain language is adequate to speak about the world tells us something about the world itself. That is the gist of Bergmann’s philosophical method. Besides that of Carnap, another obvious influence was that of Wittgenstein’s
73
Tractatus Logico-Philosophicus, with its distinction between what can be said and what only shows itself. One is immediately reminded of Wittgenstein’s remark according to which tautologies tell us nothing about the world, but the fact that certain sentences (and not others) are tautologies tells us something about (the structure of) the world. An important difference between Bergmann’s view and that of Wittgenstein is that according to the former what cannot be said in the ideal language can be said, at least after a fashion, by speaking about it in a metalanguage. Thus, according to this method, the main task of a philosopher is that of putting forward a language, which is hopefully the ideal one, and then of showing how, by means of it and of metalinguistic discourse about it, one can explicate philosophical problems. Concerning more specifically ontology, its purpose is traditionally conceived as that of answering the question ‘What is there?’ or ‘What exists?’, by listing the categories of entities; each ontological category should correspond to a syntactic category of the ideal language. To be an entity means, according to Bergmann, to have an ontological status, i.e., to exist, in a peculiar sense of the verb ‘exist’. The peculiar sense of existing in which the ontologist is interested is the philosophical one, to be distinguished from the ordinary (commonsensical) one. Everyone knows what he means when – commonsensically speaking – he says that in this room there are (or exist) three chairs. No explication or clarification is required. But when a philosopher says that chairs do not (really) exist, it should be obvious that he is not using the verb ‘exist’ in a commonsensical way, otherwise he would be simply saying something false, or he would be mad. What the whole machinery of ideal language and metalinguistic discourse about it aims at is the explication of the philosophical use of ‘exist’ and of other similar philosophical uses. Bergmann thinks that the main philosophical use of ‘to exist’ can be paraphrased as ‘to be a simple’; such a paraphrase must in turn be explicated by means of metalinguistic discourse about the ideal language. In his early works Bergmann held that to exist (to be a simple) is to be the referent of an undefined descriptive constant of the ideal language, where ‘descriptive’ is opposed to ‘logical’. According to this early view, logical constants (such as logical connectives, quantifiers, etc.) do not refer. Later Bergmann came to believe that they refer as well, and that therefore one can speak of the existence (or subsistence) of their referents7; of course, since logical constants belong to different syntactical categories from those 7
74
Cf. especially Bergmann (1960) and Bergmann (1962).
of the descriptive constants, their referents belong to different ontological categories. What a traditional phenomenalistic metaphysician says when he asserts that only sense data (really) exist (and that therefore chairs do not exist) can be reformulated, according to Bergmann’s method, by saying that the undefined descriptive constants of the ideal language refer to phenomenal entities. The fact that chairs do not “exist”, in this peculiar sense of ‘exist’, does not mean that there are no chairs. The commonsensical statement ‘There are chairs’ can be reformulated in the ideal language by means of complex logical constructions (possibly along Berkeleyan lines). The philosophical statement that only phenomenal entities (really) exist cannot be formulated in the ideal language at all (and that is as it should be), but its sense can be explicated by speaking (in the ordinary language) about its interpretation. In the same way, what a traditional materialistic metaphysician says when he asserts that only physical objects (really) exist (and that therefore mental states do not exist) can be reformulated by saying that the undefined descriptive constants of the ideal language refer to physical objects. The fact that mental states do not “exist”, in this peculiar sense of ‘exist’, does not mean that there are no mental states. The commonsensical statement ‘There are mental states’ can be reformulated in the ideal language by means of complex logical constructions (possibly along behaviouristic lines). The philosophical statement that only physical objects (really) exist cannot be formulated in the ideal language at all (and that is as it should be), but its sense can be explicated by speaking (in the ordinary language) about its interpretation. The real philosophical problem is trying to understand whether one of these two languages (or some other language) can be the ideal one. The main ontological categories acknowledged by Bergmann are the following: (A) things (1) particulars (2) characters (i) properties (ii) relations (B) facts (C) subsistents (3) nexus (4) other kinds of subsistents The category of subsistents is rather variegated; roughly, the subsistents are the referents of the logical constants and some other kindred entities. Intuitively, they are responsible for what Wittgenstein would call the
75
“form” or the “structure” of the world. By contrast, things correspond to the world’s “content” (they are the referents of the descriptive constants). The category of things is further divided into two subcategories: that of particulars (referred to by individual constants), and that of characters (referred to by predicates). The terms ‘character’ is more or less equivalent to the more usual ‘universal’. As predicates may be monadic or polyadic, so characters may be properties or relations. The whole question of higher order characters is not taken into consideration here8. Characters are conceived of by Bergmann in a rather traditional way. All characters possess a qualitative aspect, which Bergmann calls ‘nature’; such qualitative aspect is what distinguishes, for instance, red from blue. Particulars are conceived of as bare, i.e., as devoid of any nature. That means that two distinct particulars differ only numerically. The only task particulars are called to perform is what is traditionally known as individuation. For instance, two distinct spots which are exactly the same with respect to their characters are nevertheless two and not one, since the same characters are exemplified by two (numerically) distinct particulars. As particulars account for individuation, characters account for identity (in a certain respect). For instance, two red spots are identical with respect to colour because the two relevant particulars exemplify (literally) the same character (i.e., redness). Bergmann holds that the acknowledgment of both characters and (bare) particulars is the only adequate way to account for the old problems of the one and the many. The arguments purporting to get to such a conclusion do not concern us here. All things – be they characters or particulars – are simple entities. Contrary to things, facts are complex entities; hence in a sense they should not exist. But, as for the case of subsistents, Bergmann eventually came to relax his criterion of existence, so that facts as well as things may be said to “exist”. Of course that must not make one forget that facts are not simples. The complexity of facts consists in their having constituents; such constituents are particulars and characters, which may be said to be “in” facts. But in addition to particulars and characters, subsistents are involved in facts as well. In connection with facts, the most significant subsistent is the nexus of exemplification, which ties together the particular(s) and the character (property or relation) which make up the fact. Exemplification does not need a further nexus to tie it to what it ties, otherwise an endless regress would start, as F.H. Bradley showed. Ordinary objects, such as chairs, are not to be conceived of as things in Bergmann’s ontology, but rather as 8
76
But cf. especially Bergmann (1957).
facts. A red round spot, for example, might be analysed – or assayed, as Bergmann says – as a particular exemplifying two characters, i.e., redness and roundness. According to Bergmann, only entities with which we have a direct acquaintance can be admitted as referents of the undefined descriptive constants of the ideal language. This principle, known as the Principle of Acquaintance, unpacks what is usually meant with the label ‘empiricism’, so that in this sense Bergmann’s ontology may be said to be an “empiricist” one. Yet it is not completely clear what the correct interpretation of the principle is, and the whole question of acquaintance and its role in ontology has always been one of the most widely debated concerning Bergmann’s philosophy. However, we can leave these problems on one side. 3. What is spatial (or temporal)? It is now time to come to Bergmann’s analysis of space and time. I will not try – except occasionally – to explain why Bergmann chose to analyse space and time in the way he did. What is important here is simply to grasp the general “pattern” of such an analysis, in order to make a comparison with alternative views possible. First of all, it must be noticed that what Bergmann tried to analyse is primarily the perceptual space and time, not the physical ones: physical space and time are what physicists, not philosophers, are interested in, and they should be left to physicists9. In two papers, Some Reflections on Time and Synthetic A Priori, Bergmann put forth two different taxonomies of views about space and time10. The details of the two taxonomies are not exactly the same, but both of them are based on the traditional distinction between absolute and relativist (or better, relational) conceptions. Bergmann’s own view definitely belongs to the relational brand. To put it very roughly, Bergmann thinks that space and time are accounted for by relations holding between objects, not by a special kind of objects. To be more precise, and using Bergmann’s own terminology, space and time are accounted for by relations (therefore things) holding between (bare) 9
Cf. Bergmann (1964b), p. 338. On the relationship between physics and ontology cf. Bergmann (1961). 10 Cf., respectively, Bergmann (1958b), pp. 227, 230, and Bergmann (1964a), pp. 285-288; for the peculiarity of Leibniz’s view cf. Bergmann (1956).
77
particulars. Being earlier than and being to the left of are two typical examples of temporal and spatial relations respectively. The main reason to think that there are space and time relations is that we need them in order to explain certain facts concerning space and time (here ‘fact’ is not used in Bergmann’s technical sense, but in its ordinary one). For instance, we know that if a is earlier than b and b is earlier than c, then a is earlier than c: relations are required in order to explain order. If we chose to admit space or time properties, we would nevertheless need some space and time relations as well, to explain the order among such properties. Economical considerations would therefore favour a view which only admits space and time relations. An example will make the issue clearer. Let us take the case of two spots; one of them is red and square and the other is blue and round; the former is to the left of the latter. How does Bergmann analyse the whole situation? The situation involves at least the following elements: (i) two bare particulars (let us call them a and b); (ii) four properties, that is four monadic characters (red, square, blue and round); (iii) one relation (dyadic character), i.e. the relation of being to the left of (let us call it R); (iv) the nexus of exemplification (a subsistent), occurring five times. The assay of the situation is the following: (1) red (a); (2) square (a); (3) blue (b); (4) round (b); (5) R (a, b). Each sentence represents a fact. The exemplification of the characters by the particulars is “symbolized” by the juxtaposition of the names of the characters and of those of the particulars in each sentence. For clarity’s sake one can make the nexus of exemplification explicit, by writing ‘a ε red’ (where ‘ε’ stands for the nexus), instead of ‘red (a)’; yet such a notation – as Bergmann noticed11 – would be redundant. One more notation is possible: one can write (6) ‘ν (a, red, square)’ (where ‘ν’ is once again a name for the nexus), which is a sort of abbreviation for ‘red (a) ∧ square (a)’. The sentence (6) represents the fact that the particular a exemplifies both the character red and the character square; since this fact is – according to Bergmann – the first spot of our example, we can say that the sentence represents the spot itself (an 11
78
Cf. Bergmann (1960), Bergmann (1962).
ordinary object). Now the question is: What constituents, if any, among those mentioned can be regarded as spatial? And in what sense? In a sense, spatiality concerns the relation R; were it not for such a relation, nothing peculiarly spatial would have entered the situation. Let us remember that every absolutist conception of space is excluded: space is built by the spatial relations holding among objects. As red and blue are monadic colour characters, and square and round are monadic shape characters, so R (being to the left of) is a dyadic space character. The world, or some parts or aspects of it, may be said to be coloured thanks to colour characters; in the same way it, or some parts or aspects of it, may be said to be spatial thanks to space characters. What holds for space holds, mutatis mutandis, for time as well. Are the space (or time) relations themselves spatial (or temporal)? The question may seem idle, and actually it is mainly a terminological one. If calling them spatial or temporal sounds well to anyone, why not? We have seen in what sense a relation such as being to the left of can be called a space relation; nothing seems to prevent us from calling it a spatial entity as well, if we like. Yet there are reasons to answer the question in a negative way, since an affirmative answer could be subtly misleading. Let us consider the case of colour properties. Do we say that they are themselves coloured? Or, to take a more specific example, do we say that red itself is red? To be sure, in this case as well we can settle the question with a terminological choice. But it seems that the sense in which red may be said to be red is not exactly the same in which an apple is said to be red. If this distinction is missed, arguments such as the Third Man may threaten the soundness of the whole analysis. Therefore, to avoid any misunderstanding, it is perhaps better to say that red is a colour property, but that it is not itself coloured. In a similar way, it seems better to say that space relations are not themselves spatial, and that time relations are not themselves temporal. But then what is properly spatial or temporal? We can turn again for inspiration to the case of a red apple. According to Bergmann’s analysis the apple belongs to the category of facts. It can be analysed as: (7) ν (a, P1, …, Pn, red), where a is the bare particular which “individuates” the apple, and P1, ..., Pn are the other properties exemplified by a in addition to red (for instance, the shape of the apple, its smell, etc.). Now, let us imagine that we point to the apple and say: ‘This is red’. What does ‘this’ refer to in this sentence?
79
Or, what entity can properly be said to be red in this situation? (A) One could say that ‘this’ refers to the bare particular a, and that it (i.e., the bare particular) exemplifies a certain property. (B) Alternatively, one could say that ‘this’ refers to the apple, which is not a, but ν (a, P1, …, Pn, red); in this case, by saying ‘This is red’ we would really say that the character red is part of a certain complex (the apple itself). According to the analysis (A), the copula ‘is’ connects a and red, which are external to each other. According to the analysis (B) it connects red and ν (a, P1, …, Pn, red), where red is part of ν (a, P1, …, Pn, red). One could also say that according to (A) what is properly red is a, whereas according to (B) it is ν (a, P1, …, Pn, red). Bergmann has some general reasons to accept the first analysis and reject the second, and such a preference plays an important role in his philosophy. Yet his arguments have little to do with the issue of space and time, and they can be safely ignored here. Let us come back, then, to the case of the relation R. If we choose (A), the entities which can be said to be spatial will be particulars; if we choose (B) they will be facts. In the first case they will be a and b. In the second case the paradigm of the red apple could suggest that the entities which are spatial are the spots, i.e., ν (a, red, square) and ν (b, blue, round). The main reason favouring this choice seems to be that, commonsensically speaking, the space relation concerns the spots. Yet in the two facts in question the space relation R does not occur at all, and that detracts from the appeal of the proposal. It remains another candidate fact: R (a, b), in which the relation R does occur, but in this fact no spots are present (it must be remembered that a and b are not spots, but bare particulars, i.e., mere individuators), and the suggestion coming from the case of the red apple loses much of its strength. As in the previous paragraph, the question could be settled with a terminological choice. As long as we know what we are saying, there is nothing bad in saying that either (i) a and b, or (ii) ν (a, red, square) and ν (b, blue, round), or (iii) R (a, b) are spatial. Yet we have seen that (ii) and (iii) do not seem to be very “natural” solutions. If all is taken into account (and chiefly Bergmann’s independent reasons for preferring (A) rather (B)), the best solution seems to be that of regarding the bare particulars a and b as the proper spatial entities. The apparent oddity of this solution (a and b are bare particulars) may perhaps be lessened by reflecting on the fact that a and b are the very terms of the space relation R. The notions of concreteness and abstractness do not play an important role in Bergmann’s ontology. In fact, that is probably an understatement: as Bergmann says in a note to Russell’s Examination of
80
Leibniz Examined, “Let me say, then, that ‘concrete’ and ‘abstract’ are banned from this essay. For I have found not only that they are a pair of troublemakers but also that in philosophy I can do nicely without them”12. Of course, that does not come as a surprise: it is one of the main theses of this paper that the notions of space and time, and therefore those of concreteness and abstractness, do not play an important role in Bergmann’s ontology. Nevertheless, that does not mean that there is no room at all for them. Bergmann thinks that the only clear sense in which one can speak of concrete entities at all is that in which they can be said to be spatial or temporal, along the lines which have just been set forth (abstract entities are obviously those which are not spatial or temporal in the aforementioned sense). But we have seen that according to Bergmann only particulars may be properly said to be spatial or temporal; therefore only particulars can be regarded as concrete, whereas entities belonging to other ontological categories are necessarily abstract. In Some Reflections on Time an additional argument is put forward concerning the concreteness of particulars13. The argument is rather complex, but the gist of it is that in a world in which there is no lowest type of things (for instance, a world in which the hierarchy of logical types is …, - 2, -1, 0, 1, 2, …) there cannot be space and time as we ordinarily understand them (more exactly, the epistemology of time would conflict with the principle of acquaintance). That means that the existence of particulars (defined as the entities belonging to the lowest type available) is strictly connected to our notions of space and time; in Bergmann’s view, ‘concrete’ is merely a label that can be attached to particulars by virtue of such a peculiarly strict connection. Thus Bergmann can conclude that an entity is concrete if and only if (i) it is named by a descriptive constant (of the ideal language) belonging to the lowest logical type, and (ii) it exemplifies at least one space and/or time relation (i.e., it is spatial and/or temporal). To sum up, the conception according to which particulars (and only particulars) can be said to be concrete, i.e., spatio-temporally localized, seems to depend on two main arguments. On the one hand, it depends on a rather technical consideration about the hierarchy of logical types, which must have a lowest level (i.e., particulars) if we want to speak of space and time in the way we are used to. On the other hand it depends on a somewhat arbitrary linguistic convention as to the least unnatural use of the adjectives ‘spatial’ and ‘temporal’: that is what the whole discussion 12 13
Bergmann (1956), p. 158. Cf. also Bergmann (1958a).
81
concerning spots and apples amounts to. Bergmann’s stipulation concerning the meaning of ‘concrete’ and ‘abstract’ is intended to capture – at least approximately – the Harvard meaning: for instance, according to such a stipulation, universals come out as abstract, and that agrees with Harvard use. Yet the approximation is rough indeed. Let us consider again the case of a red apple: what may properly be said spatial and therefore concrete according to Bergmann is not the apple itself – which is a fact –, but rather the particular that is “in” the apple. That has momentous consequences. One who holds that the apple is a particular, and that what is concrete is therefore the apple, may reasonably be induced to argue that what (primarily) exist are objects such as apples, and that universals, qua abstract, do not exist. That does seem to have an appealing common sense ring, and it is probably one of the motives underlying, for instance, Quine’s nominalism14. But if one holds (as Bergmann does) that what is concrete is a bare particular – i.e., a mere individuator, something which is far removed from common sense –, there does not seem to be any good prima facie reason to attribute existence to “concrete” entities and deny it to “abstract” ones. Thus in Bergmann’s ontology the notions of space, time, concreteness and abstractness do not play any significant role in the determination of what there is or exists, which is governed by other criteria. On the contrary, in his opponents’ views, such a fundamental ontological question is closely connected with those notions. Other considerations contribute to the playing down of the significance of the notions of spatiality (/temporality) and concreteness (/abstractness) in Bergmann’s ontological scheme. We have seen that, as long as it exemplifies space relations, a particular is regarded by Bergmann as spatial and therefore as concrete. Yet there is a clear sense in which the characterization of a particular as spatial is not an ontological one. According to Bergmann the classification of a certain entity as belonging to a certain group may be considered ontological only if it depends on intrinsic aspects of that entity. Particulars, being bare, are by themselves devoid of any quality, property, nature, etc. Therefore all particulars, by themselves, are alike: they are – as it is usually said – only numerically distinct. Strictly speaking a particular, by itself, is not an apple; the apple is really a fact. To express that in symbols, in (8) ν (a, P1, …, Pn), 14
For Bergmann’s diagnosis of the motives underlying Quine’s nominalism cf. Bergmann (1954) and Bergmann (1961).
82
where P1, …, Pn are the several properties belonging to an apple, it is not ‘a’ by itself which stands for the apple, but rather the whole ‘ν (a, P1, …, Pn)’; a may be said to be an apple – if at all – only in a derivative sense, inasmuch as it exemplifies P1, …, Pn. As Bergmann says, a is an apple contextually, not ontologically, since its being an apple does not depend on intrinsic aspects of a, but on the context in which it is situated. Analogously, if we consider a context assayed as (9) R (a, b), where ‘R’ stands for a space relation, the particular referred to by ‘a’ is not “ontologically”, but only “contextually” spatial. In other words, being spatial is not an ontological category any more than being an apple is one. It is possible to object to such a conclusion on the ground that space (or time) relations cannot be put in the same basket as other relations or other properties, such as being red, or being an apple, or being heavier than. Space and time relations would differ from other characters more than the other characters differ among themselves; or, to be more exact, the difference between space and time characters on the one hand and other characters on the other would be – contrary, for instance to the difference between red and blue – an ontological difference, and that would account for the ontological difference between concrete (spatio-temporal) entities and abstract entities. The point is that in Bergmann’s views there is no room for such a “difference among differences”. Space and time relations are referred to by polyadic descriptive predicates, and from this point of view they are exactly on a par with all other relations. Since ontological features, according to the ideal language method, must be reflected in syntactic features, and no significant difference is distinguishable at the syntactic level, no ontological difference can be recognized. Of course, Bergmann is ready to acknowledge that space and time relations are more “pervasive” – as it were – than other relations or properties. Space and time are definitely fundamental features of our world, but they are not ontological features. As Bergmann says, drawing from Wittgenstein’s terminology: since space and time relations are symbolized in the ideal language by descriptive constants, not by logical ones, they are not part of the world’s form, but of its content. Or, to say the same thing in a different way: “The world is not in space and time; space and time are in the world”15.
15
Bergmann (1953), p. 51.
83
REFERENCES Armstrong (1988) = David M. Armstrong, “Can a Naturalist Believe in Universals?”, in Edna Ullmann-Margalit, Science in Reflection, vol. III, Dordrecht, Kluwer, 1988, pp. 103-115. Armstrong (1989) = David M. Armstrong, Universals. An Opinionated Introduction, Boulder (Col.), Westview Press, 1989. Armstrong (1997) = David M. Armstrong, A World of States of Affairs, Cambridge, Cambridge University Press, 1997. Bergmann (1953) = Gustav Bergmann, “Logical Positivism, Language, and the Reconstruction of Metaphysics”, Rivista critica di storia della filosofia, VIII, 1953, pp. 453-481; then in Gustav Bergmann, The Metaphysics of Logical Positivism, New York, Longmans, Green and Co., 1954, pp. 30-77 (2nd ed. Madison (Wis.), University of Wisconsin Press, 1967). Bergmann (1954) = Gustav Bergmann, “Particularity and the New Nominalism”, Methodos, VI, 1954, pp. 131-147; then in Gustav Bergmann, Meaning and Existence, Madison (Wis.), University of Wisconsin Press, 1959, pp. 91-105. Bergmann (1956) = Gustav Bergmann, “Russell’s Examination of Leibniz Examined”, Philosophy of Science, XXIII, 1956, pp. 175-203; then in Gustav Bergmann, Meaning and Existence, Madison (Wis.), University of Wisconsin Press, 1959, pp. 155-188. Bergmann (1957) = Gustav Bergmann, “Elementarism”, Philosophy and Phenomenological Research, XVIII, 1957, pp. 19-27; then in Gustav Bergmann, Meaning and Existence, Madison (Wis.), University of Wisconsin Press, 1959, pp. 115-123. Bergmann (1958a) = Gustav Bergmann, “Individuals”, Philosophical Studies, IX, 1958, pp. 78-85; then in Gustav Bergmann, Meaning and Existence, Madison (Wis.), University of Wisconsin Press, 1959, pp. 124-131. Bergmann (1958b) = Gustav Bergmann, “Some Reflections on Time”, Archivio di filosofia, 1958, n. 1, pp. 49-82; then in Gustav Bergmann, Meaning and Existence, Madison (Wis.), University of Wisconsin Press, 1959, pp. 225-263. Bergmann (1960) = Gustav Bergmann, “Ineffability, Ontology, and Method”, Philosophical Review, LXIX, 1960, pp. 18-40; then in Gustav Bergmann, Logic and Reality, Madison (Wis.), University of Wisconsin Press, 1964, pp. 45-63. Bergmann (1961) = Gustav Bergmann, “Physics and Ontology”, Philosophy of Science, XXVIII, 1961, pp. 1-14; then in Gustav Bergmann, Logic and Reality, Madison (Wis.), University of Wisconsin Press, 1964, pp. 108-132. Bergmann (1962) = Gustav Bergmann, “Generality and Existence”, Theoria, XXIX, 1962, pp. 1-26; then in Gustav Bergmann, Logic and Reality, Madison (Wis.), University of Wisconsin Press, 1964, pp. 64-84. Bergmann (1964a) = Gustav Bergmann, “Synthetic A Priori”, in Gustav Bergmann, Logic and Reality, Madison (Wis.), University of Wisconsin Press, 1964, pp. 272-301. Bergmann (1964b) = Gustav Bergmann, “Realistic Postscript”, in Gustav Bergmann, Logic and Reality, Madison (Wis.), University of Wisconsin Press, 1964, pp.
84
302-340. Bergmann (1967) = Gustav Bergmann, Realism. A Critique of Brentano and Meinong, Madison (Wis.), University of Wisconsin Press, 1967. Van Inwagen (2001) = Peter van Inwagen, Ontology, Identity, and Modality. Essays in Metaphysics, Cambridge, Cambridge University Press, 2001.
85
Sellars et Bergmann Lecteurs de Leibniz La querelle des particuliers JEAN-BAPTISTE RAUZY § 1 Introduction : la querelle des particuliers Une des questions autour desquelles la querelle des Universaux a retrouvé de la vigueur dans la période récente a été : Est-ce que l'on peut dire exactement ce que l'on dit par l'énoncé : (1)
Le rouge est une couleur
en se passant de l'universel auquel réfère « couleur » ? En disant, par exemple, (2)
tout ce qui est rouge est coloré ? 1
Une question analogue a agité la métaphysique analytique des années 1950. Il s’agissait également d’une question portant sur la réductibilité des énoncés. Mais ce qu’on essayait de faire disparaître – ou ce dont on tentait de se passer – n’était pas les noms des universaux mais ceux des particuliers. Est-ce que l'on peut dire exactement ce que l'on dit par l'énoncé : (3)
Fido is angry
si le particulier auquel réfère « Fido » n'est plus considéré comme une entité dont le statut ontologique est réglé une fois pour toutes par la notion vénérable de la substance. La « querelle des particuliers », comme concept historiographique, désigne le devenir de cette question dans les textes de Russell, de Gustav Bergmann, de Goodman, de Strawson, etc. et, bien entendu, de Wilfrid Sellars. De la géographie de cette querelle se dégagent principalement deux axes distincts : appelons le premier l'axe de l'individuation et l’autre l'axe de l'exemplification. Frank Jackson (1977) « Statements about Universals », in : Mind 86 (1977) p. 427429. La réponse de Jackson est négative. Aucun des prétendants particularistes ne peut jouer le même rôle linguistique que l’énoncé comportant un universel, aucun n’a, pour valeur, la même valeur de vérité pour tous les contextes (arguements). 1
Selon l'axe de l'individuation, la querelle des particuliers est une conséquence de la décision prise par Russell et par Carnap, chacun pour des raisons différentes, de se passer de la notion métaphysique de substance. Mais elle n’est pas à proprement parler une querelle à propos de la substance ou de ce qui doit en tenir lieu dans la nouvelle métaphysique. Elle porte plutôt sur la possibilité de décrire les particuliers dont on a besoin dans un framework dans lequel on essaie autant que possible de se passer des particuliers. On trouve une expression exemplaire de cette querelle dans la note de Gustav Bergmann : « Russell on Particulars2 ». Bergmann souligne que le langage des sense data est conçu comme un langage particular-free. Appelant « particuliers », avec un petit « p », tous les éléments du langage qui permettent de désigner les Particuliers avec un « P ». Il entend montrer : « Que les énoncés usuels dont on a besoin pour décrire un champ de ce genre [le champ visuel] c’est-à-dire des énoncés dans lesquels les particuliers réfèrent à des Particuliers, ne peuvent pas être reconstruits adéquatement dans un langage dans lequel les particuliers réfèrent aux universaux qui font occurrence dans le champ3 ».
Dans le détail de la démonstration, on trouve des exemples dont, le moins qu’on puisse dire, est qu’ils partagent un air de famille avec ceux qui sont connus du lecteur de l'Aufbau et de Structure de l'apparence4. On se donne deux relations, l’une est un peu plus faible qu’une relation d’équivalence (non transitive) et l’autre un peu plus faible qu’une relation d’ordre (non irréflexive). On montre alors que le problème de la détermination « structuralement extensive » des particuliers du champ admet plus d’une solution et contrevient par conséquent au principe d’indiscernabilité des identiques. Je cite la note de Bergmann de préférence aux passages cor2 Bergmann, G. (1947) « Russell on Particulars », The Philosophical Review, 56/1, p. 59-72. 3 « […] The usual statements which are needed to describe such a field, that is, statements whose particulars refer to Particulars, cannot be adequately reconstructed within a language whose particulars refer to the Universals that present themselves in this field. », art. cit. p. 64 (The Metaphysics of Logical Positivism, ch.3, Collected Works, Vol. I, p. 69). 4 Ce que j'appelle l'axe 1 recouvre la discussion inaugurée par le manuscrit de Carnap daté de janvier 1923 et intitulé Die Quasizerlegung.
88
respondants de Structure précisément parce qu'elle est citée par Sellars, dans l'article de 1952 intitulé « Particulars ». Si l'on admet avec Carnap et Russell que les particuliers peuvent être obtenus à partir des universaux présents dans le champ, alors c'est la distinction et, pour ainsi dire, le face à face des particuliers et des universaux qui risque de s'évanouir complètement. Non seulement l'universel perd alors son caractère d'entité partagée, mais il perd aussi un autre de ses traits fondamentaux, à savoir d'être une entité exemplifiée. Sellars écrit donc en 1952 : « On doit avoir présent à l'esprit que l'argument du présent article est entièrement placé sous la présupposition générale que la distinction entre les universaux et les particuliers est ultime et irréductible et que la tentative pour faire des particuliers des 'complexes d'universaux' est aussi discutable que la notion des bare particulars5. »
Puis il renvoie aux « critiques incisives de Gustav Bergmann » sur cette tentative. Il y a quelque chose d’assez singulier dans ce passage. L’objet de l’article est une critique des particuliers nus. Et Bergmann est un défenseur des particuliers nus. Mais la situation stratégique autorise une sorte de trêve : Bergmann est considéré comme un opposant sous un certain rapport, mais comme un allié sous un autre. En 1952 existe cette ligne de partage très forte qui procède de l’approche structurale des descriptions de relation. Cette approche tend à relativiser la distinction du particulier et de l’universel au mode de description d’un champ d’objets. Si, comme le pense Carnap, Russell selon Bergmann – dans certaines versions de sa métaphysique – et comme le pensera Goodman, « être un particulier » est une propriété relative à une certaine apparatus, – perceptif, descriptif ou systémique – alors on ne peut plus se demander s’il y a des particuliers nus parce qu’on doit craindre qu’il n’y ait plus de particulier tout court. C’est pourquoi Bergmann et Sellars se retrouvent dans le même camp selon l’axe 1 de la querelle des particuliers. Ce que j'appelle l'axe 2 de la querelle porte exactement sur le point qui est établi dans la note que je viens de citer. On se demande si, une fois qu'on a accepté la distinction des universaux et des particuliers comme 5 « It must be borne in mind that the argument of the paper moves within the framework of the assumption that the distinction between universals an particulars is ultimate and irreductible, and that the contention that particulars are 'complexes of universals' is as unsound as the notion of bare particulars. » Sellars, W. (1952) p. 190.
89
distinction primitive – une fois qu'on a définitivement écarté les diverses tentatives d'un monisme structural de l'axe 1 – on a encore le moyen de faire l'économie des bare particulars dans l'ontologie. La situation est à peu près la suivante : il y a des entités, les bare particulars qui ont, plus ou moins à juste titre, mauvaise réputation. On a pu croire que ces entités étaient devenues inutiles, mais l'échec ou la défaite supposée du monisme structural nous oblige à refonder le particularisme ontologique. Il convient d'achever cette refondation en évitant de réintroduire post limino – comme disait Leibniz à propos des formes substantielles – les entités en question. Une part importante du travail de clarification du statut des particuliers passe pour Sellars comme pour Bergmann par l'analyse de l'exemplification et par la distinction des particuliers et des faits. Lorsqu'on objecte que tout dualisme de l'universel et du particulier revient à distinguer entre un this-factor et un such-factor et que c'est exactement ce qui est requis par la doctrine des bare particulars, Sellars répond que l'objection est fondée sur la confusion des particuliers et des faits. Si le particulier a exemplifie Φ, alors a est une instance de Φ mais Φ n'est pas un composant de a. Φ est un composant du fait 'que a est Φ'. Mais le fait 'que a est Φ’ n'est pas lui-même une instance de Φ. Je cite Sellars : « Dire qu'un particulier bleu consiste dans Bleu et un particulier, c'est certes dire quelque chose qui n'a pas de sens, mais il s'agit d'un non-sens qui procède non pas du dualisme des particuliers et des universaux mais de la confusion des particuliers et des faits6. »
Et il ajoute, c'est le point qui est plus difficile à comprendre, lorsqu'on imagine un particulier qui exemplifie plus d'un quale, on est coupable de la même confusion. La thèse de Sellars sur les particuliers est donc la suivante : il est possible de rejeter les bare particulars à condition (i) de montrer par une analyse précise que les particuliers ordinaires ou « choses » sont des « particuliers complexes » d'ordre élevé et (ii) que les particuliers de base exemplifient un universel et un seul. Les avantages de cette solution sont multiples : la distinction de l'universel et du particulier est conservée – les universaux, par exemple les qualia, demeurent des entités partagées parce qu'ils peuvent être exemplifiés par plusieurs particuliers de base. 6 « To say that a blue particular consists of Blue and a particular is indeed to talk nonsense, but it is nonsense which arises not out of a dualism of particulars and universals, but out of a confusion between particulars and facts. » Ibid. p. 188.
90
L'exemplification demeure aussi une relation authentique entre deux types d'entités différents. Le calcul des fonctions est amendé mais le fonctionnalisme ontologique n'est pas renversé – ces deux points sont importants pour le rapport avec Bergmann. Enfin et surtout, la conception inférentielle des contenus conceptuels se donne l'ontologie qui lui convient. Mon objet n'est pas de développer tous ces points mais d'insister sur quelques aspects du particularisme de Sellars dans sa relation avec la tradition leibnizienne. À l'occasion de cette querelle des particuliers, Sellars a, en effet, été amené à préciser une doctrine logico-métaphysique de l'ingrédience et des concepts de choses (θ-concepts) ; cette doctrine n'est pas sans rapport avec le Leibniz qui a habité la philosophie du XXe siècle depuis la Critical Exposition de Russell et que j'appelle Leibnizxx. Il s'agit principalement d'une rétroversion de la théorie des types dans la métaphysique de l'inesse. Sur ce point, l’usage de la théorie des types dans la métaphysique de l’exemplification, Sellars et Bergmann se font étrangement échos dans leur relation à Leibniz. C’est le premier point que je voudrais établir. Un autre aspect m’a paru intéressant : faire voir que, dans l'argumentaire de Sellars, la distinction déterminable/déterminé – dans la version défendue par Prior en 1949 – jouait un rôle discret mais décisif, au point que la logique des particuliers complexes peut être interprétée comme une extension de l'ontologie des déterminés. §2. La notion de Nature : Sellars et Bergmann lecteurs de Russell lecteur de Leibniz. Dans la Critical Exposition du système de Leibniz, Russell insiste volontiers sur les difficultés que soulèvent les énoncés leibniziens. Et, bien souvent, il évoque les énoncés approchant de celui qu'il appelle toujours « Mr Bradley ». À propos de la substance et de l'identité, Russell note soigneusement que le principe d'identité des indiscernables (PII) n'a rien à voir avec la doctrine de l'identité contentuelle prêtée, justement, à Bradley. Si tel était le cas, PII devrait être un principe absolument premier antérieur à la théorie de la substance. Or on sait qu'il n'en est rien puisqu'il est en général évoqué comme une conséquence. On doit, écrit Russell, distinguer entre la diversité de contenu et la diversité matérielle ou numérique. La 7 seconde est la différence entre « un sujet et un autre sujet ». Il poursuit : le 7
« Diversity of content proper is the difference between one content and another.
91
principe d'identité des indiscernables n'a aucun sens philosophique si l'on ne tient pas ensemble ces deux distinctions. Il ne porte ni sur la seule identité des sujets, ni sur la seule identité des contenus, mais bien sûr le rapport des deux. Ce que dit le principe porte ainsi sur la possibilité d'une inférence. On peut conclure d'une identité des contenus à une identité des sujets. Dans l'article de 1959 sur Leibniz, Sellars revient sur la notion de « nature » et sur l'idée d'un contenu. La nature est un certain contenu qui permet d'inférer l'identité d'une chose. Toute la question porte sur ce que contient, si l'on ose dire, ce contenu. Que doit-il y avoir dans une nature pour qu'on puisse conclure, comme on le fait lorsqu'on admet une version très forte de PII, de l'identité des natures à celle des particuliers ? Sellars reconnaît que, dans la tradition, Leibniz n'est pas le premier à concevoir l'individualité par la nature, mais il ajoute : « Il est le premier à avoir vu clairement que l'individualité de la substance ne peut être conçue qu'en termes d'épisodes de son histoire et à conclure que, si la nature doit expliquer l'individualité, elle doit expliquer des épisodes et non pas seulement des capacités, des pouvoirs et des dispositions, toutes choses qui sont en principe répétables et étaient dans la tradition attachées à la nature des choses8 ».
La nature n'est pas ce par quoi nous identifions un individu. Elle est quelque chose de beaucoup plus considérable. On peut identifier un individu par une de ses propriétés remarquables lorsque celle-ci suffit, alors que nous demandons plus à la nature. Nous lui demandons une explication du comportement de l'individu lorsqu'il est placé dans telles ou telles circonstances. Sellars ajoute que, pour cette raison, la nature de la substance doit au moins être donnée sous forme de faits hypothétiques pour
Material or numerical diversity is the difference between one subject, or one substance, and another. Leibniz's doctrine is, that two things which are materially diverse, i.e. two different substances, always differ also as to their predicates. » Critical Exposition p. 55. 8 « He was, however, the first to see clearly that the individuality of a substance can only be understood in terms of Episodes in its history, and to conclude that if the nature of a substance is to account for its individuality, it must account for episodes and not merely the capacities, powers, dispositions – all, in principle, repeatable – which were traditionally connected with the natures of things. » Sellars, W. (1959) « Meditations Leibnitziennes », p. 154.
92
9
lesquels il propose la formulation suivante : (HF)
Si, à un certain moment, la substance S était impliquée dans un épisode 10 du genre E1, elle serait impliquée dans un épisode du genre E2 .
La nature ainsi conçue devient une fonction qui, pour un épisode dans lequel l'individu a été impliqué, donne le ou les épisode(s) dans le(s)quel(s) il doit également avoir été impliqué et qui constituent la raison du fait en question (que S est impliqué dans E). On obtient ainsi une série d'inférences selon le schéma : (RS)
S a été impliqué dans E2 parce qu'il était d'abord impliqué dans E1.
Le propre de l'approche leibnizienne de la nature des choses, conclut Sellars, est que celle-ci fournit non seulement des faits hypothétiques comme dans (HF), mais également les épisodes-prémisses à partir desquels on peut, à l'aide des faits hypothétiques, établir ce que Hegel a appelé des « syllogismes in re ». Si l'on dresse le cahier des charges de la nature ou « notion » – et celui-ci est très largement dérivable d'une position forte de PII – on s'aperçoit que cette nature doit à la fois contenir des éléments de type fonctionnel comme HF et des éléments de type objectuel comme sont les « épisodes-prémisses ». Elle doit à la fois nous renseigner sur ce qui a lieu si S est impliqué dans x et sur les x dans lesquels S est impliqué. On pourrait dire, dans les termes de Frege, que les natures sont des conjonctions de fonctions et de parcours de variables, ce qui en fait des entités d'un genre délicat. Je trouve très remarquable que Sellars soit parvenu à une analyse si fine de la notion leibnizienne de nature au point de pouvoir montrer qu'une conception fonctionnaliste de la raison suffisante porte en elle une tendance à une forme de panlogisme idéaliste – le syllogisme in re hégélien. D'un autre côté, je trouve aussi très étonnant qu'il n'ait pas fait intervenir l'apparatus de la théorie des types dans son analyse de la nature leibnizienne dans la mesure où il l'avait fait dix ans plus tôt dans l'analyse 9
Il s'agit d'une formulation empruntée à McTaggart. Cf. Broad, C. D. (1933) Examination of McTaggart's Philosophy, Cambridge U.P,, v. 1 p. 264-278. Même référence, ce n'est pas fortuit, dans « On the Logic of Complex Particulars », p. 313. 10 « If at any time S were to be involved in an episode of kind E1, it would be involved in an episode of kind E2. » Ibid.
93
des particuliers. Je note – j'ignore si c'est une explication – que, quelques années auparavant, on trouve une analyse de ce genre chez Gustav Bergmann. En 1956, Gustav Bergmann prononce une conférence sur le Leibniz de Russell11. Son commentaire porte également sur la notion de substance. Il appelle « character » ce que nous appelons plus volontiers un universel ou une propriété. Les caractères sont exemplifiés par les particuliers. Entre ces caractères et les particuliers, il y a un lien ou « nexus ». La nature, demande Bergmann telle qu'elle est employée par Leibniz, est-elle un caractère ? La nature est le lien dont on peut présumer qu'il relie différents caractères si et seulement s'ils sont exemplifiés par le même individu. Voilà ce dont il faut rendre compte. On est évidemment enclin à penser que le lien entre les différents caractères qui sont exemplifiés par un même individu consiste dans les différents caractères en tant qu'ils sont exemplifiés par le même individu, faisant ainsi de la nature et du particulier une seule et même chose. Seulement, note Bergmann, la nature est désignée par un prédicat et elle est, aux yeux des philosophes de la tradition, plutôt un caractère qu'un particulier. D'où, évidemment, le problème. Si la nature est un caractère, elle doit être un caractère de second ordre, qui relie entre eux les caractères exemplifiés par un individu. Mais qu'en est-il alors de son lien à l'individu en question ? Peut-elle être partagée par plusieurs individus ? Imaginons un monde raréfié dans lequel il n'y aurait qu'un seul individu i exemplifiant deux caractères, f1 et f2. Dans ce monde, la nature de i est le caractère de second ordre R1(f1, f2) sur le statut duquel on s'interroge. Ceux qui, note Bergmann, tirent argument d'une interprétation de R1 comme : R1(f1, f2) ≡ (x) (f1(x) ⊃ f2(x)), selon laquelle R1 serait une relation logique définie, ceux-là doivent convenir que (x) (f1(x) ⊃ f2(x)) représente au mieux un lien humien « c'està-dire, pour un substantialiste, tout sauf un lien12 ». Revenons à l'expression « la nature de a ». Un des usages possibles des Principia Mathematica sur cette question est évidemment la théorie des types et j'ai dit que le nerf de la lecture de Bergmann était est une sorte de rétroversion du symbolisme des Principia, considéré comme un
11 12
Gustav Bergmann (1956) « Russell's Examination of Leibniz examined ». Art.cit. p. 163.
94
13
« langage idéal », dans la Critical Exposition. Selon les PM, avec f et g de type 2, x de type 1 et F de type 3 : (I)
(f = g) ≡ (x) [f(x) ≡ g(x)] . (F) [F(f) ≡ F(g)]
et (E)
(x) [f(x) ≡ g(x)] ⊃ (F) [F(f) ≡ F(g)]
sont analytiques. (I) est l'expression typée de l'identité des indiscernables, (E) de l'extensionalité. Or, note Bergmann, (I) permet de dériver une description définie : (N)
(ιX) (f) [X(f) ≡ f(x)]
Il y a un caractère et un seul qui est exemplifié par tous les caractères qui sont exemplifiés par un individu. (N) permet de capturer une conception de la nature. Car il ajoute: si f1 est un caractère d'un certain type, f1(a) et A(f1) sont bien entendu synthétiques, mais, par (N) : f1(a) ≡ A(f1) est analytique. L'usage réfléchi de l'anachronisme tel que Bergmann le pratique s'avère ici extrêmement intéressant et productif. Il permet de faire apparaître plusieurs points qui ne sont jamais notés par les historiens. Premier point : un langage idéal dans lequel on doit exprimer ce que Leibniz veut exprimer doit contenir au moins trois types : pour les particuliers, pour les prédicats et pour les natures. Or le langage de Leibniz contient un seul type de catégorèmes correspondant aux prédicats des PM. Il ne contient pas de nom pour les particuliers et construit les natures comme des prédicats. Deuxième point : lorsque Leibniz dit que le concept du prédicat est contenu dans celui du sujet, que dit-il exactement ? A(f1) ? f1(a) ? ou bien ni l'un ni l'autre ? Leibniz précise : « a, b, c, d, e, … etc. est 13
Sur la position des problèmes philosophiques comme problèmes métalinguistiques à partir d'un langage idéal, cf. Hochberg, H. : « Though his writings from the mid 1940s on consistently criticised Carnap, one of the characteristic features of what Bergmann took to be his own innovative « philosophical method », the reconstruction of philosophical claims as meta-linguistic claims about an ideal language, was clearly derived from Carnap’s Aufbau conception of « reconstruction » and Carnap’s later distinction between external and internal questions – a distinction Bergmann took to epitomize what he saw as Carnap’s nihilistic positivism ». The Positivist and the Ontologist, Bergmann, Carnap and Logical Realism, Rodopi, Amsterdam (p. 36).
95
a ». Et Bergmann commente : « L'erreur consiste dans le changement opéré par le etc. […] Un prédicat de second ordre peut être défini ou indéfini. […] Mais aucun prédicat de ce genre, s'il est exemplifié par un nombre infini de prédicats de premier ordre, ne peut être défini par énumération. Donc Leibniz commet deux erreurs : une erreur de type et l'erreur de la définition par énumération infinie. Commettre la première erreur prédispose à commettre la seconde14 ».
Russell aurait dû voir, et il nous a finalement fait voir, que Leibniz n'aurait pas pu affirmer que le prédicat est dans le sujet s'il avait disposé du langage (PM) dans lequel il est possible de dire quelque chose comme le prédicat est dans le sujet. Du point de vue métaphysique, les remarques de Bergmann sur Leibniz s'inscrivent de plein droit dans la querelle des particuliers et dans le prolongement de la note que j'ai citée en commençant : « Russell on particulars ». Je suis néanmoins frappé de l'absence de Sellars dans cette affaire dans la mesure où celui-ci s'était attaché à produire, dans l'article de 1949 « On the Logic of Complex Particulars », une logique de l'ingrédience dans laquelle les erreurs de type auraient été soigneusement évitées. §3 La logique des particuliers complexes. Dans la section VI de l'article de 1949, la thèse selon laquelle les particuliers ne peuvent pas exemplifier plus d'un universel est justifiée d'une manière qui ne fait pas appel au statut des bare particulars. Auparavant, je résume à grands traits et en simplifiant, les principaux points de l'article. A. Ingrédience et coingrédience La règle selon laquelle les particuliers de base ne peuvent pas exemplifier plus d'un universel est établie en vue de distinguer des niveaux de langage :
14
« The error lies in the positional shift of 'and so on'. […] A second-order predicate may be defined or undefined. The natures of our own schema are, as it happens, defined. But no such predicate, provided it is exemplified by an infinite number of first-order predicates, can be defined by enumeration. Leibniz thus makes two mistakes, the type error and the error of "definition by infinite enumeration". Making the first may facilitate making the second. » Art. cit. p. 172-173.
96
« Dans la reconstruction rationnelle d'un langage dans lequel on parle du monde, une distinction très forte doit être faite entre le niveau des énoncés qui ne comportent que des constantes individuelles et des prédicats primitifs ou non-dérivés et le niveau où interviennent des constantes15. »
L'analyse porte sur un type d'énoncés où l'on a, en position de prédicat, non pas des universaux sortaux (comme dans « Fido is a dog ») mais des universaux caractérisants, par exemple : (1)
Fido is angry.
Contrairement à ce qui se passe avec les sortaux, un énoncé comme (1) ne dit pas quelque chose de Fido pris comme un tout mais de quelque chose qui est dans Fido, un de ses ingrédients, par exemple un état émotionnel de Fido. On utilisera donc une relation d'ingrédience (notée I) pour le schéma qui explicite (1) : (2)
∃y I(y,x) & f(y)
Il y a un y tel que y est un ingrédient de x et y est une instance ou un « specimen » de f. Sellars note qu'on ne peut pas considérer (2) comme un schéma permettant de saisir le choses qui sont des f ou des f-things, en raison du caractère existentiel de (2). Selon nos intuitions en effet, un universel est quelque chose dont l'analyse ne doit pas comporter d'opérateur existentiel. Si l'on admet que (2) explicite un trait de tous les universaux, alors : « it follows that in so far as it is possible to speak of concepts or universals of the form f-things, they must be recognised to have a most unusual character. They are concepts or universals which require for their analysis the use of existential operators. » (LCP, p. 315).
L'analyse doit porter sur le statut de cette notion d'ingrédience et sur le rôle que celle-ci joue dans l'exemplification. (1) devient : (3)
Fido exemplifies anger,
où « Fido » est le nom d'un particulier complexe. L’ingrédience enveloppe 15
« In the rational reconstruction of a language in which one speaks about the world, the strongest of distinctions must be drawn between a level of statements involving only underived or primitive individual constants and predicates, and the level on which derived individual constants ("things-names") and predicates put in an appearance. » LCP, p. 307.
97
des relations que Sellars appelle « réelles » (ce sont principalement des relations spatio-temporelles et causales). Elle est elle-même fondée sur une relation antérieure, la coingrédience Φ : I(y,x) =def x = Φ(……, y, ……), y est un ingrédient de x ssi x est un ensemble de coingrédients qui contient y. Sellars conclut que les énoncés par lesquels est attribué un universel sortal (type B) portent sur des particuliers qui sont analysables dans leurs ingrédients et la vérité de ces énoncés est analysable en termes de vérité sur les ingrédients. B. Typologie des énoncés fonctionnels L'analyse de l'ingrédience permet d'établir une hiérarchie des types d'énoncés fonctionnels f(x) : – Type I : les énoncés atomiques qui ne peuvent pas être réduits. – Type II : les f(x) qui sont de la forme ∃y I(y,x) & f(y), où f est un prédicat primitif ; on les appelle les énoncés qui attribuent une propriété non-définie (undefined property). – Type III : les f(x) pour lesquels on ne peut pas produire ce genre de réduction. Dans ce cas, on a une forme dérivée dans laquelle f est un prédicat dérivé. Ce sont les énoncés qui attribuent une propriété définie (defined property) et dont l'analyse est quelque chose comme : ∃y ∃z … I(y,x)& I(z,x) & g(y) & h(z) … – Type IV : est un cas spécial de type III dont l'analyse contient en outre le schéma de complétude : ∼ ∃w I(w,x)& w ≠ y & w ≠ z … Il n'y a pas d'ingrédient de x qui n'est pas mentionné dans le schéma de coingrédience. Les f(x) de type IV sont les θ-predicats qui désignent des θ-concepts (les concepts de chose). Il s'agit d'une classe de fonctions descriptives à une place dont les arguments sont des noms de chose. Si t est la variable sur les choses. θ(t) se lit : « le particulier complexe t exemplifie le caractère complexe θ ». Nous retrouvons quelque chose d'assez proche de 16 la notion leibnizienne de nature : une nature dont on aurait conservé 16
« A q-concept specifies the complete and determinate nature of any complex particular which exemplifies it. » (p. 323)
98
uniquement les éléments fonctionnels. Se dégagent alors deux thèses remarquables : Thèse 1 :
les concept-θ surviennent sur les prédicats primitifs du langage. θit =def ∃y ∃z … & g(y) & h(z) … & t = Φ(y,z,…)
où Φ est la coingrédience. Autrement dit : l'exemplification du θ-prédicat est un compendium loquendi pour l'exemplification des prédicats primitifs par les ingrédients de t. Thèse 2 :
θi(t) & θj(t) n'est pas plus légitime au niveau des particuliers
complexes ou moléculaires que ne l'est f(x) & g(x) au niveau atomique. Autrement dit : le propre d'un particulier de base, c'est qu'il ne peut pas exemplifier plus d'un universel non-défini de même que le propre d'un particulier complexe (une chose) 17 est qu'il ne peut pas exemplifier plus d'un concept de chose .
C. Déterminables et déterminés La section VI de l'article de 1949 mérite, il me semble, qu'on s'y attarde. Sellars donne la raison pour laquelle il a choisi cette position assez inconfortable, à savoir la position selon laquelle on doit réformer le calcul des fonctions de telle sorte que, si a est une constante individuelle primitive et non dérivée, alors f(a) & g(a) n'a pas de signification. En 1949, la question agitée n'est pas encore celle des bare particulars mais plutôt celle des faits négatifs. Je donne un résumé très resserré de l'argumentation de cette section : (i) On pose, ou repose, la difficulté classique soulevée par la vérité des énoncés négatifs. Quel est le quelque chose en vertu de quoi ~Φ(b) est vrai et, ce quelque chose, peut-on l'appeler Ψ(b) ? (ii) On répond, comme l’avait fait Leibniz en son temps, par la notion d'incompatibilité : Φ et Ψ sont incompatibles. (iii) Mais l'incompatibilité ne permet pas de se passer entièrement des faits négatifs. Dire que ~Φ(b) est vrai ssi Ψ(b) lorsque Φ et Ψ sont incompatibles ne dispense pas de répondre à la question : 17
On voit le gain analytique de la typologie mise ne place. On peut par exemple comprendre angry-thing dans les termes des q-concepts. « Angry-thing » réfère à une classe de q-concepts : la classe des q-concepts which contain the concept Anger as a constituent. Autrement dit : « f-thing(t) » ne dit pas de t qu'il appartient à la classe des q-concepts qui ont f comme ingrédient mais dit de t qu'il exemplifie un des concepts qui appartiennent à cette classe (p. 326).
99
qu'est-ce que la vérité de Ψ(b) a à voir avec celle de ~Φ(b) ? (iv) On ajoute alors que l'incompatibilité doit se comprendre comme une relation fondamentale : la relation qui existe entre des déterminables de la même famille – les déterminables d'un même déterminé – de sorte que ce qui devient primitif est la structure déterminable/déterminé. Sellars est convaincu que la distinction déterminable/déterminé (i) permet d’éviter les faits négatifs extralinguistiques et (ii) constitue le modèle d'une solution 18 inférentielle . (v) Nouvelle difficulté. Si la structure en familles est admise, alors on a des contradictions liées à l'autoréférence : il est impossible de dire qu'un particulier n'exemplifie aucune qualité d'une famille donnée. Soit K un prédicat d'ordre 2 désignant une famille de déterminables ; dire que c n'a pas de qualité de la famille de K doit s'écrire : ∀f, K(f) ⊃ ~ f(c), Mais dire ~ f(c) quand f appartient à une famille de prédicables revient à dire : ∃f, K(f) & f(c). Ces deux énoncés sont contradictoires. (vi) Pour sortir de cette difficulté, on ajoute, comme une solution ad hoc, ou comme une contrainte supplémentaire, qu'un particulier 18
Les familles sont constituées des déterminés d'un même déterminable. Si P1 et P2 sont les deux déterminés de P on a (avec T et F les prédicats de vérité et de fausseté) T[P1(x)] entails ~[P2(x)] ~T[P1(x)] entails [P2(x)] qui sont des règles d'inférence et non des faits extra-linguistiques. Johnson insiste sur cette relation fondamentale des déterminables d’une même famille. Cf. le commentaire qu’en donne Prior : « […] la relation des déterminés qui tombent sous un déterminable est la relation d'incompatibilité. Les déterminés pris comme tels – par opposition aux déterminés sous un certain déterminable – formeraient selon cette conception une classe d'adjectifs, définis par la propriété d'être des termes d'une relation d'incompatibilité, exactement de la même façon que les personnes mariées forment une classe d'objets définie par la propriété d'être des termes de la relation le mariage. Mais, de même qu'une paire particulière de personnes mariées ne constitue pas une classe mais un couple, de même les déterminés qui tombent sous un déterminable donné ne constituent pas une classe mais un ensemble d'incompatibles. » Prior, A. N. (1949), p. 11.
100
ne peut exemplifier plus d'une qualité. On introduit la otherness : « Je suggère que lorsque la valeur de x est un particulier de base, la forme f(x) & g(x) ne soit pas légitime. Cela revient à dire que ∼Φ(b) est entraîné par Ψ(b) non pas en vertu du fait que Ψ appartient à la même famille que Φ – même si c'est éventuellement le cas – mais en vertu du simple fait que Ψ est une qualité différente de Φ. Dans le jargon de la tradition on dirait : la 19 réponse n'est pas l'incompatibilité mais l'altérité . »
Ce passage est à la fois remarquable et équivoque. Il y a deux manières assez différentes de le comprendre et de comprendre, par suite, toute l'argumentation de la section VI. La logique des déterminables offre quelque chose de précieux : la possibilité de considérer l'incompatibilité comme un trait inférentiel. Mais elle porte avec elle la structure en familles qui présente des inconvénients. Ces inconvénients résident, bien entendu, dans le fait que les règles d'incompatibilité ne peuvent pas être entièrement a priori. Ce n'est pourtant pas sur cela que Sellars insiste mais sur une difficulté dans la grammaire de c n'est pas un f. Et il propose, pour contourner la difficulté, la thèse des particuliers premiers qui exemplifient un universel et un seul. Ou bien l'argument est une sorte de généralisation, ou bien plutôt une manière de trancher la difficulté. Ou bien on admet que la relation qui existe dans les familles de déterminables est la relation qui existe en général au niveau des particuliers premiers. Ou bien la relation qui existe entre les déterminés d'un même déterminable est l'objet d'une imitation ou émulation pour tous les prédicats atomiques, les prédicats moléculaires étant de leur côté des familles recomposées. J'insiste sur cette équivoque parce qu'il me semble qu'on n'aboutit pas au même résultat dans l'un et l'autre cas. Dans le second cas, les déterminables sont l'occasion de la formulation d'une règle inférentielle qui ne doit rien à l'ontologie (mais alors sa justification risque d'être un peu légère, et l'on est entièrement renvoyé au débat sur les bare particulars). Dans le premier cas au contraire, les particuliers complexes de Sellars doivent être compris dans la tradition de Johnson : mais alors, est-ce qu'ils ne requièrent pas plus que de 19
« I suggest that where the values of 'x' are basic particulars, the form 'f(x) & g(x)' is illegitimate. This amounts to saying that '∼Φ(b)' is entailed by 'Ψ(b)' not in virtue of the fact that Ψ belongs to the same family as Φ, should it do so, but rather merely by virtue of the fact that Ψ is a different quality than Φ. To use traditional jargon, otherness rather than incompatibility is the answer". Art. cit. p. 318-319. »
101
simples règles d'inférence ?
REFERENCES BIBLIOGRAPHIQUES : Bergmann, G. (1947) « Russell on Particulars », The Philosophical Review, 56/1, p. 59-72, repris dans : Collected Works, Vol. I, The metaphysics of logical Positivism, ch. 3. — (1956) « Russell's Examination of Leibniz Examined », Philosophy of Science, 23/3, p. 175-203, (repris dans : Meaning and Existence, 1960, The University of Wisconsin Press, p. 155-188). Jackson, F. (1977) « Statements about Universals », Mind, 86, p. 427-429. Johnson, W. E. (1892) « The Logical Calculus », Mind, I, p. 3-30 & 235-250 & 340347. — (1921-1924) Logic, Cambridge U.P. Prior, A. N. (1949) « Determinables, Determinates and determinants », Mind, 58/229 p. 1-20 & 58/230, p. 178-194. Russell, B. (1900) A Critical Exposition of the Philosophy of Leibniz, Cambridge University Press, Cambridge. Sellars, W. (1948) « Concepts as Involving Laws and Inconceivable without Them », Philosophy of Science, 15/4 (repris dans : Pure Pragmatics and Possible worlds. The Early Essays of Wilfrid Sellars, J. Sicha (ed.), Ridgeview Publishing Company, 1980, p. 95-124). — (LCP) (1949) « On the Logic of Complex Particulars », Mind, 58, p. 306-338 (repris dans Pure Pragmatics and Possible worlds. The Early Essays of Wilfrid Sellars, J. Sicha (ed.), Ridgeview Publishing Company, 1980, p. 157-196). — (1952) « Particulars », Philosophy and Phenomenological Research, 13/2 p. 184-199 (repris dans : Science Perception and Reality, Ridgeview Publishing Company, (19631 & 19912) p. 282-297. — (1957) « A Critique of Mr. Wilson Strawson's Contribution », Philosophy and Phenomenological Research, 17/4, p. 458-472. — (1959) « Meditations Leibnitziennes », Philosophical Perspectives; History of Philosophy, Ridgeview Pub. Comp.
102
Time and Existence: A Critique of Degree Presentism L. NATHAN OAKLANDER One of the most fascinating and puzzling aspects of our ordinary language, thought and experience of time is its passage. In some sense, time seems to move from the future to the present and then from the present into the more and more distant past. We use tensed language to reflect this allegedly sui generis form of change. For example, it is now true to say that “I will retire within the next 10 years,” and in a few years it will be true to say that “I am now retiring,” and some years later it will be true that “I have been retired for x number of years.” Furthermore, we think differently about events and experiences that appear to be moving toward or away from us. Thus, an unpleasant future event is thought of with dread whereas the same event when it becomes past is thought of with relief. Finally, there is some plausibility in the claim that we experience the passage of time since we seem to experience the present as having a reality that the past and future do not have. How, then, are we to understand this enigmatic feature of reality –temporal becoming or passage—in all its various manifestations? To give an adequate answer to that question is one of the most fundamental problems in the philosophy of time. At a minimum, an adequate answer to the question, what is temporal becoming? must be able to specify what there is in reality that provides an ontological ground for those aspects of language, thought and experience that purportedly reflect the passage of time. B-theorists maintain that a logically coherent account of passage cannot be gleaned from a superficial examination of ordinary language, thought or experience. For, to suppose that our use of tensed language represents the non-relational temporal A-properties of pastness, presentness and futurity, or that our thoughts (and attitudes) about events reflect their movement through time, or that our experience implies that the present has some special ontological status not bestowed on the past or future is, B-theorists allege, fraught with insurmountable dialectical difficulties, such as McTaggart’s paradox1. A-theorists, on the other hand, 1
See J. M. E. McTaggart, The Unreality of Time, Mind, 17 (1908): 457-74.
maintain that a careful examination of our ordinary language, thought and experience is a sure path to the truth about time, and that B-theorists are mistaken in thinking that all A-theoretical accounts of temporal becoming are internally inconsistent. Although A-theorists universally agree that difficulties, such as McTaggart’s paradox, pose no lasting threat to our socalled “intuitive” conception of time, they do debate amongst themselves over which version of the A-theory is best capable of avoiding the charge of incoherence levied against them by B-theorists. Although the “A-theory/B-theory” terminology was coined after Bergmann’s writings on time he unquestioningly falls within the B-theory camp broadly conceived. For Bergmann the only intrinsically temporal entities are the external B-relations of earlier/later than and simultaneity (or perhaps only prior and simultaneity). The terms that tenselessly exist and exemplify those relations are bare particulars. In Bergmann’s ontology of time, there are no monadic temporal qualities of pastness, presentness and futurity since we are not acquainted with them and dialectically they are unnecessary. The present is what is simultaneous with a particular presented to us in an intentional act. The past is what is earlier than a particular “in” my present intention and the future is what is later2. To my knowledge, Bergmann did not discuss McTaggart’s paradox in his published writings other than to note, in Realism, that “McTaggart’s calamities [of taking presentness to be objective non-relational quality of what is perceived] are, thanks to Broad, notorious3.” After this brief digression on Bergmann’s views on time, of which I shall say no more, I return to the main topic of this presentation. In a recent paper, Quentin Smith puts forth a novel version of the Atheory of time he calls, “degree presentism” that he believes is perfectly consistent and supported by experience4. According to this view, there are See, “Some Reflections on Time,” in Meaning and Existence (Madison, Wisconsin: University of Wisconsin Press, 1960): 225-263. 2
Realism: A Critique of Brentano and Meinong (Madison, Wisconsin: University of Wisconsin Press, 1967), p. 324. For my interpretation of McTaggart see, L. Nathan Oaklander, “McTaggart’s Paradox Defended,” Metaphysica: International Journal of Ontology and Metaphysics 3, no. 1(2002): 11-25. Reprinted in L. Nathan Oaklander, The Ontology of Time (Amherst, NY: Prometheus Books, 2004): 51-62. 3
Quentin Smith, “Time and Degrees of Existence: A Theory of ‘Degree Presentism’,” in Time, Reality & Experience, ed. Craig Calender (New York: Cambridge University Press, 2002), pp. 119-136. 4
104
tensed facts (specifying what is earlier/later or simultaneous with the present), but there are no tensed or A-properties. Although A-properties do not exist, Smith refuses to follow Prior in maintaining that the past and the future do not exist or the B-theorist in maintaining that the past, present and future exist equally5. The past and future do exist, but to a lesser degree than the present. Accordingly, the present is not understood in terms of what exemplifies the property of presentness, but rather in terms of what exists to the highest degree or, as he puts it, what has “maximal existence.” What is simultaneous with the present is maximally existent; what is earlier than the present is past and has a lesser degree of existence, and what is later than the present is future and also has a lesser degree of existence than the present. Temporal passage involves states or particulars gaining (as they approach the present) and then losing (as they recede from the present) varying degrees of existence. Smith maintains that “These degrees of existence are immediately given in our phenomenological experience6,” and that “the intuitively plausible degrees of existence theory can be defended with respect to its logical coherency7...” The aim of my paper is to argue, contrary to what Smith asserts, that degree presentism is not phenomenologically grounded and that its logical consistency cannot be defended. I want to begin my critique of degree presentism by questioning the ontological principle upon which this new A-theory rests, namely, that reality or existence comes in degrees (of being more or less real). Smith claims that “Most (but not all) philosophers from Plato to Meinong have 5
A. N. Prior, Time and Tense (Oxford: University Press, 1968). For recent proponents of the B-theory see, L. Nathan Oaklander, The Ontology of Time (Amherst, NY: Prometheus Books, 2004). D. H. Mellor, Real Time II (London: Routledge, 1998), Michelle Beer, “Prior's 'Thank Goodness that's Over' Objection to the B-theory,” in Philosophy of Time: Critical Concepts in Philosophy, ed. L. Nathan Oaklander, vol. III (London, Routledge, 2008): 28-34. Robin Le Poidevin, Change, Cause and Contradiction: A Defence of the Tenseless Theory of Time (New York: St. Martins Press, 1991). Joshua Mozersky, “A Tenseless Account of the Presence of Experience,” Philosophical Studies 129 (2006): 441-476, and Heather Dyke, “McTaggart and the Truth About Time,” in Time, Reality and Experience, ed. Craig Callender (Cambridge, Cambridge University Press, 2002), pp. 137-152 to name just a few.
6
Smith, Degrees of Existence, p. 120.
7
Ibid. p. 136.
105
held doctrines of degrees of existence8” and that the difference in degree of existence is not a different kind of mode of being (such as subsistence, as in Meinong, the early Russell and the early Moore9) but of the relation of one and the same entity to the present. As he puts it, “The degree to which an item exists is proportional to its temporal distance from the present; the present, which has zero temporal distance from the present, has the highest (logically) possible degree of existence10.” I will deal with his phenomenological claim shortly, but it seems to me that the appeal to the history of philosophy to defend his theory of degreed existence is questionable at best. For there is an important difference between Smith’s theory of degrees of existence and others who have held the doctrine of degreed existence, e.g., Meinong and Descartes (in Meditation III). For Smith, it is one and the same entity that differs in degree of existence as time passes since, for example, it is the particular Socrates, or the state Socrates (tenselessly) being bald, that gains and loses degrees of existence with the passage of time. For others who have maintained that existence is a matter of degree, different kinds of entities have degreed being. I am not suggesting that simply because Smith’s theory of degrees of existence is not isomorphic with Meinong’s or Descartes’ that it is mistaken, but only that his appeal to such theories as representative does not give his version of the doctrine a clear meaning or a defense. Smith does give one argument for degree presentism, namely, that it is phenomenologically obvious that what is present has more reality than what is not. In other words, we experience those things that are present as having the greatest degree of reality and as they become further and further past we experience them as being less and less real. To the extent to which it is true, and I shall argue that it is not true, Smith’s point seems to me to blur a subjective feature of our experience of the world, with an objective characteristic of the world itself. To use Hume’s terminology, while it may be the case that our experience of events that are happening now sometimes appear more forceful, lively and vivid than our recollection of 8
Ibid. p. 119.
See Bertrand Russell, The Principles of Mathematics, 2nd ed. (London: Allen and Unwin, 1937, originally published in 1903). G.E. Moore, “The Nature of Judgment,” in Selected Writings, ed. Thomas Baldwin (London: Routledge, 1993; originally published in 1899). 9
10
Smith, “Degree Presentism,” p. 120.
106
past events, or our anticipation of future events, which are faint copies of them, it does not follow that the experience in question reflects an objective difference in the degree of existence of the events themselves. Furthermore, Smith’s phenomenological claims about degrees of existence seem to be mistaken. He claims that “we experience existence, as something with degrees, and thus that degree of existence = distance from the present accurately describes our immediate acquaintance with existence and time11,” but this strikes me as false. For example, it seems clear that events which are further in the past than others can seem to be phenomenologically more real than events that are closer to the present. For example, an uneventful event that occurred one year ago can seem less real than an event that occurred 45 years ago if the older event had a greater emotional impact. The experiences of being told of my father’s death and later of my mother’s death, events that each occurred 47 and 46 years ago are more vivid and lively than an uneventful happening of one year ago such as giving a lecture in Philosophy 101. Thus, if our acquaintance with existence and time accurately reflect distance from the present, it would follow that events that happened 47 and 46 years ago are closer to the present and have more reality than an event that happened, say, one year ago. That phenomenological point contradicts the theory that degree of existence = distance from the present describes our immediate acquaintance with existence and time. Of course, sometimes people say things like “It seems to me as if it happened yesterday”, if they want to express that they have a very vivid and intense memory of a certain event that happened in the farther past. At first sight, it might seem as if this linguistic datum is a point in favor of my criticism of Smith’s “degree presentism”. But is it really? A critic could argue that utterances like this one reflect the conviction that there usually is a connection between distance to the present and vividness and intensity of memory (such that it is an exception if something that happened long ago is remembered in a vivid and intense way, an exception for which a particular explanation is needed). In other words, couldn’t one interpret utterances like this as expressing the conviction that in certain cases our experience of vividness and intensity of a memory deceives us with respect to the temporal distance of the remembered event, and would that not imply that we assume that in normal cases there is a relation between the 11
Ibid., p. 122.
107
phenomenology of our memories and the temporal distance of the remembered events?12 Of course, even if this was the case, that there is a relation between the phenomenology of our memories and the temporal distance of the remembered events, that would still not conclusively support degree presentism, but I don’t think it is the case. Consider our experience of the future. We often anticipate events farther into the future with a greater vividness than those events that will occur much closer to the present perhaps because they are more important to us. Such experiences don’t deceive us with regard to how close they are to the present or to their degree of reality but do demonstrate that there is no clear phenomenological correlation between distance from the present and degrees of existence. Although degrees of existence is not conclusively established by how we experience existence, one may still argue, as Smith in fact does, that “degree presentism is a logically consistent …new version of the tensed theory of time13” clearly different from (Prior’s) solipsistic presentism (according to which only the present exists, the past and the future are species of unreality), and the tenseless B-theory of time. In what follows I shall argue even that modest thesis cannot be sustained. To help clarify the ontology of degree presentism I want to distinguish it from Smith’s earlier version of the A-theory, according to which there are A-properties and tensed exemplification ties. Thus, it is not the case that, say, event e is (tenselessly or timelessly) past, present and future, but rather, e is past, was present and future, or e is present, was future and will be past, or e is future and will be present and past. On this view, e is now past, for example, is analyzed as follows: e exemplifies1 pastness, and the exemplification1 of pastness by e exemplifies2 presentness, and the exemplification2 of presentness (by the exemplification1 of pastness by e) exemplifies3 presentness and so on ad infinitum. Similarly, e was future is analyzed such that e exemplifies1 futurity, and the exemplification1 of futurity by e exemplifies2 pastness, and the exemplification2 of pastness (by the exemplification1 of futurity by e exemplifies3 presentness, and so on ad infinitum. As I have shown elsewhere, and shall not repeat here, countenancing A-properties and tensed exemplification in this way leads to a contradiction unless one 12
13
I am indebted to Maria Elisabeth Reicher for making me aware of this objection. Ibid. pp. 122, 125.
108
claims that a term can presently exemplify a property at a time at which it does not exist. In that case, however, one can neither provide an adequate ontological ground for past and future tensed states of affairs nor account for the direction of becoming and time14. Smith’s new A-theory of degree presentism explicitly acknowledges these criticisms by abandoning the key ontological commitments that underlie them. According to degree presentism, exemplification is tenseless (not in the omnitensed sense that it was, is or will be present but) in the sense that it has no A-properties and stands in no B-relations. Furthermore, the monadic properties of pastness and futurity are eliminated. What, then, is the degree presentist analysis of statements about the past (and future)? Smith sets forth the analysis of the past in the following passage that I shall quote at length: “Having been alive is analyzable into the property of aliveness and the state S of the thing tenselessly being alive being earlier than the present time. It is the whole complex state S that stands in this relation to the present, not the thing’s exemplification of being alive. ‘Pastness’, ‘was’, ‘have been’, and the like are analyzable into the exemplification of the property F that the thing possessed at the time it was present, and the complex state consisting of the thing’s exemplification of this property being related to the present time by the relation of being earlier than it. … Socrates’ having been alive is analyzable into tenseless exemplifying the property of aliveness, such that the state of Socrates tenselessly exemplify[ing] this property is over 2,000 years earlier than the present15.”
The point I want to highlight in this analysis, because it gets Smith into trouble, concerns the notion of “the whole complex state S.” One obvious interpretation is that a “whole complex state,” consists of a thing or a particular, (tenselessly) exemplifying a monadic property. Thus, the past tensed state of affairs Socrates having been alive is analysed in terms of the whole complex state, Socrates (tenselessly) being alive, standing in the earlier than relation to the present. It is not the thing’s exemplification (of being alive) that stands in relation to the present (since exemplification is tenseless it does not have B-relations to the present), and it is not the thing See, L. Nathan Oaklander, “McTaggart’s Paradox and Smith’s Tensed Theory of Time,” Synthese 107 (1996): 205-221. Reprinted in L. Nathan Oaklander, The Ontology of Time (Amherst, NY: Prometheus Books, 2004), pp. 153-169. 14
15
Smith, “Degree Presentism,” pp. 126-27; my emphasis.
109
or particular (Socrates) per se that has this relation to the present (since it is the whole complex state S (and not a particular constituent of it) that stands in the earlier than relation to the present). Rather, a state is a particular exemplifying a monadic property (such as being alive), and the more distant a state is from the present (in either temporal direction) the less reality it possesses and consequently the further past (or future) it is. As Smith puts it: “The degrees of existence theory imply that no nonpresent items presently exemplify properties. Rather past or future items tenselessly stand in relations to the present of being earlier or later than it to a certain degree16.”
My interpretation of a “complex state” and the analysis of statements about the past are reinforced by his account of the order of past tense facts, for example, of the fact that say, Thales died before Socrates was born, which he expresses in the following passage: “The complication of the tenses still preserves this relatedness to the present. For example, if I say that Thales had been dead before Socrates was born, we have two past tense expressions, each of whose semantic content includes a relation of being earlier than the present time. The state S composed of Thales (tenselessly) being dead is earlier than the present time and is earlier than the state S’ composed of Socrates’ birth; in addition, the state consisting of Socrates’ being born is earlier than the present17.”
Clearly, Smith is claiming that Socrates (tenselessly) being born and Thales (tenselessly) being dead are each complex states that stand in temporal relations to each other and to the present. Unfortunately, given this understanding of a “whole complex state,” his analyses of the past and the order of past tense states are inconsistent with degree presentism. For the essence of degree presentism is that what is present has maximal existence, but what is it about present items that give them maximal existence? The answer is that present items have maximal existence because only present particulars have non-relational properties. As evidence, consider the following passages: “The fact that past and future individuals lack non-relational properties reflects their ontological status as not fully real beings; in a sense, they are partial 16
Smith, Degree Presentism, p. 131.
17
Ibid. p. 127; italics added.
110
beings. [A past particular is partially nonexistent and] Its partial nonbeing consists in its lack of nonrelational properties and its lack of full existence18.”
Or again, “So we have this result: maximal existents have nonrelational monadic properties and also stand in relations. But particulars that exist to less than the maximal degree only stand in relations. This is one sense in which they are only partly real and partly unreal. Let us ask ourselves again; does it involve a logical contradiction19?”
Assuming the analysis of a complex state given earlier, an affirmative answer can be given to that rhetorical question. Since the complex state, Socrates (tenselessly) being alive is earlier than the present, it is past, but since being alive is a monadic property of Socrates, and only present particulars have non-relational monadic properties, it follows that Socrates being alive is present. Furthermore, in virtue of exemplifying being alive Socrates partakes of maximal existence, but in virtue of being earlier than the present Socrates partakes of existence to a lesser degree. Thus, on Smith’s analysis the complex state of Socrates being alive is both past and present, and the particular Socrates both does and does not exist to a maximal degree. Moreover, the order of past tense facts is lost. Since a state is a particular exemplifying (tenselessly) a monadic property, and given that all such states are present, they exist to a maximal degree, and for that reason Socrates (tenselessly) being alive and Thales (tenselessly) being dead are both present, and therefore one cannot be earlier than the other. Of course, one might then maintain, as some certainly have, that the ground of statements about the past and temporal relations between such states lies in the present. In that case, however, degree presentism reduces to what Smith calls “solipsistic or modal presentism,” Le Poidevin calls “temporal solipsism” and Tegtmeier calls “Solpräsentismus,” a view strongly rejected by each of those philosophers20. 18
Ibid. pp.133 and 129.
19
Ibid. p. 132.
See Smith, “Degree Presentism,” pp. 123-24. Robin Le Poidevin, Change, Cause and Contradiction, pp. 36-57, and Erwin Tegtmeier, Zeit und Existenz, Parmenideische Meditationen, (Mohr Siebeck: Tübingen, 1997). 20
111
Alternatively, since one could say instead that since complex states such as Socrates’ being alive and Thales being dead, and all other complex states each have maximal existence, and do stand in temporal relations to each other, degree presentism reduces to the B-theory. Each state is present and thus has maximal existence at the time at which it occurs, regardless of what time it is, and that is a B-theoretic conception of time. To say that a state stands at a certain temporal relation (or distance) to the present, i.e., from what exists maximally, is, on this analysis of a state and maximal existence, to say that states (tenselessly) exist in temporal relations to each other and that is the B-theory since the resulting facts are B-facts and not A-facts. There is a final problem I wish to mention with Smith’s analysis that is closely related to the preceding. The degree presentist analysis of “Thales was dead before Socrates was born” involves the state or fact of Thales (tenselessly) being dead earlier than the state or fact of Socrates (tenselessly) being born. Is that a B-fact or an A-fact? If it is a B-fact, then that that contradicts Smith’s claim that, “For me there are only tensed facts (where ‘tensed’ now has the nonlinguistic, ontological sense of Afacts, as distinct from B-facts). Every fact includes a relationship to the present21.” On the other hand, if Thales being (tenselessly) dead is (tenselessly) earlier than the state S’ composed of Socrates’ birth, is an Afact, then it must have a relation to the present. However, it is difficult to understand what that relationship could be. If it is earlier or later than the present, then there must be some time when it is simultaneous with the present. That is, if the complex state lacks existence to a certain degree (since it is at a temporal distance from the present), then it must have had maximal existence at one present time or another. Since, however, both of the relata cannot have maximal existence at the same time, the entire relational complex could not be present with maximal existence at any time. Thus, the fact of a state being earlier than another state cannot be a tensed fact since if it cannot be simultaneous with the present then it cannot stand in any other temporal relation to the present. If the complex state in question is a fact that has no relation to the present it is a B-fact and not an A-fact. This criticism will emerge again later. The objections I have been discussing rest on the assumption that a “complex state” is a particular (tenselessly) exemplifying a property, including monadic properties, and that complex states that stand in earlier/later and simultaneity relations to the present. However, perhaps it 21
Ibid. p. 129.
112
is particulars (and not states) that stand in relations to the present and to each other. Perhaps it is particulars that undergo temporal becoming by changing their relation to the present by gaining or losing a greater degree of existence. There is some evidence that Smith holds such a view when he says: “The unusual feature of degree presentism is summarized as this: Past (or future) particulars do not have nonrelational, monadic properties, but only stand in relations or have relational properties. Thus they are ‘bare particulars’ in the sense that they lack nonrelational, monadic properties22.”
And, “When the particular becomes past, it possesses an essentially relational property, one that is the past-time version of the presently possessed property. Instead of it being true that x has the essentially nonrelational property of being human, it is now true that x tenselessly has the essentially relational property of having been a human over 100 years earlier than the present23.”
These passages support the interpretation that it is particulars and not states that stand in earlier/later than relations to the present and that past and future particulars have only relational properties, and not monadic properties. Does that enable him to avoid any of the objections previous delineated? In correspondence Smith has pointed out that there is a difficulty with degree presentism even if “monadic predicates … of past and future events are abbreviations of relational predicates24,” and thus the proper reading of “the whole complex state S” is the entire fact consisting of the thing’s exemplifying a property being temporally related to the present, (and not a particular exemplifying a monadic property). For, he says, “You probably could refute [my criticism] by saying (a) Socrates (tenselessly) is alive is earlier than the present Ibid. p. 132; emphasis added. Of course this is a very unBergmann like use of “bare particular” since for Bergmann there are no bare particulars that lack non-relational monadic properties. Since, however, that difference makes no difference for what I am about in this paper, I shall ignore it. 22
Ibid. p. 135. I will ignore the potential difficulty that is, the circularity, of incorporating the tensed expression “having been” into his analysis of the past.
23
24
Ibid., p. 126.
113
entails (b) Socrates is (tenselessly) alive And we do not know whether it is true or false unless we value it with respect to a temporal circumstance of evaluation (to use Kaplan’s terminology), such as 420 bce25.” If the proposition (b) does not possess a truth value unless it is evaluated with respect to some (B-time) circumstance of evaluation, for example 1 June, 420 BCE, then we arrive at the B-theory. Certainly that is a difficulty with degree presentism, but in what follows I wish to attend to a different set of problems. The first question to ask concerning this revised analysis of the past is what constitutes the passage of time? Alternatively, what makes degree presentism an A-theory of time? Suppose that a is present when a is F is simultaneous with the present, and that a becomes past when it exemplifies the relational property ‘being F earlier than the present.’ Presumably, in that case, a first has maximal existence and then has less than maximal existence. To see why this is problematic I want to explore the following question: What is the ontological status of the fact (or complex state) that a (tenselessly) exemplifies being F earlier than the present time? In particular, how is that fact related to time and existence? Is it a timeless (atemporal, eternal) fact or temporal fact26? If the fact (call it “P”) that, say, a (tenselessly) exemplifies being F (one minute) earlier than the present time is timeless, that is, it does not exist in B-time or A-time (and thus the proposition that expresses it is freely repeatable), then degree presentism is either contradictory or reduces to the B-theory. To see why consider another timeless fact, (call it “Q”), that a (tenselessly) exemplifies being F (two minutes) later than the present time. “The present time,” in P and Q denote the same present time or a different present time. If “the present time” denotes the same present time in P and Q then a (tenselessly) exemplifies being F one minute earlier and two minutes later than the same present moment, and that is absurd for then a would be past and future at the same present moment, and it would exist to different degrees at the same present moment. 25
E-mail correspondence from Quentin Smith to Nathan Oaklander on 1/10/2007.
To put the question linguistically, Is the proposition that represents the fact that a (tenselessly) exemplifies being F earlier than the present time, freely repeatable, that is, does it have an unchanging truth value, or does its truth value vary depending upon the time at which the proposition is expressed?
26
114
On the other hand, if P and Q have different present times as constituents then a contradiction is avoided, but then degree presentism collapses into the B-theory. Clearly, there is no contradiction in a’s tenselessly exemplifying being F one minute earlier than the “present” time t1 and two minutes later than the “present” time t2. However, in that case, there is no moment ontologically distinguishable as the present, and thus a never changes its relation to the present or undergoes temporal becoming. In short, if P and Q are timeless facts (and thus expressed by freely repeatable tenseless sentences) and the present denotes a different time in each such fact, then P and Q are really B-facts and not A-facts, and thus we arrive at a B-theory, not an A-theory. Suppose we deny that P is timeless and maintain instead that P, Q and all other facts of that form exist in time, then either a contradiction ensues, statements about the past (and future) lack an ontological ground, or there results a vicious infinite regress. For if P: a exemplifies being F earlier than the present time t1, exists maximally at the present time t1, then P is simultaneous with the present t1. However, P cannot be simultaneous with the present time (t1), since it contains a constituent, a, that exists earlier than the present time t1. Alternatively, if P including a, exists maximally at the present time t1, then a is simultaneous with the present time and a is earlier than the present time t1 which is impossible. Nor can P (including a) exists maximally at the some other present time ti, since then it would not be the case that a exemplifies being F earlier than the present time t1. Thus, P cannot exist maximally at any time, and thus cannot exist in time at all, and thus if we construe P, Q and all other facts of that form as existing in time, then statements about the past (and future) do not have an ontological ground. Moreover, to assert that P, Q and all other facts of the same form exist in time gives rise to a vicious infinite regress. If P exists in time, and there is temporal passage, then (ignoring the preceding argument), P first exists to the maximal degree and then it diminishes in existence as it recedes from the present into the past. That is, P is first simultaneous with the present time (call this fact (P’), and then P is earlier than the present time (call this fact (P’’). What, then, is the ontological status of (P’) and (P’’)? Are they timeless or temporal facts? I have already shown that if P and Q are timeless, then we can avoid a contradiction only by reducing this version of the degree presentist theory to the B-theory. And, if P and Q are in time, we can only avoid a contradiction if we fail to give an adequate ground of statements about the
115
past (or future). What goes for P and Q goes equally for P’ and P” whether those facts are timeless or in time the result is a contradiction, a reduction to the B-theory or a failure to ground statements about the past (and future). Hence the appeal to P’ and P’’ to provide an A-theoretical account of temporal becoming or passage will lead to a vicious infinite regress. I conclude, therefore, that that degree presentism is not grounded in the phenomenology of time and existence and that its logical consistency cannot be defended27.
I have benefited greatly from comments by Michelle Beer, Quentin Smith, and Erwin Tegtmeier on earlier drafts of this paper. The research for this paper was supported by a grant from the faculty development fund of the University of Michigan-Flint.
27
116
La théorie des faisceaux et la théorie des substrats1* JIRI BENOVSKY La théorie des substrats s'explique aisément en termes de la relation qui unit les particuliers et leurs propriétés : il y a des particuliers et il y a des propriétés qui sont exemplifiées par ces particuliers qui en sont les porteurs. Un tel porteur de propriétés, dont l'identité est indépendante des propriétés qu'il porte, est souvent appelé "substrat" ou "particulier nu". Un objet comme une table ou une personne est ainsi fait de deux différents types de composants : les propriétés et un substrat qui les soutient et les 'colle' ensemble pour faire un objet. À l'inverse de la théorie des substrats, la théorie des faisceaux nie l'existence et le besoin théorique du substrat : en tant que composants fondamentaux de la réalité, il n'y a que des propriétés. Prenons à titre d'exemple mon voisin Cyrano : il est d'un certain âge, il a un grand nez, il a une certaine taille, et ainsi de suite. L'énumération de ses propriétés constitue tout ce qu'il y a à savoir sur Cyrano et également tout ce qu'est Cyrano. Selon la théorie des faisceaux, un objet est alors un faisceau (ou un "bouquet") de ses propriétés. Il n'y a pas besoin, et il n'y a pas de place, à l'intérieur de la théorie des faisceaux, de postuler deux types de composants pour faire des objets, plutôt, ceux-ci sont des faisceaux de propriétés qui sont les composants ultimes de la réalité et qui sont maintenues ("collées") ensemble par une propriété spéciale (une relation n-adique où n est le nombre de propriétés de l'objet) souvent appelée comprésence. Selon comment l'on conçoit la relation de comprésence et la nature des propriétés, la théorie des faisceaux se présente alors sous l'une de ces différentes variantes :
*1 J'aimerais remercier les participants à la Conférence Bergmann à Aix-en-Provence en décembre 2006 pour leurs commentaires très utiles qui m'ont permis d'améliorer cet article. J'adresse des remerciements particuliers aux organisateurs de la conférence Lynda Gaudemard et Jean-Maurice Monnoyer; ainsi qu'à Bruno Langlet et Lynda Gaudemard pour leur aide précieuse en ce qui concerne la langue française dans laquelle j'ai peu l'habitude d'écrire.
comprésence
une seule et même relation numériquement identique pour tous les objets
une seule et même relation variablement polyadique numériquement identique pour tous les objets
différentes relations (une par objet)
les propriétés sont des universaux
3
5
4
les propriétés sont des tropes
1a
1b
2
théorie des faisceaux
Les numéros dans ce tableau indiquent simplement dans quel ordre je vais examiner ces différentes variantes; les variantes réellement intéressantes étant la 2, la 4 et la 5. Qu'en est-il de la théorie des substrats ? Faut-il également dessiner une telle table ? Pas vraiment : bien sûr, la théorie des substrats existe aussi en différentes versions selon que les propriétés sont conçues comme étant des universaux ou des tropes, mais il n'y a pas ici lieu de faire une distinction parallèle à celle concernant la relation de comprésence – le substrat est, par définition, numériquement différent pour chaque objet. Les variantes (1a) et (1b) ne sont pas vraiment des variantes possibles. Premièrement, simplement parce que les tropes, contrairement aux universaux, ne peuvent pas être localisés de manière multiple dans différents objets et donc il n'est pas possible ici qu'une seule et même relation de comprésence fasse son travail de "colle" dans différents objets. Deuxièmement, (1a) n'est pas une variante acceptable pour une autre raison : tous les objets n'ont pas le même nombre de propriétés, et quelque chose comme une relation variablement polyadique (comme "x, y, z, … sont comprésentes les unes avec les autres") est requis. (Ceci sera aussi le cas pour (3).)
118
Ainsi, la seule variante véritablement intéressante de la théorie des faisceaux avec les tropes en ce qui concerne la relation de comprésence est (2). Comparons alors à présent cette théorie, la TFT (Théorie des Faisceaux avec Tropes) à son opposant, la TST (Théorie des Substrats avec Tropes). Pour faire des objets avec des propriétés, la TFT utilise une relation de comprésence qui est généralement considérée comme étant nonanalysable et ontologiquement primitive. Elle n'est ainsi pas définie et individuée par sa nature ou par ses traits intrinsèques sur lesquels TFT ne nous renseigne pas beaucoup, mais plutôt par son rôle théorique : il s'agit d'un outil d'unification, un outil qui prend des propriétés pour faire des objets. Comparez ceci à TST (et voyez à quel point il m'est facile d'écrire une paraphrase presque exacte du paragraphe précédent). Pour faire des objets avec des propriétés, la TST utilise des propriétés et un porteur de propriétés, le substrat, qui est généralement considéré comme étant nonanalysable et ontologiquement primitif. Il n'est ainsi pas défini et individué par sa nature ou par ses traits intrinsèques sur lesquels TST ne nous renseigne pas beaucoup, mais plutôt par son rôle théorique : il s'agit d'un outil d'unification, un outil qui prend des propriétés pour faire des objets. Les deux théories possèdent ainsi un outil d'unification, qui est primitif et non-défini, une entité dont le seul but est de "coller ensemble" des propriétés d'un objet. En paraphrasant Locke, dans les deux cas cet outil d'unification est un "we-know-not-what" … mais en revanche c'est un "we-knowwhat-it-does", c'est-à-dire, nous savons quel est son rôle théorique. Il est souvent objecté à la TST : "En vertu de quoi un substrat est-il distinct d'un autre substrat ? Aucun attribut ou propriété ne permet de les distinguer !". Mais la même question peut également être soulevée à propos de la relation de comprésence : "Tu quoque : en vertu de quoi une relation de comprésence (qui est la colle d'un objet A) est-elle distincte d'une autre (qui est la colle d'un objet B) ? Aucun attribut ou propriété ne permet de les distinguer !". Les deux théories répondent à ces interrogations par une thèse primitiviste. C'est en réfléchissant à ce qui précède qu'il m'est apparu pour la première fois que TFT et TST ne sont peut-être pas très différentes l'une de l'autre. Mais bien sûr cela ne suffit pas (encore) pour même commencer à justifier une thèse d'équivalence métaphysique entre les deux théories. Je vais faire cela en examinant d'abord une différence possible entre les deux théories en ce qui concerne le statut de l'outil d'unification (pour voir qu'il
119
n'y en a aucune), et ensuite en examinant quelques objections à TFT et TST où l'outil d'unification joue un rôle crucial et où je vais défendre un tu quoque dans tous les cas. Une différence possible entre TFT et TST pourrait être que la comprésence est juste un des éléments parmi ceux du faisceau (juste une des propriétés parmi celles de l'objet) alors que le substrat doit être considéré à part des propriétés qu'il porte. Ainsi, la comprésence et le substratum n'auraient pas le même statut, ils ne joueraient pas le même rôle théorique dans la composition d'un objet. Mais ceci n'est pas vrai, comme l'a montré Ehring (2001) : les propriétés incluses dans un faisceau sont comprésentes. La relation de comprésence, toutefois, n'est pas un membre du faisceau comme les autres propriétés et relations car si nous incluions la comprésence dans le faisceau, alors elle devrait elle-même être comprésente avec les autres propriétés : la comprésence comprésente avec F, G, H, … Mais ceci soit ne fait pas sens, soit mène à un regressus à l'infini. Ce que cela montre est que la comprésence, exactement comme le substrat, doit être considérée à part des autres propriétés de l'objet; son statut en tant qu'outil d'unification est ainsi différent de celui des propriétés et est alors le même dans TFT et dans TST. Considérons à présent la première de deux objections classiques faites à TFT, et examinons comment TST est censée les éviter, en faisant appel au substrat. La première objection peut être trouvée par exemple dans Van Cleve (1985, p. 122) : "If a thing were a set of properties, it would be incapable of change. For a thing could change its properties only if the set identical with it could change its members, but that is impossible; no set can change its members." En prenant un exemple d'un individu dont les propriétés sont censées changer à travers le temps, il ajoute : "[…] what we have here is replacement of one individual by another, not change in the properties of one and the same individual." (Van Cleve (1985, p. 124)). L'idée ici est simple : si un objet est identifié avec un faisceau de propriétés, alors si l'une de ces propriétés change, le faisceau n'est plus le même, et par conséquent, l'objet qui est ce faisceau n'est pas le même – il a simplement cessé d'exister, alors qu'un autre objet a pris sa place. Ainsi, selon TFT, rien ne peut subir un changement en ce qui concerne ses propriétés. Comment est-ce que TST évite cette objection ? Elle ne l'évite pas ! Il n'y a pas plus de changement véritable dans la théorie TST que la théorie TFT : le substrat (le "particulier nu") ne peut évidemment pas changer (dans la
120
mesure où il n'a aucune propriété, il ne peut pas subir un changement qualitatif), et le particulier pris en entier (le substrat + les propriétés qu'il porte) ne peut pas changer non plus pour exactement la même raison que dans le cas de TFT. Il n'y a pas de changement, mais plutôt un remplacement d'un particulier par un autre : prenez un tel particulier à t1 et à un temps postérieur t2 alors qu'il a subi un changement qualitatif, il s'agit tout simplement de quelque chose de différent. Bien sûr, l'amie de la théorie TST va objecter : "J'ai quelque chose que toi, TFT, tu n'as pas : un substrat qui reste le même à travers les différents temps, et ceci me garantit que l'objet, alors qu'il change ses propriétés, reste le même objet." Mais si ceci était une réponse acceptable, TFT pourrait l'utiliser également : "J'ai une relation de comprésence qui reste la même à travers les différents temps, et ceci me garantit que l'objet, alors qu'il change ses propriétés, est le même individu. Tu as ton outil d'unification primitive, moi j'ai le mien." Ainsi, il apparaît que les deux théories peuvent répondre à l'objection de la même manière, en faisant appel à l'outil d'unification, et le fait de nommer cet outil par des termes différents ("comprésence" ou "substrat") ne fait aucune différence, dans la mesure où les deux outils d'unification jouent le même rôle de la même manière. Pour être plus précis, examinons à présent les deux stratégies principales qu'il y a pour faire face au problème de la persistance à travers le temps – une version du perdurantisme et une version de l'endurantisme – et voyons s'il y a là ou non une différence entre l'utilisation d'un substratum ou de la relation de comprésence. (Il n'y en a pas.) La version traditionnelle de perdurantisme (le "worm view"), étant donné TFT, est une théorie de faisceaux de faisceaux. Pensez encore à mon voisin Cyrano, comme à un cas d'un objet qui change de propriétés intrinsèques. A t1, il a un grand nez mais il subit une opération de chirurgie esthétique et à t2 il a un petit nez.
121
Cyrano
t1
t2
Selon le perdurantisme, Cyrano est un agrégat de ses parties temporelles, qui sont des objets numériquement et qualitativement distincts. Le compterendu perdurantiste du changement affirme ainsi que Cyrano change à travers le temps en ayant différentes parties temporelles à différents temps. En termes de TFT, l'image que nous avons à l'esprit est alors la suivante (où "C" désigne la relation de comprésence) : Cyrano t1-partie
t2-partie
t3-partie
F, G, H
F, G, K
………..
C
C
C
Cyrano est ici un faisceau de faisceaux instantanés de propriétés (un faisceau de faisceaux de propriétés). Qu'en est-il de la version perdurantiste de TST ? Elle n'est pas très différente de l'image du perdurantisme TFT, sauf que j'ai dû changer "C" en "S" (pour "substrat") :
122
Cyrano t1-partie
t2-partie
t3-partie
F, G, H
F, G, K
………..
S
S
S
J'aimerais alors suggérer qu'il semble que la seule différence que nous avons ici est terminologique : dans les deux cas, il y a un outil d'unification inclus dans les parties temporelles instantanées de Cyrano, et dire que dans un cas il s'agit de la comprésence et dans l'autre cas c'est un substrat ne fait aucune différence dans la manière dont cet outil d'unification arrive à jouer son rôle théorique. Regardons à présent l'alternative endurantiste. L'endurantisme est la théorie selon laquelle Cyrano persiste à travers le temps et à travers le changement en existant entièrement à différents moments, plutôt qu'en ayant des parties temporelles. Pour l'endurantiste, une et la même personne numériquement identique existe entièrement à t1 et t2 et a les deux propriétés incompatibles d'avoir un grand nez et d'avoir un petit nez. Pour éviter d'affirmer une contradiction, les endurantistes vont typiquement utiliser une stratégie d'indexation temporelle, et c'est la version standard avec les propriétés temporellement indexées que je vais utiliser maintenant2. Selon cette théorie, l'indexicalisme, Cyrano n'a pas des propriétés incompatibles, car au lieu d'avoir des propriétés comme "avoir un grand nez", il a des propriétés temporellement indexées comme "avoirun-grand-nez-à-t1" et "avoir-un-petit-nez-à-t2" qui sont parfaitement compatibles et non-contradictoires. Ainsi, puisque l'indexicaliste va affirmer que toutes les propriétés sont toujours indexées, aucune contradiction ne peut apparaître dans des cas de changements intrinsèques d'un objet qui est numériquement un et le même à différents instants, comme le soutient l'endurantisme. Le schéma de cette version de la TFT que l'on est tenté de dessiner ici est le suivant (comme avant, "C" est la relation de comprésence) : 2
Je parlerai de l'adverbialisme plus loin.
123
F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
Cyrano
Cyrano
Cyrano C
F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
t1
C
F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
C
t3
t2
Mais ceci est une mauvaise image de la théorie. Ce n'est pas l'image que l'endurantiste devrait avoir à l'esprit et dessiner. Premièrement, pourquoi dessiner trois fois la même chose ? Ces trois faisceaux de propriétés ne sont pas trois faisceaux localisés à trois localisations temporelles différentes, plutôt il s'agit d'un seul et même faisceau. Dessiner ici trois fois la même chose est tout à fait redondant et inutile. Deuxièmement, l'axe du temps est également redondant – nous n'en avons pas besoin pour savoir ce qui est arrivé à Cyrano à différents moments, puisque toutes les spécifications temporelles sont déjà incluses dans les propriétés de Cyrano. Donc, après avoir enlevé les éléments redondants, voici l'image endurantiste correcte : Cyrano F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
C
Cyrano est simplement un faisceau de propriétés temporellement indexées. Si vous trouvez cela étrange et surprenant … vous ne devriez pas ! Si vous trouvez cela étrange et surprenant, c'est peut-être parce que vous êtes habitués à penser en termes d'images perdurantistes. Mais j'espère que vous ne trouvez pas cela étrange et surprenant – après tout, c'est exactement l'image endurantiste que Peter Van Inwagen dessine (voir Van Inwagen (1985, p. 195)), sauf qu'il ne le fait pas en termes de la théorie des faisceaux.
124
Et de nouveau, sans aucun effet de surprise, l'image endurantiste TST est comme l'image précédente : Cyrano F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
S
Le résultat des considérations ci-dessus est de nouveau simplement le suivant : le substrat et la relation de comprésence jouent exactement le même rôle théorique. Ainsi, parce que TFT et TST utilisent leur outil d'unification de la même manière, elles ont exactement les mêmes moyens de faire face à l'objection d'une manière parallèle et il semble que la différence entre elles est simplement terminologique – l'une a un outil d'unification appelé "C" et l'autre a un outil d'unification appelé "S" mais puisque les deux outils sont des entités théoriques (elles sont là pour faire le même travail théorique) qui sont individuées par leur rôle théorique, et puisqu'elles jouent leur rôle théorique de la même manière, elles semblent être une seule et même chose sous différents déguisements. Et si ceci est bien le cas, alors il ne semble y avoir aucune véritable différence entre TFT et TST. Je vais à présent examiner plus rapidement la seconde des deux objections classiques à TFT, qui est l'analogue modal de la première, et qui est structurellement similaire. Prenons Cyrano, qui est un faisceau de faisceaux de propriétés (ou un faisceau de propriétés temporellement indexées, si l'on préfère l'approche endurantiste). L'objecteur remarque alors, comme auparavant, que l'identité des faisceaux dépend de leurs constituants – un faisceau doit avoir les constituants qu'il a, autrement il ne serait plus le même faisceau. Ainsi, il semble que les composants d'un faisceau lui sont essentiels. Mais ceci signifie que l'amie de TFT doit faire face à une conséquence indésirable de sa théorie qui est que n'importe quelle propriété de n'importe quel objet en est une propriété nécessaire. Si l'on prend Cyrano, qui a un grand nez, nous avons un faisceau de faisceaux de propriétés parmi lesquelles il y a la propriété d'avoir un grand nez. Mais puisque Cyrano est ce faisceau, et puisque les faisceaux ont leurs composants de manière
125
nécessaire, il est impossible pour Cyrano d'avoir d'autres propriétés que celles qu'il a actuellement – même les plus insignifiantes et contingentes, comme la taille de son nez, ou le nombre exact de cheveux qu'il avait ce matin à sept heures. Si cela est vrai, alors la théorie des faisceaux n'apparaît certainement pas comme très séduisante. Comment TST évite cette objection ? Elle ne l'évite pas ! Le substrat (le particulier nu) ne peut pas avoir différentes propriétés que celles qu'il a actuellement, car il n'en a aucune, et l'objet entier (substrat + les propriétés qu'il porte) ne peut pas non plus avoir différentes propriétés que celles qu'il a actuellement pour exactement la même raison que nous avons vu dans le cas de TFT. Si ce qui individue l'objet Cyrano est seulement le substrat (ce qui serait une théorie plus qu'étrange) alors il a toutes ses propriétés de manière nécessaire parce qu'il n'en a aucune, et si ce qui individue l'objet Cyrano est le substrat et ses propriétés, alors si on lui enlève une propriété, il n'est plus le même objet, exactement comme dans le cas de TFT. Bien sûr, l'amie de TST va, comme avant, objecter : "Moi, j'ai quelque chose que toi, amie de TFT, tu n'as pas : un substrat qui me permet de dire qu'un objet peut avoir différentes propriétés parce que le substrat va garantir que ce sera toujours le même objet même si certaines propriétés changent." Mais si cela était une réponse acceptable ici, l'amie de TFT pourrait l'utiliser également : "J'ai une relation de comprésence qui reste la même quelles que soient les propriétés qu'elle lie ensemble, et ceci me permet de dire qu'un objet peut avoir différentes propriétés parce que la relation de comprésence va garantir que ce sera toujours le même objet même si certaines propriétés changent. Tu as ton outil d'unification, moi, j'ai le mien." Comme avant, il apparaît que les deux théories peuvent donc répondre à l'objection de la même manière en faisant appel à leur outil d'unification, et le fait de désigner cet outil par différents termes ("comprésence" ou "substrat") ne change rien, puisque les deux outils d'unification jouent le même rôle théorique de la même manière. Pour être plus précis, exactement comme dans le cas temporel, toutes les deux théories ont dans le cas modal la possibilité de choisir leur théorie préférée de la persistance à travers les mondes possibles (l'identité transmondaine, la théorie des contreparties, …) et l'utiliser pour répondre à l'objection. Comme avant, mon point n'est donc pas ici de défendre TFT contre cette objection, mais plutôt de montrer que quels que soient les moyens que TST a à disposition pour y répondre, ceux-ci peuvent être utilisés de la même manière par TFT, tout simplement parce que les deux théories contiennent un outil d'unification
126
qui semble être tout simplement une seule et même chose. L'argument pour cette dernière affirmation, que j'ai déjà introduit plus haut, est le suivant : A.
TFT et TST sont équivalentes parce que B. l'outil d'unification appelé "substrat" dans TST et l'outil d'unification appelé "comprésence" dans TFT sont identiques (métaphysiquement équivalents) parce que C. ils jouent le même rôle théorique de la même manière et D. ce sont des entités théoriques (c'est-à-dire, elles sont individuées par leur rôle théorique). Mais peut-être que certains ne vont pas se sentir à l'aise avec (D) parce qu'ils vont penser que le substrat et la relation de comprésence, même s'ils font le même travail théorique, ne sont pas 'juste' des entités théoriques mais sont des entités métaphysiquement différentes. (Je leur demande alors, non pas de manière purement rhétorique mais avec un réel intérêt, de montrer pourquoi ils pensent cela.) Ou peut-être que certains ne vont pas être d'accord que (C) et (D) sont suffisants pour justifier (B), et par conséquent pour justifier (A), parce qu'ils pensent que jouer le même rôle théorique n'est pas suffisant pour justifier qu'il y a une équivalence métaphysique. Selon le point de vue que l'on adopte face à ces soucis, l'on peut alors être tenté d'accepter l'une ou l'autre des conclusions suivantes : • Conclusion forte Thèse : TFT et TST sont métaphysiquement équivalentes Argument : C et S sont des entités théoriques, ce qui veut dire qu'elles sont individuées par leur rôle théorique. Et puisque leur rôle théorique est le même, ce sont les mêmes entités théoriques. • Conclusion faible Thèse : il est épistémiquement sous-déterminé laquelle des deux théories TFT ou TST nous devrions choisir. Argument : C et S sont des entités métaphysiques différentes, mais elles jouent le même rôle théorique de la même manière, et TST et TFT ont le même pouvoir explicatif (les deux font le travail que nous voulons qu'elles fassent).
127
Pour ma part, je suis personnellement tenté d'accepter la Conclusion Forte, en suivant l'argument (A)-(D) ci-dessus, mais parce que je ne sais pas comment éliminer la Conclusion Faible, je me limite donc à cette seconde affirmation. Il est temps à présent d'examiner la ligne du haut de ma table au début, et de comparer la Théorie des Faisceaux avec Universaux (TFU) et la Théorie des Substrats avec Universaux (TSU). Comme il a déjà été noté, dans cette table, (3) est à éviter pour exactement la même raison que (1a) : les objets n'ont pas tous le même nombre de propriétés, et donc si l'on souhaite avoir une seule et même relation comme relation de comprésence pour tous les objets (ce qui est ici plausible car nous sommes ici amis des universaux), quelque chose comme une relation variablement polyadique telle que "x, y, z, … sont comprésentes les uns avec les autres" est requis. Mais tout d'abord, regardons de plus près (4) où différents universaux jouent le rôle de la relation de comprésence pour différents objets; un universel par objet. En ce qui concerne les deux objections que nous avons vues dans le cas de TFT et TST, la situation est ici la même, et la même conclusion concernant leur équivalence peut être tirée. Mais il y a une autre objection traditionnelle à la théorie des faisceaux avec universaux qui pourra peut-être faire une différence entre TFU et TSU : celle qui vient du principe de l'Identité des Indiscernables. Examinons donc cette objection. TFU est obligée d'accepter le principe de l'Identité des Indiscernables. Mais ce principe est faux3. Donc, TFU est fausse. [Id.Ind.] (∀x) (∀y) ((∀F) (Fx ↔ Fy) → (x = y)) Selon TFU, les objets matériels sont des faisceaux de propriétés. Prenons maintenant le cas de deux objets qui ont les mêmes propriétés, par exemple, comme dans le monde de Max Black, deux sphères parfaites de la même taille, même masse, même composition, même couleur, etc. Les deux sphères sont des faisceaux des mêmes propriétés (universaux4) – et sont donc les mêmes faisceaux. Mais alors, TFU doit accepter que les deux sphères sont numériquement identiques – c'est-à-dire qu'il n'y a qu'une seule sphère. Et ceci est exactement ce que le principe de l'Identité des Indiscernables soutient. 3 4
Ou vrai seulement de manière contingente. Bien sûr, TFT n'est pas vulnérable à cette objection.
128
Mais ce principe est faux, car il est bien possible qu'il y ait deux objets numériquement distincts qui ont exactement les mêmes propriétés (ce sont des duplicata qualitatifs). L'exemple des deux sphères ayant exactement les mêmes propriétés est possible. Comment est-ce que TFU peut s'en sortir ? Distinguer entre les deux sphères par l'utilisation de propriétés de localisation spatio-temporelle ("être à gauche de Cyrano") ou par l'usage de propriétés haecceistiques ("être identique à la sphère A") n'est pas, comme le montre la littérature abondante à ce sujet, une stratégie très séduisante. Mais en ce qui concerne mon but, ce n'est pas le moment maintenant de se demander comment la TFU peut faire pour répondre à cette objection. Plutôt, ma question est : comment TSU lui fait face ? Pourquoi est-ce que cette objection n'apparaît même pas contre TSU ? La réponse est obvie : les deux sphères sont distinguées non pas par leurs propriétés (ce sont des duplicata qualitatifs), mais par ce qui les porte, c'est-à-dire le substrat qui les unifie et les 'colle' ensemble pour faire un objet. Les substrats étant numériquement distincts, les deux sphères sont numériquement distinctes également. Mais quelle est la raison pour affirmer que les substrats sont numériquement distincts ? Il ne peut pas s'agir d'une différence qualitative entre eux, donc qu'est-ce que c'est ? Comme nous avons déjà vu, il n'y a pas beaucoup de choix, et donc les amies de la théorie des substrats affirment que la différence numérique entre deux substrats est un fait primitif. (Très bien, toute théorie a ses primitives.) Comme avant, TFU peut alors utiliser exactement la même stratégie – rappelez-vous que nous avons ici des relations de comprésence différentes, une par objet, et donc que deux objets, même qualitativement identiques, seront numériquement distincts parce que la relation de comprésence qui lie leurs propriétés ensemble est un universel différent – exactement comme dans le cas de TSU c'est un substrat différent. On peut alors soulever un tu quoque : en vertu de quoi est-ce qu'une relation de comprésence donnée est numériquement distincte d'une autre relation de comprésence ? Et il n'y a aucune réponse à cette question que dans le cas des substrats : la seule bonne option est de les distinguer de manière primitive. Si ce qui est précède est correct, il semble alors que la chose qui joue le rôle d'outil d'unification dans TSU (le substrat) et la chose qui joue le rôle d'outil d'unification dans TFU (la relation de comprésence) sont toutes les deux également bien équipées pour faire le travail : une fois que nous sommes prêts à accepter que l'objection de l'Identité des Indiscernables est évitable grâce à une distinction primitive entre substrats, pourquoi ne pas
129
aussi joyeusement accepter que l'objection est évitable grâce à une différence numérique primitive entre deux relations de comprésence ? Après tout, les amies de la théorie des faisceaux parlent souvent de la relation de comprésence comme étant une primitive sous-expliquée et sous-définie, tout comme le substrat, alors pourquoi ne pas la laisser faire le travail pour lequel est si naturellement faite ? Ainsi, de même que dans le cas de TFT et TST, les deux théories ont les mêmes moyens d'éviter toutes difficultés liées au principe de l'Identité des Indiscernables, car les deux contiennent un outil d'unification qui le leur permet de la même manière, et appeler cet outil par différents noms ("substratum" ou "comprésence") ne fait aucune différence autre que terminologique. Si nous nous en tenons à un vocabulaire neutre (comme "outil d'unification") et que nous reformulons les deux théories en ce qui concerne l'objection, nous dirons que les deux peuvent affirmer que la sphère A et la sphère B peuvent être distinguées par le fait qu'il y a un outil d'unification primitivement distingué pour A et pour B. Par conséquent, comme avant, la Conclusion Forte ou la Conclusion Faible peuvent être tirées. Note : de la même manière, TFU peut éviter d'autres objections qui reposent en fin de compte sur le principe de l'Identité des Indiscernables, comme par exemple "le problèmes des anges" (Hawthorne et Cover (1998, p. 216)) ou "le problème avec le temps circulaire et l'espace circulaire" (Hawthorne et Cover (1998, p. 218)). En effet, ces objections ne sont en réalité 'que' des variantes d'un même problème avec le principe de l'Identité des Indiscernables. Mais bien sûr, Hawthorne et Cover utilisent une autre version de TFU – celle que je vais examiner à présent. Dans le reste de cet article, je vais considérer des points où la théorie des faisceaux et la théorie des substrats ne sont pas équivalentes, et je vais montrer que la raison pour laquelle elles ne sont pas équivalentes est aussi la raison pour laquelle elles sont bien moins attrayantes5. Un premier tel point est le cas (5) de ma table au début : TFU où la relation de comprésence est un seul universel numériquement identique qui est une relation variablement polyadique, et qui joue le rôle de la relation 5
Un autre tel point est, je crois, la théorie "nucléaire" défendue par Peter Simons (1994) qui est basée sur la théorie de Husserl. Cette théorie ne rentre pas dans ma table au début car elle a une structure bien différente des théories des faisceaux et des substrats standards, puisque, comme Simons lui-même le dit "it combines aspects of both bundle theory and substratum theory". Je crois que précisément à cause de sa structure inhabituelle la théorie est peu performante, mais je n'offre pas d'arguments en faveur de cette affirmation dans cet article.
130
de comprésence pour tous les objets (appelons cette théorie "TFU2"). Ce trait rend TFU2 plus vulnérable à l'objection de l'Identité des Indiscernables et fait que cette théorie est différente et non-équivalente à TSU (et à la version de TFU examinée plus haut (disons "TFU1"). Il est aisé de voir comment et pourquoi : puisque l'outil d'unification de TFU2 est un et le même pour tous les objets (plutôt qu'un par objet comme dans TSU ou TFU1), il n'est pas étonnant que nous ayons des soucis avec l'Identité des Indiscernables puisque les deux sphères ne partagent pas seulement toutes leurs propriétés (universaux) mais elles partagent également ce qui fait qu'elles sont des objets, elles partagent un seul et même outil d'unification ! De ce point de vue, il n'est pas difficile de voir pourquoi des soucis apparaissent concernant l'affirmation qu'il y a deux sphères plutôt qu'une ! Ainsi, ce qui rend TFU2 différente des autres théories est également ce qui la rend plus faible. John Hawthorne défend TFU2 contre cette objection dans Hawthorne (1995), et sa défense est aussi ingénieuse que simple : puisque, selon TFU2, les objets sont des faisceaux d'universaux, ils peuvent avoir le comportement d'universaux – un faisceau d'universaux (par exemple le faisceau qui est une sphère dans le monde de Max Black) peut être, exactement comme un universel, bi-instantié, et bi-localisé. Le monde de Black peut ainsi être re-décrit en termes de TFU2 comme un monde où il y a une sphère bilocalisée à une certaine distance d'elle-même. (Et ceci est étrange, soutient Hawthorne, seulement dans la mesure où la notion même d'un universel bilocalisé est étrange.) Une conséquence de cela est que les objets matériels, comme une sphère ou Cyrano ou vous-même, se comportent comme des universaux. Une telle théorie dissout ainsi la distinction entre les objets et les propriétés – en fait, il semble que les objets sont simplement éliminés de l'ontologie. Ceci constitue, en résumé, une objection mise en avant par Vallicella (1997). Mais il est difficile d'évaluer la force dialectique de cette objection, car l'amie de TFU2 pourrait très bien tout simplement accepter cette conséquence. Mais Vallicella offre une seconde objection qui me semble plus forte, tout en étant plus simple : pour être localisé de manière multiple, un faisceau d'universaux devrait être instantié, mais ceci n'as pas de sens pour TFU2. Voici une citation presque exacte de Vallicella (1997, p. 94) que j'ai seulement un peu modifiée pour utiliser la même terminologie que j'utilise dans cet article : A universal U is (multiply) located if and only if it is (multiply) instantiated. So if a bundle B of universals is itself a universal then it is (multiply) located if and
131
only if it is (multiply) instantiated. But what could account for B's (multiple) instantiation ? On TFU2, universals are instantiated by being bundled together with other universals. But it makes no sense to suppose that B is bundled together with other universals; for B is a complete bundle of universals. […] But if B is not bundled together with other universals, then it is not instantiated. For on TFU2, a universal is instantiated just in case it enters into a bundle. And if B is not instantiated, then it cannot be multiply instantiated. But if B cannot be multiply instantiated, it cannot be multiply located. So Black's world cannot be given Hawthorne's reading : it cannot be construed as a single sphere at a nonzero distance from itself. For the sphere cannot be doubly located without being doubly instantiated, and it cannot be instantiated at all, for the simple reason that a bundle of universals is not a universal but a particular, and no particular can be instantiated.
Mon objectif ici n'est pas d'essayer de réfuter TFU2, même s'il est vrai que je partage l'opinion de Vallicella. Mon point, comme je l'ai déjà mentionné, est simplement de voir que la raison pour laquelle TFU2 est nonéquivalente aux autres théories est également la raison pour laquelle elle soulève des problèmes sérieux, là où ses concurrentes n'ont aucun souci à se faire. Pour rendre mon affirmation plus forte, je vais encore noter un autre point d'insatisfaction avec TFU2 et la manière dont elle traite la persistance à travers le temps. Supposons d'abord que l'amie de TFU2 est une endurantiste. Rappelons-nous l'objection des intrinsèques temporaires à l'endurantisme, et prenons encore une fois le cas de Cyrano : à t1 il a un grand nez, ensuite il subit une opération de chirurgie esthétique (pour des raisons d'exposition disons un peu plus tard, à t5) et par conséquent il a un petit nez à t6. Pour l'endurantiste, cela signifie qu'une et la même personne existe entièrement à t1 et t6 et a les propriétés incompatibles d'avoir un grand nez et d'avoir un petit nez. Comme nous avons vu, pour éviter d'affirmer une contradiction, l'endurantiste fait ici appel à une stratégie indexicaliste et c'est donc ainsi que sa théorie se présente : Cyrano F-à-t1 G-à-t1 F-à-t2 K-à-t2 …
132
C
Le problème avec cette conception est le suivant. A t1 Cyrano a un grand nez. A t2, il a un grand nez. A t3, il a un grand nez. Et ainsi de suite, jusqu'à l'opération. La chose intuitive à dire ici est clairement que Cyrano garde une certaine propriété pendant un certain temps. Mais l'endurantiste indexicaliste ne peut pas accommoder cela. Selon l'indexicalisme, à n'importe quel moment durant l'intervalle t1-t4, Cyrano doit perdre toutes ses propriétés et doit gagner de nouvelles propriétés : il a d'abord la propriété "avoir-un-grand-net-à-t1", ensuite la propriété "avoir-un-grand-nez-à-t2", ensuite la propriété "avoir-un-grand-nez-à-t3", et ainsi de suite. Selon cette théorie, parce que Cyrano ne peut pas simplement (simpliciter) avoir la propriété d'avoir un grand nez, il doit changer ses propriétés tout le temps et ne peut en garder aucune – il ne peut jamais rester le même. Et puisque la propriété "avoir un grand nez" ne lui est pas disponible, l'endurantiste n'a aucun moyen théorique de dire que toutes ces différentes propriétés temporellement indexées ont 'quelque chose en commun' – ce sont tout simplement des propriétés complètement différentes. Mais plutôt que d'objecter à l'endurantisme, mon point ici est que les propriétés temporellement indexées sont des tropes. Dans le monde de l'indexicaliste il n'y a pas de place pour une propriété qui soit instantiée de manière multiple, puisque toute propriété est toujours liée à un temps et ne peut pas être instantiée à différents temps, et donc il n'y a pas de place ici pour des universaux (des propriétés instantiées de manière multiple). Et bien sûr, non seulement les propriétés doivent être liées à un temps, mais également à une localisation spatiale. La nécessité de cela apparaît le plus clairement dans le cas d'un scénario de voyage dans le temps. Supposez qu'à t6 Cyrano voyage dans le passé pour dire à son moi plus jeune que l'opération va bien se dérouler et qu'il n'a pas à s'en inquiéter. Selon l'endurantisme, Cyrano a alors les propriétés "avoir-un-grand-nez-àt1" et aussi "avoir-un-petit-nez-à-t1" : une apparente contradiction, aisément évitée en disant que les propriétés sont toujours liées à une location spatiotemporelle, car bien sûr "avoir-un-grand-nez-à-l1-t1" et "avoir-un-petit-nezà-l2-t1" ne sont pas contradictoires6. En bref : l'endurantiste doit faire quelque chose pour éviter l'objection lewisienne à propos d'intrinsèques temporaires, et si ce qu'elle fait est d'accepter la stratégie indexicaliste, ses propriétés doivent être liées à une location spatio-temporelle et ne peuvent pas être instantiées de manière 6
Je ne dis pas que ceci résout tous les problèmes que l'endurantiste pourrait avoir avec un scénario de voyage dans le temps.
133
multiple – ce sont des tropes. Ainsi, il apparaît que ceci n'est pas une voie que TFU2 peut emprunter. Mais elle a peut-être d'autres options : elle peut soit choisir d'essayer d'être perdurantiste ou bien essayer de rester endurantiste mais abandonner l'indexicalisme en faveur de l'adverbialisme. Examinons donc à présent ces deux options. Le perdurantisme ne semble pas être une option disponible pour TFU2. En tout cas, il serait particulièrement étrange pour elle d'opter pour cette voie car sa thèse centrale est que les objets se comportent comme des universaux et peuvent être localisés de manière multiple, alors que la thèse centrale du perdurantisme dit que tous les objets sont liées à une localisation spatio-temporelle et que rien (aucun objet) ne peut être localisé de manière multiple. Donc même si le perdurantisme ne l'oblige pas à accepter les tropes (contrairement à l'endurantisme indexicaliste), cela ne semble pas être une bonne option pour l'amie de TFU2. La dernière option est la version endurantiste connue sous le nom de adverbialisme. Mais en est-ce vraiment une ? En effet, nous allons maintenant voir que l'adverbialisme est disponible uniquement pour la théorie des substrats, et non pas pour la théorie des faisceaux, et donc que cette stratégie ne peut pas être utilisée pour aider TFU2 non plus. La solution adverbialiste au problème des intrinsèques temporaires propose non pas de modifier temporellement la propriété mais son instantiation. Ainsi, une adverbialiste va dire que "Cyrano a un grand nez à t1" doit être analysé comme "Cyrano a-à-t1 un grand nez" ou, un peu plus élégamment, "Cyrano a t1-ment un grand nez" (voir Johnston (1987, p. 129)). Ainsi, selon l'adverbialisme, l'instantiation d'une propriété est toujours temporellement modifiée. Ceci va résoudre le problème des intrinsèques temporaires car Cyrano a un grand nez à t1 et un petit nez à t2, et donc il a les deux propriétés incompatibles, mais il a la première t1-ment et la seconde t2-ment et c'est ainsi que la menace d'une contradiction est évitée. Pour pouvoir être une adverbialiste, l'amie de la théorie des substrats peut dire qu'il y a un troisième composant dans sa théorie – une relation d'exemplification entre le substrat et ses propriétés – et que cette relation est temporellement indexée (au lieu d'avoir les propriétés temporellement indexées, comme le voudrait l'endurantiste indexicaliste). L'image endurantiste-adverbialiste-substratiste est alors la suivante :
134
Cyrano F
t1-ment
G t -ment 1 F t2-ment
S
K t2-ment
L'amie de la théorie des faisceaux ne peut, bien évidemment, pas (et ne veut pas) utiliser une telle stratégie, car elle n'introduit pas un substrat qui doit être relié par une relation spéciale à ses propriétés; sa théorie ne requiert pas de tels intermédiaires – et par conséquent, elle ne peut pas être une adverbialiste car il n'y a aucun endroit adapté où l'on pourrait mettre l'index temporel adverbialiste7, et donc TFU2 ne peut tout simplement pas faire appel à cette stratégie. Pour résumer : Premièrement, TFU2 n'est pas compatible avec l'endurantisme indexicaliste parce que cette théorie requiert des tropes plutôt que des universaux; de plus, il ne semble pas, au moins prima facie, que TFU2 pourrait être compatible avec le perdurantisme; et l'endurantisme adverbialiste n'est tout simplement pas une option disponible. Ajoutez à cela les objections de Vallicella, et il apparaît clairement que TFU2 est une théorie nonéquivalente aux autres seulement à cause de traits qui la rendent en fin de compte très peu acceptable. Deuxièmement, il y a un autre endroit où la théorie des faisceaux et la théorie des substrats ne sont pas équivalentes : seulement la théorie des substrats est compatible avec l'endurantisme adverbialiste. Mais : ceci est vrai seulement pour une certaine version de la théorie des substrats, à savoir celle qui insiste sur l'existence d'un troisième composant dans la théorie, une relation (temporellement indexée) entre le substrat et ses propriétés. Et même les amies de substrats elles-mêmes sont d'accord sur l'idée que ceci est une mauvaise version de leur théorie (entre autres, à cause du regressus de Bradley, et de problèmes reliés). On voit bien l'importance de ce type de considérations dans le changement opéré par Gustav Bergmann entre la version de son ontologie dans Realism (Bergmann (1967)) et la 7
Indexer la relation de comprésence serait une théorie perdurantiste.
135
version plus tardive dans New Foundations of Ontology (Bergmann (1992)) : alors qu'il a au départ considéré les liens (non-relationnels) comme des entités centrales de son ontologie, il a ensuite modifié celle-ci de telle manière qu'elle fait plutôt appel aux complexes dont l'unité ne nécessite pas de liens. L'importance de la possibilité de "liens non relationnels" est ici centrale, et transparaît notamment chez Sider (2006) qui, dans sa défense de la théorie des substrats, insiste sur le fait qu'il ne faut pas mettre trop de poids sur la relation d'exemplification, et endosse la vue souvent acceptée que l'exemplification n'est pas une relation, mais est un "lien non-relationnel", et qu'il ne faut pas lui donner le statut d'une chose/entité (Lewis (1983, p. 351-355)). J'imagine que l'idée ici est d'inclure déjà dans le substrat la fonction théorique d'être capable de lier des propriétés (de 'coller à elles') en plus de sa fonction d'être un unificateur (exactement comme la comprésence !), plutôt que de le voir comme uniquement un unificateur qui a besoin d'un autre outil théorique pour rendre compte de la nature des objets. Et c'est seulement si l'on accepte cette seconde version, que l'on aura une théorie différente de la théorie des faisceaux (et ici encore, la raison qui fait qu'elle est différente est également la raison pour laquelle elle est peu acceptable).
RÉFÉRENCES BIBLIOGRAPHIQUES Bergmann, Gustav. 1967. Realism: A Critique of Brentano and Meinong. Wisconsin. Bergmann, Gustav. 1992. New Foundations of Ontology. ed. by William Heald and Edwin B. Allaire. Wisconsin. Ehring, Douglas. 2001. "Temporal Parts and Bundle Theory", Philosophical Studies, 104:163-168. Hawthorne, John. 1995. "The bundle theory of substance and the identity of indiscernibles". Analysis 55.3, pp. 191-196. Hawthorne, J. and Cover, 1998. J. A. "A World of Universals", Philosophical Studies, 91:205-219. Johnston, Mark. 1987. "Is there a problem about persistence ?", The Aristotelian Society 61:107-135. Lewis, David. 1983. “New Work for a Theory of Universals”, Australasian Journal of Philosophy 61: 343–77. Lewis, David. 1986. On the plurality of worlds. Oxford : Blackwell Publishers. Loux, M. J. ed. 2001. Metaphysics contemporary readings. Routledge.
136
Sider, Ted. 2006. "Bare Particulars", Philosophical Perspectives, 20: 387-97. Simons, Peter. 1994. “Particulars in Particular Clothing : Three Trope Theories of Substance,” Philosophy and Phenomenological Research, 54:553-575. Vallicella, William. 1997. "Bundles and indiscernibility : a reply to O'LearyHawthorne". Analysis 57.1, pp. 91-94. Van Cleve, James. 1985. "Three versions of the bundle theory". Reprinted in Loux 2001. Van Inwagen, Peter. 1985. "Plantinga on trans-world identity". Reprinted in Van Inwagen 2001. Van Inwagen, Peter. 2001. Ontology, Identity and Modality. Cambridge University Press.
137
Bare Particulars and Persistence in Bergmann LUCA ANGELONE GIULIANO TORRENGO I. Introduction The notion of bare-particular is one of the distinctive marks of Bergmann’s philosophy. In the sixth chapter of his main work Realism, Bergmann gives us a clarification of this notion, which is carried about on the background of a radical criticism of the traditional notion of substance as continuant. The bottom line of Bergmann’s argument is that the notion of substance should be given up in metaphysics, because it is too messy to be of any theoretical use. Besides, and more importantly, Bergmann extends most of his anti-substantialist arguments to the conclusion that it is the notion of identity over time, in the first place, which has to go. Bergmann’s hostile stance towards identity over time may raise perplexities. Bergmann deals hastily—too hastily—with the connection between the notion of substance and the notion of identity over time. Even granting that the arguments that Bergmann provides in this chapter actually proof that the notion of substance is messy, it does not seem to be selfevident that we should thereby conclude that identity over time has to be given up as well. Secondly, even granting the soundness of this inference, it is still not clear in which sense the notion of identity over time must be abandoned. All over Realism, as almost everywhere else in his mature work, Bergmann stresses that one of the fundamental goal of philosophical research is to provide an ontological ground to phenomenology, namely to account for phenomenological evidences in ontological terms. If there is no entity explaining why the phenomenon is there, the account is bogus. Why, then, identity over time is an exception to this methodological stance? Or, to put it more dramatically, how comes that Bergmann doesn’t seem to care much about the “ontology of the identity-over-time situation” (to mimic his own words), although our experience of this phenomenon is almost ubiquitous?
The situation is even more puzzling if one considers the few words Bergmann offers as an account of what identity over time amounts to. His account has two parts. The first part of the account is unproblematic: ordinary objects are nothing over and above sequences of independent bare particulars, each exemplifying possibly different properties, and none of which lasting over time. Bergmann rejects the idea that ordinary objects are continuants: an ordinary object is not an entity that is wholly present at each time it exists, and that exemplifies possibly different properties at different times. In his ontology: ordinary objects are sequences of momentary bare particulars existing at different times and exemplifying possibly different properties. However (and here the second part of the account comes), those sequences do have phenomenological saliency, and thus they cannot be just collections of unrelated things, there has to be something that grounds them. Bergmann admits that the objects in the sequence, although independent from each other, are linked. But then what, namely what entity according to the fundamental principle of ontology, does link them? Bergmann claims that the right solution to this problem was offered by Berkeley’s anti-realist philosophy: the link on which the identity over time of ordinary objects rests on is ultimately psychological, and thus the ontological ground of identity over time is the human mind. This may sound mystifying if one thinks to Realism’s overall philosophical project. In this paper we purpose to reconsider the Bergmannian solution to the problem of identity over time. We are persuaded that, even remaining within the basic guide lines of Bergmann’s ontology, it is possible to offer an account of identity over time more satisfying than Bergmann’s own explicit account. In order to reach this aim, we will focus on the notion of bare particular, and question the thesis (to be found in Bergmann) that bare particulars are necessarily momentary entities. We maintain that Bergmann does not furnish sufficient reasons to rule out the contrary, and indeed the notion of a non-instantaneous bare particular is not at odds with most of Bergmann’s tenets––insofar as for a bare particular to exist over a longish span of time is not tantamount to existing like a substance enduring over time. We stick to the idea that there are no continuants, be them bare or not. Bergmann’s arguments exclude only that the same bare particular can exemplify different properties at different times, they do not force us to uphold that the same bare particular cannot exemplify the same set of properties at different times. The idea that a bare particular can exemplify a set of properties throughout a time span goes along with another idea we
140
are proposing in this paper: bare particulars have temporal parts and temporal parts of a bare particular are bare particulars too––roughly as an event may have a part, and this part may be, in turn, an event. Our bet is that these two theses can provide the base for a solution to the problem of identity over time, which is better than the solution Bergmann offers in Realism, but still fully Bergmannian in spirit––so to speak. II. Universals and Particulars Why does Bergmann introduce bare particulars? The main reason is strictly linked to the thesis that properties are not particulars but universals. According to the realist view on universals, when two objects exemplify the same properties, the same universal is in both objects. As Bergmann says, “literally the same” thing is present both in the first and in the second object. “Literally the same” here stands for the relation grounding the phenomenological situation that Bergmann refers to as “exactly sameness”. This is the fundamental characteristic of universals: they can be at the same time in different places. But could universals be the only constituents of objects? The answer to this question is no. In order to see why, let us image a world in which the only constituents of objects are universals. Let us suppose that in this world there are two objects located in different places, which are phenomenally exactly the same. Taking Bergmann’s example, there are two spots, located in different places, and each of them is both red and round. Call the first a and the second b. If neither a nor b have further constituents (as we have assumed), one of the basic tenets of Bergmann’s ontology, namely that entities the have exactly the same constituents are identical, lead us to the conclusion that a is identical with b. But this conclusion is absurd because a is not where b is, and therefore a is not identical with b. In order to avoid the reductio ad absurdum, we should reject one of the three assumptions of the argument, namely: 1. Properties are universals 2. If x and y have literally the same constituents, then x is identical to y 3. The only constituents of an object are properties The first thesis is at the very core of Bergmann’s philosophy, and thus it is no surprise he does not want to discharge it. The same holds for thesis 2,
141
which Bergmann, as we will soon see, upholds as even more certain than the more famous principle that it recalls: the principle of identity of indiscernibles. Hence, the only option left is rejecting 3 and admitting constituents of objects different from properties. III. Particularity and Bareness What are the characteristics of these further elements of Bergmann’s ontology? First of all, we should look at what characteristics could solve the problem we have just seen. The problem rises because of the repeatability of universals. Since universals are repeatable, if the only constituents of objects were universals, then it would be possible that two objects are constituted by the same constituents, contrary to assumption 2. If we were to introduce a further constituent that is repeatable as universals are, then, we couldn’t solve the problem. It follows that the further constituent must not be repeatable, and since only particulars are not repeatable, it will be a particular. While universal can exist at the same time in different places, particulars can be in one place only per time. This is what ‘being a particular’ means. However, particularity is not the only characteristic the further constituents should have. The second characteristic that Bergmann introduces is a consequence of the assumption that properties are universals, and thus repeatable. If each property is repeatable and particulars cannot be repeatable, then particulars cannot be constituted by properties only. But neither can they be constituted by properties together with some other elements that are not properties, because those further elements would not be repeatable, and thus particulars, and this would start a vicious regress. Hence, particulars cannot be constituted by any property. This is the thesis of bareness. It should be straightforward, then, why Bergmann names the further constituents ‘bare particulars’. We will add just a few words here on the thesis of bareness. The question we want to address is this: may anything exist without exemplifying any property? According to Bergmann we should answer yes to this question, but we should also hasten to add that bare particulars and universals depend on each other. The dependence Bergmann is talking about goes in both directions. The first side of his thesis is the well known anti-platonic posit to the effect that non-instantiated universals do not exist, and it is quite unproblematic. The second side, however, seems much more questionable: a bare particular would not exist, if it did not exemplify any
142
universal. But how it is possible for a particular to be bare and standing in need of exemplifying at least one universal? Several bare particulars’ critics, for instance Sellars, Loux and Mertz1 have seen here a plain contradiction, or closely so. According to Bergmann their criticism is based on a confusion. Bare particulars are bare because there are no properties in them, namely among their constituents, but this does not prevent properties to be tied to bare particulars by exemplification, since exemplification is an external relation. What Bergmann means here is even clearer once we keep in mind another (deeper, in a way) fact, namely that there is a sense in which bare particulars and universals are independent of each other. Specifying, pedantically, Bergmann’s thesis in Bergmann language, bare particulars and universals are categorically independent of each other, although they are logically dependent on each other. In a more plain talk: it is not necessary for a bare particular to exemplify the properties that it actually exemplifies, neither it is necessary for an universal to be exemplified by the bare particular(s) that actually exemplify it. What is necessary is that a bare particular exemplifies at least one universal, and that a universal is exemplified by at least one bare particular. It follows that bareness has an important general consequence: “Being bare, a particular provides no cue whatsoever as to which universal or universals it may or may not exemplify”. This feature distinguishes bare particulars and substances: the only job of a bare particular consists in providing individuation to universals, and not providing anything about the nature of the complex entity it is a constituent of. IV. Momentary Entities Now, there is a further feature of Bergmann’s notion of bare particular that distinguish them from substances, and which is of central interest for our paper. While the notion of substance is usually exploited to give an account of the identity of objects through time and change, Bergmann never exploits his notion of bare particular for this purpose. Bergmann states straightforwardly why bare particulars couldn’t account for identity 1
See W. Sellars, Particulars, “Philosophy and Phenomenological Research”, XIII/2 1952, 184-199; M. J. Loux, Substance and Attribute, Reidel, Dordrecht, 1978, pp. 146149; D. W. Mertz, Individuation and Instance Ontology, “Australasian Journal of Philosophy”, LXXIX/1, 2001, pp. 45-61.
143
over time. They are momentary entities, and momentary entities obviously cannot account for identity through time. The first question we want to address, thus, is whether Bergmann has good reason to maintain that bare particulars are momentary. The idea that bare particulars are momentary – Bergmann says – follows from the very notion of bare particular. To say that bare particulars are momentary amounts to make explicit something that is implicit in the very notion. Where exactly? At first sight, the notion of particularity together with the notion of bareness do not “analytically beget”––to put it almost poetically––the notion of being momentary. Thus the origination ought to be more indirect. The only notion that is involved in bareness, as we have seen, is the lack of properties among constituents, and this latter notion has nothing to do with temporal length, and so cannot be responsible for such an offspring. Particularity comes from the opposition to the main feature of universality: while universals are repeatable, particulars aren’t. But repeatability sways between two definitions in Bergmann’s Realism: 1. Something is not repeatable if and only if it can be just in one place per time and in one time per place. To be in two places is a repetition, and to be at two times is a repetition. 2. Something is not repeatable if and only if it can’t be in more than one fact. As long as we accept the first definition, bare particulars need be momentary––at least if we read ‘one time’ as meaning ‘one moment’ rather than ‘one stretch of time’. But if we accept the second definition, bare particulars don’t need be momentary: something can be necessarily in one fact only, but it can last long enough to last longer than a moment. For instance, in case it is the “temporal constituent” of the fact that a certain thing or event has been having a certain property for a certain amount of time. In Realism, Bergmann adopted both the first and the second definition, as they were equivalent. But the two definitions are not equivalent because the first implies, while the second does not imply (still being compatible with the thesis) that bare particulars are momentary. In order to see that the second definition does not entail momentariness, consider the case of an entity lasting over time, which does not undergo changes during its existence. Armstrong, in A World of States of Affairs, supplies the example of an entity lasting through time without
144
change2: atoms, in classical physics, do not change their properties. If atoms do not change, we have no reason to suppose that different individuators are in them at different times. It is one and the same fact at different times we are dealing with: the fact that the atom has been having such and such properties. Hence, the bare particular in individuating the atom at a time t may be seen as a temporal part of a “longer” bare particular that individuates it through an interval encompassing t. Why, then, does Bergmann insist on the fact that bare particulars must be momentary? Is it because he thinks that the notion of a non-momentary bare particular is incoherent? The problem of identity through time of ordinary objects is strictly linked to that of their individuation, namely the problem of individuating what distinguishes, at a certain time t, an object x from everything else. Being identical through time may be spelled out indeed as being individuated as the same at different times. As we have seen, Bergmann maintains that bare particulars are the ground of identification: by exemplifying properties, bare particulars identify facts. However, bare particulars cannot individuate through their properties: they haven’t any. What they individuate is rather a spatiotemporal location, the where and when a universal property is actually exemplified. Now, if an ordinary object x persists for a while, we may say––in ordinary talk and without committing ourselves to a particular metaphysics of persistence––that x is the same at different moments of time. But if x, in some sense, is the same at t and at t', then what individuates it at t individuates it also at t'. Bergmann assumes that the principium individuationis––as the tradition has it––of something does not change through time. This assumption is not only very plausible––indeed it seems to follow from the platitude that if x and y are the same, then what individuates x individuates y––, but it is ontologically very innocent. It is indeed prima facie compatible with there being different things grounding the individuation of x at different times, in so far as the principle is the same, and therefore it allows us to individuate x as the same. However, Bergmann’s reading is stronger than that, as a consequence of another general assumption of his: only entities may be ontological grounds (incidentally, this is why he cannot accept functions as ontological grounds). If the principle of individuation of an entity x is the same at every moment in which x exists, then there is one entity that grounds what the principle accounts for. This means that what grounds the 2
D. M. Armstrong, A World of States of Affairs, Cambridge University Press, Cambridge, 1997, pp. 103-104.
145
individuation of x at different moments has to be one entity––otherwise the ground of individuation would not be an entity but simply a collection of entities. And collections of entities (another basic principle of Bergmann’s ontology declares) are not themselves entities. Obviously, if we assume that entities are individuated only by their spatiotemporal location, it follows that there is no one single entity that is the same at different moments, and therefore there is no one single entity which is both the same at different moments and which individuates the same entity by being “in” the object at every moment of its existence––that is indeed the gist of Bergmann’s critique to the notion of substance. If a moment (and a spatial location) individuates an object x, then that very moment (together with that very spatial location) cannot individuate the same thing later on––if we assume that the ontological ground of individuation of something existing at t has to exist at t. Shortly: no thing exists for more than a moment, in Bergmann’s world. However, Bergmann’s critique of substance goes through as well in case we take the temporal location to be a longish time interval, rather than a moment––therefore, even though no thing exists for more than a moment in Bergmann’s world, no thing needs to exist only for such a short span in his world, after all. What is relevant is that the entity grounding the individuation of x is temporally (and spatially) delimited: this means simply that at different spatiotemporal locations necessarily we find different ontological grounds, viz. distinct bare particulars. In other terms, what is relevant is definition 2 of repeatability. Whether those spatiotemporal locations are small or large, whether what temporally individuates x is a moment rather than a period, such a minute or a century, is immaterial to the argument. What is relevant to the argument is that once we consider a temporal location different from the location individuating an object x, what individuates x is necessarily no longer there to individuate it later on, and therefore x cannot be at this new temporal location either. If we assume, with common sense, that objects are not necessarily linked to the time of their existence, whereas temporal section of objects––namely periods of the life of an object––are, then we may say that Bergmann’s bare particulars are the pricipium individuationis of temporal sections of objects, rather than of objects simpliciter. Indeed, he is quite explicit in claiming that what individuates an ordinary object is not a bare particular, but rather a series of facts concerning the distinct bare particulars individuating its temporal sections. But, again, those sections need not be momentary sections. A moment individuates a section of the
146
life of an object as much as a longer span of time does. The right conclusion of Bergmann’s argument, then, is that there are not entities that are individuated by something that is, or better “continues” to be, the same through distinct temporal locations––what Bergmann calls ‘substances as continuants’. Continuants are contradictory because according to Bergmann only an entity whose spatiotemporal boundaries are necessarily delimited, namely an entity that cannot swift through distinct temporal locations can be an “individuator”, therefore there is no ontological ground to continuants’ individuators, and therefore to continuants either. V. Continuants The thesis that bare particulars are momentary has wide metaphysical consequences. Firstly, as we have just seen, it implies the following: (C) there are no continuants By using a modern terminology, we say that continuants are entities that are wholly present at each moment of their existence. Something that is wholly present at each time of its own existence is something that is individuated by an entity that is literally the same in every temporal section of it. Secondly, given that bare particulars are constituents of facts, and that they individuate spatiotemporally the facts they are in, it follows that according to Bergman there are only momentary facts. Notice, however, that one of the main tenets of Bergmann’s Realism is that ordinary objects are complex entities, namely facts, rather than things. Therefore, Bergmann seems compelled to say that ordinary objects are momentary, namely to hold (P) There are no persisting objects Obviously, if you believe that being wholly present is the only way an object can persist, then thesis (C) would imply thesis (P)3. Indeed, since
3
This is van Inwagen’s position, who takes living entities (and hence persons) to be the only persisting things, and has an eliminativistic stance towards artifacts, namely “ordinary” objects. See P. van Inwagen, Material Beings, Ithaca and London, Cornell University Press, 1990 and T. Merricks, Objects and Persons, Oxford, OUP, 2001.
147
persisting through time seems to be an essential feature of ordinary objects, (C) would imply: (P') There are no ordinary objects But one may believe that there are persistent objects––and thus ordinary objects––while she thinks there are no continuants, for instance if she believes that ordinary objects are like events in having temporal parts, or they are sequences of momentary slices, “linked” somehow by a similarity relation4, or by some external relation5. Surely Bergmann upholds (C); does he also upholds (P) or (P')? Who had read even only the first pages and the title of his main work, Realism, would strongly suspect he did not. One of the refrains of Realism is that ordinary objects, like chairs, flowers and persons, are not “things”, namely simple entities, but facts––and not that there aren’t any. But if Bergmann does not provide any other account of persistence, it is hard to see how he can maintain (C) without thereby being committed to (P) as well. Here we need to be careful. Bergmann talks of “temporal crosssections” of ordinary objects in Realism: a temporal cross-section is a complex entity constituted by all the things (i.e. simple entities) constituting an ordinary object at a given time. For methodological reason, Bergmann’s temporal cross-sections are momentary, and so are the bare particulars “in” them. In no pair of distinct temporal cross-sections we find the same things, therefore we do not find the same facts either. However, there may be facts concerning the relations among bare particulars in different cross sections of an object. Indeed, according to Bergmann, identity through time of objects is grounded in “temporal series of bare particulars, held together […] by laws connecting the cross sections of objects” (Realism: 120). These laws express relations between bare particulars that are “in” cross sections of objects (a bit loosely, we will talk also about relations between cross-sections). Our target question is now: what kind of relations are those?
4
As in T. Sider, Four-Dimensionalism. An Ontology of Persistence and Time, Oxford, OUP, 2001. 5 As in K. Hawley How Things Persist, Oxford, OUP 2001.
148
VI. Bareness and Temporal Mereology Bare particulars are not parts––unless in a very loose and misleading sense––of the temporal sections they are “in”, but they are constituents of these temporal sections. The two notions are distinct, and this distinction is crucial for Bergmann. Complex entities, namely facts, have constituents: particulars exemplifying properties, universal properties exemplified by particulars, and nexi connecting them. Being a constituent of is a relation between a fact and the simple things that are “in” it, and it is a fundamental primitive notion of Bergmann’s complex-style ontology. Being part, on the other hand, is an ordinary spatiotemporal relation between bare particulars––as being on the left of and coming five minutes later are. Obviously the two notions extensionally part company: consider for instance two areas one of which is part of the other. The area of a temporal cross section is nothing else than the bare particular exemplifying the properties that we find in the cross section; in a way, they are the cross sections themselves considered without any of their properties. Therefore, the fact that the two areas are one part of the other amounts to two bare particulars standing in an ordinary spatial relation. But one bare particular is obviously not a constituent of the other, since bare particulars are simple things, and simple things do not possess constituents at all, even though they may be in part-whole relation with other simples. For the same reason, Bergmann distinguishes between a bare particular α whose only parts are β and γ and the sum of β and γ. What is exactly a sum here is not important, what is important is that facts concerning α and facts concerning the sum of β and γ are distinct facts.
__________ |_____|_____|
For instance the fact that the sum exemplifies being rectangular and the fact that α exemplifies being rectangular are distinct. This is trivial, given that the constituents of those facts are distinct, and facts are individuated by constituents and the arrangement of those. Notice moreover, that if the two parts are of two different colors, for instance blue and green, then even though the two parts of α, namely β and γ do exemplify a color, blue and
149
green respectively, α does not exemplify any color. Again, this is trivial, since α is simple, and therefore it cannot be partly green or partly blue, it cannot partly exemplify anything, not even those properties that its parts–– that is the entities that stand in the part-whole relation with it––exemplify. Now, being part is a spatiotemporal relation that may hold between bare particulars––as with our (which is also Bergmann’s) example, between two bare particulars individuating distinct spatial locations in the same temporal cross section. What about the bare particulars constituting a series of temporal cross sections of an object? Is part-hood a relation that may also hold between bare particulars individuating temporally distinct cross sections? Given that bare particular are momentary, according to Bergmann, the answer is ‘no’. However, this is thus for a rather shallow reason––with respect to Bergmann’s parameters––namely for some a-priori fact about the world, where the a-priori here is neither a logical impossibility nor a question of analyticity. Indeed, if bare particulars are momentary, then there are no two non-contemporary bare particulars standing in the part-whole relation, namely there are no periods whose parts are shorter periods or moments. Obviously, there are not periods whose constituents are shorter periods or moments, simply because if periods are bare particulars, they are simple, and simple things do not have constituents. However, Bergmann seems to rule out also the possibility for them to have parts (that is to stand in a mereological relation with bare particulars that are not in the same cross-temporal section). More precisely, he rules out the possibility of there being periods at all, and therefore of there being periods whose parts are shorter periods or moments. In other term, continuants are not the only non momentary alleged entity lacking an ontological ground, but also facts whose temporal extension is longer than a moment, by being individuated by a non-momentary bare particular, share this destiny. There is no room in Bergmann’s ontology for any kind of entity that lasts for longer than a moment, which is a simple, and that may have other simples as parts. Let us now try to understand this last claim and see why the existence of non-momentary “bare” simples is compatible with Bergmann’s ontology. Consider first, in analogy with what Bergmann says for spatial parts, two moments β and γ whose sum is the period α. Both β and γ are parts of α, even though they are not constituents of α. The three of them are simples that possibly exemplify properties. In order to make the example more vivid, imagine we are talking about a computer screen that is on from 11.00 AM to 11.03 AM. β individuates the first minute,
150
whereas γ individuates the second one, and α individuates the whole span from 11.00 to 11.03. β and γ are parts of α, even though they are not constituents of it, since α is simple. ||---------------------|---------------------||
For this reason, if the screen stays blue from 11.00 to 11.02, then turns green from 11.02 to 11.03, we may say that β exemplifies blue, γ exemplifies green, whilst α does not exemplify any color at all. The analogy is straightforward: bare particulars are extended not just in space, but also in time. This is not to mean that bare particulars are or may be continuants, but simply that they may individuate spatiotemporal locations that are wider than moments. In a more recent terminology: nonmomentary bare particulars individuate not only spatial parts in a momentary stage, i.e. cross-section, but also temporal parts stretching longer than a momentary stage. By dismissing non-momentary bare particulars, Bergmann dismisses the notion of temporal (proper) part too: the cross-temporal sections of an ordinary object are not themselves the parts of any larger whole. The “laws” holding them together are not grounded on any mereological facts. Bergmann quotes Berkeley as the ancestor of this approach to the problem of identity through time. VII. Bergmann’s Ontology Revisited We have thus showed that the notion of a temporally extended bare particular neither is incoherent as such, nor––if we exclude the explicit claims by Bergmann to the contrary––is inconsistent with Bergmann’s ontology. In showing how such a notion may be spelled out we have simply extended Bergmann’s notion of part from space to time. Bergmann is not compelled to limit his ontology to momentary bare particulars by the very notion of bare particular: neither their bareness, nor their particularity or simplicity implies their being momentary. In short, he needs only definition (2) of repeatability, and not the stronger (1) to get along. Bergmann actually claims that it is “part of the notion” of bare particular that it is momentary, but what he means here for being “part of the notion” is not clear, as we have just seen. At any rate, there is nothing incoherent or unintelligible in the notion of a particular “barely” individuating a span of time rather than a moment. Indeed, we think that Bergmann’s ontology, at
151
least in a sense, suggests the existence of temporally extended bare particulars. Recasting Bergmann’s terminology: their existence is a sort of “dissatisfied” structural suggestion of his ontology. Let us explain this more precisely. Consider the two claims (BMOM) There exist only momentary bare particulars (BEXT) There exist non-momentary bare particulars Bergmann explicitly maintains (BMOM) although (BEXT) would have allow him to ground persistence through time without giving up his critique to the notion of substance. Why? Bergmann often claims even if a theory does not imply a certain thesis it may nonetheless “structurally suggest” it. The notion of structural suggestion is anything but clear; however, the following definition will suffice for our present purposes. (SS) A theory T structurally suggests a thesis φ iff (a) T does not imply either φ nor ¬φ, (b) there are theses in T that together with ¬φ imply “philosophically awkward” consequences (in particular, with respect to T). What about the structural suggestions in favor of each of them? It may be thought that Bergmann choose (BMOM) rather than (BEXT) because his ontology structurally suggests the former, but not the latter. In this last part of the paper we will argue that Bergmann’s ontology structurally suggests both theses, and––indeed––it even more strongly suggests (BEXT) over (BMOM). Taking the risk of being even more pedantic than Bergmann, we will distinguish between local structural suggestions and global structural suggestions. Roughly, a thesis φ––from a cluster F of theses whose members are φ together with all the thesis of T that are about the same (or a closely related) topic than that of φ (therefore φ ∈ F ⊂ T)––is locally structurally suggested by T, iff φ is structurally suggested by some theses in F, namely by features of the ontology that concerns issues close to what φ is about. Whereas, φ is globally structurally suggested if φ is structurally suggested by a thesis in T that is not in F, namely a feature of the ontology that is either a very general one, or one that does not concern issues close
152
to φ. Structural suggestions may be weak or strong, in different degrees. We will assume that a thesis that is both globally and locally structurally suggested is suggested more strongly, to some degree at least, than a thesis that is only globally suggested. Our claim may now be more clearly stated: temporal parts are both locally and globally suggested, while momentary bare particulars are only globally suggested, and thus Bergmann’s ontology can consistently be reinterpreted as a ontology of temporal parts. Firstly, we need to dismiss a “seemingly” reason for thinking that Bergmann’s ontology structurally suggests (BMOM), namely the idea that we need avoid the “illusion” that there might be substances as continuants; an illusion that “cluster” ontologies––Bergmann’s nemesis––foster. This is Bergmann’s reasoning: although even in a cluster ontology spatiotemporal individuators (and properties bearers) are plausibly not entities that are literally the same at every moment of their existence, we cannot rule out the possibility of a world where no spatial change (movement) occurs. In such a world the place where an object is would be a continuant, since it is a quality (a ‘coordinate quality’ in his terminology), which is the same in every cross-section of the object. If bare particulars, on the other hand, are momentary, then they are not continuants even in such a possible world. The assumption that they are momentary entities blocks this possibility at the outset, and thereby it hiders the slippery slope towards the flickering idea of a substance “hidden” underneath the sensible qualities. Taking this to be a reason to have only momentary bare particular is a theoretical mistake. Bergmann does not need such a strong assumption about bare particulars to reach this conclusion. It is obviously true that a momentary entity cannot be a continuant (but in a degenerated sense). But it is as much as true that a spatiotemporal location (no matter how temporally wide) cannot be a continuant: claiming that a spatiotemporal location keeps on being the same through time is tantamount to claiming a nonsense; and that is all what it takes to block the “illusion”. In other terms, a temporally extended rather than momentary bare particular is not a continuant either: it is not the same at different times of its existence, because it does not exist at different times, only its parts––one after another––do. What suggests (BEXT) and therefore the existence of temporal parts is, in the first place, a local factor: the analogy between space and time Bergmann upholds. In particular, the contrast between the axiomatic “space”/”time” composed of un-extended points and the phenomenological space/time, which is discrete and not dense, and whose points are never unextended. Indeed, Bergmann explicitly maintains that moments––that he
153
actually calls time points (since ‘moment’ is the term he reserves for the temporal coordinate quality in cluster ontologies)––are extended simples. Being simple is not incompatible with having parts (only with having constituents): areas have parts while being simple, no matter how extended and scattered they may be. Being extended, once we drop the assumption of atomicity of space-time as Bergmann explicitly does (Realism: 71), not only is compatible with, but actually suggests it: what is extended and not atomic is divisible in parts, that is why Bergmann explicitly drops the assumption of atomicity of space-time once he takes (spatial) mereology into account. (Besides, as we will see soon, a global structural suggestion holds too.) The reasons to uphold that bare particulars are momentary, on the other hand, are in the first place global: the analogy between space and time does not let out any chance for there being local suggestions towards the atomicity of time. The first suggestion towards (BMOM) is methodological. A general tenet of ontological analysis for Bergmann is that almost everything in ontology can be done in a cross-temporal section, namely in a momentary cross-temporal section. And even what cannot be accounted for simply by remaining in a momentary section has a “core” that can be accounted for already there. The second one is phenomenological: only momentary bare particulars exemplify (interesting) phenomenological properties. In order to account for what properties objects of perceptual knowledge exemplifies at different times we seldom need to go further. Remember that a bare particular whose temporal parts exemplify different properties of a kind (a color, for instance) would not in turn (always) exemplify any property of that kind–– not even “partially”. Objects of perceptual knowledge do not persist as ordinary objects do, what we perceive are momentary cross sections of ordinary objects. Bare particulars are there to individuate those cross sections and to exemplify the property that the objects “have” at the corresponding time. VIII. Stages and The Ground of Persistence In a up to date terminology, Bergmann’s general picture is that of a series of stages, namely of momentary cross-sections of an object, standing in a gen-identity relation, i.e. the relation linking two distinct temporal parts of the same object. Now, and again, why does Bergmann consider only momentary temporal sections of objects, and therefore only relations
154
among momentary stages of objects? Does not this let the risk that genidentity have a psychological ground come in. Obviously one may still maintain that even though there are not mereological facts grounding the unity of the persisting objects, there are mind-independent relations that are grounded in facts of similarity, or causal facts, between the stages. However, this vaguely nominalist maneuver would not be very convincing, and moreover, we do not think that this is Bergmann’s position. Indeed, Bergmann probably thinks of psychological connections, when he talks about the laws connecting the temporal cross sections, and this is why he quotes Berkeley just right after. At the same time, he is not concerned about these connections impinging on the general realist stance he professes. According to his view, what one has to do to avoid idealism, even in a “hidden” form, is to avoid “adding” some ontological ingredients to the objects of knowledge once she passes from an ontology devoid of minds to a world with knowing subjects in it. However, since Bergmann maintains that objects of perceptual knowledge are always confined in cross temporal sections, he is not concerned about the ontological ground of persisting objects. Persisting objects are a sort of cognitive illusion, in his view, and this is why they are grounded in psychological laws of associations. What is not constructed by the mind (since it is grounded in facts we find in the world before considering the existence of minds) are the momentary sections of the world we perceive. Nonetheless, although it may be true that persisting objects are not object of perception, it seems to us that persisting objects can be object of knowledge––that is why a constructionist or eliminativist stance towards persisting objects is usually view as a anti-realistic attitude. As we said at the beginning, persistence seems to be a phenomenological evidence on a par of those Bergmann does endeavor to explain seriously throughout his work––by providing them with an ontological ground, an assay. The temporal part solution would provide a more “realistic” ground for persisting objects––while being fully compatible with the rest of his ontology. This is a general reason to prefer them to momentary particulars: it is a global suggestion that adds to the local ones and makes the temporal part’s solution of the problem of persistence a solution more strongly suggested by Bergmann’s ontology than the “stage-view”-like position he actually maintains.
155
Bergmann et l’ontologie de la connexion FREDERIC NEF L’idée de connexion est centrale dans les touts organisés. Dans la syntaxe, qu’elle soit linguistique ou discursive; elle assure l’unité de la phrase dans sa fonction de prédication (Mertz 2003) ; elle assure la cohérence de la valence verbale ; par de multiples types d’anaphores l’unité du discours. Elle est également centrale dans la constitution des objets et prend de multiples formes, clouage, collage ou liage. On distingue donc deux types de connexion, formelle, par exemple entre un verbe et son sujet, une anaphore et son antécédent, et matérielle, par exemple l’assemblage de deux morceaux de matière pour en faire un seul dans un artefact. L’ontologie de la connexion1 s’est développée parallèlement à l’ontologie de la relation. Elle a été présentée par exemple sous la forme de lien et même de ‘lien non relationnel’ (quoique d’après Bergmann cette expression, comme nous le rendrons évident, n’aurait pas grand sens). C’est à propos du lien à établir entre les universaux et les particuliers que cette notion de lien distinct de la relation a été proposée par Strawson, puis reprise par Armstrong. En effet si c’est une relation qui unit le particulier et les universaux qu’il instancie, il faut une relation pour unir cette relation au particulier, et ainsi de suite à l’infini (argument du troisième homme). Plus important pour notre propos actuel, l’idée de connexion est présente dans l’analyse de la structure ontologique des complexes, faits ou particuliers épais. On laissera ici de côté l’intervention de ce concept dans la théorie des faisceaux, des particuliers épais conçus comme des faisceaux, car j’entends discuter des métarelations compliquées entre connexions et relations à partir d’une lecture de l’ontologie bergmanienne qui présente l’intérêt de s’y être consacrée avec constance, sinon toujours avec clarté. L’analyse des faits, qu’elle soit celle de l’atomisme logique primitif ou revisité par Armstrong bute effectivement sur la nature de la connexion qui assure à la fois leur unité relative et leur indépendance. Je propose par une exégèse assez classique de Bergmann d’exposer ce 1
On trouve chez Wolff, dans sa cosmologie, une distinction de deux types de connexion, le nexus elementarum et le nexus rerum.
problème. On verra que pour retirer du fruit de cette lecture, il faudra lever une difficulté centrale : la caractérisation de ce philosophe du côté du constituant ou de la relation. 1 Quelle place occupe l’ontologie bergmanienne dans l’ontologie contemporaine ? Cette question n’est pas simplement une question historique, qui touche l’interprétation d’un philosophe encore mésestimé ou au moins sous-estimé. C’est également une question de méta-ontologie qui touche à l’essentiel de la question sur la forme que peut revêtir une ontologie. En effet, l’ontologie peut être soit une ontologie de constituant, soit une ontologie de relation2, et pour déterminer la place de l’ontologie bergmanienne, il faut savoir si elle est du premier ou du second type. Une hypothèse qui vient immédiatement à l’esprit est que Bergmann serait passé d’une ontologie de constituants, exposée dans Realism, à une ontologie de relation dans les New Foundations of Ontology. Il est effectivement tentant d’interpréter l’évolution d’un auteur à partir de cette grille d’analyse. Nous montrerons cependant que si cette hypothèse nous met effectivement sur une piste intéressante, il faut la modifier substantiellement, principalement à cause de la différence qui existe entre les concepts ontologiques de relation et de connexion. L’ontologie de Gustave Bergmann a traversé trois stades3 et je ne m’intéresserai qu’aux deux derniers4, la middle ontology de Realism (1967) 2
Dans la classification de Loux (2006), héritée de Wolterstorff (1970), qui est un compte rendu de Bergmann (1967). 3 Cette classification n’est pas incompatible avec l’éclairage donné par Edwin Allaire dans l’ouvrage collectif The Ontological Turn : « the world of the early Bergmann is a desert, the world of the later Bergmann is a jungle. » (op. cit. p. 38) E. Allaire distingue des ontologies de cigales (lavish ontologies) et des ontologies de fourmis (frugal ontologies) et il interprète l’évolution de Bergmann comme l’abandon progressif de la frugalité. Il voit dans l’adoption de la méthode du langage idéal la cause de la prodigalité de son ontologie tardive et contraste son destin avec celui de Wittgenstein : ce dernier abandonne la méthode et se convertit à l’austérité, tandis que Bergmann méthodiquement renonce à la frugalité. 4 La première ontologie correspond à la période des années 50-60 où Bergmann critique le positivisme logique et se dégage de l’ontologie de Frege (cf. « Frege’s hidden nominalism » 1958). Recueil d’articles de cette époque : The metaphysics of logical positivism (1954) spécialement : « Logical positivism, language and the
158
et la late ontology des New Foundations (publié en 1992, après la mort de Bergmann survenue en 1987). Les « Notes on Ontology », une réponse importante à des critiques de Hochberg appartiennent à cette période (ces notes furent publiées en 1981. L’essai capital « Diversity » (1968) appartient à la middle ontology. La middle ontology est une ontologie qui distingue des choses (things), des faits5 (facts) et des subsistants. Les choses et les subsistants sont simples et les faits sont complexes. Les faits sont composés de qualités liées par des nexus. Les qualités sont des choses simples et les nexus sont des subsistants6. Dans la late ontology les deux piliers de l’ontologie sont les circonstances et les classes. Les circonstances sont complexes et les classes sont intermédiaires entre le simple et le complexe. Les circonstances sont des complexes qui n’ont pas besoin de nexus. Les circonstances sont de trois types : diversité ((X,Y)), élémentarité ( X est un élément de Y), nexus intensionnel (αM.β) Une circonstance existe par le fait même de la connexion et cette dernière agit eo ipso7, c’est-à-dire par son acte même, sans rien réclamer en plus. Pour le nexus intensionnel la conséquence est que la pensée n’est pas une pensée de quelque chose en vertu d’un lien qui l’attache à ce quelque chose, mais que eo ipso l’unité est faite du nexus de la pensée et du fait. Les circonstances sont considérées comme des contreparties structurales des relations internes fondées dans la nature de leurs relata (1992, p. 139). Les circonstances ne sont donc pas des sommes d’éléments et ontologiquement on ne peut identifier classes et circonstances (bien que comme Carnap et d’ailleurs le premier Bergmann on puisse réduire les circonstances à des classes, si on s’intéresse uniquement à l’énumération lexicographique et non à la structure). Les deux types d’ontologie distinguées sont les ontologies de constituants, que Loux fait dériver d’Aristote et les ontologies relationnelles qu’il fait remonter à Platon, via le Russell des Problems of Philosophy (Loux, op. cit. p. 209-210). Les premières s’efforcent de livrer
reconstruction of metaphysics » (1953). 5 Bergmann identifie les faits et les choses. Wolterstorff (1970) rejette cette identification (cf . p. 116-117). 6 La distinction entre exister et subsister provient sans doute de Meinong. 7 Eo ipso signifie « par l’acte même ». C’est un terme de droit et de philosophie. Par exemple le fait que je suis n’implique pas eo ipso le fait que je pense, mais l’inverse est vrai.
159
la structure profonde du monde et des particuliers en termes de constituants : matière et forme essentiellement8. La thèse de l’hylémorphisme est typiquement une thèse de ce genre d’ontologie. Les secondes posent des relations fondamentales entre les universaux et les particuliers, relations essentiellement de participation et d’imitation9. Les particuliers sont des sommes de relations. Bergmann distingue les deux types d’ontologie de cette manière également. Dans la middle ontology l’ontologie de constituants correspond à l’ontologie ε (ontologie avec un nexus) et l’ontologie relationnelle à l’ontologie γ (ontologie avec un pseudo nexus, c’est-à-dire une fonction prise à tort pour un nexus : cf 1967, p. 31). Dans la late ontology la première correspond à l’ontologie du complexe et la seconde à l’ontologie de la fonction. L’ontologie du complexe est une ontologie des particuliers nus, tandis que l’ontologie de la fonction est une ontologie des particuliers parfaits. Un particulier nu est un particulier qui n’a comme propriété que d’être particulier, tandis qu’un particulier parfait (particulier épais d’Armstrong) est doté de ses propriétés qui le particularisent complètement. Un particulier nu est un pur individuateur (individuator, 1967, p. 25) et les particuliers nus s’apparient invariablement avec les universaux10. Bergmann remarque que les nominalistes ont toujours refusé de combiner les particuliers nus et parfaits. Dans l’ontologie ε « Ceci est vert » est analysé comme un particulier nu lié par un nexus fondamental et non homogène, nexus binaire qui d’un côté lie le ceci et de l’autre un universel qui est un caractère (op. cit. p. 47). De même dans « Ceci est à gauche de cela », il y a deux particuliers nus liés à un universel, qui est une relation (la relation « être à gauche de »), par un nexus inhomogène (ibid). Dans une ontologie γ, il y aurait là une relation entre deux particuliers parfaits, une relation externe, les relations 8
On peut se demander si la potentialité et l’actualité sont aussi des constituants. Dans la mesure où les propriétés sont presque toutes dispositionnelles, une théorie des constituants qui prend la forme d’une théorie des choses comme faisceaux de propriétés, prend la forme d’une théorie de constituants comme mixtes de potentialité et d’actualité. La difficulté est que les dispositions sont des propriétés relationnelles et donc que l’on peut être tenté de penser que le dispositionnalisme va dans le sens des ontologies relationnelles. 9 Au niveau sémantique la relation de paronymie survient sur la participation et l’imitation. Ces deux relations sont fondées sur la relation d’émanation, que l’on pourrait être tenté de nommer une « archi-relation ». 10 C’est effectivement le cas chez Armstrong.
160
entre les propriétés et les individus étant internes. Dans l’ontologie epsilon « ceci est à gauche de cela » le nexus agit sur deux particuliers nus, les propriétés étant extérieures aux individus qui les exemplifient : « un particulier et un universel qu’il exemplifie sont complètement externes l’un à l’autre.» (ibid). Nous avons tendance à assimiler profondeur et intériorité et ici Bergmann nous conduit à les distinguer11. Bergmann voit un rapport nécessaire et profond entre cette externalité des propriétés à leurs supports individuels et l’attribution au nexus d’un statut ontologique profond : « Un métaphysicien (ontologist) dont les « individus » et les « propriétés » sont complètement externes l’un à l’autre est virtuellement forcé à reconnaître le statut ontologique du nexus fondamental entre eux, ce qui à son tour lui rendra la tâche plus aisée pour reconnaître que chaque connexion doit avoir un fondement ontologique. Nous sommes arrivés à une intuition structurale fondamentale. L’affirmation qu’il y a des particuliers nus et le fait de reconnaître le statut ontologique de toutes les connexions s’impliquent l’un l’autre. En d’autres termes, le réisme et les particuliers nus entrent en conflit l’un avec l’autre. » (p. 47)
La division en ontologies de constituants et relationnelles n’épuise pas la totalité des ontologies ; elle a été choisie ici car elle éclaire la nature de l’ontologie bergmanienne. En effet, cette division s’applique aux ontologies réalistes et de plus elle n’est pas totalement exclusive : il peut y avoir des ontologies mixtes (on devra se demander si l’ontologie bergmanienne n’est pas mixte, finalement). L’arbre de classification des ontologies, suivant ce critère, est donc le suivant :
11
Ce n’est pas pour autant que l’on pourrait conclure que la connexion n’est pas plus qu’une relation, mais moins. La connexion est à la fois plus, dans la mesure où elle unifie, mais elle est moins dans la mesure où elle n’a pas à exprimer la nature de ce qu’elle lie, comme la relation interne (mais peut-être ceci est un plus également : la connexion de dépend pas de l’essence de ce qu’elle connecte et en ce sens elle jouit d’une sorte d’indépendance ontologique). Donc la connexion ne peut être comparée en termes de profondeur de la fondation ontologique avec la relation. De plus la connexion est aussi spatiale (et temporelle) et comme on le sait pour la plupart des auteurs les relations spatiales sont externes (pour les relations temporelles, c’est beaucoup plus compliqué, cf. Whitehead et un commentaire de Nef 2005). Je remercie François Clementz (Aix-en-Provence) qui a attiré mon attention sur ce point.
161
Ontologies
Nominalistes radicales12
Réalistes
Mixtes
Non mixtes Relationnelles
de constituants
En ce qui concerne l’ontologie de Bergmann, les deux points cruciaux sont le statut des particuliers (comment les individuer) et la nature des connexions (comment lier les particuliers entre eux et les particuliers avec les universaux). Les particuliers sont comme on vient de le voir au moins de deux sortes : parfaits (ontologie de fonction) et nus (ontologie de complexe). Une ontologie de fonction est une ontologie de ‘makers’, le terme fonction remplaçant ce dernier terme pour des raisons de standardisation du langage idéal : « Le facteur (maker) ou peut être le faire (making) est une « relation » entre le matériau et le produit. Appelez ceci la thèse du relationnisme. Ceux qui l’affirment relationnistes. Le relationnisme, je vais le montrer, conduit au désastre. (…) Ce que j’appelai un facteur, je l’appelle maintenant une fonction13. Ce que j’appelai matériau ou matériaux je l’appelle maintenant un argument (…). » (1992, p. 125-126)
On peut remarquer que Bergmann rejette un certain type de relationnisme
12
Cf. Loux 2006, p. 211. Ces ontologies considèrent les faits comme atomiques et ne les analysent donc pas en constituants, ni ne discernent les relations structurantes, comme l’exemplification. 13 Et en fait le terme maker avait été substitué à lien (tie). Bergmann a cherché à gommer l’investissement métaphorique (anthropocentrique, comme il le dit). Il hésite entre un langage simple et imagé (exemple : foil) et un langage technique (le langage idéal IL, cf. Carnap, Montague).
162
dans la late ontology tout comme il a rejeté un certain type de réisme14 dès la middle ontology. L’erreur du relationnisme (1992, p. 130) est de postuler une relation cachée entre le fait et le nexus. Or, Bergmann insiste la connexion entre les particuliers ne réclame pas de relation, elle se fait d’elle-même eo ipso, sans rien réclamer de supplémentaire : « … je peux seulement répéter que le fait n’est ni la collection, ni la classe, ni la ‘ somme’ ni quoique ce soit d’autre du premier et du second [le particulier nu et la qualité], mais plutôt le troisième [le fait =la qualité + le particulier nu qui est eo ipso si et seulement si le premier et le deuxième sont là. Ou, avec une torsion, la nature du fait est complètement épuisée par le premier satisfaisant la restriction qui va avec le second et par le fait que le second est inséparable15 du premier. » (op. cit. p. 128-129)
La position de Bergmann est claire : le réiste réduit le fait à une somme, une classe … de choses et le relationniste pense rendre compte de l’inséparabilité des éléments du fait par une relation cachée, alors que cette inséparabilité est le fait d’une connexion qui par soi même la réalise et rend inutile une relation supplémentaire (qui conduirait d’ailleurs à une régression à l’infini). C’est la nature de cette connexion qu’il faudra préciser, dans les deux phases de son ontologie. Rappelons que pour Bergmann les concepts de connexion, de connecteur et de fait sont liés : Une connexion entre deux ou plus de deux entités est un fait.. Le fondement ontologique en addition aux entités connectées (…) est une entité connectante (connector entity) ou, plus simplement un connecteur. Tous les connecteurs que nous avons rencontrés jusqu’ici sont quelques nexus (et pseudo nexus). (1967 p. 42)
Les faits sont indépendants, alors que les choses ne le sont pas, ce qui pour Bergmann est identique au principe d’exemplification, qui est donc identique à la dépendance ontologique essentielle des choses. Dans l’exemplification il y a nécessairement arrêt du processus de régression de
14
Dans New Foundations, il reprend la critique du réisme. C’est cette inséparabilité ontologique qu’il faut penser. On pourrait faire l’hypothèse que l’inséparabilité est un autre nom de la dépendance mutuelle. On peut même se demander si le choix du particulier nu contre le particulier parfait, russellien ou goodmanien, ne vient pas de cela : le particulier nu dépend de la qualité, tout comme la qualité dépend du particulier nu, alors que la dépendance est unilatérale de la qualité au particulier parfait.
15
163
Bradley, car un nexus, ou un subsistant en général ne réclame pas un lien supplémentaire pour être attaché à ce qu’il exemplifie. L’ontologie de Bergmann est systématisée grâce aux ‘catégories’ de déterminé et de sous déterminé :
Subdéterminés
Connecteurs
Déterminés
quantificateurs
particuliers
universels
Deux-en-un
pas deux-en-un
Choses
Classes
(1992, p. 17). Le deux-en-un (two-in-one) est ce qui a un lien (tie). Il faut distinguer relation et lien. La sémantique (ou l’ontologie, tout dépend de l’interprétation) frégéenne16 qui est relationniste, selon Bergmann, au sens qu’il donne à ce terme, établit une relation entre objet et fonction, les deux catégories de base, relation cachée de complétion (ou de saturation). Une relation met en présence deux choses sans forcément les lier : par exemple une relation de comparaison n’implique pas qu’il y ait un lien entre les termes de la comparaison. Habituellement, on déclare que dans ce cas il y a une relation externe, une relation mentale projetée sur les choses, mais même pour les relations internes la question du lien est notoirement compliquée. Pour Bergmann il y a toute une série d’« aversions » qui caractérisent l’ontologie réiste17 et nominaliste (ou plutôt la non ontologie qu’il entend 16
Bergmann insiste sur le relationnisme frégéen, op. cit. p. 131-132 pour le critiquer. 17 En ce sens Bergmann n’est pas un brentanien, si le réisme est une sorte de telos de l’œuvre de Brentano et non un accident ou un élément adventice.
164
détruire et qui est caractéristique du positivisme logique). Ces aversions sont : les subsistants, les entités nues, les relations et les connexions externes (op. cit. p. 48). Les nexus sont externes : Les nexus sont complètement externes aux entités qu’ils connectent. Il en va de même évidemment des relations. De ce point de vue il n’y a pas de différence entre les deux. Cela laisse seulement une issue pour ceux qui à cause de leurs aversions fondamentales rejettent les connexions externes. Ils doivent admettre le principe des connexions internes (Habituellement on l’appelle le principe des relations internes. Pour les réistes cette étiquette est correcte ; nous l’évitons bien sûr, au sens habituel relations et connexions sont externes). (1967 p.48)
Il serait donc erroné de supposer que Bergmann, s’opposant à une ontologie de clusters ou de classes, de simples collections, ait mis le balancier du côté d’un relationnisme extrême dans lequel les relations internes occuperaient une place centrale. En particulier, on ne peut dériver du caractère nécessaire des relations internes le caractère de lien : la nécessité de la relation interne et celle du lien ne sont pas forcément de la même espèce, on le verra. On y reviendra. On peut remarquer que les classes ne sont pas des deux-en-un et donc qu’il n’y a pas de lien entre les éléments d’une classe : la relation d’appartenance des co-éléments à une même classe n’est pas un lien entre ces co-éléments. Il existe une analyse plus fine des déterminés : on peut admettre des complexes entre les choses et les classes. Les complexes sont, on l’a déjà dit, des faits ou des circonstances. On aboutit alors à la classification suivante des entités déterminées (en donnant à ‘entité’ un sens absolument général, entitatif et non forcément objectif ou objectal) : DETERMINES
Simples
Choses
Complexes
Faits
entre Simples et Complexes
Circonstances
On peut poser un principe de variation ontologique
Classes
qui établit une
165
corrélation entre degré d’indépendance et degré de connexion. E. Allaire (op. cit. p. 16) établit ainsi cette corrélation +
déterminés DEGRE D'INDEPENDANCE subdéterminés
+ deux-en-un (choses) DEGRE DE CONNEXION
-
faits
Si cela est correct, il y a une corrélation entre indépendance et connexion : une entité déterminée est indépendante et elle est connectée, tandis qu’une entité moins déterminée est moins connectée. Après ces remarques je rappelle ce qu’il en est de l’ontologie du nexus dans la middle ontology (2) avant de m’attaquer à l’ontologie de la connexion dans la late ontology (3).
2 Connexion et relation Tout le début de Realism est consacré à la nature du lien ontologique et de la connexion, qui est pensée à partir d’une réflexion sur la nature du nexus : Une chose ordinaire est un complexe ; ses qualités sont parmi ses constituants ; mais ils ne n’en épuisent pas la nature. (…) le troisième constituant je l’appelle un nexus. (Realism, p. 9)
Par exemple une tâche lumineuse rouge et ronde existante est un complexe qui lie rouge et rond par un nexus. La tâche ovale et bleue qui n’existe pas et donc les qualités ovale et bleue ne sont pas connectées18. Bergmann symbolise le complexe par α et le nexus binaire par ν (ternaire par µ) Les qualités ont besoin de nexus pour les connecter dans les choses
18
On voit là une différence essentielle avec l’ontologie meinongienne des complexes. Pour Meinong un complexe non existant est un complexe et donc dans le cas où par exemple je m’imagine un ovale bleu, il y a une connexion entre le bleu et l’ovale.
166
ordinaires. Un nexus ne réclame pas une entité supplémentaire pour lier (tie) ce qu’elle lie, sinon on entrerait dans une régression infinie. La différence entre qualités et nexus est ontologique : les qualités sont des choses parmi des choses ; les nexus des subsistants parmi les subsistants. Le triple des deux qualités et du nexus ν n’est pas un fait, ni une entité, il faut donc un nexus ternaire, µ, pour lier les deux qualités et le nexus ν. Bergmann résume son argument en quatre étapes : 1. Celui qui ne reconnaît pas le statut ontologique du nexus ne peut avoir une notion claire du complexe, 2. Avoir une telle notion et distinguer clairement choses et faits est virtuellement une seule et même chose, 3. Les choses ordinaires ne sont pas des choses, mais des faits, des complexes. 4. Il n’y a pas de choses qui sont des complexes exactement au même sens où le sont les faits. (Realism p. 12)
On peut faire ici deux remarques historiques. Tout d’abord le réisme du dernier Brentano et dans une certaine mesure de Meinong (que rejette Bergmann) consiste à affirmer qu’il n’y a pas de nexus, seulement des choses. Pour Bergmann le nominalisme est une prémisse structurale du réisme qui est la cible principale. Ensuite, Goodman d’après Bergmann analyse le nexus comme un chevauchement (overlapping). D’un point de vue temporel un chevauchement est donc un nexus. L’ordre temporel est fortement connecté (contre Armstrong et le Principe d’Indépendance qui stipule qu’il n’ y a pas de relation nécessaire entre les instants) et cette connexion n’est pas assurée par des relations internes entre instants19 mais des liens ou nexus fondés dans la nature même des instants qui est relationnelle (cf. la théorie du continu de Peirce). Si deux taches de couleur ont exactement les mêmes qualités, connectées par un nexus ν., il y a deux éventualités : du point de vue du réalisme des universaux, il faut quelque chose de plus pour les différencier, alors que d’un point de vue nominaliste ce n’est pas forcément nécessaire, bien que beaucoup d’entre eux, d’après Bergmann le fassent. Quel est le constituant supplémentaire ? Le principe d’individuation peut être l’espace
19
cf. Whitehead pour l’asymétrie : relation interne présent/passé et externe futur/présent, Nef 2005, cf. la notion de ‘connexion indépendante’ chez Whitehead.
167
ou le temps, mais selon Bergmann ils n’engagent à rien d’ontologique. Il faut donc un constituant supplémentaire qui ne soit pas l’espace ou le temps. Distinguons deux sens de possible (et donc d’impossible) : possible a priori (possible1) et possible a posteriori (possible2). Il est possible1 pour plusieurs choses d’avoir la même propriété, mais pas possible2. Dans notre monde avoir et ne pas avoir une même propriété, bien que possible1 est impossible2. C’est ce problème que j’ai essayé de résoudre avec la notion de nécessité post hoc : Une fleur peut être fanée ou non fanée, mais une fois qu’elle est fanée, cette propriété appartient nécessairement à cette fleur ; on ne peut plus dire qu’il est possible qu’elle soit fanée, au sens d’une possibilité réelle dans notre monde. Cf. Wolterstorff 1970 p. 127. La middle ontology de Bergmann comprend des universaux, des particuliers parfaits (perfect particulars) et des particuliers nus (bare particulars). Les nexus assurent l’unité et l’indépendance des faits et des complexes. Les choses dépendent des faits et donc le réisme en partant des choses, ce qui conduit soit à scotomiser le nexus, soit à en donner une version ad hoc sous la forme d’un pseudo-nexus, est dans une erreur complète. Il faut un renversement qui parte, comme chez Russell, des faits, mais pas des mêmes faits au sens ontologique. Pour Bermann dans le foil tous les complexes sont des « faits » et tous les « faits » sont des complexes ; tous les simples sont des « choses » et toutes les « choses » sont des simples. Mais les choses ne sont pas les seuls constituants des « faits ». Chaque « fait » a au moins un constituant de la sorte ontologique que j’appelle « nexus », qui est à son tour une sous sorte de ce que j’appelle « subsistants ». (1967, p. 4) 3 Dans la late ontology des New Foundations, Bergmann entreprend à nouveaux frais de décrire la structure des faits atomiques. Bergmann refuse à la fois les faisceaux russelliens et l’ontologie frégéenne des fonctions et objets. Le premier de ces refus l’a conduit dans la midle ontology à admettre des particuliers nus et le second à y poser un nexus de signification (meaning nexus). Dans les « Notes on Ontology » (1981) Bergmann livre une esquisse de représentation formelle du nexus intensionnel. Soit g1 (a) un fait atomique (a une constante, g1 une qualité), par exemple cette tache de couleur est verte. Soit g2 une pensée (par exemple la perception de cette
168
tache colorée). Alors g2 M g1 (a) est un paradigme de connexion – Bergmann insiste : de connexion – « pas de relation ! ». (op. cit. p. 134). M est le symbole du nexus de signification. Pour Bergmann g2 M g1 (a) est analytique. M ne représente rien et surtout pas une relation entre g1 et g2 On peut donc à première vue affirmer qu’ici on n’introduit pas un élément relationnel. Cependant il y a un élément relationnel du côté de la dépendance. Les faits sont indépendants au sens fort, les choses au sens faible et les subsistants sont dépendants. Il y a donc bien deux relations de dépendance : les subsistants dépendent des choses et les choses dépendant des faits. Toutefois, c’est dans la différence des ontologies γ et ε, dans un retour à une distinction de la middle ontology, qu’il faut chercher ce qui touche à la nature du relationnel. Dans un monde epsilon (i .e. dans un monde représenté dans une ontologie epsilon) les choses ordinaires sont des faits, tandis que dans un monde gamma (même remarque) les choses ordinaires sont des clusters. Donc γ est toujours un pseudo nexus, ce que Bergmann appelle « un monstre » (Bergmann déclare aussi que le réisme est monstrueux). Cette ontologie est une ontologie de composants qui sont des pseudo-constituants. Dans l’ontologie epsilon le nexus est non homogène, en effet la différence entre les particuliers est numérique et la différence entre les universaux est qualitative. Dans l’ontologie gamma au contraire le nexus est homogène, la différence entre particuliers et entre universaux est qualitative. Les différences entre les deux ontologies sont les suivantes. Dans un monde epsilon il y a de pures entités, c’est-à-dire des particuliers nus dans un monde gamma il n’y en a pas. Dans un monde epsilon les connexions ont toutes un statut ontologique, dans une ontologie gamma non. Pour Bergmann la résistance réiste aux particuliers nus est à la base de tout atomisme qui est présent dans la dialectique ontologique classique. Selon lui, l’aversion pour les particuliers nus va de pair avec l’aversion pour les connexions externes. L’aversion pour les particuliers nus et les relations externes est caractéristique pour Bergmann du nominalisme qui doit prendre de manière cohérente la forme d’un réisme. Dans sa late ontology Bergmann s’efforce de penser le lien à l’aide du maker. Est-ce une relation ? Une connexion ? Le maker est un non déterminé, alors que l’exemplification est un déterminé et comme on l’a vu les connecteurs (connectives) et les quantificateurs du langage logique idéal des sous déterminés. Les non déterminés sont au nombre de trois : les
169
makers, les sortes ultimes et les modes ; ils sont non séparables de ce avec quoi ils se combinent ; leur unicité ne réclame aucun support ontologique. Tous les makers (à l’exception des exemplifications) font à partir d’eux des faits supplémentaires. Les exemplifications combinent les choses en faits atomiques. Un texte difficile et dense des New Foundations expose la machinerie ontologique : Les faits et les circonstances sont de beaucoup de sous sortes. Mais ils sont tous des termes de circonstances autant que des éléments de classes ; ils ont tous des modes et les connecteurs autant que les quantificateurs, c’est-à-dire tous les makers à l’exception des exemplifications, font d’autres faits à partir d’eux. La connexion entre les choses et les exemplifications est particulièrement étroite. Ces dernières sont parmi tous les makers les seuls qui combinent les précédents en ce que l’on appelle des faits atomiques. Les classes, finalement, ne sont pas des matériaux pour quelque maker que ce soit. Cependant, comme les choses, ils sont des éléments de classes et des termes de circonstances ; tout comme les choses ils n’ont pas de modes. Cela sont les similarités principales et différences entre les quatre sortes de déterminés. Deux ont des modes ; deux ne le sont pas ; (…) Les déterminés (…) sont séparables. Les trois sortes d’entités n’ont pas un si grand nombre d’associations. Chaque item se combine avec un et seulement un d’une sorte ultime. Les modes se combinent de manière large avec les faits et les circonstances. Cependant aucune entité ne réclame un maker pour se combiner avec ce avec quoi il se combine, tout comme aucune des deux sortes de deux-en-un dans lesquels ils se combinent (…) a elle-même un mode. Donc je référerai à ces caractéristiques en appelant toutes les entités totalement inséparables. Les makers sont entre les deux. Un maker bien sûr ne réclame pas de lien pour lier ce qu’il lie. C’est leur essence même. (op.cit. p. 113-114)
La structure du nexus est donc assez complexe. Dans une formule comme g1M g2 il y a un exemplification de g dans g1 et g2, unité eo ipso du deux-enun, et signification du lien par M. Dans les « Notes on ontology » (op. cit. p. 149) Bergmann entreprend de donner une représentation non linéaire de η (g1, a) c’est à dire du fait composé du nexus η, de la qualité g1 et de la constante a ? Bergmann donne des instructions qui aboutissent à la représentation non linéaire suivante :
170
η g1 a où dans le plus petit cercle nous avons < g1, a > et dans le plus grand le nexus η. Relativement à la question posée au départ de ce texte : faut-il ranger l’ontologie bergmanienne parmi les ontologies de constituants ou de relations, la réponse est nuancée. M ne signifie pas une relation ni ne la représente. En ce sens il n’y a pas de relationnisme, pour reprendre un terme de Bergmann. Mais en même temps le schéma d’exemplificationunification-connexion est relationnel dans sa nature même. On pourrait peut-être s’en tirer par une pirouette et conclure que Bergmann est un philosophe du constituant et que son originalité est d’intégrer la connexion comme constituant ontologique, à la fois comme relation externe et fondée sur la nature de ce qu’elle relie. En ce sens on peut se demander s’il s’agit du comble de l’audace métaphysique pour échapper au réisme, ou s’il s’agit d’une tentative qui expose à l’incohérence, les tentatives cohérentes étant par exemple un nominalisme atomiste ou un réisme méréologique. Si dans les années 70 l’ontologie marquée à la fois par Carnap et la phénoménologie se trouvait devant un choix entre réisme et ontologie du nexus, on peut, dans un coup d’œil rétrospectif, interpréter cette situation comme une véritable antinomie. Bergmann a repris l’héritage de l’ontologie autrichienne du complexe, mais en la débarrassant du nominalisme brentanien et du relationnisme meinongien. Cette tentative a au strict minimum l’intérêt de nous permettre une lecture non historique profonde de ce moment de l’ontologie intensionnelle, ontologie qu’un Roderick Chisholm ou un Hector-Neri Castañeda reprendront à nouveaux frais.
171
REFERENCES Bergmann G. 1953, « Logical positivism, language and the reconstruction of metaphysics ». Egalement dans Bergmann 1954. Bergmann G. 1954, The metaphysics of logical positivism. Bergmann G. 1958, « Frege’s hidden nominalism », E. Klemke ed. Essays on Frege, University of Illinois Press. Bergmann G. 1967, Realism. A Critique of Brentano and Meinong, The Wisconsin University Press. Bergmann G. 1968, « Diversity » Proceedings and Addresses of the American Philosophical Association, 51, p. 21-37. Bergmann G. 1981, « Notes on Ontology », Nous, 15, 131-164. Bergmann G. 1992, New Foundations of Ontology, W. Heald & E. Allaire éds., The Wisconsin University Press. Gram M. & Klemke E.D. eds., 1974, The Ontology of Turn : Studies in the Philosophy of Gustav Bergmann, University of Iowa Press. Loux M. 2006, « Aristotle’s Constituent Ontology », Oxford Studies in Metaphysics, vol. 2, Dean Zimmermann ed., Clarendon Press, Oxford, p. 207-250. Mertz D.W. 2002, « Combinatorial Predication and the Ontology of Unit Attributes » The Modern Schoolman, LXXIX, p. 163-197, également dans Mertz 2006, p. 141. Mertz D.W. 2006, Essays on Realist Instance Ontology and its Logic. Predication, Structure and Identity, Ontos Verlag, Francfort. Nef F. 2006, « Abstraction, objets éternel et occurrence actuelle » p. 363-383 in F. Beets, M. Dupuis et M. Weber éds., La science et le monde moderne d’Alfred North Whitehead, Ontos Verlag Francfort. Wolterstorff N. 1970, « Bergmann’s Constituent Ontology », Nous, 4, p. 109-134.
172
Modes of Exemplification E. J. LOWE In the neo-Aristotelian system of ontology that I have been developing in recent years,1 exemplification is not regarded as a primitive formal ontological relation but, rather, as obtaining in two different varieties or modes, the dispositional and the occurrent, each of which is analysable with the aid of two formal ontological relations that are regarded as primitive: instantiation and characterization. In the present paper, I should like, amongst other things, to clear up some difficulties that a number of my critics have claimed to find in my account of these matters. My treatment will not make direct reference to Gustav Bergmann’s philosophy, but it should be evident that there are many points of contact between my concerns and some of his, not least in connection with the logic of generality and its relation to the ontological structure of the world. In my system of ontology there are, unquestionably, general facts, in the form of laws, and in my view formal logic itself needs to be reformulated to reflect this fundamental truth. 1. The four categories of the four-category ontology The four basic ontological categories of my system are these: (1) substantial universals, (2) non-substantial universals, (3) substantial particulars, and (4) non-substantial particulars — or, less long-windedly and more memorably, kinds, attributes, objects, and modes. The latter terminology, although convenient, is, I concede, not entirely perspicuous. For one thing, the term ‘attribute’ might be thought to embrace only monadic universals, whereas I am happy to include relational universals in my ontology. For another, the terms ‘object’ and ‘kind’ both have common uses in metaphysics that are much broader than mine is intended to be. Thus, ‘object’ is sometimes used as a synonym for the all-purpose term ‘entity’, while ‘kind’ is often used as an alternative to ‘type’, in the sense in which the latter figures in the so-called ‘type–token distinction’. (In other words, ‘type’ is often used in a way in which it is pretty much 1
See especially my The Four-Category Ontology: A Metaphysical Foundation for Natural Science (Oxford: Clarendon Press, 2006).
interchangeable with ‘universal’, thus prescinding entirely from my own distinction between substantial and non-substantial universals.) Finally, ‘mode’ may strike many as archaic-sounding in comparison with the currently more fashionable term ‘trope’ — although I would urge that the former is in fact more, rather than less, perspicuous than the latter, since it is appropriately suggestive of the idea that properties, whether they are universals or particulars, are rightly to be thought of as ways of being. (The term ‘trope’ has, furthermore, a standard literary use which has no connection whatever with its current use in metaphysics — a fact that can only serve to render the latter use confusing.) Since, however, any choice of terminology in this area of metaphysics is bound to involve some departure from common usage, it may help at this point if I supply some everyday examples of items that at least appear to fall into the four categories that I have in mind when I present my ontological system. I must stress that these can only be taken, at this stage at least, to be apparent examples because, although I certainly want to defend the four-category ontology as a foundation for metaphysical inquiry, I don’t want to be committed without argument to supposing that everyday language and thought provide us with incontestably correct illustrations of its applicability to the real world. We should be prepared to allow that the four categories are best illustrated, in fact, only by entities postulated in advanced scientific theories, rather than by those assumed in our ‘common-sense’ ontology. So, with this caveat in mind, here are some putative examples of the four categories. A particular table, rock, or dog would, then, be an example of something belonging to the category of object, as I conceive of it. Such items are more traditionally known as individual substances. Corresponding examples of the category of kinds, as I conceive of it, would be the kinds table, rock, and dog of which the foregoing objects are, respectively, particular instances. Examples of the category of attribute would be the properties, conceived as universals, of brownness, hardness, and furriness that are exemplified, respectively, by those objects. And, finally, examples of the category of mode would be the particular instances of those universals possessed by those objects: the table’s particular brownness, the rock’s particular hardness, and the dog’s particular furriness.
174
2. The Ontological Square It may be noticed that, at least implicitly, I have already drawn on both of my aforementioned primitive formal ontological relations — instantiation and characterization — in introducing and describing the foregoing examples. For, first of all, kinds and attributes are both instantiated by, respectively, the objects and modes that are their particular instances — for example, the kind table by particular tables and the attribute brownness by particular brownnesses. And, secondly, in my terminology, to say that an object possesses a mode of a certain attribute is just another way of saying that that mode characterizes — or, if one prefers an older expression, inheres in — the object in question. Indeed, I want to go further and say that, likewise, characterization is a relation in which, at the level of universals, attributes stand to kinds. I shall say much more about this in due course, but already we have the materials to construct, in Figure 1 below, a version of the Ontological Square: a diagram that is enormously useful for the purposes of depicting the formal ontological relationships in which items belonging to the four different categories stand to one another. Kinds
instantiated by
Objects
characterized by
exemplified by
characterized by
Attributes
instantiated by
Modes
Figure 1: The Ontological Square (Version I
3. Universal versus particular and subject versus predicable It will be noted that the upper level of the Ontological Square, occupied by kinds and attributes, is the level of universals, while the lower level of the square, occupied by objects and modes, is the level of particulars.
175
Similarly, we can call the lefthand side of the square, occupied by kinds and objects, the side of subjects and the righthand side of the square, occupied by attributes and modes, the side of predicables — subjects being entities that are characterized in various ways and predicables being entities that characterize in various ways. Indeed, using this terminology, we could speak of the four fundamental ontological categories depicted in the square, beginning at the bottom lefthand corner, as being those of particular subjects, universal subjects, universal predicables and particular predicables. In that case, however, it should be clearly understood that the expressions ‘universal’, ‘particular’, ‘subject’ and ‘predicable’ do not themselves signify ontological categories as such but are, rather, cross-categorial terms, just as the all-purpose ontological term ‘entity’ is. Here, below, is another version of the Ontological Square depicting this aspect of the four-category ontology: Universals Kinds
Attributes
Subjects
Predicables
Objects
Modes Particulars
Figure 2: The Ontological Square (Version II)
4. Logical symbolism for four-category metaphysics At this point, it will be useful for me to introduce some logical symbolism, for in what follows we shall be concerned quite as much with the logic as with the metaphysics of the four-category ontology. Standard first-order
176
predicate logic with identity deploys only a single class of constants and variables — objectual ones — and a way of representing explicitly only a single formal ontological relation, identity. In a logic that is capable of representing perspicuously all of the metaphysically important features of the four-category ontology, we need, however, four different classes of constants and variables, together with ways of representing explicitly three different primitive formal ontological relations — not just identity, but also instantiation and characterization. So this is what I propose. We shall use, as is already customary, lower case letters from the beginning of the Roman alphabet, a, b, c, ... as object constants and lower case letters from the end of the Roman alphabet, x, y, z, ... as object variables. For kind constants we shall use lower case letters from the beginning of the Greek alphabet, α, β, γ, ... and for kind variables we shall use lower case letters from the end of the Greek alphabet, ϕ, χ, ψ,... For attribute constants we shall, in mimicry of existing custom, use the upper case Roman letters F, G, H, ... and for attribute variables we shall use upper case letters from the end of the Roman alphabet, X, Y, Z, ... Finally, for mode constants we shall use the lower case Roman letters f, g, h, ..., while for mode variables we shall use the lower case Roman letters r, s, t, ... As for the three primitive formal ontological relations, we shall, as is customary, represent identity by the equality sign, ‘=’, and supplement this with the slash, ‘/’, to represent instantiation. Finally, we shall represent characterization by simple juxtaposition of appropriately chosen constants or variables. Thus, for example, ‘a/β’ and ‘f/G’ say, respectively, that object a instantiates kind β and that mode f instantiates attribute G, while ‘af’ and ‘βG’ say, respectively, that object a is characterized by mode f and that kind β is characterized by attribute G. The foregoing proposals are presented in more convenient tabular form below. Objects
Kinds
Attributes
Modes
constants
a, b, c, ...
α, β, γ, ...
F, G, H, ...
f, g, h, ...
variables
x, y, z, ...
ϕ, χ, ψ, …
X, Y, Z, …
r, s, t, …
Instantiation: a/β, f/G
Characterization: af, βG
177
5. Dispositional versus occurrent exemplification Since I don’t regard exemplification as being a primitive formal ontological relation, I don’t need an undefined symbol to represent it. Exemplication, as Figure 1 implies, is a relation between objects and attributes. Or, more exactly, there are two different relations of exemplification between objects and attributes, corresponding to the two different routes from the bottom lefthand corner of the Ontological Square (the object corner) to the upper righthand corner (the attribute corner). For reasons which will become more apparent in due course, I call these two different species or modes of exemplification dispositional and occurrent exemplification, which may be depicted on the Square as follows:
dispositional exemplification Kinds
Objects
Attributes
Modes
occurrent exemplification Figure 3: The Ontological Square (Version III)
Using the expressions ‘Dis[a, F]’ and ‘Occ[a, F]’ to say, respectively, that object a exemplifies attribute F dispositionally and that object a exemplifies attribute F occurrently, I propose that we may define these two modes of exemplification as follows: Dis[a, F] =df (∃ϕ)(ϕF & a/ϕ) Occ[a, F] =df (∃r)(ar & r/F)
178
In other words, an object a exemplifies an attribute F dispositionally just in case a instantiates some kind that is characterized by F, while an object a exemplifies an attribute F occurrently just in case a is characterized by some mode that instantiates F. 6. Ontological dependency relations and categorial uniqueness Before I say more about the crucial issue of the two different modes or species of exemplification, I want to present yet another version of the Ontological Square, this time one which represents the metaphysically significant relationships of ontological dependency that are characteristic of entities belonging to the four different ontological categories.2 First of all, then, I need to point out that, true to the Aristotelian spirit in which the Square is conceived, the four-category ontology embodies an immanent realist view of universals, according to which it is an essential feature of any universal that it has particular instances, which provide the ground of its existence. According to this view, there are, then, no uninstantiated universals and every universal stands in a relationship of existential dependence to its particular instances. I shall call the relationship in question weak existential dependence. I call it ‘weak’ for the following reason: although, according to this view, a universal must have particular instances, which constitute the ground of its existence, a universal does not depend for its very identity on the particular instances that it happens to have, which is a purely contingent matter. Thus, for example, although the attribute brownness (assuming there to be such an attribute) must have particular instances in form of the particular brownnesses of various brown objects, that very same attribute could have existed even if those particular brownnesses had not, provided that other particular brownnesses had existed — for example, if other objects had been brown. Next, I want to say that the modes of an object stand to that object in another relationship of existential dependence, which I shall call strong existential dependence — ‘strong’ because in this case the modes do depend for their very identity on the object that they characterize. Thus, for example, what distinguishes one particular brownness from another exactly resembling brownness is, precisely, the object that possesses it, or in which it ‘inheres’ — the implication of this being that modes cannot be 2
For more on the notion — and the varieties — of ontological dependence, see my The Possibility of Metaphysics: Substance, Identity, and Time (Oxford: Clarendon Press, 1998), ch. 6.
179
‘transferred’ from one object to another and cannot exist ‘unattached’ to any object. Clearly, since one object must possess many different modes, whereas the same mode cannot be possessed by many different objects, the relationship of strong existential dependence between modes and their object is a many-one relation. Finally, I want to say that a kind stands in a one-many relationship of strong existential dependence to its various attributes: that is, that one kind must be characterized by many different attributes and that it depends for its very identity on the attributes that characterize it. Thus, for example, I would want to say that the kind electron depends for its identity on the specific attributes of charge, spin, and rest mass that characterize that kind of fundamental physical particle. (Electrons carry unit negative charge, have a spin of one half, and a certain specific rest mass, differing from all other kinds of fundamental physical particle in at least one of these respects.) Here below, then, is a version of the Ontological Square which represents the foregoing relationships of ontological dependency between items located at the different corners of the Square, where a solid-headed arrow signifies strong existential dependence and an open-headed arrow signifies weak existential dependence: Kinds
Attributes one-many
many-one Objects
Modes
Figure 4: The Ontological Square (Version IV)
It will be noticed that, in Figure 4, each corner of the Ontological Square differs from every other corner in respect of the number and type of
180
arrows that proceed from or lead to it. The object corner has two arrows, one of each type, leading to it. The kind corner has two arrows, one of each type, proceeding from it. The attribute corner has one solid-headed arrow leading to it and one open-headed arrow proceeding from it. And the mode corner has one open-headed arrow leading to it and one solid-headed arrow proceeding from it. Thus, objects are represented as being the least dependent of all entities, in keeping with another Aristotelian sentiment, namely, that individual substances are ontologically more basic than entities belonging to other ontological categories, insofar as they constitute the ultimate ground of all being. Moreover, because the four corners of the Ontological Square can be unambiguously identified in terms of the number and type of arrows leading to or proceeding from them — in other words, because the four categories can be unambiguously differentiated from one another in terms of the characteristic relationships of ontological dependency in which their members stand to the members of the other categories — the four-category ontology is not subject to the kind of objection that Frank Ramsey and others have raised against simpler systems of categorial ontology that appeal merely to the subject–predicable or universal–particular distinctions.3 This objection is that the advocates of these systems cannot explain, in a non-question-begging way, what distinguishes each of the categories in such a system from another category in the same system and thus what renders each category unique. We could call this the problem of categorial uniqueness — and the important point is that the four-category ontology has the distinct advantage over some of its apparently more parsimonious rivals of being able to solve it. Incidentally, before I move on from the subject of the Ontological Square, here is a useful mnemonic to remind one of the order of its four corners, starting, in appropriately Aristotelian fashion, with the bottom lefthand object corner: OcKhAM — Objects, Kinds, Attributes, and Modes. There is some irony in this mnemonic, of course. William of Ockham himself, who did not favour realism concerning universals, would certainly not have approved of the four-category ontology. And, indeed, many present-day ontologists would no doubt attempt to wield Ockham’s notorious razor against my system, accusing me of multiplying entities beyond necessity. However, one of the objectives of my work on the fourcategory ontology has been precisely to show how that ontology, while it is 3
See F. P. Ramsey, ‘Universals’, in his The Foundations of Mathematics and Other Logical Essays (London: Kegan Paul, 1931). For further discussion, see my The FourCategory Ontology, ch. 7.
181
undoubtedly richer than many of its popular rivals — such as the pure trope ontology, the ontology of objects and attributes, and the ontology of objects and modes, each of which limits itself to only one or two corners of the Ontological Square — deserves our allegiance on account of its superior capacity to resolve a wide range of metaphysical problems, such as the problem of categorial uniqueness, which we have just discussed, and the so-called inference problem concerning the relationship between laws of nature and particular matters of fact. This latter problem, indeed, is the one that I shall focus on next. 7. Laws of nature and the inference problem As I have already indicated, in my view all kinds are characterized by certain attributes — and laws of nature, by my account, consist precisely in such facts of characterization solely involving universals.4 All objects, on the other hand, are characterized by modes that instantiate certain attributes — and it is in their being so characterized that particular occurrent facts or states of affairs consist. By contrast, particular dispositional facts or states of affairs consist in objects instantiating kinds that are characterized by certain attributes — in other words, they consist in objects being subject to certain laws governing the kinds to which they belong. What Bas van Fraassen and others call ‘the inference problem’ does not afflict this account of laws, as it does those of David Armstrong, Michael Tooley, and Fred Dretske — accounts which, like mine, locate laws in the domain of universals.5 The complaint, as it applies to Armstrong’s theory, is just this. According to Armstrong, a law of nature (at least, in the simplest sort of case) consists in the obtaining of a secondorder relation of natural necessitation, N, between first-order universals or attributes — for example, F and G — which may be represented by a statement of the form ‘N(F, G)’ (‘F-ness necessitates G-ness’, as we might express it in English).6 And then he contends, crucially, that ‘N(F, G)’ entails the corresponding universal generalization concerning particulars, ‘(∀x)(Fx → Gx)’ (in English, ‘Everything that is F is G’ or ‘All Fs are Gs’). But now the problem is that no account is provided of how ‘N(F, G)’ 4
See further my The Four-Category Ontology, especially Part III. See Bas C. van Fraassen, Laws and Symmetry (Oxford: Clarendon Press, 1989), ch. 5. 6 See D. M. Armstrong, What is a Law of Nature? (Cambridge: Cambridge University Press, 1983). 5
182
entails ‘(∀x)(Fx → Gx)’, since there appears to be no appropriate formal relationship between two sentences of these forms. As I have just remarked, the inference problem does not arise for my account of laws. This is because, according to my account, what laws entail regarding the domain of particulars are solely dispositional, not occurrent, facts or states of affairs — and such entailments are formally valid in the system of sortal logic that I endorse. In the simplest case, a law has the form ‘βF’, where ‘β’ denotes a kind and ‘F’ an attribute. A corresponding particular dispositional fact has the form ‘(∃ϕ)(ϕF & a/ϕ)’, where, as was explained earlier, ‘ϕ’ is a variable ranging over kinds, ‘a’ denotes an object, and ‘/’ signifies the formal ontological relation of instantiation. (The quoted formula here expresses, thus, the fact that a is disposed to be F — in other words, that a exemplifies F dispositionally. Recall here our earlier definition of ‘Dis[a, F]’ as being equivalent to ‘(∃ϕ)(ϕF & a/ϕ)’.) The crucial point, then, is that ‘βF’ entails ‘(∀x)(x/β → (∃ϕ)(ϕF & x/ϕ))’ — and thus, via the definition of ‘Dis[a, F]’, entails ‘(∀x)(x/β → Dis[x, F])’ — the proof of which is elementary: see Proof [1] in the Appendix below. In other words, if the law that kind β is characterized by attribute F obtains, then it follows that every object instantiating β has the disposition to be F. But, crucially, it does not follow that every such object is occurrently F. Laws, then, fully determine how objects are disposed or tend to appear or behave, but not how they actually appear or behave. Consequently, they do not entail universal regularities (Humean ‘constant conjunctions’) amongst such appearances or behaviour, in the way that the Armstrong–Tooley–Dretske account of laws proposes and which critics like van Fraassen profess — rightly, I think — to find mysterious. 8. Conditional laws and the problem of idiosyncratic dispositions However, some further complications now need to be discussed, in the light of criticisms of my system alluded to earlier. As some critics have pointed out, it seems implausible, at first sight, to suppose that every particular dispositional fact is grounded in some law governing a kind to which the object possessing the disposition belongs.7 After all, objects of 7
Amongst the people who have raised this point, either in print or in correspondence, are Ryan Wasserman, Ludger Jansen, and David S. Oderberg. I am particularly indebted to David Oderberg in this regard.
183
the same kind often seem to have different and sometimes even entirely idiosyncratic dispositions. For instance, a particular dog, Fido, might conceivably be the only dog in the world disposed to drink whisky: but it is clearly not a law governing canine behaviour that dogs drink whisky. This apparent difficulty can, however, be overcome by recognizing the irreducibly conditional character of many laws. That is to say, the fundamental form of many laws is not, as we have hitherto been supposing, simply something like ‘βF’, but rather something like ‘β(G → F)’. Consider the following simple example. A body of frozen water — a piece of ice — is not disposed to evaporate, but a body of boiling water is. (Change the example if you don’t like it: I use it purely for illustrative purposes.) Yet both are bodies of the same kind — water. Ice and boiling water are not, after all, different kinds of stuff, just the same kind of stuff in different physical forms. The change that happens when ice melts is a phase change, not a substantial change. Consequently, however, we should not affirm it as a law governing the kind water that water evaporates — rather, what we should affirm is that water evaporates if, or when, it is boiling, and this is consistent with our also affirming that water does not evaporate if, or when, it is frozen. Both laws apply at all times to any body of water, whether or not it happens to be frozen or boiling (or indeed neither). What entitles us to say that a piece of ice is not disposed to evaporate, whereas some boiling water is, is not, then, that they instantiate different kinds governed by different unconditional laws, but rather that the single kind that they both do instantiate — water — is governed by two different conditional laws, and that while one of these bodies of water happens to satisfy the antecedent condition of one of these laws, the other body happens to satisfy instead the antecedent condition of the other law. Casually speaking, we may indeed say that the sentences ‘Boiling water evaporates’ and ‘Frozen water does not evaporate’ express laws. But it is at least potentially misleading to express the laws in question in those ways, since the complex sortal terms ‘boiling water’ and ‘frozen water’ are not denotative of different kinds. We do better to express the laws in question conditionally, as earlier: ‘Water, when it is boiling, evaporates’ and ‘Water, when it is frozen, does not evaporate’. An adequate system of sortal logic will reflect this by admitting logical operations on predicates as well as on whole sentential formulas.8 8
See further my Kinds of Being: A Study of Individuation, Identity, and the Logic of Sortal Terms (Oxford: Blackwell, 1989), ch. 10. Note especially, in this connection, that ‘β(G → F)’ must not be taken to be logically equivalent to ‘βG → βF’: for
184
9. Essential versus accidental dispositions So how do the foregoing considerations enable us to deal with the example of Fido, the dog that is idiosyncratically disposed to drink whisky? In the following way. It is surely reasonable to suppose that Fido’s peculiar condition is lawfully explicable, rather than just being miraculous. Presumably, Fido — perhaps on account of some feature of his past experience or training — has some property, X, such that it is a law that dogs, if they have X, drink whisky. Fido just happens to be peculiar in being the only dog to have X.9 It may be helpful at this point to introduce a distinction between essential and accidental dispositions — and to say that Fido possesses only an accidental disposition to drink whisky. Our earlier definition of ‘Dis[a, F]’ is, in these terms, really only a definition of essential dispositionality. So let us make that explicit by using the formula ‘DisE[a, F]’ to say that a is essentially disposed to be F, defining this exactly as we earlier defined ‘Dis[a, F]’: DisE[a, F] =df (∃ϕ)(ϕF & a/ϕ) Then let us use the formula ‘DisA[a, F]’ to say that a is accidentally disposed to be F, defining this as follows: DisA[a, F] =df (∃X)(Occ[a, X] & ¬DisE[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ)) In other words, we say that a is accidentally disposed to be F just in case there is some attribute, X, such that a occurrently exemplifies X — but is not essentially disposed to be X — and for some kind, ϕ, that is instantiated by a, it is a law that ϕs are F if they are X. Drawing on the preceding definition of essential dispositionality, we can simplify this to: DisA[a, F] =df (∃X)(Occ[a, X] & ¬DisE[a, X] & DisE[a, (X → F)]) example, ‘Water evaporates if, or when, it is boiling’ is plainly not equivalent to ‘If water boils, then water evaporates’. 9 Note that I am assuming here that Fido’s peculiarity is not the upshot of some genetic mutation, because in that case there might well be grounds for saying that Fido differs in kind from other dogs and that he belongs to a kind, ϕ, such that it is a law that ϕs drink whisky. I am grateful to Frédéric Nef for raising this issue in conversation.
185
That is to say, a is accidentally disposed to be F just in case a occurrently exemplifies some attribute, X, such that, although a is not essentially disposed to be X, a is essentially disposed to be F if it is X. Here it may be asked why I include the conjunct ‘¬DisE[a, X]’ in the definiens of ‘DisA[a, F]’. The answer should be evident. It is plausible to suppose that ‘DisE[a, X]’ and ‘DisE[a, (X → F)]’ together entail ‘DisE[a, F]’, so that without this conjunct our definition would allow an object a to be both essentially and accidentally disposed to be F, which seems absurd. In fact, this entailment holds only under certain assumptions — notably, under the assumptions that an object cannot instantiate more than one kind and that two laws of the form ‘βG’ and ‘β(G → F)’ together entail one of the form ‘βF’. See Proof [2] in the Appendix for a proof that the entailment holds under these particular assumptions, which I shall take to be correct for present purposes. This, however, raises a further question. It might be supposed, prima facie, that ‘Occ[a, F]’ entails ‘DisE[a, F]’ — that if an object a exemplifies an attribute F occurrently, then a is essentially disposed to be F, rendering our definition of ‘DisA[a, F]’ inconsistent. Clearly, we must for this very reason deny that such an entailment holds — and, indeed, it seems reasonable to maintain instead that ‘Occ[a, F]’ entails only the disjunction ‘DisE[a, F] ∨ DisA[a, F]’: that if an object a is occurrently F, then a is either essentially or accidentally disposed to be F. But aren’t we now faced with a regress problem? For, if the latter entailment holds, then, clearly, given our original definition of accidental dispositionality, so does the following: ‘DisA[a, F]’ entails ‘(∃X)(DisA[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ))’. The implication seems to be that an object can possess one accidental disposition only on condition of possessing another one. Fortunately, it seems clear that this potential regress is not inevitably infinite. For notice that we don’t have that ‘DisA[a, F]’ entails ‘(∃X)(X ≠ F & DisA[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ))’. So it is not in fact strictly true, given our assumptions so far, that an object can possess one accidental disposition only on condition of possessing another one. We can, then, have a case in which ‘DisA[a, F]’ is true even though the following is also true: ‘(∀X)((DisA[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ)) → X = F)’. In such a case, it would clearly follow that this is true: ‘DisA[a, F] & (∃ϕ)(ϕ(F → F) & a/ϕ))’, which is equivalent to ‘DisA[a, F] & DisE[a, (F → F)]’. For a proof of this, see Proof [3] in the Appendix. However, there is no
186
contradiction lurking here, because ‘DisA[a, F] & DisE[a, (F → F)]’ is not at all absurd, in constrast with ‘DisA[a, F] & DisE[a, F]’. For, presumably, it is a merely trivial logical truth that any object a is essentially disposed to be F if it is F. In other words, our definitions of accidental and essential dispositionality allow that sometimes, at least, it is true that an object a is accidentally disposed to be F simply because a truth of the following form obtains: ‘Occ[a, F] & ¬DisE[a, F]’ — for this certainly entails the definiens of ‘DisA[a, F]’, given the triviality of the aforementioned truth. See again the Appendix, Proof [4], for a proof of this. In other, and presumably more usual cases, of course, ‘DisA[a, F]’ will be true even though ‘Occ[a, F]’ is not true. 10. Conditional laws and the principle of instantiation Finally, I want to remark that an important advantage of recognizing the prevalence of conditional laws is that it renders more plausible the principle of instantiation that is central to a neo-Aristotelian immanent realist view of universals like mine — especially for those philosophers, such as Armstrong and myself, who not only adhere to this principle but also hold that we should decide which universals to admit into our ontology largely by seeing which universals science invokes in the laws that it postulates to explain empirical phenomena. According to the principle of instantiation, every existing universal must have, at least at some time and some place, a particular instance: there are no uninstantiated universals. Now, for a conditional law of the form ‘β(G → F)’ to obtain, the principle of instantiation requires only that each of the universals β, F, and G should have instances — not that there should be an instance of β that also occurrently exemplifies either G or F. For example, the principle allows that it may be a law that water, when it is boiling, evaporates — that boiling water evaporates — without requiring there to exist, at any time or place, a body of water that is boiling. Imagine, thus, a universe in which water exists but temperatures never rise high enough to reach its boiling point. Provided that some liquids sometimes actually boil and evaporate in this universe, it can still be true, consistently with the principle of instantiation, that water, if or when it is boiling, evaporates. Now, clearly, many laws are expressed by sentences which do not reveal their underlying conditional form — ‘Ice does not evaporate’ is a simple example of this. ‘Water boils’ is, of course, another — for this must be taken as an elliptical way of stating the conditional law that water boils
187
if, or when, its temperature and the atmospheric pressure reach certain levels (for instance, when its temperature reaches 100 degrees Centigrade at sea-level pressure).10 But it may well be that very many more laws than we intuitively suppose are really conditional in form, thereby reducing considerably the number of universals that we need to include in our ontology and, correspondingly, reducing the burden that is imposed by the principle of instantiation. Here, then, is another example of how the fourcategory ontology, despite being — or, rather, in virtue of being — more complex than some of its rivals, can deal better than they do with certain longstanding metaphysical problems.
10
Note, indeed, that if we held ‘Water boils’ to be a law, then, given our previous assumption that ‘Water, if it is boiling, evaporates’ is a law, we would be committed — via the principle that ‘βG & β(G → F)’ entails ‘βF’ — to holding that ‘Water evaporates’ is a law, which we have already denied to be the case. This confirms that ‘Water boils’ should be taken to be an elliptical expression of a law which is really conditional in form.
188
APPENDIX [1] Proof that ‘βF’ entails ‘(∀x)(x/β → (∃ϕ)(ϕF & x/ϕ))’ βF ¬(∀x)(x/β → (∃ϕ)(ϕF & x/ϕ)) | (∃x)¬( x/β → (∃ϕ)(ϕF & x/ϕ)) | (∃x)( x/β & ¬(∃ϕ)(ϕF & x/ϕ)) | (a/β & ¬(∃ϕ)(ϕF & a/ϕ)) | a/β ¬(∃ϕ)(ϕF & a/ϕ) | (∀ϕ)¬(ϕF & a/ϕ) | (∀ϕ)(ϕF → ¬a/ϕ) | βF → ¬a/β | ______________________ | | ¬βF ¬a/β x x
189
[2] Proof that ‘DisE[a, G] & DisE[a, (G → F)]’ entails ‘DisE[a, F]’ DisE[a, G] & DisE[a, (G → F)] ¬DisE[a, F] | (∃ϕ)(ϕG & a/ϕ) (∃ϕ)(ϕ(G → F) & a/ϕ) | ¬(∃ϕ)(ϕF & a/ϕ) | βG & a/β γ(G → F) & a/γ | βG a/β a/γ | β=γ | β(G → F) & a/β | β(G → F) | βG & β(G → F) | βF | βF & a/β | (∃ϕ)(ϕF & a/ϕ) x
190
[3] Proof that ‘DisA[a, F]’ and ‘(∀X)((DisA[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ)) → X = F)’ together entail ‘(∃ϕ)(ϕ(F → F) & a/ϕ)’ DisA[a, F] (∀X)((DisA[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ)) → X = F) ¬(∃ϕ)(ϕ(F → F) & a/ϕ) | (∃X)(Occ[a, X] & ¬DisE[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ)) | Occ[a, G] & ¬DisE[a, G] & (∃ϕ)(ϕ(G → F) & a/ϕ) | Occ[a, G] ¬DisE[a, G] (∃ϕ)(ϕ(G → F) & a/ϕ) | | | DisA[a, G] DisE[a, G] | x | (DisA[a, G] & (∃ϕ)(ϕ(G → F) & a/ϕ)) → G = F) | | ¬(DisA[a, G] & (∃ϕ)(ϕ(G → F) & a/ϕ)) | | ¬DisA[a, G] x
| ¬(∃ϕ)(ϕ(G → F) & a/ϕ) x
| G=F | (∃ϕ)(ϕ(F → F) & a/ϕ)) x
[4] Proof that ‘Occ[a, F] & ¬DisE[a, F]’ entails ‘(∃X)(Occ[a, X] & ¬DisE[a, X] & (∃ϕ)(ϕ(X → F) & a/ϕ))’ — the definiens of ‘DisA[a, F]’ Occ[a, F] & ¬DisE[a, F] ¬(∃X)(Occ[a, X] & ¬DisE[a, X] &(∃ϕ)(ϕ(F → F) & a/ϕ)) | (∃ϕ)(ϕ(F → F) & a/ϕ) | Occ[a, F] & ¬DisE[a, F] & (∃ϕ)(ϕ(F → F) & a/ϕ) | (∃X)(Occ[a, X] & ¬DisE[a, X] &(∃ϕ)(ϕ(X → F) & a/ϕ)) x
191
On Ties and Copulae within the Ontological Square LUC SCHNEIDER Introduction The basic thesis of the present article is that, given a minimal correspondence between logical and ontological form, the adoption of a nexus-style ontology in the sense of Bergmann [7] entails the rehabilitation of the traditional notion of copula in formal logic. The paper is divided in two parts. In the first part I propose and motivate a nexus-orientated account of the so-called “Ontological Square”, a categorial framework that combines the distinction between universals and particulars with that between substrates and characters. In the second part, I compare predicate-based and copulabased assays of predication and argue that under the conception of logic as language, the latter are to be preferred to the former if the ontology described in the first part is true. Furthermore I will show that a logic for the Ontological Square requires a manifold of copulae and is therefore richer than standard predicate logic. Thus the discussions in this article constitute the prolegomena for a new formal calculus of copulae developed in [36]. 1 The Ontological Square 1.1 Overview In Categories [1, 1a20–1b10], Aristotle suggests two orthogonal distinctions between things: types vs. their tokens on the one hand and characters vs. their substrates (or bearers) on the other hand. The resulting categorial scheme, which Angelelli [3, p. 12] [4] has called the Ontological Square, consists of the following sorts of things [27, chap. 2] [39]: Types Tokens
Substrates Kinds (Man) Substances (Socrates)
Characters Attributes (Wisdom) Moments (Socrates’ wisdom)
Substances, i.e. common-sense objects such as organisms and artifacts, are tokens of kinds, e.g. Man or Chair. The characters or features of substances are moments, such as states (e.g. the wisdom of Socrates or Mary's love for Sam), events (e.g. Nelson's victory at Trafalgar), powers or dispositions (e.g. the brittleness of a glass or Mary's fickleness), and spatial boundaries (such as surfaces and edges) [30]. Moments in their turn are tokens of attributes such as Wisdom or Brittleness which (may) be generic characters of certain kinds like Philosopher or Glass. The Ontological Square, which arguably embodies common-sense intuitions as to how things are to be classifed, is articulated around two main ties or nexus, namely instantiation, which holds between types and tokens, and characterisation, which holds between characters and their bearers [27, pp. 22, 40, 60, 79, 93 & 111]. Besides instantiation and characterisation, there are two (respectively three) other ties that make up the structure of the Ontological Square, namely subsumption (which holds between types) and (specific or generic) exemplification (which holds between substances and attributes). I will review these ties in more detail below. 1.2 The fundamental ties of the Ontological Square 1.2.1 Ties and facts Ties or nexus do not fall into one of the four categories of things mentioned above; in particular they are not universals. Indeed, they do not belong to the content or matter of reality as represented by the Ontological Square, but to its form, since they constitute the structure of the world. Nonetheless, they are full-fledged entities of a distinct ontological category. Nexus do not require further entities to connect them to the things they tie together, since otherwise there would be an infinite regress [7, pp. 9, 43 – 44]. One could object that if ties have to be their own ontological ground in order to avoid a Bradleyan regress, it would be more parsimonious to assume that some things, i.e. characters respectively universals, are such as to be saturated by others, e.g. substrates respectively instances, and that the coordination between saturated and unsaturated things is its own ontological ground [7, pp. 7 – 8]. However, I agree with Bergmann that what it means for such a coordination to be ontologically basic must remain a
194
complete mystery [7, p. 8]. Indeed, one could conceive of unsaturated things as functional entities. But the mathematical notion of a function is commonly understood either extensionally as a set of tuples or intensionally as an operation that for a given input returns a specific output or result. The extensional conception of functions is not adequate to grasp the idea of unsaturatedness, since class abstraction is unrestricted (in the limits of consistency) and hence cannot explain coordination between unsaturated and saturated entities. As to the intensional conception of functions, while the application of a concept can be accounted for as an operation of the human mind, and thus concepts can indeed be assayed as functions, this approach is obviously inappropriate for characters or universals as denizens of a reality that exists independently of any human activity of construction. Alternatively, and more promisingly, unsaturatedness could be understood in terms of ontological dependence. Yet ontological dependence or foundation itself is a primitive tie or nexus that cannot be simply analysed away as the (relevant) necessitation of the founder by that of the foundee, in case necessary beings are admitted in the ontology [13, chaps. 3 & 4]. Hence it seems that Bergmann is right in preferring a nexus-style ontology to a function-style ontology [7, p. 8]. A tie combining things yields a fact, i.e. a complex consisting of a nexus and the things it holds together [7, p. 4]. Facts are the building blocks of the structure of reality that is generated by ties applying to things. Hence the Ontological Square explicitly or implicitly comprises not only material categories, such as substances, kinds, moments and attributes, but also formal categories, namely nexus and facts. Some philosophers defend the view that all nexus are internal, i.e. that they hold of their terms in virtue of the nature of their terms [7, p. 54]. Adepts of this view also assert that internal ties are no addition to reality and thus internal facts collapse into collections of things [7, p. 61]. Leaving aside the issue whether or not this further assertion makes sense, there are two plausible objections to the view that all nexus are internal. The minor one is that the notion of a thing's nature itself may seem in need of clarification; however, it is only fair to admit that all ontological explanations have to stop somewhere. The major worry is that if nexus hold of their terms in virtue of the latters' essence, they hold of their terms essentially. Nonetheless it seems intuitive that at least some facts are accidental, i.e. the fact that Socrates instantiates the kind Greek and the fact that Socrates is a bea-
195
rer of an instance of Wisdom. For this reason, one may concede to Bergmann [7, p. 53] that at least some ties are external. I have pointed out that ties are not universals; this may seem odd, especially since nexus may be viewed as patterns of facts – hence why not regard them as types ? Thus, Wolfgang Degen [16, 17] has proposed to extend the Ontological Square into an Ontological Hexagon by adding fact types and fact tokens. Yet such an extension would blur the distinction between matter and form in ontology. Furthermore, there is a decisive difference between facts and instances, whether substances or moments: in contrast to the latter, facts are complexes that have nexus among their constituents. Therefore, ties and facts should be regarded as (formal) categories sui generis. 1.2.2 Instantiation Instantiation binds types to their tokens or instances: kinds are instantiated by substances and attributes by moments as particular features of substances. The bond between a type and its tokens or instances may super_cially seem akin to that between a class and its members: we do often say that instances „belong to“ or are „members“ of a certain type. However, contrary to classes, types are generally considered to be intensional: there may be distinct types sharing all their instances, e.g. the types Cordate (animal having a heart) and Renate (animal having kidneys) [5, pp. 25 – 26]. The extensional identity criterion would even fail in the case of necessarily co-extensive types such as Three-sided Figure and Three-angled Figure [5, p. 26]. The distinction between types and tokens is usually conceived of as that between universals and particulars [5, pp. 1–2, 6–7]. Universals and particulars differ in the following way: while it is metaphysically possible that there are two distinct particulars that are qualitatively, though not necessarily relationally undistinguishable (like the whiteness of a rose and the whiteness of a lotus), there cannot be several qualitatively indiscernible universals (like two Whitenesses) [53]. In other words, particulars may be merely numerically distinct, while distinct universals always differ qualitatively [7, p. 46]. It is usual to differentiate between abundant and sparse universals [5, p.
196
87]; sparse universals provide an explanatorily minimal partition of the domain of particulars, while abundant universals are arbitrary boolean combinations of sparse universals. It is also often argued that there are only sparse types. Now, if one adopts an Aristotelian view of universals, one cannot remain wholly neutral on this issue. Indeed, according to the Aristotelian conception, types are „immanent“ in the sense that there are no uninstantiated universals, just as there are no untyped tokens. This certainly is a serious constraint on the abundancy of universals. Ultimately, if all universals are immanent, i.e. are instantiated, then the distinction between types and tokens amounts to a difference in the method of counting or thrawling through reality: a universal is tantamount to a plurality of tokens counted as one [53]. In other words, a type is the unity underlying the manifold of its instances: a universal stands to its instances as the one stands to the many. 1.2.3 Characterisation Characterisation links features to their bearers: moments characterise substances and attributes characterise kinds. While all moments, with the debatable exception of process-things like avalanches, ashes, and snow falls, are particular features of substances, some (but not all) attributes may be generic features of kinds, such as Rationality with respect to Man. In order to distinguish specific from generic characterisation, I will henceforth refer to the former as inherence and to the latter simply as attribution. Attribution does not simply reflect inherence: that a type is an attribute of some kind does not imply that all substances of this kind have a moment that is an instance of the attribute. For example, Sixleggedness is an attribute of the Insect, but there are some insects which have more or less than six legs due to some mutation or accident. That characterisation between types does not amount to a generalisation of characterisation between (their) instances lends support to the distinction between universals and particulars. Evidence for a commonsensical commitment to moments like states or events may be the use of verb nominalisations such as „Mary's song“ or „Sam's sadness“ in ordinary English [49]. The commitment to moments is paramount for so-called „event-based“ accounts of natural language se-
197
mantics [14] [31, chap. 2]. Moments are often said to be „ways“ of substances [35, chap. 4] [5, p. 116]. Indeed, except maybe the controversial case of process-things, moments do not exist in abstraction from substances as their bearer(s) [28]. Thus, in general, moments are untransferable in the sense of being specific to their bearers or relata [5, pp. 117 – 118], i.e. a moment inheres in the same substance(s) in each world it exists at all. A substance, by contrast, may not only be the bearer of more than one moment, but may also have or lack certain moments it actually has, namely those that are not essential to it. The distinction between substances and moments can be fruitfully understood in terms of Strawson's conception of identifiability-dependence [43, p. 17], which is meant to capture Aristotle's [1, 16b 10–15] notion of ontological priority. According to this assay of ontological priority, the items of a category A are identifiability-dependent on the items of a category B, respectively the Bs are ontologically prior to the As, if, and only if the As can only be identified provided the Bs have already been singled out. Substances are identifiability-independent for they are suitably dimensioned to ground synchronic spatial frameworks of reference and are also invariant enough to be re-identifiable and thus to support a diachronic framework of reference. Moments, including putatively bearerless process-things, cannot provide such frameworks of reference and for this reason are identifiability-dependent on substances. The ontological priority of substances over moments is not just a figment of armchair philosophy, but is corroborated by empirical psychology. Indeed, experiments as to how members of different linguistic communities nonverbally represent events have shown that there is a languageindependent preference for the following ordering: first the substances as participants (i.e. the agent followed by the theme), then the event itself [22]. This seems to confirm that in order to single out events and other moments, one has to identify substances first and thus that the latter form the background for the individuation of the former. The criterion of identifiability-dependence presupposes that, just like substances, moments are located in space and time. Therefore moments are not „abstract particulars". The neologism „trope” is predominantly used in exactly this sense, which is why I prefer to avoid this term altogether. Moreover, friends of „tropes“ as abstract particulars, such as Stout [41, 42],
198
Williams [52], Campbell [10], Bacon [6] and Simons [38], defend reductionist assays according to which universals are equivalence classes of exactly resembling „tropes” and substances are bundles of co-located „tropes“. The term „moment” has the advantage to be free of such philosophical associations; I use it to refer to dependent concrete particulars, a.o. states and events. In order to avoid the conundrum of what a substance could be in abstraction of its moments, some ontologists, such as Carl Stumpf [45, pp. 22 – 31], have found it more useful to think of a substance as the unity underlying the manifold of its moments rather than a mere support for particular features. That is, a substance would be the plurality of the states and events (as well as other moments) it participates in counted as one. Hence, just as the type-token distinction, the substance-moment dichotomy would amount to a difference in the method of establishing the inventory of the world's furniture, i.e. the difference between the one and the many, the singular and the plural. 1.2.4 Exemplification The diagonal that holds between particular substances and universal attributes is the only nexus expressed by the atomic logical form in conventional predicate calculus. According to Lowe [27, pp. 30 – 32], there are two distinct, but not mutually exclusive, sorts of exemplification, which can be accounted for in terms of the previously mentioned bonds: 1. Generic exemplification holds between a substance and an attribute iff the latter is an attribute of the kind which is instantiated by the former. E.g. since Sixleggedness is an attribute of the Insect, every particular insect generically exemplifies Sixleggedness, even if it happens to have more or less than six legs. 2. Specific exemplification holds between a substance and an attribute iff the former has a moment that instantiates the latter. E.g. a particular insect that has lost two legs in an accident, specifically exemplifies Fourleggedness, though as a member of the kind Insect it generically exemplifies Sixleggedness. Note that attributes may be exemplified by many substances, while mo-
199
ments are bearer-specific; in other words, attributes are repeatable, while with the universal-particular dichotomy, since there are types which are not attributes, namely kinds, and instances which are not moments, namely substances. One may think that if specific exemplification were considered to be a primitive tie, moments could simply be assayed as facts that consist in substances specifically exemplifying attributes [5, pp. 132 – 133] [29, pp. 15– 16 & 98 – 100]. However, I have already argued that facts and instances are distinct categories: moments and substances are simple, while facts are complexes of ties and things. Furthermore, states and events have a temporal location, while facts are outside of space and time. The distinction between events and facts is also supported by linguistic analysis: in English, events are referred to by „perfect nominals“, i.e. verbs that are fully nominalised, such as in „John's sudden revelation of the secret“, while facts are denoted by „imperfect nominals“, i.e. verbs that are nominalised but still retain verbal features, such as in „John's suddenly revealing the secret“; perfect nominals and imperfect nominals are not substitutable for each other in every linguistic context [49, pp. 131 – 142]. 1.2.5 Subsumption The hierarchical tie of type-subsumption is crucial in Aristotelian syllogistics and in object-oriented design. Under the assumption that types are not extensional, subsumption cannot be accounted for in terms of coinstantiation and hence has to be accepted as primitive. Nonetheless one may conceive of subsumption as a form of overlap between types, i.e. as the subsuming universal's being contained in the subsumed one. Hence the attribute Blue is contained in the subsumed attribute Turquoise and the kind Insect is contained in the subsumed kind Ant. Note that the direction of containment between types is opposite to the direction of the inclusion between their extensions. Thus Blue is contained in Turquoise, but the extension of Turquoise is a subclass of the extension of Blue. For the sake of generality, I will make but few assumptions as to the formal structure of subsumption. First, I will assume that it is an ordering relation between types of the same adicity (kinds being types of adicity 0). Furthermore, subsumption and attribution (i.e. characterisation on the level of types) interact with each other in the sense that attributes of the sub-
200
sumer are inherited by the subsumee. Hence if Sixleggedness is an attribute of the Insect and the Ant is a subkind of the Insect, then Sixleggedness is also an attribute of the Ant. 2 Predication in the light of the Ontological Square 2.1 Two views of logic In order to motivate the impact of formal-ontological choices on formallogical ones, it may be helpful to distinguish between two perspectives on logic, namely the view of logic as language and the view of logic as calculus [48] [12, p. 118], and to clarify my position with regard to them. According to the view of logic as abstract calculus, a logical system is a formalism without a content of its own, but which can be given various interpretations over arbitrary set-theoretical domains [12, pp. 118–119, 123]. This conception is predominant today, not the least due to Tarski's [46, 47] initiation of model-theoretic semantics, which provides an account of logical consequence and logical truth that most logicians and philosophers find compelling [12, p. 123]. According to the view of logic as language, by contrast, logic amounts to formal ontology inasmuch as (atomic) logical form is meant to reflect ontological form, i.e. predication is supposed to mirror the various ontological nexus as ways in which items of the different ontological categories can stick together [12, p. 124] [11, p. 27]. Now, I agree with Bochenski [8], Cocchiarella [12, p. 123] that the view of logic as language is by no means obsolete, for the simple reason that the idea of a nexus as it underlies formal ontology in general and the Ontological Square in particular is not the same as set membership. Of course, this does not mean that set-theory and Tarskian semantics should be jilted, for they provide useful mathematical models of the intended formal ontology [12, p. 114], and are necessary tools for establishing soundness and completeness results. However, embracing the view of logic as formal ontology does mean to be guided by ontological considerations in the choice or design of a logical calculus. 2.2 Three assays of predication 2.2.1 Predicates versus copulae
201
If the view of logic as language is adopted, then, for the sake of a minimal correspondence between logic and reality, atomic logical form should mirror the structure of atomic facts. Because the Ontological Square involves ties or nexus as a distinct category of being, and atomic facts result of the combination of things by nexus, it would seem to be mandatory to rehabilitate the logico-grammatical category of the copulae as ties of predication. Indeed, since Frege [18, p. 30] [19, p. 54], the standard account of predication is that an atomic formula of first- or higher-order calculus results of the application of a predicate to a sequence of terms, where a predicate is a sentence-forming operator that takes names and returns a sentence and whose nominalisation denotes an attribute, i.e. a universal feature. The latter in turn is conceived of extensionally as a set and intensionally as a characteristic function mapping the referents of the terms into the pair {True, False}. Thus according to the standard view of atomic logical form, predicates act both as syncategoremata and categoremata, i.e. both as syntactical operators and as expressions having semantical content. In order to disentangle this ambiguity, Wiggins [51, p. 319], anticipated by Quine [33, p. 95], has slightly amended the standard view of predication as follows: type-names are primitive and predicates are derived from typedenoting names by means of a copula. According to this amendation, a copula is an operator that transforms a type-name into a predicate. Just as in the standard account, the unity of the proposition results of the application of an unsaturated predicate to one or more saturated terms yielding a sentence. By contrast, according to the traditional view of the structure of atomic propositions, which has been recently advocated by Wolfgang Künne [24], copulae are the actual nexus of predication that combine singular and general names into sentences. It is this traditional assay that is supported by a nexus-based ontology given a minimal logical realism: just as ontological nexus unify facts, copulas unify atomic sentences; copulas are the linguistic counterpart of ontological ties. 2.2.2 A comparative analysis using categorial grammar The comparison between the different assays of predication can be made
202
clearer by using a slightly adapted version of categorial grammar [2] [20, chap. 3]. Let S stand for the grammatical category of sentence; assuming that the two pairs of indices {+u, –u} and {+s,–s} represent the distinctions underlying the Ontological Square, namely universals vs. particulars and substrates vs. characters, the expressions N–u,+s, N+u,+s, N–u, –s, N+u,–s respectively denote names of particular substrates (i.e. substances), names of universal substrates (i.e. kinds), names of particular features (i.e. s) and names of universal features (i.e. attributes). Let C1,...,Cn and C be arbitrary grammatical categories, then C1,...,Cn / C corresponds to the category of operators that take items of C1,...,Cn and return an item of C. Such operators are referred to as syncategoremata; items of the categories, i.e. N–u,+s, N+u,+s, N–u, –s, N+u,–s and S, are called categoremata. According to the the standard account of first-order) predication, then, predicates are of the category: N–u,+s,..., N–u,+s / S i.e. they take names of substances and return a sentence; the unity of the latter is grounded on the saturation of the predicate by the substance names. Moreover, names of attributes are derived from predicates through the application of nominalisers (like the suffix „-ness“ in „Whiteness“), i.e. operators of the category : (N–u,+s,..., N–u,+s / S) / N+u,–s In the Quine-Wiggins amendation of the standard view, however, attribute names are primitive, and predicates are derived from the latter through the application of the copula (like the third person indicative present of the verb to be in \is white"), which is treated as an operator of the category: N+u,–s / (N–u,+s,..., N–u,+s / S) The traditional assay of predication does away with the intermediary step of transforming attribute-names into predicates and treats the copula in
203
first-order atomic formulae itself as a sentence-forming operator which combines an attribute name with one or several substances names into a sentence: N+u,–s, N–u,+s,..., N–u,+s / S This approach is certainly a great improvement, inasmuch as it is much more direct. However, it does not go far enough. On the one hand, the copula is mute in conventional predicate calculus, i.e. it is expressed by concatenation only. On the other hand, it exclusively expresses exemplification, i.e. the diagonal tying universal characters to particular substrates [40, p. 153]. More precisely, its ontological meaning is specific exemplification, which holds between a substance and an attribute iff the former has an instance of the latter as a moment. 2.3 A cornucopia of copulae for the Ontological Square Now, a logic that is faithful to the Ontological Square and its basic ties has to be much richer than ordinary predicate logic. It should not only provide names for all categories of things, including moments and kinds, but also four copulae instead of only one, namely: 1. the copula of instantiation, that ties an instance name to a type name: N-u, N+u / S 2. the copula of inherence, that ties a moment name to one or more substance names: N-u,–s, N–u,+s,..., N–u,+s / S 3. the copula of attribution, that ties an attribute name to one or more kind names: N+u,–s, N+u,+s, ..., N+u,+s / S 4. the copula of subsumption, that ties two type names together: N+u, N+u / S Accordingly, the copulae for generic and specific exemplification, which are both of the grammatical category N-u,+s,..., N-u,+s, N+u,–s / S can be introduced by definition.
204
The resulting calculus should be called a copula logic rather than a predicate logic, for the simple categorial reason that copulae are not predicates in the sense of standard predicate calculus. As argued above, copulae are mere syncategoremata, mere sentence-forming operators, that cannot be nominalised into general terms, i.e. names of types, nor derived from the latter, for they reflect ontological ties or nexus, not universals. Moreover, in contrast to conventional predicates, some copulae, namely the ones for inherence and attribution, are multigrade, i.e. have no fixed arity, since they hold of a character and arbitrarily many substrates or bearers. Hossack [23, pp. 46–47] has pointed out that the available strategies to paraphrase multigraduality in standard predicate logic fail to account for our ontological intuitions regarding ties. On the one hand, one may adopt the view that there is an inherence or attribution tie for each arity – bu this leads to a multiplication of inherence and attribution ties, while intuitively there is only one inherence or attribution tie. On the other hand, one could conceive of characterisation as holding between a feature and a tuple of bearers. This solution, however, would not do justice to the intuition that the feature (moment or attribute) is immediately tied to it(s) bearer(s). Finally, leaving aside ontological considerations, both approaches are arguably less perspicuous than the approach to atomic logical form represented by a copula logic. However, out of fairness it should be conceded that a satisfactory treatment of multigrade relations could be provided by a plural logic of so-called „vectors“ [23, pp. 47–55], i.e. plural terms that denote linearly ordered manifolds of arbitrary length. The decisive argument for distinguishing copula logic from predicate logic is the fact that predicates and copulae are to be analysed differently in terms of categorial grammar. Conclusion Thus a logic for the Ontological Square has to be designed as a copula logic similar to Lowe's [26] proposal of a sortal logic. As such it is equivalent to a calculus of restricted many-sorted second-order logic, inasmuch as any of its terms can be bound by a quantifier. A logic for the Ontological Square can be given a model that is similar to a first-order semantics in the sense of Shapiro [37, pp. 74–75] and with respect to which its soundness and completeness can be established. Further details regarding a formal calculus that is based on the Ontological Square as well as its meta-theory
205
can be found in my [36]. Bibliography [1] Ackrill, J. L., Aristotle's Categories and De Interpretatione, Oxford, Oxford University Press, 1963. [2] Ajdukiewicz, K., `Die syntaktische Konnexität’, Studia philosophica 1 (1935), 1– 27. [3] Angelelli, I., Studies on Gottlob Frege and Traditional Philosophy, Dordrecht, Reidel, 1967. [4] Angelelli, I., `Accidents III: The Ontological Square’, in [9], vol. 1, 1991, 12–13. [5] Armstrong, D. M., Universals. An Opinionated Introduction, Boulder, London, Westview Press, 1989. [6] Bacon, J., Universals and Property Instances. The Alphabet of Being, Oxford, Blackwell, 1995. [7] Bergmann, G., Realism: A Critique of Brentano and Meinong. Madison/WI, University of Wisconsin Press, 1967. [8] Bochenski, I. M., `Logic and Ontology’, Philosophy East and West 24 (1974), 275–292. [9] Burkhardt, H., and B. Smith, Handbook of Metaphysics and Ontology, 2 vols., Munich, Vienna, Philosophia, 1991. [10] Campbell, K., Abstract Particulars, Oxford, Blackwell, 1990. [11] Cocchiarella, N., `Conceptual Realism as Formal Ontology’, in [32],1996, 27–60. [12] Cocchiarella, N., `Logic and Ontology’, Axiomathes 12 (2001), 117–150. [13] Correia, F. Existential Dependence and Cognate Notions, Munich, Vienna, Philosophia Verlag, 2005. [14] Davidson, D., `The Logical Form of Action Sentences’, in [15], 1980, 105–148. [15] Davidson, D., Essays on Actions and Events, Oxford, Oxford University Press, 1980. [16] Degen, W., `Das onto-logische Sechseck’, Grazer Philosophische Studien 32 (1988), 113–112. [17] Degen, W., `Accidents IV: The Ontological Hexagon’, in [9], vol. 1, 1991, 13–15. [18] Frege, G., `Function and Concept’, in [21], 1960, 21–41. [19] Frege, G., `On Concept and Object’, in [21], 1960, 42–55. [20] Gardies, J.-L., Esquisse d'une grammaire pure, Paris, Vrin, 1975. [21] Geach, P., and M. Black, M. (eds.), Translations from the Philosophical Writings of Gottlob Frege, Oxford, Blackwell, 1960. [22] Goldin-Meadow, S. et al., `The natural order of events: How speakers of different languages represent events nonverbally’, Proceedings of the National Academy of Sciences 105 (2008), 9163–9168. [23] Hossack, K., The Metaphysics of Knowledge, Oxford, Oxford University Press, 2007. [24] Künne, W., `Properties in abundance’, in [44], 2006, 249–300. [25] Landesman, C. (ed.), The Problem of Universals, New York, London, Basic Books, 1971.
206
[26] Lowe, E. J., Kinds of Being. A Study of Individuation, Identity and the Logic of Sortal Terms, Oxford, Blackwell, 1989. [27] Lowe, E. J., The Four Category Ontology. A Metaphysical Foundation for Natural Science, Oxford, Oxford University Press, 2006. [28] Martin, C.B., `Substance Substantiated’, Australasian Journal of Philosophy 58 (1980), 3–10. [29] Moreland, J.P., Universals, Chesham/Bucks., Acumen, 2001. [30] Mulligan, K., P. Simons and B. Smith, `Truth-Makers’, Philosophy and Phenomenological Research 44 (1984), 287–321. [31] Parsons, T., Events in the Semantics of English: A Study in Subatomic Semantics, Cambridge/MA, MIT Press, 1990. [32] Poli, R., and P. Simons (eds.), Formal Ontology, Dordrecht, Kluwer Academic Publishers, 1996. [33] Quine, W. O., Word and Object, Cambridge/MA, MIT Press, 1960. [34] Reicher, M.E., and J.C. Marek (eds.), Experience and Analysis, Vienna, öbv&htp, 2005. [35] Seargent, D. A. J., Plurality and Continuity: An Essay in G. F. Stout's Theory of Universals, Dordrecht, Martinus Nijhoff, 1985. [36] Schneider, L. `The Logic of the Ontological Square’, Studia Logica 91 (2009), 25 – 51. [37] Shapiro, S., Foundations without Foundationalism: A Case for Second-Order Logic, Oxford, Oxford University Press, 1991. [38] Simons, P., `Particulars in Particular Clothing: Three Trope Theories of Substance’, Philosophy and Phenomenological Research 65 (1994), 553–575. [39] Smith, B., `On Substances, Accidents and Universals: In Defence of a Constituent Ontology’, Philosophical Papers 26 (1997), 105–127. [40] Smith, B., `Against Fantology’, in [34], 2005, 153–170. [41] Stout, G. F., `The Nature of Universals and Propositions’, Proceedings of the British Academy 10 (1921). Reprinted in [25], 1971, 153–166. [42] Stout, G. F., `Are the Characteristics of Particular Things Universal or Particular ?’, Proceedings of the Aristotelian Society suppl. vol. 3 (1923), 114–122. Reprinted in [25], 1971, 178–183. [43] Strawson, P. F., Individuals: An Essay in Descriptive Metaphysics, London, Routledge, 1959. [44] Strawson, P F. and Chakrabarti, A. (eds.), Universals, concepts and qualities: new essays on the meaning of predicates, Ashgate, 2006. [45] Stumpf, C., Erkenntnislehre: Erster Band, Leipzig, Verlag Johann Ambrosius Barth, 1939. [46] Tarski, A., `Der Wahrheitsbegriff in den formalisierten Sprachen’, Studia Philosophica 1: (1935), 261–405. [47] Tarski, A., `The Semantic Conception of Truth: and the Foundations of Semantics’, Philosophy and Phenomenological Research 4 (1944), 341–376. [48] van Heijenoort, J., `Logic as Language and Logic as Calculus’, Synthese 17 (1967), 324–330. [49] Vendler, Z., `Facts and Events', in [50], 1967, 122–146.
207
[50] Vendler, Z., Linguistics in Philosophy, Cornell, Cornell University Press, 1967. [51] Wiggings, D., `The Sense and Reference of Predicates: A Running Repair to Frege's Doctrine and a Plea for the Copula in Frege', Philosophical Quarterly 34 (1984), 311–328. [52] Williams, D. C.,`On the elements of being', Review of Metaphysics 7 (1953), 3– 18 & 171–192. [53] Williams, D. C., `Universals and existents', Australasian Journal of Philosophy 64 (1986), 1–14.
208
Gustav Bergmann et les complexions meinongiennes BRUNO LANGLET & JEAN-MAURICE MONNOYER « You believe, not only that the phenomenological and the dialectical context must be clearly separated, but also that an accurate grasp of their interplay is indispensable for doing ontology. I agree. », Realism, p. 69
G. Bergmann a entretenu avec Meinong des rapports qui ne sont pas faciles à débrouiller. En 1967, Meinong est présenté comme un interlocuteur « historique » : ce qui explique le sous-titre de l’œuvre dédiée à la « glorious memory » du philosophe de Graz, Realism : a Critique of Brentano and Meinong (notamment dans les chapitres 18 et 20, le chapitre 18 reprenant la thèse du chapitre 13 et la condamnation du « réisme » chez les deux auteurs)1. Bergmann y garde ses distances vis-à-vis de Findlay2, le gardien de la doctrine, p. 417 ; il se livre à un examen presque exhaustif de ses premiers essais, et prioritairement de Sur les objets d’ordre supérieur, mais sans tenir compte aucunement de l’évolution de Meinong ; il discute même des changements entre les deux versions de Über Annahmen (1902-1910), à défaut de les expliquer vraiment. Comme l’a très bien montré Venanzio Raspa, Bergmann a lu Meinong « sans méthode interprétative », et il lui fait « violence »3. Bien que reconnaissant son « flair incomparable », Bergmann critique son style « diffus » ; il ironise sur la poursuite de ces « papillons phénoménologiques » qu’il chasserait pour son seul plaisir (III, ibid.. p. 342) ; de même, se plaint-il de sa terminologie « excentrique », sauf qu’il la réutilise et la dé1
: Collected Works, Vol. III, Ontos Verlag, 2005. : J. N. Findlay, Meinong’s theory of objects and values, Oxford Clarendon Press, 1963. 3 : Venanzio Raspa, « ‘ …The most memorable Don Quixote of a great cause’, Bergmann’s critique of Meinong », in Fostering the Ontological Turn, Rosaria Egidi, G. Bonino (eds), Ontos Verlag, Heusenstamm, 2008, pp. 201-228. 2
tourne. Bergmann appelle la Relation de Meinong « connection », et se réserve justement le terme de « relation » (p. 344). Pourtant, la fin du livre adopte un verdict beaucoup plus nuancé. Bergmann révèle que sur le fond Meinong a compris la « structure relationnelle » présentée dans l’expérience. The most memorable Don Quixote of a great cause entraîne soudain son affection et son admiration (pp. 440-41). Le tout dernier § de Realism annonce effectivement une autre lecture, et paraît dessiner une forme anticipée de retour en grâce. Cela ne signifie en rien, selon nous, que Bergmann soit devenu ensuite à proprement parler « meinongien ». Mais il est vrai que certaines des thèses de New Foundations of Ontology (rédigé en 1977, et édité en 19924), ou développées dans son époque tardive, pourraient être lues comme proches de celles que Meinong a préfigurées ou comme des conséquences objectives de la pensée de Meinong. Nous suivrons donc l’hypothèse générale qu’a soutenue Rosaria Egidi (« Bergmann intends to carry out the Meinongian programm »5) : pour elle, la critique du « représentationalisme », qui suit celle du réisme et du nominalisme (un résultat où Bergmann dit lui-même que « Meinong [realist] came agonizingly close », p.340), ne pouvait que le conduire à revenir vers Meinong, un retour qui celui-là n’est plus historique. La position de Bergmann est intéressante et topique, indépendamment de sa sophistication. Elle se distingue de celles des autres interprètes qui ont entrepris par la suite de relire Meinong (R. Routley, R. Chisholm, T. Parsons, E. Zalta, R. Grossmann, K. Lambert, F. Nef, D. Jacquette). 1/ Le « Meinong » de Bergmann L’essentiel dans cette réhabilitation inattendue ne peut pas se dire en quelques mots. Ou plutôt, l’article lumineux qu’a donné E. Tegtmeier : «Meinong’s Complexes » pourrait servir de guide contre d’éventuels égarements dans cette « reconstruction rationnelle » de Meinong qu’a tenté de faire Bergmann6. On peut néanmoins entrer en matière par un autre biais, à cause simplement d’une formulation audacieuse et d’un revirement 4
: New Foundations of Ontology, Edited by William Heald, Foreword by Edwin E. Allaire, The University of Wisconsin Press, 1992 (=NFO). 5 : Rosaria Egidi, « Il Meinong di Gustav Bergmann », in C. Barbero, V. Raspa eds, in Il Pregiudizio a favore del Reale, Rivista di Estetica, XLV, Ns, n°30, Torino, 2005, pp. 54-70. Ce faisant nous ne donnons pas tort à l’examen minutieux des distorsions qu’a relevées V. Raspa en défenseur vigoureux de Meinong. 6 : E. Tegtmeier, « Meinong’s complexes », The Monist, 83 :1, 2000, pp. 89-100.
210
qu’effectue Bergmann dans les « early seventies », comme il le déclare très ouvertement devant H. Hochberg. Le point n’est pas seulement exégétique et terminologique. L’affaiblissement tardif du rôle crucial qu’aura joué, dans l’ontologie de Bergmann, le nexus d’exemplification — lequel se transforme du tout au tout dans les New Foundations, puisqu’il n’est plus un « constituant » comme il l’était encore dans Realism7 —, est solidaire d’un changement de « prototype » et des nouveaux canons de l’ontologie bergmannienne. Ce changement l’entraîne à l’adoption des fonctions dites « meinongiennes » (les M-Functions) (NFO, p.131) : un « tribut » que Bergmann veut rendre « à cet esprit tortueux, aussi profond que subtil » (id.). Il désigne ainsi, en opposition aux fonctions « frégéennes » insaturées, ces fonctions qui conservent une relation basique avec des existants, par deux actes simultanés de prise de conscience directe de leur diversité. Ainsi renommées et correspondant à un meaning-nexus, les fonctions meinongiennes engageraient donc à l’existence des complexes, leurs fondements — entre les « choses » (les faits), et les « classes » — , mais en un sens il est vrai exactement contraire à celui que Bergmann a longuement défini dans Realism, quand il faisait la critique de la « complétion » de la complexion. Que s’est-il passé ? Bergmann n’avait-il pas opposé l’ontologie des fonctions et l’ontologie des complexes ? H. Hochberg avoue qu’il lui semble que cette nouvelle ontologie est « baroque ». Une propriété du « contenu » formerait une circonstance donnée, couplée avec son intention ou son Objectiv. Les déterminés (couleurs, sons) s’accrocheraient ensemble (clinging to), hors du making de l’exemplification pour constituer un complexe. On pourrait certes, comme l’a indiqué Venanzio Raspa, suspecter d’emblée l’usage très libre des mots : « Komplex » et « Relation », qui avaient été transposés de l’allemand en anglais, et qui ne retrouvent pas ici leur vrai place. Mieux vaut s’en tenir à l’illustration directe qui est proposée de la M-Function. Par exemple, une quinte entre deux notes existe sans conjonction, sans terme supplémentaire. L’exemple peut sembler trop pauvre pour explorer les Two-in-ones (inseparable compound of two simples), mais Bergmann — comme d’ailleurs Meinong — y reviennent souvent8. 7
: H. Hochberg, The Positivist and the Ontologist, Bergmann, Carnap and Logical realism, Rodopi, Amsterdam, 1999, p. 160. 8 : Il faut déjà noter la différence avec Realism où la tierce et la quinte sont des universaux (Ontos Verlag, T. III, p. 344). Mais on peut ensuite lire chez Bergmann une longue analyse de ces questions qui rejoint la problématique des « tropes », puisque l’exacte similarité entre deux paires de quintes différentes se distingue de la
211
Si ces M-functions jouent un rôle d’abord assez allusif dans l’économie générale d’une ontologie réaliste, elles sont donc en affinité étroite avec ces circumstances mystérieuses, qui sont effectivement présentées ailleurs, et principalement dans Notes on Ontology (1981), comme devant se substituer à la notion du non-relational Tie. Les circonstances « se fondent » dans le complexe tout en demeurant des éléments du complexe sous un autre rapport : celui de leur « diversité » (p. 139). Si nous comprenons bien, chaque note est une circonstance désignée (telle note), l’intervalle de quinte est lui aussi une circonstance, l’intention mentale qui vise la propriété de relation complexe : « être une quinte », en est une autre9. L’indication diacritique du « M » dans les Notes on Ontology qui signale le dernier mot de Bergmann sur le sujet, semble également être une extrapolation des M-functions. C’est une indication infiniment plus souple ; elle ne tient lieu de « rien d’existant dans le monde » : ses arguments sont ceux d’une « fonction de signification » (p. 331), comme il en va dans le Tractatus pour le sich verhalten des faits « subsistants » qui ne sont pas présentables dans le langage idéal. Cette fonction ne réfère pas, mais elle peut aussi valoir cependant pour les dyades de toutes sortes réunissant un universel et un particulier : elle apparaît donc assez éloignée, à première vue, de cette allusion qui est faite cursivement dans New Foundations aux « objets d’ordre supérieur » comme à leur modèle structural (NFO, 131). Des fondements (les notes) au fondé (la quinte), Bergmann dit alors, suivant ici directement Meinong, « si les premiers sont là, la seconde est là eo ipso ». Quand l’on y regarde de près, il est clair cependant que les présentations signifiantes sont d’abord intentionnelles (puisqu’il y a une dyade séparable
9
similarité entre deux quintes obtenues à partir de deux notes identiques. : Ce mode d’accrochage oblique (« slantwise ») des 3 plans dépend de ces deux sortes fondamentales d’entités que Bergmann isole dans sa dernière période : les determinates (indépendants et séparables), et les subdeterminates, qui ne sont pas séparables et ne sont pas explicitement des objets de l’acte intentionnel. Les subdeterminates ne peuvent être visés que slantwise, dans leur union avec d’autres entités formant un complexe qui est signifié par un acte intentionnel. Les déterminés sont des choses qui ne sont pas absolument simples, des complexes ou des classes. Les entités absolument simples sont sous-déterminées ou sont des subsistants, tels les connecteurs et les quantificateurs dans le langage idéal (Bergmann, 1981, Ontos Verlag, Collected Works, Vol. II, p. 322). Dans l’exemple donné, les circonstances sont imbriquées dans le complexe sous-déterminé par ces dernières.
212
et une fonction inséparable d’elle, de même que chez Meinong il y a une différence entre le complexe donné et la pensée du complexe factuel qui le fait subsister), en sorte qu’une problématique déterminée du fondement (foundation) conditionne le mode de la liaison en dehors d’une relation linéaire ou temporelle10. Plus subtilement encore, Bergmann ajoute que le meaning-nexus est lui même une nouvelle sorte de circonstance. Mais il faut d’abord revenir à la critique de Meinong par Bergmann dans Realism avant d’anticiper sur l’occurrence d’idées meinongiennes dans New Foundations. La position du concept de nexus, dans l’œuvre majeure de 1967, jouait un rôle décisif en regard des points de l’ontologie de Meinong sur lesquels Bergmann faisait porter sa critique. Le nexus devrait occuper en principe dans Realism le rôle d’un « fondement » ontologique des complexes (grounding) : mais c’est un fondement subsistant (pp. 7-9), qu’il considèrera justement ensuite comme un « sous-déterminable ». De plus, en tant que « constituant », comme on l’a vu ci-dessus, le nexus n’est pas un composant. Il remplace la forme de collection propre aux fonctions qui ne peuvent pas revendiquer un tel statut et ne seraient rien d’autre que des règles de « coordination »11. Le nexus nous dispense d’une autre liaison pour être coordonné aux autres constituants du fait, alors que ces derniers ont toujours besoin de lui. Dans cette définition minimale, il n’y a rien en droit qui « supporte » les complexes réductibles à des circonstances ou assimilables à des faits. D’autant que les choses (ordinaires) sont des complexes improprement dits. En quoi le nexus ou le tie pourrait-il alors servir de fondement pour des complexes « authentiques » ? La question mérite d’être posée. Ce qui est sûr est que dans les New Foundations, les fonctions (meinongiennes) servent à rétablir la notion de « fait »12, alors que Bergmann avait admis que le monde de Meinong ne contenait pas de faits. Si les positions assumées par le Bergmann de la late ontology ne correspondent plus vraiment à celles qu’il avait critiquées chez Meinong dans Realism, notre hypothèse permettrait de penser que Bergmann a réincorporé à sa philosophie des traits meinongiens qui avaient été plus ou moins passés sous silence, voire même évincés de la reconstruction fournie 10
: Frederic Nef (dans ce même volume) jette une lumière nouvelle sur la différence entre « lien » et « relation ». 11 : G. Bergmann, Realism, The University of Wisconsin Press, 1967, p. 7. Selected Works II, Ontos Verlag, p. 7. 12 Ibid, pp. 128-129.
213
pas Realism. Nous ne prétendons pas offrir ici la seule clé de la lecture de l’évolution de Bergmann, car elle ne gouverne ni l’évolution du philosophe de l’Iowa, ni la permanence de certaines de ses positions. Son intransigeance sur la question des universaux et la critique persistante adressée à « Meinong-le-nominaliste » en sont la preuve. Comme on le verra, cette même question ne peut être rendue intelligible que sous l’éclairage de la pensée de Russell ; elle s’enracine aussi dans le rapport lui-même tortueux que Bergmann entretient avec Carnap. La réprobation de Bergmann dans Realism concerne bien, à coup sûr, le sujet le plus central : la question de la nature de la relation qui fait qu’un collectif « objectif » [Objektiv Kollektiv] se distingue d’un complexe réel. Meinong aurait une ontologie « inadéquate » à cet égard, d’abord à cause de son refus des universaux. Bergmann s’étonne ensuite de voir dans les solutions de Meinong une ontologie implicite de la fonction, dont il se surprend que le philosophe de Graz ne l’exploite pas, malgré sa finesse ontologique. Le caractère implicite de cette ontologie de la fonction la distingue déjà de l’ontologie de la fonction (explicite) de Frege. Comme le dit Tegtmeier, il est très douteux qu’elle soit fidèle à Meinong (id., p. 94). D’abord, si les complexes de Meinong sont des objets d’ordre supérieur, ils sont dépendants et incomplets (unselbstängig et unfertig) : ils restent à mi-chemin de leur actualisation et de leur remplissement potentiel. (Ainsi une hauteur de ton pour une note peut se combiner avec une autre hauteur de ton pour produire une autre quinte ou produire une tierce). Ensuite, leur défaut d’unité est patent : la manière dont Meinong envisage le « fait » que le superius soit là eo ipso dès que les inferiora sont présents, est obtenu dans Realism selon une conception de l’intégration opérée par des relations internes qui ne peuvent en réalité rien connecter du tout. On sait qu’elles n’ont de support ontologique en tant que relations que dans la « nature » de leurs termes. Que la tonique et la dominante se rencontrent dans la quinte, ne crée en rien, pour Bergmann, l’entité-quinte : ce sont là des relations externes à la hauteur pure des notes, et c’est pourquoi elles définissent, par exemple — anecdotiquement, mais au sens littéral du mot — le style musical hérité de Haydn. Qu’une quinte soit plus haute qu’une autre quinte, forme aussi, entre deux quintes, une collection, et même une classe. Mais les fondements qui nous sont présentés ne sont jamais que les quintes (les complexes donnés avec les notes), en plus des rapports subsistants entre les intervalles.
214
Dans New Foundations le renversement est total (presque « dramatique », souligne Hochberg) : les fonctions dites meinongiennes disqualifient les fonctions frégéennes et sont émancipées du modèle des relations internes. Plus important sans doute, ce que Bergmann y appelle le « principe de présentation », est doté d’une importance majeure en prenant de fortes résonances meinongiennes : ce principe énonce que tout ce qui fait l’objet d’une présentation « existe » (en incluant le Sein et le Sosein). Dédoublant le principe de présentation avec le principe d’acquaintance, hérité de la lecture de Russell — et d’après lequel toute acquaintance avec un universel est intimement solidaire d’une acquaintance avec un particulier, un item ou un « individuateur » — Bergmann pense aussitôt que les universaux doivent, eux aussi, faire l’objet d’une « présentation ». Il adopte ce faisant et transforme un principe meinongien, même si l’accusation de nominalisme envers Meinong persiste. 2/ Le problème de la relation « ajoutée » Pour comprendre ce changement de lecture, il nous faut d’abord souligner quel a été son point de départ. Dans Realism, Bergmann affirme que Meinong aurait prétendu fonder les objets de son ontologie en ajoutant un particulier spécial à une « collection de particuliers ». L’ajout en question est celui de la relation, qu’en l’occurrence Bergmann décide d’identifier comme un particulier de plus. Il est vrai que cette relation est désignée par Meinong comme ce qui « s’ajoute13 » (hinzukommt), mais de manière très caractéristique, c’est-à-dire lorsqu’un complexe est déjà unifié ou actuel ou lorsque que la relation « fait partie d’un tout » qui ne se résorbe pas en elle : bref, quand nous avons un complexe duquel les constituants sont réellement connectés et non pas seulement juxtaposés ou réunis14. 13
: « Über Gegenstände höherer Ordnung und deren Verhältnis zur inneren Wahrnehmung », Zeitschfrit für Psychologie und Physiologie der Sinneorgane, 21 (1899), pp. 187-272 : Gesammtaushabe, Vol II, ed. by R. Haller, Graz, Akademische Druck-u.Verlaganstalt, p. 391. Trad. Fr. G. Frechette, « Sur les objets d’ordre supérieur et leur rapport à la perception interne », in G. Frechette & D. Fisette (dirs), A l’école de Brentano, Vrin, 2007, p. 272. 14 : On retrouve dans Über Annahmen, cette même différence dans l’identification des complexes qui se démarque de la juxtaposition des composants. Les choses sont aussi beaucoup plus compliquées dans l’évolution de Meinong, car Meinong regarde d’abord à la fois le complexe [aRb] et la relation individuelle (a) R (b) comme des objecta. Il y a une relation d’équivalence logique entre l’objectif : que [aRb] existe et l’objectif : que (a) R (b) subsiste. Mais encore faut-il, selon Mei-
215
Au regard de la distinction existant entre un collectif objectif (cluster) — une chose ou une collection — et un complexe réel (complexion), Meinong est déjà plus proche de la problématique de Bergmann que ce dernier n’entend y consentir. S’ils traitent d’un problème apparenté, il reste que la « solution » de Meinong reste pour Bergmann inadaptée. Bergmann conteste la pertinence de « l’ajout », absolument impropre selon lui à fonder ontologiquement la distinction entre les deux types de complexités. Il soutient classiquement que si une relation est ce qui fait qu’un complexe « est » connecté, il faut expliquer comment cette relation se relie à ses relata. Or, elles ne peuvent faire cela par elles-mêmes : il faudrait toujours une relation supplémentaire, un connecteur de plus, pour justifier de la « relationnalité » de la relation (tel que s’exprime alors Bergmann). La solution de Meinong, exposée de la sorte, ne permettrait pas d’écarter le problème de Bradley : il est impossible d’y voir le fondement ontologique de la connexion adéquate entre constituants d’un complexe — soit un fondement qui serait ce qui « fait » que le complexe existe comme tel. Selon Bergmann toutefois, Meinong échapperait — en principe — à cette critique. La relation « ajoutée », pourrait recevoir le statut d’une fonction distincte si elle pouvait être considérée de manière séparée du tout qu'elle unifie en tant que fonction (simple), et donc en prenant les éléments du complexe comme arguments. Or, si le statut de la fonction permet ainsi d’échapper au danger bradleyien de la régression, c’est néanmoins précisément cette seule fonction comme « entité » (supposée toujours implicite chez Meinong) qui l’empêcherait d’avoir une complex-ontology, puisque la fonction devrait être « à elle-même » son propre fondement. D’une confusion (blurred grasp), on passe à un gap (une absence). Cette absence de fondement ontologique explique que Meinong échoue à connecter réellement les entités dans un « fait » : le nexus lui fait défaut, et seules les propriétés de ce dernier sont décisives pour obtenir un fait. Bref, si la seule « relationnalité » disponible pour Meinong est interne, c’est un échec selon Bergmann. Par conséquent, la relation supra-additive ou complexive entendue au sens de ce dernier — si l’on peut user de ce terme contradictoire nong, « que a soit dans la relation R avec b » — ce qu’il appelle justement Komplexion : celle-ci étant à la fois une réunion par la pensée et une conscience de cette réunion. Meinong oppose déjà Relation et Komplexion au § 5 de Sur les objets d’ordre supérieur.
216
pour tenter de la comprendre — ne saurait avoir de fondement ontologique adéquat (tandis qu’elle peut en avoir un chez Frege). Lorsqu’elle est capable en effet de colliger ou de lier et une fois qu’elle est entendue comme une fonction, elle perd son grounding et se soustrait au modèle pertinent de la complex ontology. En réalité, cette critique échoue selon nous pour une seconde raison. Elle échoue parce que Meinong ne se met pas en quête d’un fondement ontologique lorsqu’il met en avant l’ajout d’une relation caractéristique pour les complexions. Ces dernières sont présentées sous l’aspect d’objets d’ordre supérieur. Sur le plan strict du fondement, rien n’autoriserait à retraduire dans un autre idiome, un idiome fonctionnel, l’ontologie de Meinong. Dans Realism, Bergmann procède ainsi à une analyse très fine de l’ontologie que Meinong « devrait » assumer. Son verdict est que les particuliers peuvent bien s'agglutiner, mais sans jamais former un complexe intégré : « Let α be an object ; { a1,…, an}, the collection of its particulars. Meinong accepts the ultimate premiss that the collection as such does not exist. Thus he knows that unless he does something about it, he will be left without objects. This is his strength. His weakness lies in what he does. Roughly speaking, he adds a (n +1)st particular, of a very special status. But he fails to see that, even though the particular he adds exists if and only if α exists, α itself literally still doesn’t exist, even if the particular does 15».
Il est difficile de distinguer dans cette présentation si le particulier « au statut spécial » de Meinong est la relation elle-même ou l’objet d’ordre supérieur. Le résultat pour lui « ne fait pas de différence ». Il n’y aurait pas ici d’objet α (existant factuellement) qu’on puisse nommer complexe. La thèse de Bergmann est que Meinong s’en tient à une tentative de coordination entre des éléments et qu’il ne dispose pas ontologiquement de la possibilité d’avoir des objets. La relation serait quelque chose de superflu pour la collection des particuliers : elle serait « ontologisée » comme une internal connection (p. 338) Or, on ne saurait en toute rigueur lire cet ajout comme un geste gratuit de Meinong et comme s’il s’agissait pour lui de « construire » des objets. Ce n’est pas le fondement ontologique que re15
: Realism, p. 337.
217
cherche ici Meinong. Le caractère impressionnant et fouillé de la critique de Bergmann laisse apparemment peu de place pour une contestation de son bien-fondé. Il ne nous reste que la possibilité d’y voir une modification du sens de « l’ajout » dont Meinong fait mention. La reconstruction de Bergmann est telle cependant qu’elle écrase complètement les traits saillants du Meinong historique : la relation dont parle Meinong n'est pas une nouvelle entité introduite en vue d'unifier de nouveaux éléments « dans » une entité hétérogène. Elle ne vise pas à créer la connexion : elle est plutôt ce qui est appréhendé comme étant là lorsqu’il y a une complexion. Elle lie des constituants lorsqu’ils sont unifiés dans leur présentation et n’est donc pas ce qui permet de les unifier. L’ajout est analytique. Meinong va de la théorie de l’objet vers la postulation des objets d’ordre supérieur. Un complexe unifié ou intraconnecté est un objet d’ordre supérieur ; son contraire étant une collection objective. Les modèles inspirateurs et critiqués par lui d’objets d’ordre supérieur ont bien été (et demeurent en arrière-plan) les Gestaltquälitaten, depuis l’article Zur Psychologie der Komplexionen und Relationen (1891)16. Il est fort difficile de penser ce genre d’objet ou d’entités comme ne manifestant pas une certaine connexion où est présumée subsister en acte une complexion qui la fonde. Le fait qu’elle soit là est dépendant de ceci que ses éléments sont connectés de façon adéquate dans la « présentation » qu’on en forme. La preuve en serait que des ina-déquations sont d’ailleurs « possibles » de par la structure du superius, que ce soit dans l’imagination et la mémoire. Dans l’école de Graz, Rudolf Ameseder a développé cette optique de la « production » qui fait « apparaître » la présence de la connexion sous la forme du Realkomplex. Pourtant le Relat, dit-il (la relation ajoutée) n’est pas un objet de plus.17 Les objets d’ordre supérieur supposent qu’une part de leur être soit un « être-présenté » (relevant typiquement d’une Erfassung, dit Meinong). Ils sont des objets appréhendés, bien qu’il ne s’agisse pas de l’appréhension directe au sens que lui donneront ses objecteurs pour la déconsidérer empiriquement. Ce qui vient de leur présentation intentionnelle entre dans leur 16
: Nous avons montré cette différence entre la complexion gestaltique et le complexe ressemblant dans Jean-Maurice Monnoyer, « Formes de complexion, types de connexion », Philosophiques, Vol 26, n°2, août 1999, pp. 245-261. 17 : « Über Vorstellungsproduktion », in A. Meinong (ed.), Untersuchungen zur Gegenstandstheorie und Psychologie, Leipzig: Barth, 1904, P. 483.
218
constitution et ainsi s’apparente — dirait Bergmann — à un « pseudonexus ». Mais il ne s’agit nullement d’une constitution arbitraire ou psychologique de représentations mentales, même si la problématique est héritée des puzzles psychologiques qu’étudie Meinong dans le cours de ses analyses antérieures sur le nominalisme et les relations depuis les HumeStudien.18 Ce qui donne leur relief et leur couleur propre à ses positions l’est en vertu de son intérêt persistant pour l’abstraction et l’analyse : elle lui permet d’interroger sous plusieurs aspects différents la manière dont doivent être obtenus des « contenus » — de par leur nature et leur structure — et ce afin d’expliquer la possibilité des opérations qu’ils permettent. Des questions multiples suggérées par la notion débattue de Gestaltqualität est née une ontologie des objets d’ordre supérieur : elle suppose des fundamenta ou inferiora « donnant lieu » à la production nécessaire d’un superius. Cette opération sans doute ne se fait pas sans l’esprit, mais elle ne se fait pas non plus réductivement dans l’esprit : elle n’est jamais une représentation mentale « associée ». Il s’agit d’une sorte d’objet pour l’esprit, bien qu’il ait aussi des propriétés contraintes par l’imposition de qualités du réel, et qu’il acquière comme tel une existence objective. L’être de tels objets est dépendant de la possibilité de leur saisie, qui fait qu’à proprement parler ils sont des « être-objets ». C’est bien là ce que Findlay19 a pu appeler un fort tropisme phénoménologique, tout en ne voulant pas dire que Meinong fasse une phénoménologie descriptive : il est plutôt le premier à reconnaître que la dimension qu’on prétend exclusivement descriptive risque de devenir futile (comme il le constate chez Husserl lui-même) et qu’il faut y associer une dimension « dialectique ». Ces entités sont à l’évidence moins caractérisées par une « existence » superflue, au sens que Bergmann peut conférer à ce terme, que par ce que Meinong appelle primitivement « l’être-donné » (Gegebenheit)20, soit la propriété universelle des 18
: Hume Studien I. Zur Geschichte und Kritik des modernen Nominalismus Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften. Philosophischhistorische Klasse, vol. 78, Wien 1877 pp. 185-260. (Gesamtausgabe vol. I, pp. 172) ; Hume Studien II. Zur Relationstheorie - Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften. Philosophisch-historische Klasse, vol. 101, Wien 1882 pp. 573-752. (Gesamtausgabe vol. II, pp. 1-172). 19 J. N. Findlay, « Meinong the phenomenologist », Revue Internationale de Philosophie 27, pp. 161-177, 1973. 20 : Meinong, Über Gegenstandtheorie, in Untersuchungen zur Gegenstandstheorie und Psychologie, op.cit., p. 20. Gesamtausgabe II, p. 500.Trad. Fr. par J.-F. Courtine et M. de Launay, in A. Meinong, Théorie de l’objet et présentation personnelle, Vrin, 1999, p. 83.
219
objets d’être des objets.21 La différence d’avec le jargon de la « donation » est que ce soit des objets et d’eux seuls dont nous parlons. Bergmann, qui reconnaît le tempérament phénoménologique de Meinong, le traduit assez fidèlement en examinant ses propositions dans leur visée ontologique, tout en s’inquiétant de savoir comment l’on peut ainsi fonder l’être des objets qui peuplent son monde. Or les entités dont traite Meinong ne peuvent aucunement jouer le rôle de « constituant » ontologique que leur confère Bergmann. La relation chez Meinong est certes bien là : elle est ce qui se trouve être co-présent lorsqu’il y a un complexe intégré, mais elle est aussi ce qu’elle est par surcroît, comparée à une collection objective, précisément parce que cette relation fait défaut à la seconde. Cet ajout n’est rien que la conséquence analytique qui apparaît dans l’ordre d’une complexion, ou qui « doit » être là pour que l’objet soit « donné » en tant qu’objet unitaire ou comme l’un de ceux qui ne subsistent pas. L’objet révèle la complexion ou bien, si ce n’est pas le cas qu’une complexion soit obtenue, elle en fait son objet « inexistant ». La relation n’est donc pas, en résumé, une entité posée dialectiquement en vue de déployer l’ontologie d’un complexe. Cela reviendrait à pouvoir rendre raison par construction du caractère intégré ou dissociable d’un complexe, comme le fait Bergmann lorsqu’il souligne dans Realism qu’un complexe existe s’il se compose d’un universel, d’un particulier brut, et éventuellement de relations entre entités qui seraient des particuliers parfaits, plus un nexus d’exemplification qui les « lie ». Quand l’ensemble est ainsi inventorié en une série d’entités dont on décrit les rôles, l’ontologie se fait sous forme dialectique. Mais la manière dont Meinong déduit la présence de la relation en question ne repose pas sur ce genre d’approche : la relation « de plus » est présentée via le complexe donné dans une appréhension phénoménologique. Elle n’est en rien la condition ontologique qui fait qu’il y a une connexion, au sens de ce qui la produit. Meinong ne se place pas dans une optique où la position d’une entité est là pour expliquer dialectiquement ce qui doit être pour contribuer à faire exister une entité. La manière dont les objets sont appréhendés par l’esprit ne résulte pas du processus faisant que le superius surgisse — dans certaines conditions pré-existantes — « sur » des 21
: Ce point a été étudié soigneusement par F. Nef dans l’objet quelconque, Paris, Vrin, 1999.
220
inferiora. La complexion n’est pas produite ipso facto, a-t-on envie de dire. On comprend à cet égard que Bergmann puisse se référer plus tard à Meinong à propos des circumstances. Comme la connexion est toujours supposée pour que l’objet supérieur soit tel qu’il est en tant qu’il est, et parce qu’il est intended to, la seule disponibilité dialectique est de parler d’une certaine relation entre les constituants. Mais elle ne coïncide que « partiellement » avec la connexion, nous rappelle Meinong. Le geste dialectique qui marque la présence de la relation est là pour traduire le fait que l’entité avec laquelle elle se donne (c’est-à-dire par laquelle elle est pensée) existe dès lors que les constituants sont connectés. La Komplexion n’est pas là pour fonder ontologiquement la connexion elle-même. C’est bien ce qui explique que Meinong se dispense de l’introduction d’universaux. Si l’analyse nous rend présente une « relation » entre constituants, celleci ne se confond pas avec une disposition référentielle : en revanche, si le complexe n’était pas intraconnecté ou pseudo-connecté par la pensée et s’il n’était qu’une collection objective, la relation en question ne serait pas disponible pour l’analyse et elle ne serait pas présentée comme l’ajout caractéristique des complexes intégrés. 3/ Le problème de Bradley et la confusion des contextes Pour confirmer l’hypothèse précédente, considérons l’attitude de Meinong face au problème de Bradley. S’il pensait que la relation « ajoutée » était le facteur ontologiquement pertinent pour fonder la connexion, il aurait dû y renoncer, car loin de permettre de résoudre — dans l’ontologie — le problème de Bradley, elle le conditionne. La solution de Meinong n’est pas de le résoudre au moyen de la détermination de certaines entités (comme un non-relational tie, ou un nexus). Dans le cas de Meinong, le problème est évité parce que le statut de la relation ne requiert pas qu’il soit résolu. On sait que dans les textes visés par Bergmann,22 Meinong introduit le problème en question à partir de la « coïncidence » entre relation et complexion (dont le principe est écarté par Bergmann d’un revers de main23). 22
: Dans ce cas surtout, Über Gegenstände höherer Ordnung und deren Verhältnis zur innehren Wahrnehmung, op.cit. 23 : Précisément parce qu’il lui paraît faire fi du problème qu’il se pose en termes dialectiques : comment la relation peut-elle connecter ? On a vu que le principe de coïncidence signale que s’il y a relation, alors il y a connexion, mais c’est dans un
221
Elle signale une double dépendance : les relata sont dépendants envers la relation comme la complexion est dépendante envers la relation. Meinong refuse le caractère problématique de la régression à l’infini, ce n’est pas qu’il pense qu’elle n’est jamais vicieuse : il l’utilise lui-même comme argument dans d’autres textes 24. Pour garantir que la liaison en question est organisée et intégrative, Meinong affirme que les relata se tiennent dans ce que Armstrong appellerait une certaine « direction » de la relation. C'est ce qu'il exprime en disant que A et B sont en rapport avec R d'une manière précise, impliquant qu'il y ait une relation r' entre A et R, et une relation r'' entre B et R : ces relations r' et r'' pouvant éventuellement être identiques. Leur fonction est de garantir que A et B sont liés correctement à R, autrement dit, que la relation R « accroche » ses termes d’une manière propre à organiser le complexe comme il doit l'être. Meinong remarque que la même question se reposerait pour les nouveaux complexes relatifs qui sont formés par Ar'R et Br''R : eux-mêmes pourraient être supposés impliquer des relations supplémentaires. Le problème de la régression dès que posé, se trouve donc écarté. La relation en question possède des relata dont la nature est moins de jouer un rôle « constitutif » dans la connexion que de signifier indirectement que la complexion existe, ou qu’il y a un ordre dans lequel elle consiste. Cet ordre structural est le seul corrélat pensable de la connexion. Bergmann reprend tout autre sens que celui que cette proposition recevrait à l’intérieur de la dialectique bergmannienne. Voir par exemple Realism, p. 370. 24 : Comme par exemple dans « Abstrahieren und Vergleichen » (1900, Gesamtausgabe I, pp. 477-8) où il réfute un argument de H. Cornelius qui consistait à montrer que l’appréhension de propriétés — d’une manière voisine de la thèse basique du nominalisme de la ressemblance — suppose de détecter des ressemblances en vertu desquelles la propriété peut être obtenue. Meinong y soutient en effet que si l’appréhension suppose la ressemblance, appréhender le résultat de la ressemblance suppose que ce résultat lui-même soit donné en vertu d’une autre ressemblance, laquelle a lieu entre ce résultat et une autre, ce qui, pour donner lieu à l’appréhension ce nouveau résultat, suppose encore qu’une ressemblance de plus soit à nouveau l’opérateur de donation, et ainsi ad infinitum. Dans ce cas, la régression est considérée comme vicieuse. Dans Über Gegenstände höherer Ordnung, le fait que la régression soit réputée non-vicieuse implique quelque chose de très différent. G. Bonino, «Why There Are No Facts in Meinong’s World (according to Gustav Bergmann », Meinong Studien, Vol. II, Ontos Verlag, pp. 239-266, a tenté une reconstruction hardie dans la seule ontologie de Realism, pour justifier herméneutiquement de la position de Bergmann.
222
justement cette idée dans New Foundations, à propos des questions centrales de la diversité et de l’ordre internes aux « canons », dont la structure est toujours celle des circumstances. Meinong soutient que ces relations (ordre et diversité) peuvent bien être vues comme essentielles sans que cela ne fasse d’elles des parties constituantes de la complexion. Et il est clair que l’ordre des constituants peut être compris comme supposant ontologiquement une infinité de relations essentielles en ce qu’elles construisent la nature des constituants. Mais alors le tout de ces relations ne peut fonder ontologiquement la complexion dans sa diversité. Une autre possibilité est d’envisager que la démultiplication des relations trahisse une difficulté de la fondation ontologique, auquel cas nous n’aurons plus besoin de trancher l’alternative précédente. L’argument de Meinong consiste à dire que le caractère inadéquat de la présomption des relations supplémentaires ou constituantes se manifeste dans la « nature » de ce qu’elles tentent de relier : elles ne peuvent être conçues dans ce cas autrement que comme reliant d’autres relations. Elles ne peuvent plus jouer le rôle « constituant » que l’on est tenté de leur conférer lorsqu’on les multiplie. Meinong perçoit que ces relations supplémentaires ayant toutes pour termes une autre relation font disparaître l’objet. Ici se signale le point de retournement décisif : soit on comprend que la relation, pensée à partir de l’entité qu’est l’objet d’ordre supérieur, manifeste la connexion par le biais d’un ordre, soit on cherche à fonder ontologiquement cet ordre (et la connexion qui en est solidaire) en multipliant les relations. Le deuxième cas suppose un geste dialectique. Le premier traduit bien que Meinong ne pense pas la relation en question en termes de fondement ontologique. La relation R est comprise comme dépendant directement de l’ordre des parties constituantes. Cet ordre fait que le superius est là. Que certaines relations « supplémentaires » puissent être essentielles sans être constituantes traduit le sens meinongien (ou la métonymie inspirée du Daseinsfrei) affecté à la notion du « constituant ». Multiplier les relations signifie vouloir déployer à partir du donné l’ensemble des relations pensables : ce serait en faire des relations « réellement » constituantes dont l’exhaustion permettrait de fonder l’être du complexe, autrement fonder ontologiquement la connexion. Meinong choisit une autre alternative, dans laquelle on peut poser l’existence d’une relation, sans penser pour autant qu’elle « fonde » ontologiquement l’unité de
223
ce qu’elle relie. L’erreur de la démultiplication vicieuse procède bien du souci de dialectiser « ontologiquement » ce qui ne peut pas l’être au sein d’un contexte phénoménologique, comme le reconnaît finalement Bergmann avec ces M-functions. 4/
Dualité des complexes
Bergmann critique Meinong en affirmant de son « monde » qu’il est vide d’objets, que les choses n’y sont pas des complexes, et que s’il y a des complexes, ils ne sont pas connectés (ce ne seraient que des cryptoclusters). Mais la ligne d’interprétation précédente vaut aussi pour les « objets » ordinaires qui apparaissent comme des complexes auto-connectés, tels ceux que Bergmann assimile à des faits. Meinong prend en effet lui aussi pour exemple des objets « ordinaires », lesquels sont bien à certains égards semblables à des objets d’ordre supérieur (d’après la saisie que nous en avons) et qui sont pour cela stigmatisés par Bergmann. Cette relation qui « s’ajoute », même si elle est conçue par l’esprit, l’est à partir d’une certaine présentation et elle ne serait pas comme elle est si ce qui est présenté n’était pas déjà connecté. Tel est par exemple le sens de la proposition énoncée dans Abstrahieren und Vergleichen, qui souligne que le rapport d’inhérence entre A et B est toujours représentable comme une relation entre A et B25. L’inhérence traduit l’existence d’une connexion ou d’une implexion selon les cas (dans la philosophie postérieure de Meinong). Bergmann dirait alors qu’il ne s’agit que d’une exemplification du dispositif de la prédication. Mais en partant de l’appréhension d’un objet ordinaire avec ses propriétés, ce qui est dit appartenir à un objet peut être traduit par la pensée comme une relation entre l’inhérent et ce dans quoi ces propriétés « inhérent », à l’image du principe dit de coïncidence partielle présenté supra. Ce qui traduirait la forme de l’inhérence est, comme dans le cas que nous avons vu, la marque de l’intégration d’un complexe. Dans ce cas-ci non plus, elle ne le constitue pas. Dans un collectif objectif, il ne s’y trouve pas d’inferiora qui fondent un superius : telle « la croix rouge ». Si le superius n’est pas là, les inferiora sont aussi « inexistants » que lui. Pour Meinong, il y a toujours une différence entre appréhender une croix rouge, et appréhender les propriétés comme « être rouge » et « être en forme de croix ». Bergmann dans les New foundations affirme de façon très voisine, abandonnant son credo de l’annexion spontanée des qualités 25
: « Abstrahieren und Vergleichen », in op.cit., p. 469.
224
universelles par un particulier nu, que l’objet de la présentation est la « valeur » de la M-function. Elle serait inséparable des arguments qu’elle supporte. Comme cela ne signifie pas le besoin d’un engagement envers l’existence d’une substance ou d’un substrat, le complexe meinongien réellement « intraconnecté » reprend toute sa vigueur : nous voyons ses propriétés sous l’aspect de traits inhérents ; ils se laissent analyser comme une relation entre la propriété et l’objet, et le fait qu’une qualité inhérente puisse être interprétée comme une relation hors de toute structure prédicative est de nouveau pleinement dégagé. Sous cet aspect, l’être-ensemble (togetherness) reste une propriété phénoménologique sans avoir à fonder ontologiquement la connexion. La fonction, qui reprend du service, n’y est absolument pas vue comme une « règle de coordination » anthropomorphique. Meinong admet du reste une autre forme de complexe, appréhendé cette fois par la relation qui se tient entre ses constituants, tout en excluant qu’il soit possible de lui conférer le statut d’une fonction séparée. Il s’intéresse à des complexes traditionnels comme « x ressemblant à y » ou encore « a différent de b », qui ont bien un contenu phénoménologique : comment leur contenu principal (ici la relation) est-il lié à ses constituants (les relata) ? On ne peut nier que cette question, là encore, ait fait l’objet de toute l’attention de Bergmann, après le silence qui suit Realism. Faudrait-il penser en somme que la relation de ressemblance doive être reconstruite comme une entité constitutive du complexe ? Meinong parle en effet de l’objet qu’est la relation, et surtout de l’objet d’ordre supérieur qu’elle représente. Il est vrai que l’on ne peut la séparer de ses constituants, en faire une fonction détachable qui prendrait ou ne prendrait pas des arguments. Meinong lui prête en effet une sorte d’insaturation ou d’incomplétude interne (innerlichen Unfertigkeit)26. Il y a bien ici une dépendance de ce qui est fondé en regard de ce qui fonde. Mais Bergmann la conçoit, tout au contraire, comme la marque d’une absence de fondement ontologique. Puisque la fonction implicite — qui tient lieu de la relation — est elle-même dépourvue de fondement, elle ne connecte rien. Il faut faire remarquer néanmoins que la dépendance « interne » que met en avant 26
: « Beiträge zur Theorie der psychischen Analyse », Zeitschrift für Psychologie und Physiologie der Sinnesorgane. Bd.VI. 1894. Reproduit dans Gesamtausgabe I, p. 323.
225
Meinong n’épuise pas l’être de l’objet d’ordre supérieur : il est très possible de concevoir que « l’être différent » soit inséparable de a et de b. En tant que complexe intégré, c’est encore à travers la présentation qu’on en a que Meinong valide l’identité de ces complexes fondés sur les « natures » de a et b, lesquelles permettent qu’un objet d’ordre supérieur soit appréhendé eo ipso (la formule est, on le sait, directement inspirée d’Ehrenfels). Mais si l’absence de fondement ontologique que mentionne Bergmann était pertinente, il n’y aurait nulle différence entre le complexe « a est différent de b » et « a et b ». Cette exigence meinongienne de différenciation a d’ailleurs été très largement revalorisée par H. Hochberg à la suite de Bergmann. Bergmann affirme dans Realism que les fonctions ou bien sont à ellesmêmes leur propre fondement ontologique, ou bien « représentent » ce qui par ailleurs serait leur fondement ontologique. Frege rentre dans le second cas, Meinong dans le premier. Mais pour paraphraser le Bergmann de New foundations, et si l’on tient à la désigner ainsi, la fonction meinongienne a ceci de propre qu’elle n’est jamais dissociable et de sa valeur et de ses arguments. D’où son renouveau d’intérêt dès que la puissance connectrice du tie est battue en brèche27. 27
: Ce qui importe ici est moins la dialectique de la relationnalité que l’obtention d’un fondement qui postule une connexion sans renvoyer à une entité cachée. Ce trait analytique est meinongien et relève de la mise entre parenthèses du rôle dialectique du « tying ». Dans NFO, l’échange entre Bergmann et son interlocuteur « relationnaliste » le montre bien : plutôt que de chercher ce qui fonde (grounds) « l’accrochage » (clinging) (p. 128) de la fonction et de l’argument, le relationniste est invité à considérer qu’il n’y a pas de connexion supplémentaire, et donc pas de fondement de plus (specific ground) pour obtenir ontologiquement le complexe. Accepter la circonstance eo ipso de ce même « clinging » revient à le fonder. Pour Bergmann alors, et assez brutalement : « La ‘connexion’ devient ce que le canon dit qu’elle est » (p.132). Bergmann associe l’absence de représentativité et la nécessité du fondement de façon paradoxale, à moins de ne lui accorder que l’inscription des circonstances dans le langage idéal ne se fasse sur une base « phénoménologique ». Il accepte la thèse wittgensteinienne selon laquelle « a =b », par exemple, ne représente rien du monde (elle réfléchit une « règle linguistique »). Mais ce qu’il nie, est qu’elle n’ait pas de statut ontologique : transcrire des intentions (des objets connus) dans le langage idéal ou canonique ne peut pas se faire uniquement en vertu de « règles linguistiques », il faut regarder le monde (Notes on Ontology, p. 145) pour obtenir et retranscrire les différences entre complexes. Inscrire les signes diacritiques dans le langage idéal reviendrait à cerner le concept de « circonstance » : certaines formes de connexité sont données en « personne propre » avec les constituants connectés.
226
L’objet d’ordre supérieur est dépendant des termes qui le fondent, mais il ne se réduit pas, avons-nous dit, à une relation interne, ni aux représentation mentales que l’on associe à une relation. Bergmann, en vertu de l’absence de fondement ontologique propre aux relations internes, doit alors inscrire une nouvelle entité à son inventaire lorsqu’il s’intéresse à cette sorte de complexes « meinongiens », qui ne sont pas des complexions au sens strict. C’est le constat qu’il prononce dans la conférence de 1968, Diversity. Pour rendre raison de l’aspect phénoménologique de la différence existant entre le vert et le rouge, Bergmann admet des complexes primaires comme « l’être-différent de a et b » : il ne nie pas que « l’êtredifférent » fasse l’objet d’une présentation. En étant présenté avec le rouge et le vert, mais aussi avec leur « être-divers », il y a maintenant quatre entités : le rouge, le vert, « l’être-divers », et le complexe dont ces trois entités sont les constituants. Nous avons « l’être-différent », en plus des deux entités qui sont différentes et du fait qu’elles soient différentes. Ce qui fait quatre entités. Pour Meinong le fait que ceci et cela soient différents est identique à l’appréhension de leur différence. Cette position radicale de sa part semble n’avoir pas été maintenue par Bergmann : dans New Foundations, il affirmera de nouveau — comme Meinong — que si nous sommes en rapport phénoménologique avec « l’être-ensemble » (togetherness) de a et b, nous désignons par là même le rapport que nous entretenons avec un fait. Le cas de ces complexes primaires révèle que la lecture des relations internes jadis opérée par Bergmann entre en conflit avec la phénoménologie du « tournant » qu’il souhaitait prendre. La reconsidération des fonctions meinongiennes et la manière dont Bergmann les distingue des relations internes dénote chez lui un mouvement conceptuel important. Il ne s’agit plus de dire qu’une fonction représente une relation. 5/ Universaux : Bergmann, Russell et Meinong. Une critique constante de l’ontologie de Meinong est le caractère nominaliste de ses positions. Or, pour Meinong, la façon dont un complexe est donné ne requiert en rien de théoriser la nature universelle ou non de ses propriétés. Ce qui ne signifie pas que l’on doive oblitérer ce rapport : les problématiques sur les objets incomplets apparaissent déjà dans Abstrahieren und Vergleichen. Mais même leur développement tardif, comme par exemple dans Uber Möglichkeit und Wahrscheinlichkeit se produit sur un mode tel qu’il est difficile de les assimiler aux universaux de la tradition
227
ou à ceux de Bergmann. Hormis ces approches par le biais des objets incomplets, la perspective de Meinong est qu’il convient plutôt de déterminer comment sont séparables et analysables des qualités diverses, et comment elles se partagent en inferiora et en relations. Pour Meinong, les propriétés sont d'emblée des propriétés particulières, parce que c'est ainsi qu'elles nous apparaissent : elles sont localisées dans un complexe. Il ne fait pas de différence pour savoir si elles sont universelles ou simplement ressemblantes. Elles sont aussi bien connectées dans un fait que ne le sont les entités de Bergmann, mais Meinong adopte une ontologie propre dont l’aspect dialectique est très fortement freiné par le régime de l’appréhension qui gouverne ses recherches. Sa phénoménologie réaliste est aussi la raison (curieuse) qui veut que Meinong soit jugé d’abord par Bergmann comme « nominaliste ». Meinong rappelle certes dans Abstrahieren und Vergleichen que le chercheur qui découvre que le diamant a la propriété d’être consumé en raison de sa complexion carbonique, opère sur tel « morceau particulier » de diamant. Cela n’empêcherait pas de penser que toute entité qui a les mêmes propriétés que ce morceau de diamant présenterait la même disposition à se consumer. Mais cela ne requiert pas non plus de poser un universel instancié dans un particulier brut. On n’a pas ici besoin d’un universel et du geste dialectique qui lui est associé pour construire l’objet complexe structuré par le carbone. Si cette différence est préservée entre Bergmann et Meinong, malgré l’évolution du premier, elle souligne plutôt un décalage du rapport entre contexte phénoménologique et contexte dialectique dans leurs perspectives réciproques. Les rapports du mode phénoménologique et du mode dialectique seraient bien décisifs en ontologie, commandant une partie des raisons qui motivent l’introduction d’entités ou qui la limitent. Est-ce à dire que le motif phénoménologique interdise l’appréhension d’universaux ? Cela reviendrait à dire que la postulation d’universaux est essentiellement dialectique. Mais ce n’est pas le cas. Ce problème, comme celui de l’individuation des propriétés qui en est le corollaire, relève d’une indécidabilité foncière. Sur les problèmes précédemment évoqués, où les critiques de Bergmann peuvent être relativisées en regard de l’approche meinongienne — et qui semble avoir eu, si nous lisons bien, une deuxième influence sur l’ontologie finale de Bergmann —, la question des universaux demeure quasiment inchangée pour sa part. Bergmann, à l’occasion de la publication de An Inquiry into Meaning and Truth de Russell, pou-
228
vait formuler une exigence, qui se maintiendra jusque dans NFO sous la forme d’un principe. En effet, déjà dans Russell on Particulars, il s’oppose à la théorie du faisceau d’universaux de Russell qui aurait le défaut de libérer les universaux des particuliers qui les « localisent » lorsqu’ils sont instanciés. Bergmann remarque : « There is, first, a difficulty concerning the psychological, or as I would rather say, the phenomenological status of Particulars. As it is usually put, the point is that the content of even the simplest awareness is never one or, in the relational case, several quid, but always, wether property or relation, a quale, that is, a Universal. Against such views one could defend the unqualitied particular by pointing out its indispensability in fixing the location of qualities ».28
Ce point marque un retour de Bergmann à l’acquaintance avec les universaux revendiquée par Russell, mais celle-ci est assortie de la nécessité de ce qui apparaît comme la première formulation du particulier brut, ici un « unqualitied particular », plus tard l’ « individuator » de New Foundations. Dans ce dernier ouvrage, nous retrouvons, plusieurs décennies après la formulation de ces quelques lignes, l’énoncé du principe dit « Principe d’Acquaintance » : « No act intending a universal, say, f1, occurs in a life history unless there is a particular, say, a1, such that f1(a1) is also intended, either by this act or by a second act, either simultaneous with or preceding the first, that occurs in the same life history ».29
Bergmann précise bien qu’il n’est pas identique avec le Principe de Présentation, selon lequel tout ce qui fait l’objet d’une intention existe, i.e : tout ce qui est présenté existe. Ce dernier principe est on ne peut plus meinongien. Le premier est russellien, modulé par l’exigence de la présence corrélative d’un particulier. Dans Ineffability, Ontology and Method, Bergmann présente cependant l’acquaintance et la présentation comme synonymes : « Each of us is acquainted with some things and facts (states of affairs). Synonymously, these things and facts are presented to us ».30 28
: Russell on particulars, 1947, in G. Bergmann, The metaphysics of logical positivism, Longmans, Green and co., 1954, p. 200. 29 : New foundations of ontology, p. 65. 30 : The Philosophical Review, Vol. 69, No. 1, Jan., 1960, p. 18.
229
Or la présentation et le principe d’Acquaintance sont dissociés dans New Foundations d’une manière telle que, paradoxalement, le principe meinongien de présentation « croisant » le principe russellien d’Acquaintance joue contre Meinong, en affirmant que nous sommes en rapport non pas avec des propriétés particularisées, mais avec des universaux qui sont individués par des particuliers qui n’ont pas d’autres qualités. Cette acquaintance avec un quale, c’est-à-dire avec un universel, est entrée très tôt dans la pensée de Bergman pour n’en plus sortir, malgré quelques modifications dont nous avons montré qu’elles s’effectuaient en profondeur entre Realism et New Foundations. Le « tournant phénoménologique » mentionné au chapitre II des New Foundations ne conditionne pas la mise à l’écart des universaux. Cette position de Bergmann remonte à ces rapports critiques avec Carnap sur la nature des « intensions » que nous ne pouvons développer ici. Nous avons parlé ci-dessus d’une forme d’indécidabilité de la question de l’acquaintance avec des universaux (doivent-ils être appréhendés sur une base phénoménologique ou sur une base dialectique ? et n’est-ce pas un faux débat). Ce qui est certain dans les New foundations of ontology, c’est que la confirmation de l’engagement envers l’existence des universaux et la modification de la problématique de la constitution feront encore bon ménage. Une dyade unit eo ipso un individuateur et une « sorte » ultime, en lieu et place d’un nexus assurant l’exemplification d’un universel par un particulier brut. L’ontologie du lien a changé, non pas chez lui l’ontologie des constituants. La résurgence de thèmes meinongiens n’amène donc pas à un renoncement aux universaux. Russell reste beaucoup plus influent que Meinong sous ce rapport. Sur le plan interprétatif, il resterait à prouver dans un autre travail que la position de Meinong supporte le caractère « indécidable » de la postulation d’universaux in rebus. 6/ Fondement ontologique et « phenomenological turn » Le critère de la réapparition dans la late ontology d’aspects meinongiens qui étaient parfois complètement absents de la critique de la middle ontology, révèlent-ils certains des traits « structurels » de l’évolution de la philosophie de Bergmann ?
230
D’une part, les traits de la pensée de Meinong qui résistaient à la critique de Bergmann en 1967 apparaissent maintenant comme des positions positives au sein du système de 1992. Dans Realism, ce qui se passe peut être décrit comme une traduction dans l'ontologie pure de ce qui chez Meinong ne lui est pas exclusivement redevable. Cela indique une priorité dialectique pour le Bergmann de l’époque : elle ne pouvait se traduire que par l’affirmation de la défaillance de l’ontologie de Meinong. Au contraire, les aspects que nous avons tentés de mettre en évidence (comme les ressorts du refus chez Meinong de la régression de Bradley, le mode d’intelligibilité que les objets d’ordre supérieur offrent pour la connexion, le sens de la relation dans les complexes et complexions) révèlent d’autres traits qui échappent à la critique de Realism et sont analogues avec certaines positions décisives de New Foundations. Sous cette dimension nouvelle que les positions de Meinong pourraient recevoir à l’encontre de la machinerie de la middle ontology (le monde de Meinong étant incompatible avec le foil), il n’est alors plus vraiment surprenant de cerner dans New Foundations l’intérêt que reçoivent les MFunctions. Cette apparition coïncide avec une dépréciation du tie, jugé finalement absurde et phénoménologiquement problématique : « Is one, when explicitly presented with (α, β ), implicitly presented with its terms only or also with a third existent, either a tie or whatever else it may be ? By the uncontestable witness of the given, there is no third. (….) (α, β) needs no other grounds than α and β themselves and their each being what they are ; or, as I shall also say, a diad needs no other ground thant the « identities » of its terms. Still differently, α and β being there, ( α, β ), is eo ipso there ».31
Bref, c’est la valeur de la fonction qui fait maintenant l’objet de la présentation phénoménologique. Rappelons-nous dans Realism, le désastre des function ontologies tenait à ce qu’elles étaient à elles-mêmes leur propre fondement ontologique. Dans New foundations, les fonctions meinongiennes sont présentées comme « la clé de la notion du fait » : « …the fact is neither the collection, nor the class, nor the « sum » or anything else « of » the first and the second, but rather the third which is eo ipso there if and only if the first and the second are there. Or, with a twist, the nature of the fact is completely exhausted by the first satysfying the restriction which goes
31
: New foundations, p. 68 et p. 105.
231
with the second and by the second being inseparable from the first. That is the key to the notion of fact ».32
Il y a bien une relativisation du besoin de fondement ontologique pour certaines entités dont il était précédemment essentiel qu’elles en possèdent un. Dans Realism, le tie marquait ontologiquement la différence entre le complexe formé par telles qualités de rondeur et de rougeur qui constituent un spot, et le complexe formé par ces qualités concurrentes d’ovalité et de bleu qui ne constituent rien en tant que telles, en dehors du spot. La compréhension du fondement ontologique de cette différence est présenté dés l’ouverture de Realism comme le sujet de l’ontologie33. Le nexus était l’entité-solution pour la question du fondement ontologique des faits. Bergmann soutenait déjà pourtant que l’idée d’une différence de degré dans le mode de présentation phénoménologique implique une différence catégorielle dans l’ontologie. On le voit dans la manière dont il répond à un objecteur phénoménologue imaginaire dans Realism, celui-ci pense que le lien entre ontologie et phénoménologie est trop fort : il exige des présentations phénoménologiques auxquelles il est difficile de souscrire, comme une présentation « multiple » des entités. L’objecteur affirme que ce que propose Bergmann rend la croix phénoménologique trop lourde à porter : « it is phenomenologically absurd, as absurd as anything could be, to maintain that, the two things e and c being what they are, the former would not be higher than the latter unless there were also, in addition to these two, a third thing, of the kind you call relational, and still another entity, of the kind you call nexus, and unless this fourth entity tied the other three, which you hold to be « completely external » to each other, into a « fact ». 34
A cette objection Bergmann rétorque qu’il n’en est rien, et que la phénoménologie en question est scandée par autant de différences et modulations « in feel ». Il affirme : All this is phenomenology without torture : « I distinguish sharply between the phenomenological and the dialectical context. Sometimes, when simultaneously presented with two pitches, I am also, strikingly, lucidly, and without any torture, presented with the interval as well as 32
: Ibid, p. 129. : Realism, p. 6. 34 : Ibid., p. 68. 33
232
with the fact that it connects (obtains between) the pitches. To insist on that, as I do, is not to deny, as I don’t, that the several entities presented of this sort differ in « feel ». (…) One might wish to say, for instance, that while a pitch is more « palpable » than an interval, the latter is in turn more « palpable » than exemplification (…) Dialectically, my assay reflects the phenomenological différences between the several entities by assigning them to several categories witch differ in ontological status ».35
L’exemple de Bergmann est la différence de hauteurs entre deux sons (plus palpable que celle de l’intervalle, elle-même plus palpable que celle de l’exemplification). Il le confronte avec le fait que Peter soit assis sur une chaise. Il y a toutefois une différence dans la manière dont ces deux faits sont l’objet d’une présentation. Elle se reflète dans la distinction dialectique qu’il opère : les entités sont associées a priori dans le premier cas, dans l’autre la chaise et Pierre sont « adventices » l’une à l’autre, la circonstance est contingente. Le phénoménologue qui ne pratique pas ces différences in feel à partir de la différence entre les catégories les plus simples ne peut pas rendre compte de la différence entre ces deux faits. Bergmann soutient ainsi que la catégorie de l’entité qui joue le rôle de fondement varie en correspondance avec la différence de présentation phénoménologique. Plus exactement, les éléments reliés, la relation ou le nexus, plus le « fait » lui-même ne font pas l’objet d’une présentation phénoménologique multiple, ni qui serait de même degré d’une étape à l’autre. Or, le déplacement théorique qui suit Realism corrige cette proposition. Bergmann ne pense plus que le nexus nous soit donné, fut-ce de manière peu « palpable ». La présentation d’un fait n’est pas la présentation d’un fait plus celle d’un nexus. C’est un être-ensemble qui nous est présenté (ce que Bergmann dans Realism voyait comme la présentation du nexus « et » du fait). Nous sommes renvoyés à la valeur d’une fonction dont le complexe est inséparable. C’est donc cet être-ensemble qui prend le statut d’une entité au sens plein à travers la problématique des circumstances : il n’y a plus d’autre fondement ontologique que le fait que l’entité complexe soit obtenue, sans nexus, mais parce que la connexion est circonstanciée et soutenue par un acte. C’est pourquoi dans New Foundations, la pensée de la connexion se résorbe dans celle d’une fonction, traduisant ontologiquement ce qui dans la sphère phénoménologique est présenté comme un « être-ensemble ». De 35
: Ibid.
233
même encore, dans ses Notes on Ontology36, Bergmann estime qu’il est requis de procéder à certaines notations dans le Langage Idéal de manière à éviter la suggestion d’un Tie. L’exemple porte cette fois sur un donné basique « this-being-green », qui se laisse décrire comme la présentation de deux choses avec celle de leur être-ensemble (« togetherness »). Une notation possible dans le langage idéal est celle-ci : g1(a). Mais sa formulation linéaire suggère toujours un tie. Bergmann explique qu’il convient en effet de partir de la dyade (g1, a), mais de conférer le statut d’une entité (to « ontologize ») à l’être-ensemble, non pas sous l’aspect d’un lien, mais comme un sous-détérminé (η) qui, par sa dépendance, s’accroche (« clings ») à un déterminé unique : la dyade (g1, a) étant toujours dans ce cas le seul déterminé approprié. Il faut donc écrire dans le langage idéal non pas (g1, a), mais η(g1, a). La façon dont nous est présenté l’être-ensemble demande d’éluder une entité du type tie, ou encore maker — pour recouvrer le contexte phénoménologique de la « togetherness » dont Bergmann dit de sa présentation qu’elle est identique avec celle d’un fait. En résumé, cet « être-ensemble » est donné, mais il l’est en même temps que la valeur d’une fonction inséparable de lui. Le caractère eo ipso de la circumstance est donc privilégié. Les deux contextes se chevauchent. Ontologiquement, si les fondements sont là, le fondé l’est aussi. Mais phénoménologiquement, si une entité fait l’objet d’une présentation explicite, alors ses constituants feront l’objet d’une présentation implicite. On peut ainsi avancer que « l’être-ensemble » phénoménologique répond au caractère eo ipso du rapport existant entre ce qui fonde et ce qui est fondé : un rapport que marque la fonction, en ce qu’elle dépend des constituants sans leur être réductible. Elle demeure un sous-déterminé. Elle ne peut pas faire l’objet d’une présentation. Mais elle se substitue au nexus en tant que description de la connexion. Le tie est rejeté comme « l’absurdité d’un lien qui lie ce qu’il lie à ce qu’il lie37 ». Etre un fait devient identique à être la valeur d’une fonction38. L’importance que reçoit désormais la notion de « circumstance » traduit bien ce décalage : telle entité est localisée lorsque telles autres sont là el36
: « Notes on Ontology », Noûs, vol. 15, n°2, 1981, pp. 131-154. : New Foundations, p. 128. 38 : Ibid., p. 126. 37
234
les aussi. Bergmann n’a plus besoin de théoriser ce qui connecte ces entités, mais plutôt de partir du fait qu’il se trouve que certaines entités sont dans l’area de la connexion. Il privilégie finalement la thématique de la fondation sur celle de la relation, réputée donner prise au risque de l’anthropomorphisme en ontologie. Le point nodal qui permet d’écarter le phantasme d’une dialectique « phénoménologisée » est bien la présentation d’une valeur de fonction : celle-ci est tout à la fois la présentation d’un fait et la présentation de « l’être-ensemble » des constituants, sans qu’il soit plus besoin de détailler ontologiquement le fondement de cet « êtreensemble ». La connexion cesse de valoir en tant que condition ontologique. Cette stance philosophique paraît elle-même fortement réminiscente du Meinong de Über Annahmen (qui étudie au § 40 la contraposition stricte de Zusammensetzung et Zusammenstellung). L’évolution dernière de Bergmann semble alors moins virtuose et moins futile comprise sous cet angle. La présentation de la valeur indique que ce qui sert de fondement ontologique se réplique dans le langage idéal, tandis que la fonction « meinongienne » n’implique pas conditionnellement, ou matériellement, qu’il y ait une connexion de trop.
235
PhilosophischeAnalyse PhilosophicalAnalysis 1 Herbert Hochberg Russell, Moore and Wittgenstein The Revival of Realism
8 Rafael Hüntelmann Existenz und Modalität Eine Studie zur Analytischen Modalontologie
2
9 Andreas Bächli / Klaus Petrus Monism
ISBN 3-937202-00-5 334 pp., Hardcover € 94,00
Heinrich Ganthaler Das Recht auf Leben in der Medizin Eine moralphilosophische Untersuchung
ISBN 3-937202-01-3 167 pp., Hardcover € 58,00
3 Ludger Jansen Tun und Können Ein systematischer Kommentar zu Aristoteles’ Theorie der Vermögen im neunten Buch der „Metaphysik“ ISBN 3-937202-02-1 302 pp., Hardcover € 70,00
4 Manuel Bremer Der Sinn des Lebens Ein Beitrag zur Analytischen Religionsphilosophie ISBN 3-937202-03-X 134 pp., Hardcover € 58,00
5 Georg Peter Analytische Ästhetik Eine Untersuchung zu Nelson Goodman und zur literarischen Parodie ISBN 3-937202-04-8, 332 pp. Hardcover € 94,00
6 Wolfram Hinzen / Hans Rott Belief and Meaning Essays at the Interface ISBN 3-937202-05-6 250 pp., Hardcover € 58,00
7 Hans Günther Ruß Empirisches Wissen und Moralkonstruktion Eine Untersuchung zur Möglichkeit von Brückenprinzipien in der Natur- und Bioethik ISBN 3-937202-06-4 208 pp., Hardcover € 58,00
ISBN 3-937202-07-2 189 pp., Hardcover € 58,00
ISBN 3-937202-19-6 340 pp., Hardcover € 70,00
10 Maria Elisabeth Reicher Referenz, Quantifikation und ontologische Festlegung ISBN 3-937202-39-0 ca. 300 pp., Hardcover € 89,00
11 Herbert Hochberg / Kevin Mulligan Relations and Predicates ISBN 3-937202-51-X 250 pp., Hardcover € 74,00
12 L. Nathan Oaklander C. D. Broad's Ontology of Mind ISBN 3-937202-97-8 105 pp., Hardcover € 39,00
13 Uwe Meixner The Theory of Ontic Modalities ISBN 3-938793-11-2 374 pages, Hardcover,€ 79,00
14 Donald W. Mertz Realist Instance Ontology and its Logic ISBN 3-938793-33-3 252 pp., Hardcover, EUR 79,00
15 N. Psarros / K. Schulte-Ostermann (Eds.) Facets of Sociality ISBN 3-938793-39-2 370 pp., Hardcover, EUR 98,00
16 Markus Schrenk The Metaphysics of Ceteris Paribus Laws ISBN 13: 978-3-938793-42-8 192pp, Hardcover, EUR 79,00
EditedBy • HerbertHochberg • RafaelHüntelmann ChristianKanzian • RichardSchantz • ErwinTegtmeier
PhilosophischeAnalyse PhilosophicalAnalysis 17 Nicholas Rescher Interpreting Philosophy The Elements of Philosophical Hermeneutics
25 Laird Addis Mind: Ontology and Explanation Collected Papers 1981-2005
18 Jean-Maurice Monnoyer(Ed.) Metaphysics and Truthmakers
26 Hans Bernhard Schmid, Katinka Schulte-Ostermann, Nikos Psarros Concepts of Sharedness Essays on Collective Intentionality
ISBN 978-3-938793-44-2 190pp., Hardcover € 89,00
ISBN 978-3-938793-32-9 337 pp., Hardcover € 98,00
19 Fred Wilson Acquaintance, Ontology, and Knowledge Collected Essays in Ontology ISBN 978-3-938793-58-9 XX, 726., Hardcover, EUR 159,00
20 Laird Addis, Greg Jesson, and Erwin Tegtmeier (Eds.) Ontology and Analysis Essays and Recollections about Gustav Bergmann ISBN 978-3-938793-69-5 312 pp., Hardcover, EUR 98,00
21 Christian Kanzian (Ed.) Persistence
ISBN 978-3-938793-86-2 289pp., Hardcover, EUR 79,00
ISBN 978-3-938793-96-1 306pp., Hardcover, EUR 89,00
27 Holger Gutschmidt , Antonella Lang-Balestra, Gianluigi Segalerba (Hrsg.) Substantia - Sic et Non Eine Geschichte des Substanzbegriffs von der Antike bis zur Gegenwart in Einzelbeiträgen ISBN: 978-3-938793-84-8 565pp., Hardcover, EUR 149,00
28 Rosaria Egidi, Guido Bonino (Eds.) Fostering the Ontological Turn Gustav Bergmann (1906-1987) ISBN 978-3-86838-008-8 274pp., Hardcover, EUR 89,00
ISBN 978-3-938793-74-9 198pp., Hardcover, EUR 79,00
22 Fred Wilson Body, Mind and Self in Hume’s Critical Realism ISBN 978-3-938793-79-4 512pp., Hardcover, EUR 139,00
23 Paul Weingartner Omniscience From a Logical Point of View ISBN 978-3-938793-81-7 188pp., Hardcover, EUR 79,00
24 Simone Gozzano, Francesco Orilia Tropes, Universals and the Philosophy of Mind Essays at the Boundary of Ontology and Philosophical Psychology ISBN 978-3-938793-83-1 196pp., Hardcover, EUR 69,00
EditedBy • HerbertHochberg • RafaelHüntelmann ChristianKanzian • RichardSchantz • ErwinTegtmeier