Tommaso Piazza A Priori Knowledge Toward a Phenomenological Explanation

PHENOMENOLOGY & MIND

Herausgegeben von / Edited by Arkadiusz Chrudzimski x Wolfgang Huemer Band 10 / Volume 10

Tommaso Piazza

A Priori Knowledge
Toward a Phenomenological Explanation

ontos verlag
Frankfurt I Paris I Ebikon I Lancaster I New Brunswick

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliographie; detailed bibliographic data is available in the Internet at http://dnb.ddb.de

North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 trans@transactionpub.com

United Kingdom, Ire, Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS sales@gazellebooks.co.uk

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

2007 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm nr. Frankfurt www.ontosverlag.com ISBN 10: 3-937202-92-7 ISBN 13: 978-3-937202-92-1 2007
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 ISO-Norm 970-6 This hardcover binding meets the International Library standard Printed in Germany by buch bcher dd ag

Table of Contents
Introduction v

1. A Priori, Analyticity, and Implicit Definition Empiricism, Analyticity, and the A Priori Reductive and Non-Reductive Conceptions of Analyticity Implicit Definition, Logical Truth, and the Recalcitrant A Priori Problems with Implicit Definition BonJours Objection Fodor and Lepores Objection Horwichs Objection Hale and Wrights defence of the traditional connection Logic and Convention Coda 1 3 5 8 10 13 23 31 46 52

2. Realism about Logic Introduction Logical Principles, Justification and Epistemic Relativity Objective Truth Resniks Attack Wittgenstein on the necessity of 1 inch = 2.54 cm and logical inference Dummetts Objection Rule Following considerations and the adoption of a convention Summarising Remarks Wrights Attack Conclusion 57 60 64 65 75 79 84 87 89 109

iv

TABLE OF CONTENTS

3. Objective Knowledge Introduction What the Tortoise Said to Boghossian What Boghossian would say to the Tortoise Rule-circular Arguments The Side-Argument Rejecting the Side-Argument First Horn: Simple Internalism and Rational Insight Second Horn: Epistemic Responsibility and the Lack of Epistemic Irresponsibility Realism, the A priori and Rational Insight Boghossians Argument against Relativism Epistemological Realism about Justification Conclusion 111 115 117 119 122 123 123 126 131 132 134 135

4. Phenomenology and Rational Insight Naturalism and Justification Phenomenology, Justification, and Eidetic Seeing Is Holism a Possibility for the Empiricist? Intuition of Essences and the Analytic/Synthetic Distinction Husserls Conception of the Analytic/Synthetic Distinction Eidetic Variation Passive synthesis and Concept Constitution Knowledge of Reality and Conceptual Truth Absolute vs Relative Objectivity Are Conceptual Truths True? 138 145 150 156 157 164 168 174 177 179

Conclusion References

183 185

Introduction
Do we know anything a priori, that is to say independently of experience? Do we know a priori anything objective concerning reality? Do we know logic? Do we know it a priori? This book is devoted to critically analysing these problems, to inquiring into existing answers to these questions, and to suggesting a plausible way out of the difficulties my analysis will hopefully bring to the fore. Addressing the general question about whether we do know anything a priori involves addressing two distinct sub-questions. The first one concerns the very nature of the statements allegedly known a priori, that is to say concerns the question whether these statements do express the kind of things that can in principle be known, and the question concerning what kind of knowledge one acquires when knowing them. The second subquestion deals with the way, if these statements do express the kind of things that can in principle be known, they can, and indeed are (at least sometimes and locally), known a priori. Contemporary empiricism has delivered a unified answer to both questions with the notion of analyticity. In accordance with this traditional answer, all a priori statements are analytic statements. As synthetic statements, analytic statements are truth-apt (indeed, true) statements. Therefore, a priori statements are the kinds of things that can be known. However, unlike synthetic statements, analytic statements do not have cognitive content, they do not say anything about reality. This feature of analytic statements is due to the fact that their truth, on the empiricist reading of the notion of analyticity, is entirely due to the meanings of their constituting expressions. So, a priori knowledge is not any kind of knowledge of reality. However, the distinctive nature of analytic truth yields a satisfactory account of our knowledge of it. For a natural suggestion is that if the truth of a statement is entirely determined by the meaning of its constituting expressions, then its truth can be known simply by understanding the statement. Accordingly, the truth of an analytic statement can be known merely by understanding it, therefore a priori. This empiricist solution has exerted a considerable influence among empirically minded philosophers; in fact it avoids the intuitive drawbacks of Mills solution according to which alleged pieces of a priori knowledge, like mathematical and logical knowledge, is empirical and inductive in nature. Mills solution is consistent with the empiricist prin-

vi

INTRODUCTION

ciple according to which every piece of knowledge of reality stems from and is justified on the basis of experience. However, this solution has the unpalatable feature of imposing a conception of mathematic and logic as just contingently true, for just beliefs in contingent statements can be justified, to a degree sufficient for knowledge, by inductive and empirical means. The solution based on analyticity allows the empiricist to defend her epistemological principle without loosing the necessity of mathematical or logical statements. If mathematic and logic say nothing about reality, in fact, to admit that our mathematical or logical knowledge is a priori is not to admit that our knowledge of reality has sources other than our experience of it. The viability of this empiricist solution clearly depends on the very notion of analyticity called into question. Notoriously, the empiricist notion of analyticity is not Kants notion. However, admitting that Kants notion constitutes an improvement on such notion is probably not that far from the truth. Kant held that an analytic statement is one characterized by the fact that its predicate term expresses a concept which is contained by the concept expressed by its subject term. Kants analyticity has been found wanting for two reasons. It is too narrow, for it applies just to statements of the subject-predicate form: statements like (K1) Everything is spatio-temporal or is not spatio-temporal is not analytic according to Kants definition, because its predicate concept is not contained in its subject concept. Secondly, it is too wide, for a statement like (K2) Every daughter of a professional philosopher is a professional philosopher is false, and a fortiori not analytic, yet the concept [professional philosopher] expressed by the predicate term is actually contained by the concept [daughter of a professional philosopher] expressed by the subject term. However, Kant notoriously also held that analytic statements are not ampliative with respect to our knowledge of reality, while synthetic statements are; for this reason he also claimed that the a priori epistemological status of analytic statement is not problematic: given the containment theory, to know an analytic statement it is sufficient to possess

INTRODUCTION

vii

both its predicate- and the subject-concept, and to be acquainted with the principle of non-contradiction. Both features are preserved within the empiricist notion of analyticity, because according to this notion analytic statements do not say anything about reality. Given the wider nature of the empiricist conception with respect to its applicability just to statements of the subject-predicate form, however, the epistemology of analytic statement is not anymore Kants. Rather, we might say, it is Freges. Frege held that a statement is analytic when it is either a substitutional instance of a logical principle much in the way it rains o it rains is a substitutional instance of the principle p o p or can be reduced to a substitutional instance of a logical principle with the aid of definitions much in the way every bachelor is an unmarried man can be reduced to every unmarried man is unmarried. If, pace Quine, the question whether two expressions of a natural language are synonymous is not intractably unintelligible, and if a competent speaker of a natural language is indeed in a position to answer such a question whenever it arises concerning two distinct expressions, the question about how analytic statements can and are known reduces to the question about how predicate and propositional logic is known. Though Frege considered the latter epistemological task as completely unproblematic convinced, as he was, that logical principles are simply self-evident it is clearly vital for a sound empiricist theory of the a priori to provide such an account. Unless we are told how logic is known, we cannot stay content with the contention that a priori knowledge of analytic statement is unproblematic because, at bottom, it reduces to knowledge of logical truths. A simple suggestion is that logical principles are known either because they are implicit definitions of the logical constants they contain, or because they are deducible from such principles. This is the contemporary proposal I shall consider at the beginning of the first chapter. The basic idea is that the meaning of certain expressions in the case at issue the meaning of the logical constants is determined by constraining those expressions to have whatever meaning is required for the truth or the correctness of certain basic contexts that contain them. It follows that no one understanding such context (sentences or rules of inference) can fail to appreciate that they are true (valid), if they perform the role of implicit definitions. So long as they do, understanding what they mean coincides with (because it requires) appreciating that they are

viii

INTRODUCTION

true (valid). More than this, the theory seems to make good sense of the traditional suggestion according to which logical principles are selfevident. Since accepting such principles is constitutive of the capability of understanding what they say, it follows that no one fully understanding what they say can fail to appreciate that what they say is true. An empiricist should welcome this account for a third important reason. Take the following statement: (H) If something is entirely coloured of red it is not, at the same time and under the same respect, entirely coloured of green. (H) is not the instantiation of a logical principle, nor it is reducible to the instantiation of a logical principle if synonymous are substituted by synonymous. The meaning of red, in fact, is not the same as the meaning of the (conjunctive) predicate not green and not blue and not gray . Given the potential infinity of colour discriminating expressions, the meaning of any colour term c could not be grasped in the first place if it were equivalent to the infinite conjunction of the negative predicates constructed out of each colour expression other than c. Accordingly, statements like (H1) are to be counted as synthetic under the standard (Fregean) empiricist notion of analyticity. The problem is that synthetic a priori knowledge is not consistent with the empiricist epistemological principle. In contrast with analytic statements, synthetic statements are about reality. So, the admission of a priori knowledge of synthetic statements entails that experience is not the only source of knowledge and justification. The notion of implicit definition seemingly makes it available to the empiricist a plausible way out. As it makes available a notion of logical truth according to which logical principles are just definitions of a devised sort of the logical constants they contain, it seemingly makes available the view that statements like (H1) are just definitions of a devised sort of the color predicates they contain, therefore acceptable from an empiricist epistemological point of view. The first chapter of this book will be devoted to taking into consideration several criticisms, advanced by L. BonJour, J. Fodor and H. Lepore, and by P. Horwich, against P. Boghossians idea that meaning coincides with conceptual role, and against the idea that the meaning of certain expressions is implicitly defined by the resolve to accept as true given contexts featuring these very expressions. In accordance with a

INTRODUCTION

ix

presupposition shared within the contemporary debate, both criticisms will be considered as striving against one and the same epistemological suggestion concerning logical knowledge and knowledge of statements like (H)1. Fodor and Lepore have objected to Boghossians proposal that the conceptual role semantic is the hostage of a dilemma between two unpalatable options. The unqualified identification of meaning and conceptual role entails the violation of the compositionality of meaning. If the identification is unqualified, every transition accepted within a language constitutes a conceptual role. So, for instance, does the transition from x is a brown cow to x is dangerous, constituting a conceptual role for the expression brown cow. However this conceptual role is not a function of the conceptual role of brown and cow, because neither the transition from x is a cow to x is dangerous, nor the transition from x is brown to x is dangerous are legitimate. If meaning is identified with conceptual role, this means that the meaning of brown cow is not a function of the meaning of brown and of the meaning of cow. The natural alternative is to qualify the identification of meaning and conceptual role by narrowing the scope of meaning-constituting inferences to analytic inferences. However, according to Fodor and Lepore, the analytic/synthetic distinction has been convincingly rejected by Quine. So, this second alternative also proves unviable. It follows that conceptual role semantics itself must be rejected. Against Fodor and Lepores suggestion I shall point out that Quines arguments are correctly understood as being directed against the notion of synonymy. Accordingly, such arguments do not have any immediate bearing on the suggestion that the meaning of the logical constants is constituted by the conceptual roles specified by their introduction and elimination rules. So long as it is conceded that we should allow for the advertised correspondence between rules of inference and axiom schemata, Quines argument then has a direct bearing just on the suggestion that those analytic statements which are reducible to logical principles by means of definitions can be know a priori. However, this objection
My suggestion as to the equivalence of the implicit definitional approach and the conceptual role semantics is apparently taken for granted in the literature. An example is seen in Horwich 1998: For simplicity I am focusing on the case in which implicit definition proceeds by regarding a sentence [] as true. But this discussion carries over in an obvious way to the case of implicit definition in which certain rules of inference are regarded as valid (p. 133).
1

INTRODUCTION

can be resisted also in its qualified variety. To do so I will rehearse Boghossians characterization of Quines attack against synonymy both in its error theoretic and non factualist varieties and will show how it can be resisted in both its varieties. Horwichs objection is directed against the notion of implicit definition. According to Horwich, the main problem of the traditional connection the thesis that implicit definition explains a priori knowledge is that it is undermined by the Wittgensteinean conception of meaning needed to make good sense of the model of implicit definition. Horwich contends that the model actually explains how certain expressions receive a meaning only provided that we know that the required meaning exists, that it is unique, that our resolve to accept as true the implicit definer determines that the expression comes to possess that meaning, and that we are able to explain why it is so. The identification of meaning with use helps coping with the four problems, yet at the cost of undercutting the connection with (a priori) knowledge. Since a meaning is constituted by the regularity of use centred on the resolve to accept as true the implicit definition, it constitutes a possibility that, besides our resolve, the content thereby constituted turns out to be false. Hale and Wright have recently rehearsed the point that a wittgensteinean conception of meaning is indeed needed for making sense of the model of implicit definition. However, they also suggest that the wittgensteinean conception doesnt impede establishing the traditional connection. I will end the first chapter by critically assessing Horwichs argument, and Hale and Wrights replies to it. I will not dwell upon the question whether the wittgensteinean conception of meaning is indeed needed to make sense of implicit definition. A major problem concerns the relation between meaning and reference fixing. If an implicit definition performs the role of determining a meaning for an uninterpreted expression, it is arguably by constraining the identity of its reference to be such that it makes the implicit definition true. This is why the existence problem matters: the defined expression, if reality doesnt cooperate, may suffer from reference failure. Hale and Wright grant the point when they require an implicit definition not to be arrogant, that is to say not to entail fresh existential commitments whose fulfilment could be ascertained just by additional epistemic work. Borrowing from the early debate issuing from Priors provocative proposal of tonk as a legitimate logical constant, Hale and Wright assert that an implicit definition (or a set of rules) is not arrogant, and may

INTRODUCTION

xi

be accepted as true (respectively, valid) a priori, only if it is conservative and in harmony: that is to say, if the addition to a language of the new expression (together with the linguistic apparatus which defines it) does not allow one to derive something which is underivable within the same language without the new expression. The elimination rule of an expression is not in harmony with the introduction rule if it licenses from a context containing the expression the inference of more (or less) than is required to introduce such context. In this case then it is possible to prove statements not containing that expression which it was not possible to prove before, and this amounts to a non-conservative extension of the language2. The first thing to notice is that Hale and Wrights proposal is meant to apply both to implicit definitions provided by simple sentences, and to implicit definitions provided by pairs of rules. As it is clear, however, harmony cannot constraint implicit definitions of the first kind. Accordingly, the relevant question is about whether an implicit definition of that sort can be safeguarded from reference failure by some other condition. The most natural proposal, however, is that such implicit definitions be transformed in corresponding so-called Carnap-conditionals, whose meaning constituting role is safeguarded from reference failure by their conditional formulation. If #F is an implicit definition of #, the corresponding Carnap-conditional is If xFx, then F# is true.

The introduction rule states the conditions under which a conclusion with that operator dominant can be inferred. It describes, if you like, the obligations that have to be met by the speaker in order to be justified in asserting the conclusion in question. The corresponding elimination rule states what the listener is entitled to infer from the speakers assertion. Clearly, these mutual obligations and entitlements have to be in balance. The listener must not be allowed to infer more than is required to have been put in to the justification of the assertion. Likewise, the speaker must not be able to get away with less than is required when one is giving the listener entitlement to make certain inferences from what has been asserted, Tennant 2005, p. 628-629.

xii

INTRODUCTION

As Horwich remarks, the Carnap-conditional is uninteresting from an epistemological point of view. In particular, advocating the epistemological role of implicit definitions of the conditional form does nothing to show, as intended by the theory, that their consequents are known a priori. There remain to be analyzed implicit definitions framed in terms of introduction and elimination rules. First of all, it might be suggested that all implicit definitions are to take this form if they are both not to be arrogant and to convey substantial a priori knowledge. This move would in fact prevent the foregoing problem from arising. However, take in consideration implicit definitions of individual terms, or predicates, framed in terms of introduction and rules. If the rules are in harmony, then no fresh existential commitment has been introduced by their adoption. In particular, it means that whatever existential commitments are ratified within a language after the introduction of the new expression must have been ratified before. It seems to follow that no new knowledge is made available once it is introduced by means of the new expression through an implicit definition of it. Since no new knowledge has been introduced, the same knowledge must have been available before. This result does seem to stand in clear opposition to the idea that implicit definition plays a primary role in the epistemology of the a priori. There remains the case of the logical constants. If a pair of harmonious introduction and elimination rules implicitly defines a logical constant, are we thereby justified in believing that these rules are valid (that the corresponding conditionals are true)? Unfortunately, harmony and conservative extension guarantee the validity/truth of these rules/conditionals, only against the backdrop of our accepting the validity/truth of other rules/conditionals whose status cannot be in turn explained by implicit definition. Two distinct reactions to this circularity are available. The first one is to endorse conventionalism, and justify the adoption of the conservative extension and harmony constraint not in the light of their conductivity to validity/truth, but rather as natural desiderata, in the light of the function they are to perform, for the adoption of certain conventions. If the conventional rules of inference just play the function of facilitating the exchange of information, and logical truth is a by-product of such rules, it is easy to see how non-conservative extension and harmony suffice to vindicate the view that implicit definition conveys a priori knowledge.

INTRODUCTION

xiii

The shortcoming, however, is that it is a priori knowledge just of conventional truths. The second alternative is to welcome the circularity at issue as unavoidable, and argue that it does not stand in the way of justifying our rules of inference. Since Dummetts paper on the Justification of Deduction, such justification is alleged to be necessarily circular, that is to say necessarily to employ the very rules of inference whose validity is under dispute. I shall leave open the issue about whether such a justification is enough to ensure objective a priori knowledge of the validity of such rules. I will take up the issue in the third chapter, when discussing a recent proposal of Boghossian. The second chapter will be devoted at assessing the questions whether logical truth can coherently be conceived as objective truth, and whether, so conceived, it can be known along the circular pattern described by Dummett, and recently developed by Boghossian. Before taking into consideration an influential, negative answer to this question, I will firstly pause to stress the connection between the idea that logical truth is objective and several pressing questions in general epistemology, most importantly the objectivity of normative epistemological principles. To show this I will rehearse a very general argument put forward by L. BonJour, and much in the same vain by P. Boghossian, to the effect that an anti-objectivist construal of logic entails strongly unpalatable forms of epistemic relativism and scepticism. Objective truth can be characterized as a property that sentences (or propositions) possess independently of our holding them to be true. Along with Dummetts characterisation, objective truth has been identified with a non-epistemic property. According to a recent proposal of Wright, however, evidence transcendence should be seen just as a sufficient, but not as a necessary condition for objectivity. According to Wright even an evidentially constrained truth-predicate may be shown to deserve a realist objectivist interpretation if (a) it satisfies Cognitive Command, (b) it allows for a Socratic resolution of the Euthyphro Contrast and (c) the facts allegedly reported by true sentences within the area perform a Wide Cosmological Role. The idea that logical truth is objective has been attacked in both senses of the notion. M. Resnik, on the one side, has put forward an argument (echoing Benacerrafs argument against mathematical Platonism), according to which realism and objective truth should be rejected in the case of logic. The argument concludes with the rejection of objec-