Basic Belief and Basic Knowledge: Papers in Epistemology 9783110327519, 9783110327274

Citation preview

René van Woudenberg • Sabine Roeser • Ron Rood (Eds.) Basic Belief and Basic Knowledge

Philosophische Forschung Philosophical Research Herausgegeben von / Edited by Johannes Brandl • Andreas Kemmerling Wolfgang Künne • Mark Textor Band 4 / Volume 4

René van Woudenberg • Sabine Roeser • Ron Rood

Basic Belief and Basic Knowledge Papers in Epistemology

ontos verlag Frankfurt

.

Lancaster

Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek 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 [email protected]

United Kingdom, Ire Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]



2005 ontos verlag P. O. Box 15 41, D-63133 Heusenstamm nr. Frankfurt Tel. ++(49) 6104 66 57 33 Fax ++(49) 6104 66 57 34 www.ontosverlag.com ISBN 3-937202-70-6 2005

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 Printed in Germany.

BASIC BELIEF AND BASIC KNOWLEDGE. PAPERS IN EPISTEMOLOGY Edited by René van Woudenberg, Sabine Roeser & Ron Rood

Table of Contents Introduction

7

Part I: Basic Belief and Basic Knowledge: General Issues RENÉ VAN WOUDENBERG Intuitive Knowledge Reconsidered

15

BENCE NANAY Foundationalism Strikes Back? In Search of Epistemically Basic Mental States

41

IGOR DOUVEN Basic Beliefs, Coherence, and Bootstrap Confirmation

55

Part II: Areas of Basic Belief and Basic Knowledge A: Mathematics and Philosophy RON ROOD On the Status of Axioms in Mathematics

79

BOB HALE Mathematical Knowledge. A Defence of Modest and Sober Platonism

107

STEVEN HALES A Trilemma for Philosophical Knowledge

131

B: Religious Belief CHRISTIAN B. MILLER Defeaters and the Basicality of Theistic Belief

147

DUNCAN PRITCHARD Reforming Reformed Epistemology

177

CHRISTIAN WEIDEMANN Why Basic Theistic Belief is Probably Not Warranted, Even if it is True

211

C: Morals, Testimony, and Proprioception SABINE ROESER Defending Moral Intuition

231

DAVID ENG Basic Beliefs, Testimony, and Blind-Trust

251

ANDY HAMILTON Proprioception as Basic Knowledge of the Body

269

Introduction Throughout the centuries many if not most philosophers have held that some of our knowledge is ‘basic’, i.e. is such that it does not ‘rest upon’ other things that we know. Knowledge of the axioms of mathematics, for example, has been claimed to be basic, and so has knowledge of the principles of morality. It has also been held that some knowledge of what goes on in one’s own mind is ‘basic’. These cases of knowledge were held to be basic in the sense that one can know the propositions involved, even though one hasn’t arrived at them through reasoning or argument. It was considered to be very important that we have basic knowledge. For, as the famous regress argument seemed to show, unless some of our knowledge is basic, we would be in a very bad epistemic predicament. If all our knowledge ‘rests upon’ other things we know, then, so the argument goes, for anyone to know anything at all, he must know infinitely many things. After all, if my knowledge that p, in order to be knowledge at all, needs to ‘rest upon’ prior knowledge of mine, for instance my knowledge that q, then I need to have finished an unending regress. For my knowledge that q, if it is to be knowledge at all, needs to rest on prior knowledge as well, and so forth in seculum seculi. But since I cannot finish an unending regress, I cannot know anything. Of course one could think that my knowledge that p rests on my knowledge that q, and this, in turn, on my knowledge that r, and this on my knowledge that s which, in turn, rests upon my knowledge that p. Thus all the things that I know rest upon things that I know. This, however is a nasty situation, for in that case I am moving in a circle, which ultimately implies that my knowledge that p rests upon my knowledge that p. And it seems very problematic that one’s knowledge that p could rest upon one’s knowledge that p. Many philosophers have therefore held that there must be basic knowledge, knowledge that functions as a regress stopper and circle evader at the same time. After the relatively recent introduction of belief-based accounts of knowledge in epistemology, similar issues raised their heads. If we hold that knowledge is justified true belief, where justification is that, whatever it is, which bridges the gap between mere true belief and knowledge, then we can think of basic knowledge as true belief that is justified in a basic way, i.e. such that its justification isn’t based upon further, prior, justified

8 true beliefs that one has. The traditional worry about basic knowledge can now be reformulated as follows: unless we have true beliefs that are justified in the basic way, we would be in a very bad epistemic predicament. For in that case for it to be true that we have knowledge (justified true belief) we would either have to complete an infinite justification regress or move in a circle in a way that transfers justification. Since, so the recast regress argument goes, both alternatives are impossible, we need to have true beliefs that are justified in a basic way, if we are to have knowledge. From the 1950’s onwards, the thesis that we have basic knowledge or basic beliefs (by which I mean beliefs justified in a basic way) has come under severe attack from various corners. It has been argued that knowledge and justification have no foundational structure, that therefore there is no need to stop any infinite regress, and hence no need for basic knowledge nor for basic belief. It has also been argued that the idea of ‘basic justification’ is internally incoherent, or suffers from other fatal debilities, and should therefore be given up. One historically very important epistemological theory that is committed to the idea of basic knowledge or basic belief is, of course, foundationalism. Numerous philosophers have attacked foundationalism, and as part of their attack also the notions of basic knowledge and basic belief. Subsequent discussion, however, has made it abundantly clear that foundationalism isn’t just one theory, but rather a family of theories, and that some of the family members are much worse off than others. Especially ‘classical’ foundationalism, committed to the idea that only propositions about one’s own mind and self-evident propositions can be properly believed or known in the basic way, appears to be in a very bad shape. But other forms of foundationalism appear to have been able to survive the attack. Basic Belief and Basic Knowledge contains papers that address problems about basic belief, basic knowledge and thus issues pertinent to foundationalism. Part I, in which various general issues are discussed, opens with “Intuitive Knowledge Reconsidered” in which René van Woudenberg clarifies the traditional notion of ‘intuitive knowledge’ and defends it against a number of objections. He argues that ‘intuitive knowledge’ is traditionally used to refer to two rather distinct phenomena, viz. the

9 phenomenon of direct non-propositional awareness, and the phenomenon of the direct apprehension of truths. He furthermore counters arguments stemming from, or inspired by, Peirce, Quine, Albert, BonJour, and Gadamer against the existence of these phenomena. Bence Nanay, in his paper “Foundationalism Strikes Back?”, defends what he calls Never-mind-the-sceptic-foundationalism, according to which certain beliefs are justified by certain ‘basic mental states’, what he calls ‘action-oriented perceptual states’. These states are claimed to be fallible, but not corrigible, i.e. such “that the agent cannot correct it”. Foundationalists are committed to the existence of basic beliefs (or basic knowledge). Many of our justified beliefs, the foundationalist holds, are not of the basic kind and hence he needs to explain how non-basic beliefs can be justified by reference to basic beliefs. Assuming that foundationalists need more than deductive and inductive inference if the beliefs they want to claim are justified, come out justified, Igor Douven argues that they need a notion of coherence as well. However, many clouds surround this notion. In his paper “Basic Beliefs, Coherence, and Bootstrap Confirmation”, Douven attempts to dispel them by developing a modified version of Glymour’s theory of bootstrap testing. Part II deals with various areas of putative basic knowledge and basic belief. One such area is mathematics, where axioms have often been considered to be capable of being known or believed in the basic way. Ron Rood’s paper “On the Status of Axioms in Mathematics” is a discussion of the problematic status that axioms have in mathematics. One aspect of the problematic is whether the propositions of arithmetic are synthetic and apriori. Referring to Leibniz, Kant, Euclid, Gauss, and Riemann, Rood shows how this problem has shaped ideas about axioms. Not only is there a problem of knowledge with respect to mathematical axioms, there is also a problem about mathematical knowledge in general––Benacerraf’s problem. The problem is how to give an account of mathematical statements which are both believable as an account of their meaning and truth-conditions and at the same time avoids rendering it a complete mystery how we might acquire mathematical knowledge. In his paper “Mathematical Knowledge. A Defence of Modest and Sober Platonism” Bob Hale argues that a neo- Fregean version of Platonism permits an account of reference to, and thought about, abstract objects which overcomes the epistemological difficulties that have often been taken to render Platonism in the philosophy of mathematics untenable. It is often said that philosophy itself is an area of knowledge and

10 belief. More specifically, it is widely held that rational intuition can provide us with basic philosophical beliefs in such areas as morality, modality, and metaphysics. There is, however, a fundamental problem regarding the reliability of rational intuition, argues Steven Hales. For there exist rituals in which hallucinogens are used as a method to acquire beliefs in the same areas as rational intuition does, but where it leads to incompatible results. Hales argues furthermore that there are no good reasons for thinking that the basic philosophical beliefs delivered by rational intuition are likelier to be true than those delivered by the use of the hallucinogens. Thus “A Trilemma for Philosophical Knowledge” ensues, the horns of which are nihilism (there are no philosophical truths), skepticism (we cannot have knowledge of philosophical propositions) and relativism (which philosophical beliefs are true, depends on, and is relative to, the methods used). Hales argues that we should grope for the third horn. In a number of papers and books Alvin Plantinga has developed the idea that theistic belief, i.e. the belief that there is such a person as God, can be basic, and even properly so. Three papers in the present volume take issue with aspects of Plantinga ’s ‘Reformed Epistemology’. In his “Defeaters and the Basicality of Theistic Belief” Christian Miller explores the relationship between basic belief in God’s existence and the impact of what purport to be epistemic defeaters for that belief. He argues that while theistic beliefs might arise and even be warranted in the manner described by Plantinga, they will quickly become non-basic in the presence of putative defeaters. Duncan Pritchard offers a proposal for “Reforming Reformed Epistemology”. He argues that the analogy that, according to Plantinga, William Alston and Nicholas Wolterstorff, holds between religious belief on the one hand and perceptual belief on the other, fails in important respects. Religious beliefs, Pritchard holds, are in the main much more than perceptual beliefs. He then goes on to propose a solution to the ensuing problems, thereby drawing on work done by Keith DeRose and Susan Haack––a solution that emphasizes reflexive reflective epistemic virtues. Plantinga has argued that theistic belief is probably warranted if God exists. In his contribution to the present volume Christian Weidemann challenges this claim for three reasons. First, Plantinga’s argument presupposes a piece of natural theology that Plantinga finds untenable in other contexts. Second, in order to explain the highly unequal distribution of basic theistic belief, Plantinga referes to the cognitive consequences of sin;

11 but this, Weidemann argues, is very hard to reconcile with God’s moral perfection. Finally he suggests another explanation for the unequal distribution of theistic belief: God may have a strong motive for making theistic knowledge not too easily obtainable, because such knowledge would severely restrict human freedom of choice. The idea of basic moral beliefs, or basic moral knowledge, once was very popular among the so-called moral intuitionists (such as Clarke, Price, Reid). Intuitionism is presently experiencing some sort of come-back, but not its foundationalist aspect. Sabine Roeser, “Defending Moral Intuition”, argues that this aspect it still defensible and important. More specifically she aims to show how moral intuitions can function as a foundation for moral reasoning, how they can help to bridge the is-ought gap, and how they help to assess organic wholes. In his paper “Basic Beliefs, Testimony, and Blind-Trust”, David Eng argues that testimony is epistemically basic in two ways. First in the sense that beliefs held on testimonial grounds are basic. Second, the practice of giving and accepting testimony is a basic epistemic practice, i.e. cannot be reduced to more elementary practices. He then proceeds to argue that the so-called Blind-Trust account, according to which a hearer can be justified in believing a speaker’s utterance simply in virtue of understanding the utterance, is implausible. Proprioception provides knowledge of one’s own bodily position and movements. In the final contribution to this volume, “Proprioception as Basic Knowledge of the Body”, Andy Hamilton criticizes so-called representative and perceptual models of proprioception according to which knowledge of one’s body is based on evidence. He proposes, instead, that proprioception yields direct, or basic, knowledge of one’s body. Such knowledge, he claims, is non-inferential and one doesn’t have to do anything to acquire it. He finally claims that proprioception suggests that basic knowledge, and not basic belief, is fundamental. This volume grew out of a conference, held at the Department of Philosophy at the Vrije Universiteit, Amsterdam in June 2001 which was generously sponsored by NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek) and KNAW (Koninklijke Nederlandse Akademie voor Wetenschappen). We thank Kiki Berk for her invaluable help. René van Woudenberg, Sabine Roeser, Ron Rood

Part I Basic Belief and Basic Knowledge General Issues

RENÉ VAN WOUDENBERG

Intuitive Knowledge Reconsidered Whereas many philosophers in the past have held that there is such a thing as ‘intuitive knowledge’, in more recent times many philosophers have denied this. This paper aims to show that a number of arguments against intuitive knowledge are unconvincing. This paper is organized as follows. Section 1 seeks to clarify what I will call ‘the traditional notion of intuitive knowledge’ by drawing on the works of Descartes, Locke, and Russell. Section 2 suggests what ‘intuitive knowledge’ looks like on a belief account of knowledge–an account that none of the philosophers mentioned endorse but that is stock and trade in current epistemology. This is needed in order to be able to discuss, in section 3, nine objections to the cogency and existence of intuitive knowledge, some of which presuppose such an account. 1. The Traditional Notion of ‘Intuitive Knowledge’ In the Rules for the direction of the mind, the only work in which he offers a sustained discussion of the idea of intuitive knowledge, Descartes says that the intellect is able to perform two distinct operations: deducing and intuiting. ‘We are’, he says, distinguishing mental intuition from certain deduction on the grounds that we are aware of a movement or a sort of sequence in the latter, but not in the former, and also because immediate self-evidence is not required for deduction, as it is for intuition; deduction in a sense gets it certainty from memory.1

Intuition (or, as he also calls it, ‘mental perception’, or ‘mental 2 apprehension’ ), Descartes says here, lacks the movement (or sequence) that is characteristic of deduction. The movement he has in mind is, no doubt, the movement involved in reasoning. Someone involved in reasoning moves from premise(s) –premises she has to have in mind by aid of memory– to 1 2

Rule 3. Descartes 1984: 15. Rule 3. Descartes 1984: 14.

16 conclusion(s). Deductive knowledge involves reasoning, or inference, whereas intuitive knowledge does not. Moreover, intuitive knowledge, in 3 contrast with deductive knowledge, is self-evident. (And something’s being self-evident, Descartes continues, accounts for its being certain. Deductive knowledge, to be sure, is certain too. But this certainty is due not to a thing’s self-evidence, but to the fact that it follows ‘necessarily from some other 4 propositions which are known with certainty’. ) So far I have talked about ‘things’ being intuitively or deductively known. But Descartes is more specific about the objects of intuition. Intuitive knowledge, he says, is, first, of ‘pure and simple natures’, and second, of ‘simple propositions’. ‘Natures’ are pure and simple only if ‘we know [them] so clearly and distinctly that they cannot be divided by the mind into 5 others which are more distinctly known’. But what are ‘natures’ supposed to be? This becomes clear when we consider Descartes’ threefold distinction among them as well as the examples he provides. First, there are ‘purely intellectual simple natures’; examples of this that he mentions are: knowledge, doubt, ignorance and actions of the will. Second, there are ‘purely material simple natures’, examples of which are: shape, extension, 6 and motion. Finally, there are ‘common simple natures’ that ‘are ascribed 7 indifferently, now to corporeal things, now to spirits’ , Descartes’ examples being: existence, unity, duration. But how shall we understand Descartes here? What is it that Descartes intended to say when he said that such simple natures as ‘knowledge’, 3

Rule 3. Descartes 1984: 14. I don’t think Descartes, in the quotation given in the text, intended to make a distinction between ‘immediate self-evidence’ and ‘mediate selfevidence’. He rather wished, I take it, to emphasize something that is already implicit in the notion of ‘self-evidence’, viz. that what is self-evident is known not on the basis of reasoning. 4 Rule 3. Descartes 1984: 15. 5 Rule 12. Descartes 1984: 44. 6 Does Descartes really allow for intuitive knowledge of purely material simple natures? The following passage suggests, I believe, he does. Dealing with the quantity of things that can be intuited, Descartes says ‘that there are very few pure and simple natures which we can intuit straight off and per se (independently of any others) either in sensory experience or by means of a light innate in us’ (Rule 6. Descartes 1984: 22). In a passage like this, Descartes seems to be the heir of those medieval philosophers who distinguished two modes of intuition: intellectual intuition and sensory intuition. Both modes were understood to be noninferential in character. And the reason why they thought so was that in the typical case we don’t infer that we see something, just as in the typical case we don’t infer that we feel in a certain way. 7 Rule 12. Descartes 1984: 45.

17 ‘shape’ and ‘existence’ can be intuited (to take examples of each of the three sorts of simple natures)? One way is that he wanted to say that what we intuit is (the content of) concepts, i.e. the content of the concepts of ‘knowledge’, ‘shape’, and ‘existence’. A second way to understand him is that he wanted to say that what we intuit are particular instances of knowledge, shape and existence. Construed in the former way, concepts are the objects of intuition; and to intuit a concept means to grasp the content of the concept; and to grasp the content of a concept means to grasp a property. To intuit the concept ‘knowledge’, on this construal, means to grasp the property ‘knows’, it means grasping what it is for someone to know. And to intuit the concept ‘shape’, on this construal, means to grasp the property of having shape, it means grasping what it is for a thing to have a shape, etc. Construed in the second way, the objects of intuition are particular instances of knowledge, shape, and existence. To intuit ‘knowledge’, on this second construal, means to be directly aware of the fact that one knows (or maybe Descartes’ second example in this category is more illuminating: to intuit ‘doubt’, on this construal, means to be aware of the fact that one is in doubt). And again, to intuit shape, means to be aware of the fact that a particular object one perceives, is, for example, round, or perpendicular. That ‘existence’ may be intuited, on this construal, means that it is directly known that some object, for instance the material object someone is presently perceiving, exists. For present purposes we need not decide between these alternative construals. It is of importance, though, to bear in mind that philosophers ever since Descartes have held that both of the items distinguished here, concepts and those things to which the concepts apply, can be intuited. As indicated, next to pure and simple natures, Descartes holds that 8 ‘simple propositions’ are objects of intuition. ‘Simple propositions’, he says, 9 ‘must occur to us spontaneously; they cannot be sought out’. By this he means that one’s knowledge of them is not the result reasoning or inferential activity. They occur to us spontaneously, immediately, without argument. And Descartes, I think, wants to say that of simple propositions we intuit the 10 truth. Examples of simple propositions would be self-evident propositions such as 2+2=4, whatever is red is coloured, and if a implies b, and a is the case, then b is the case. 8

Rule 11. Descartes 1984: 37. Rule 12. Descartes 1984: 50. 10 Elsewhere Descartes writes that ‘simple natures are all self-evident and never contain any falsity’ (Rule 12. Descartes 1984: 45). This statement displays, I believe, some confusion, since only propositions, not natures, can contain falsity. The confusion shows how intimately related propositions and natures are in Descartes’ mind. 9

18 By way of summary we may say that, for Descartes, there is such a thing as intuitive knowledge. The hallmark of such knowledge is (i) that it is knowledge in which no reasoning is involved -- it is, so to say, immediate knowledge, (ii) that it is self-evident (or perhaps better, that its objects are self-evident) and (iii) that the objects of intuition are either concepts or particular instances of such concepts (properties), and simple propositions. John Locke’s thoughts on intuitive knowledge resemble those of Descartes, notwithstanding Locke’s different conception of knowledge. Locke states this conception very clearly in the beginning of book IV of the Essay: Knowledge ... seems to me to be nothing but the perception of the connexion and agreement, or disagreement and repugnancy of any of our ideas. Where this perception is, there is knowledge, and where it is not, there, though we may fancy, guess or believe, yet we always come short of knowledge.11

Knowledge, on Locke’s account, consists in the mental perception of mental facts, these facts being ideas present in the mind, and the relations among them. I can, to cite one of Locke's examples, mentally perceive that the idea ‘white’ does not agree with the idea ‘black’. The perception involved is immediate in the sense that it is unmediated by reasoning. Likewise, the mind can immediately perceive that the idea ‘circle’ disagrees with the idea ‘triangle’ and that the idea ‘three’ agrees with the idea ‘one plus two’. Whatever is immediately perceived, is, for Locke, intuitively known. On Locke's account, then, mental facts are the objects of intuitive knowledge. But he also holds that it is propositions that can be intuitively known. Mental facts and propositions are not unrelated. Says Locke: ‘everyone’s experience will satisfy him that the mind, either by perceiving or supposing the agreement or disagreement of any of its ideas, does tacitly 12 within it self put them into a kind of proposition affirmative or negative’. If I affirm that one plus two equals three, then I have put together the ideas of ‘one plus two’ (which probably is a complex idea) and ‘three’ into the affirmative proposition that one plus two equals three. Like Descartes, Locke contrasts intuitive knowledge with 13 demonstrative knowledge. In cases of demonstrative knowledge the mind 11

Locke 1975, IV,i,1: 525. Locke 1975, IV,v,6: 576 13 And also with ‘sensitive knowledge’. One problem here is that it is by no means clear how sensitive knowledge (an example of which would be someone is swimming in my pool) can fit Locke’s official analysis of knowledge (viz. the perception of mental items). 12

19 does not immediately perceive the agreement or disagreement among the ideas but only through the intervention of other ideas. ‘Those intervening ideas, which serve to shew the agreement of any two others, are called proofs; and where the agreement or disagreement is by this means plainly and 14 clearly perceived, it is called demonstration’. For instance, you cannot immediately perceive that the Pythagorean Theorem is true, but some reasoning may bring you to the conclusion that it is, in which case you have demonstrative knowledge of the Theorem. Locke holds that intuitive knowledge possesses the highest degree of certainty. ‘This part of knowledge’, he says, ‘is irresistible and like the bright sun-shine forces itself immediately to be perceived, as soon as the mind turns its view that way; and leaves no room for hesitation, doubt, or examination, but the mind is presently filled with the clear light of it. ... A man cannot conceive himself capable of a greater certainty, than to know that any idea in his mind is such, as he perceives it to be; and that two ideas, wherein he perceives a difference, are different, and not precisely the same. He that 15 demands a greater certainty than this, demands he knows not what’. Demonstrative knowledge, by contrast, although still quite certain, falls short of the highest level of certainty. 16 Intuitive knowledge, says Locke furthermore, is self-evident. In the philosophical tradition ‘self-evidence’ is usually considered as a characteristic of some propositions (and not, as Locke has it, of knowledge). I won’t go into this, however, and simply state that for Locke ‘self-evident’ is more or less synonymous with ‘intuitive’. By way of summary: intuitive knowledge, for Locke, is characterized 17 by immediacy (direct mental perception), i.e. absence of reasoning , and the highest degree of certainty. The objects of intuition are mental items (ideas and relations obtaining among them) and propositions. Bertrand Russell is in many respects in agreement with both Descartes and Locke, but he was much more explicit than they were as to the objects of intuitive knowledge. In chapter V of his book The Problems of Philosophy, Russell draws the famous distinction between ‘knowledge of things’ and But I won't go into this. 14 Locke 1975, IV,ii,3: 532 15 Locke 1975, IV,ii.1: 531 16 Locke 1975, IV,vii,2: 591 17 This characterization would allow sensitive knowledge to be intuitive. Since Locke explicitly distinguishes sensitive knowledge from intuitive knowledge, this immediacy is only a necessary, not a sufficient condition for something's being intuitively known.

20 ‘knowledge of truths’. One sort of knowledge of things is ‘knowledge by acquaintance’, about which Russell says: In the presence of my table I am acquainted with the sense-data that make up the appearance of my table - its colour, shape, hardness, smoothness; all these are things of which I am immediately conscious when I am seeing and touching my table. The particular shade of colour that I am seeing may have many things said about it - I may say that it is brown, that it is rather dark, and so on. But such statements, though they make me know truths about the colour, do not make me know the colour itself any better than I did before: so far as concerns knowledge of the colour itself, as opposed to knowledge of truths about it, I know the colour perfectly and completely when I see it, and no further knowledge of it itself is even theoretically possible. Thus the sense-data which make up the appearance of my table are things with which I have acquaintance, things immediately known to me just as they are.18

There is a difference, Russell here contends, between having knowledge by acquaintance of a colour (or, alternatively, being acquainted with a colour) and knowing a truth about that colour, and that this is just one instance of a more general difference. This difference can be further clarified as follows: the mental states involved in knowing truths about things have propositional content whereas the mental states involved in one’s being acquainted with things do not. This, in turn, accounts for a further difference, namely, as Russell says, that ‘knowledge of truths’ is troubled by the problem of error, 19 whereas ‘knowledge by aquaintance’ is not. We may understand Russell here as signalling that propositions, i.e. bearers of truth value, are involved in knowledge of truths, whereas they don’t play any role in kowledge by acquaintance. What, according to Russell, are the objects of both types of knowledge? The objects of acquaintance he mentions are (i) 18

one’s own sense-data, i.e. the colours, sounds, smells, etc. that one is immediately aware of.

Russell 1948: 46-47. ‘Knowledge of truths raises a ... problem, which does not arise in regard to knowledge of things, namely the problem of error. Some of our beliefs turn out to be erroneous, and therefore it becomes necessary to consider how, if at all, we can distinguish knowledge from error. This problem does not arise with regard to knowledge by acquaintance, for, whatever may be the object of acquaintance, even in dreams and hallucinations, there is no error involved so long as we do not go beyond the immediate object’ (Russell 1948: 110). The immediate object may be a sense datum but also any of the other objects to be mentioned further on in the text.

19

21 one’s own memories, that is, my having been in Chicago and my having seen Buckingham Palace, these being things I remember. (iii) one’s own mental states, such as my feeling tired, your being in pain. (iv) universals, i.e. such general ideas as ‘whiteness’, ‘diversity’, ‘brotherhood’, ‘being to the left of’, etc. and, possibly, 20 (v) oneself. (ii)

To be sure, Russell holds that we do not have knowledge by 21 acquaintance of physical objects, nor of other minds. It should be noted, however, that there is nothing in the concept of ‘knowledge by acquaintance’ that excludes the possibility of having knowledge by acquaintance of physical objects as well as of other people's minds. It is solely that some of Russell's other philosophical convictions prevented him from allowing this possibility. We may observe that Russell’s ‘knowledge by acquaintance’ is knowledge that consists of being aware of one thing or another. And, as noted earlier, Descartes (on one construal), suggested that we have such knowledge of our own mental states and Locke that we have such knowledge of our own ideas and some of their interrelations. It would seem furthermore that Lockian awareness of ideas and Cartesian intuition of concepts include Russellian acquaintance with universals. As to the ‘knowledge of truths’ (that is: knowledge of true propositions), Russell distinguishes ‘mediate knowledge of truths’ (which is identical with Descartes’ and Locke’s demonstrative knowledge) and 22 ‘immediate knowledge of truths’, which he also calls ‘intuitive knowledge’. Truths immediately known are self-evident, that is ‘evident, as soon as understood’ and ‘not capable of being deduced from anything more 23 evident’. Russell’s category of intuitive knowledge, then, is roughly equivalent with Descartes’ and Locke’s. The objects of intuitive knowledge for Russell are, of course, propositions. He identifies several categories of propositions that can be intuitively known: (a) (b) 20

logical principles, such as the law of non-contradiction. propositions concerning perception, such as ‘there is such-and-

Russell 1948: 48-52. Russell 1948: 52. 22 Russell 1948: 109. 23 Russell 1948: 111. 21

22 such a patch of red’ and ‘that patch of red is round’. (c) propositions about one’s past that flow from one’s memory, such as I once was in Chicago, and I once saw Buckingham Palace. (d) self-evident ethical principles, such as that we ought to pursue 24 what is good. Here again Russell appears to be closely akin to Descartes and Locke. For both Descartes and Locke hold we can have intuitive knowledge of logical principles, and Locke maintained furthermore that we can have such knowledge of some moral truths. 2. Intuitive Knowledge Characterized The foregoing discussion makes it clear that what Descartes and Locke have called ‘intuitive knowledge’ are really two quite distinct phenomena – phenomena that Russell called ‘knowledge by acquaintance’ and ‘intuitive knowledge of truths’. Both phenomena comprise something that can appropriately called ‘immediate’. The one phenomenon is that of non-propositional awareness. Descartes held that we are immediately aware of our own mental states, of some concepts (viz. the simple ones) or of concrete instances to which those concepts apply. Locke held that we immediately perceive some ideas and some of their interrelations, whereas Russell held that we are immediately acquainted with sense data, memories, mental states and universals. The term ‘immediate’ pertains to the awareness, or mental perception, or acquaintance involved: the awareness, perception, or acquaintance is not mediated by other things or states that we are aware of, perceive, or are acquainted with. Nor is there any reasoning involved. The other phenomenon is immediate knowledge of propositions, or, alternatively, the immediate apprehension of truths. What sense of ‘immediate’ attaches to such knowledge? The answer is: such knowledge of propositions is not mediated by, or based on, knowledge of other propositions. Intuitive knowledge is knowledge that is not the conclusion of an inference. In what follows I will adopt Russell’s terminology and call the first phenomenon ‘knowledge by acquaintance’, and the second ‘intuitive knowledge’. And both of them I call forms of ‘immediate knowledge’. 24

Russell 1948: 112. Here again is affinity with Locke, who also held that there are moral truths that can be intuitively known; cf. Locke 1975, IV, iv.

23 Before being able to deal with various objections that have been raised against the existence and possibility of immediate knowledge, a complicating factor needs to be introduced. Descartes, Locke, Russel and most other philosophers in the Western tradition have contrasted knowledge and belief. To know and to believe, it was generally held, are two mutually exclusive mental states. Many contemporary epistemologists in the Anglo-American world, however, do not contrast knowledge with belief, but conceive of it as a species of belief. According to the so-called standard analysis, knowledge is justified, true belief. According to this conception S knows that p, if and only if 1) S believes p, 2) p is true and 3) S is justified in believing that p. To be sure, this account of knowledge was shown defective by Edmund Gettier in his celebrated paper ‘Is Knowledge Justified True Belief?’ The consensus that emerged in response to this paper, however, is that justified true belief is nearly sufficient for knowledge, and at any rate necessary for it.25 In what follows I will abstract from Gettier problems, since these problems don’t raise special trouble for intuitive knowledge. Accordingly, I will be working with a justified true belief account of knowledge, while from the outset acknowledging that Gettier problems will have to be dealt with anyway. Obviously, a belief account of knowledge affects only a possible reformulation of intuitive knowledge. For knowledge by acquaintance is nonpropositional, and hence does not involve anything like belief states, nor does the truth condition pertain to it and neither does the justification condition. What, then, will intuitive knowledge on a belief account of knowledge look like? I propose the following: for S to intuitively know that p means for that person to have the true belief that p, while her belief is justified in an immediate way. And a belief is justified in an immediate way, just in case whatever it is that justifies S in believing that p does not include any reference to other justified beliefs that S has and that embody her evidence or reasons for it. Immediate justification is to be contrasted with mediate justification, which is justification that does include reference to other beliefs that S has and that embody her evidence or reasons for it. Putative examples of mediate justification would be • S believes that there is smoke coming out of her house. This justifies her in believing that her house has caught fire. Her belief that her house caught fire is therefore mediately justified. • S believes the Pythagorean Theorem is true. This belief is based on various other beliefs that embody S’s reasons for holding the Theorem. 25

For this discussion see Shope 1983.

24 That belief is therefore mediately justified. • S believes that iron expands when heated. His belief is based on various experiments he has conducted, and the outcome of the experiments justify him in so believing. His belief is therefore mediately justified. Putative examples of immediate justification would be: • S believes that if A is taller than B, and B taller than C, then A is taller than C. But this belief is not based on other beliefs that constitute S’s reasons for this belief. S just ‘sees’ that the proposition believed must be true. Assuming that S’s belief is justified, the justification is of the immediate variety. • S believes that a big red object is approaching him. This belief isn’t the product of reasoning or inference. Therefore, if his belief is justified, it is immediately justified. These last two examples bring to light that immediate justification can pertain to both rational (non-empirical or apriori) and empirical beliefs. One final note needs to make before turning to the objections. Recent discussions have made it clear that there are many quite different concepts of justification. Some epistemologists think of justification as having adequate evidence, others as having fulfilled one's epistemic duties, still others as being coherent with a large body of beliefs and yet others as being reliably produced. I propose to sidestep this problem in the following way. I take it to be obvious that mere true belief is insufficient for knowledge. More is needed. Let us now call that additional quantity ‘justification’, whatever the precise nature of that quantity is. Of course, when entering into discussion with critics of intuitive knowledge, we should be aware which concept of justification is being used. But we need not decide in advance what that quantity is. Let us now turn to nine objections against immediate knowledge. 3. Objections Rejected (1) In his essay ‘Questions Concerning Certain Faculties Claimed for Man’ 26 [1868] C.S. Peirce asks ‘whether by the simple contemplation of a cognition ... we are enabled rightly to judge whether that cognition has been determined by previous cognitions’ or not? So his question is whether by 26

References are to Houser 1992. The essay is included in Hartshorne 1931-1935.

25 mere contemplation we are able to identify intuitions. He defines an intuition as ‘a cognition not determined by a previous cognition’ which, he says, is nearly the same as ‘premise itself not a conclusion’– nearly because premises are judgments, whereas some intuitions are not. In effect, Peirce’s class of intuitions comprises both intuitive knowledge and knowledge by acquaintance. In order to answer his question, Peirce suggests that it is reasonable to accept the following conditional: if there are such intuitions, it would be plausible to assume that ‘we can always intuitively distinguish between an 27 intuition and a cognition determined by another’. The historical fact is, however, says Peirce, that there have always been disputes as to which cognitions are intuitions. So, it is false to assume that we have an intuitive power to distinguish intuitions from other cognitions. But if it is, then the 28 antecedent of the conditional, viz. that we do have intuitions, is false. This argument, however, is unconvincing. For, why think it is reasonable to accept the conditional in the first place? Why think that if we have intuitions we will always be able to identify them intuitively (by simple contemplation)? We should not confound the issue of the existence of intuitive knowledge with the epistemological issue of our ability to identify cases of intuitive knowledge. Why can’t it be that even though we actually have some intuitive knowledge, our ability to identify it is limited and fallible? This would nicely explain why the disputes Peirce mentions, occur. But if it is unreasonable to accept Peirce’s conditional, then the thesis that we do have intuitive knowledge has not been refuted. The point I am trying to make can be illustrated by taking a closer look at one of Peirce's examples. ‘Every lawyer’, says Peirce, ‘knows how difficult it is for witnesses to distinguish between what they have seen and 29 what they have inferred’. Likewise, performances of spiritual mediums and professional jugglers show ‘that it is not always very easy to distinguish between a premise and a conclusion’. Indeed, even normal sense perception is ‘shot through’ with interpretation, and it may therefore be difficult to tell the purely perceptual from the interpretative component. In response to this is should be stressed that this fact has no tendency to support the claim that there is nothing that is immediately perceived. No lawyer would conclude from the difficulty of identifying the purely perceptual component, that there is only inferring, or only interpretation, and no perceiving! 27

Houser: 12. I interpret Peirce as advancing a reductio argument. Other interpreters, however, do not, but this seems to me mistaken. E.g. Delaney 1993: 90. 29 Houser 1992: 13. 28

26 But Peirce develops yet another line of argument. He asks whether the existence (or non-existence) of intuitive knowledge can ‘be determined upon 30 evidence’? In order to settle this matter, he considers putative cases of intuitive knowledge. One of them is the case of someone who knows he is angry (this is an example of someone’s knowledge of the mental state he is in). This knowledge, Peirce argues, is not intuitive but determined by previous cognitions. The argument he offers is contained in the following passage: It must be admitted that if a man is angry, his anger implies, in general, no determinate and constant character in its object. But on the other hand, it can hardly be questioned that there is some relative character in the outward thing which makes him angry, and a little reflection will serve to show that his anger consists in his saying to himself, 'this thing is vile, abominable, etc.' and that it is rather a mark of returning reason to say ‘I am angry’.31

Now what exactly is the argument here, and how does it establish the conclusion that the angry person's knowledge that he is angry is not intuitive? It seems that Peirce affirms the following two theses: (1) there is something external (‘some relative character in the outward things’) which makes S (the angry person) angry and (2) S’s anger consists in S’s saying to himself ‘this thing is vile, abominable, etc.’ Now I am not at all inclined to think that (1) and (2) are true. For S may be angry with himself because of something cheap or wicked he did or said, in which case it is not something external that makes him angry. He may even be angry without anything making him angry, in which case there is no external thing that makes him angry either. And it can be seriously doubted that anger consists in S’s saying to himself ‘this is vile, abominable, etc.’, for S may be saying that to himself without being angry. But even if we grant 30

Houser 1992: 18. Sometimes it is not exactly clear what Peirce is arguing; this is due to Peirce’s switching back and forth between ‘intuitions’ and ‘intuitive power to distinguish between intuitions and cognitions determined by other cognitions’. I construe Peirce here as investigating whether or not there is enough evidence available for thinking that there is intuitive knowledge. 31 Houser 1992: 23.

27 Peirce all this, how can he move from (1) and (2) to the conclusion that S cannot intuitively know he is angry? (1) states that S’s knowledge that he is angry is caused by something vile, abominable. But this does not prove (nor is it identical with) the thesis that S’s knowledge of his mental state is mediated by other knowledge S has. Acceptance of (1) does not force one into holding that S’s knowledge that he is angry is the conclusion of an argument, the premise of which is ‘that is vile, abominable’; nor does it force one into holding that S knows that he is angry because he knows that something is vile, abominable. Friends of intuitive knowledge need not be committed to denying that there are causal conditions for every cognition, nor even to denying that prior cognitions (such as knowing that this is vile, abominable) may be among the causal conditions. The only claim the intuitionist is committed to, is that some things can be known intuitively, that is, not on the basis of other things the knower knows. And surely, S’s knowledge that he is angry seems a case in point: that knowledge is not determined by S’s previous cognitions (this holds true even if S knows that his anger is caused by something that vile, abominable). By way of summary, then, Peirce set out to argue, from (1) and (2), that S’s knowledge of his own mental states is not intuitive in character. Since the argument offered is unconvincing, the claim that such knowledge is intuitive stands undefeated. (2) Descartes and Locke, as we have observed earlier on, held that intuitive knowledge is certain (and other philosophers have supposed that it is 32 incorrigible, indubitable, or infallible ). A further objection against the existence of intuitive knowledge that I want to consider is triggered by this very property. Many philosophers from very diverse backgrounds hold that nothing is really certain, that nothing we know (or think we know) is immune from doubt, etc. and therefore, they argue, there can be no intuitive knowledge. This line of reasoning can be found, among others, in the 33 34 35 writings of Peirce , Hans Albert , and Quine . It should be noted however that the concept of intuitive knowledge as I have characterized it in section 2, does not make reference to ‘certainty’, nor to any of the other immunities mentioned. As a matter of fact, moral 32

These notions have been explained in various ways; on some explanations they are nearly identical on others they are not. A very helpful discussion of these notions is Alston 1971. 33 Cf., for example, CP 1.141, and Houser 1992: 12. 34 Albert 1985: 39. 35 ‘Two Dogmas of Empiricism’ in Quine 1963.

28 36

intuitionists like G.E. Moore and Ross held that intuitively believed moral propositions may, in the face of weighty considerations, be defeated. Such propositions do, however, possess prima facie justification: they have an initial claim to being justified. Recently Laurence BonJour has similarly argued that rational insight, i.e. the insight that might immediately justify the belief in such propositions as that nothing can be red and green all over at the same time, or the belief that there are no round squares, or the belief that all cubes have 12 edges, is not infallible. There are simply too many compelling examples of propositions that once were claimed to be rationally ‘seen’ to be 37 true and that subsequently turned out to be false or mistaken. But even if rational insight is not infallible, and hence immediate justification defeasible, this does not imply that intuitive knowledge do not exist. Defeasible immediate justification may still be very good justification, and if the proposition believed is true, it may amount to knowledge–intuitive knowledge. The present point, then, is that there is no internal connection between absolute certainty, or infallibility and intuitive knowledge. (3) Another objection against the existence of intuitive knowledge can be thought of as emanating from the quarters of ‘hermeneutical philosophy’. According to hermeneutical philosophers whenever we believe something, that belief is part of a very large ‘web of beliefs’ and none of our beliefs can be held totally independent of other beliefs. And whatever knowledge we have, it is always situated within a larger ‘horizon of understanding’, to use 38 one of Gadamer's expressions. And this fact, so one might argue, counts against the existence of intuitive knowledge, because an item of knowledge is intuitive only if it in no way depends on other knowledge of the subject, only if it can be held without the subject’s knowing or being justified in believing anything else. In many repects the imagery of ‘web’ and ‘horizon’ is compelling. It seems impossible for S to intuitively know that she is sad, without having 36

Moore 1902 calls some self-evident moral propositions ‘intuitions’, which means that they are ‘incapable of proof’. But a proposition's being incapable of proof, he says, does not imply that it ‘is true because we cognize it in a particular way or by the exercise of any particular faculty: I hold, on the contrary, that in every way in which it is possible to cognise a true proposition, it is also possible to cognise a false one’. 37 BonJour 1998: 110-5. 38 And this horizon of understanding is shaped, among other things, by language. Says Gadamer, ‘we are always already biased in our thinking and knowing by our linguistic interpretation of the world’ (from “Man and Language”, in H.G. Gadamer 1976).

29 general knowledge as to what it is to be sad. Likewise, it seems impossible for S to intuitively know that 2+2=4, without having knowledge of some significant part of a larger arithmetical system. In the same vein it seems impossible to intuitively know the truth of the moral proposition ‘What is in no degree voluntary, can neither deserve moral approbation nor blame’, to 39 cite one of Thomas Reid’s examples , without knowing what it means for something to be done ‘voluntarily’ and what it means for something to ‘deserve moral approbation’, etc. In all these cases, then, the putative items of intuitive knowledge seem to depend on other knowledge. And the objection is that this is ‘hermeneutical’ evidence against the existence of intuitive knowledge. 40 But is it? I can be argued it is not. The crucial thing to see is that when the objector says that the putative item of intuitive knowledge is dependent on other knowledge, by that she means that the very existence of that item depends on other knowledge. But intuitive knowledge, as I have spelled out that notion, means that the item at hand is justified by something other than the relation it has to other justified beliefs that the subject has and that embody evidence or reasons in its favour. S’s belief that she is sad cannot come into existence, it might be argued, without S’s knowing what it means, generally speaking, to be sad. But S’s belief that she is sad is in no way justified by the relation it has to this knowledge; it is immediately justified. Likewise, S’s knowing that 2+2=4 may depend, for its existence, on other knowledge, but not for its justification. In the same vein, S’s knowing the truth of Reid’s moral proposition may depend for its existence on other knowledge (e.g. on knowledge of the concepts involved), but, again, not for its justification. The existence of a conceptual web, or a horizon of understanding, then, is, properly understood, compatible with the existence of intuitive knowledge and hence no argument against it. (4) Another objection that might be raised against its existence, is that for every putative item of intuitive knowledge we can always ask for reasons that justify it. Every belief, so it seems, is open to assessment in terms of reasons or evidence. And if, supposedly, adequate evidence can be adduced in favour of such an item, it might be argued, this counts against the item’s capability of being intuitively knowable. But does it? For to say that something is intuitively known simply 39 40

Reid 1969: 368. See Alston 1983: 63.

30 means that there is immediate justification for it. But it does not imply that there cannot be, for that item, mediate justification as well. You may remember that in January 1994 you read a paper at a memorial conference, and for that belief you have immediate justification. But a glance at your agenda may mediately justify that belief as well. To say that something is intuitively known means no more and no less than that there is immediate justification available for it. (5) Another objection derives from the concept of justification. It is widely held, and I have expressed my allegiance to this, that ‘being justified’ is a necessary requirement for a belief to count as knowledge. Without this requirement there would be no distinction between mere true belief and knowledge. But what is justification? Laurence BonJour’s answer to this question is aimed at showing the impossibility of immediate justification, and, by implication, of intuitive knowledge. So, let us consider his views about justification: One’s cognitive endeavors are epistemically justified only if and to the extent that they are aimed at (the goal of truth) - which means very roughly that one accepts all and only beliefs which one has good reason to think are true. To accept a belief in the absence of such a reason ... is to neglect the pursuit of truth; such acceptance is, one might say, epistemically irresponsible. My contention is that the idea of avoiding such irresponsibility, of being epistemically responsible in one’s 41 believings, is the core of the notion of epistemic justification.

So, BonJour says that our beliefs are justified to the extend that we have good reasons for them, i.e. reasons that indicate that the belief at hand is true. Justification, as BonJour thinks about it, then, is inferential justification, i.e. justification that consists of producing a justificatory argument. Now BonJour concedes that someone ‘for whom a belief is inferentially justified need not have explicitly rehearsed the justifying argument in question - to others or even to himself’. What is required for justification, is, rather, ‘that the inference be available to the person in question, so that he would be able in principle to rehearse it if the belief should be called into question, either by others, or by himself; and also that the inference be, in the final analysis and in a sense most difficult to define precisely, his actual reason for holding the 42 belief’. It seems clear that if justification is what BonJour says it is, immediate 41 42

BonJour 1985: 8. BonJour 1985: 19.

31 justification is impossible indeed. Then it is impossible by definition. For a belief to be justified, on BonJour’s conception, simply is that that belief has been (or, in the final analysis: can be) the target of a successful justificatory argument. But this conception of ‘justification’ is by no means obvious. As a matter of fact, it is the very conception that friends of intuitive knowledge are committed to deny. One reason the intuitionist offers for his commitment is that we may suppose that little children are not in cognitive possession of justificatory arguments for their beliefs (e.g. perceptual beliefs), nor that such arguments are available to them. On BonJour’s conception this would mean that none of the children’s beliefs are justified. But this seems absurd. Therefore, BonJour’s conception of justification is not as obviously valid as one might think it is. At this point it is helpful to bring to mind Alston’s point that the expression ‘justification’ is ambivalent between the state of being justified in believing something or other and the activity of justifying the belief at hand 43 (i.e. the activity of showing that the belief is justified). It is possible for a belief to be justified, that is, it is possible for a belief to be in the state of being justified, without its having been shown or established to be in that state by reasons, that is, without it having been the target of a successful activity of justifying. Children’s beliefs of the sort cited would be examples. And as a matter of fact, many of the beliefs we adults have, would be examples. Many of our beliefs (we suppose) are in the state of being justified, without them having been shown, by reasons, to be justified, or without such reasons even available. An example of this would be my belief that I feel dizzy (when I feel dizzy). BonJour, obviously, is working with the activity concept of ‘justification’ and with this concept in mind he is led to think that all justification is mediate. But once one sees the difference between the state en the activity concept of justification, one can freely recognize the possibility of immediate justification. A point related to the previous one now should be observed. BonJour, we saw, is working with the activity concept of justification. To justify, for him, is to produce an argument, to offer reasons for thinking that some target belief is true. But it may be doubted that all of our justified beliefs are, or can be, justified by arguments. I cannot, for instance, justify my perceptual belief that there is a green tree in front of me (when in fact there is a green tree before me) by means of a justificatory argument. In the typical case, when 43

Alston 1971: 44,47,55,70,82.

32 someone asks what I am seeing when I am in that situation, I will simply make a gesture with my hand and repeat: there is a tree, and it is green. And for my own perceptual belief to be justified, I need not even make such a gesture. I directly apprehend that there is a tree in front of me. My belief is justified not by a justificatory argument, but by a visual experience. And in a different case we might say that the justification does not stem from experience either, but from some fact. For instance, to cite one of Chisholm’s examples: Suppose I am thinking about Albuquerque and believe it is in New Mexico. And now suppose furthermore someone asking me: what is your justification for your belief that you are thinking that Albuquerque is in New Mexico (note that that person does not ask: what is your justification for your belief that Albuquerque is in New Mexico). The answer would be: what justifies me in believing that I think that Albuquerque is in New Mexico is simply the fact that I am thinking that Albuquerque is in New Mexico. So, justification may occur in epistemologically quite different situations: (A) A belief may be justified, and it can be shown by reasons that it is, in which case we have mediate knowledge. (B) A belief may be justified, but it is impossible to show that it is by reasons. In the second type of case, the reason for this impossibility is that the belief results immediately from experience, or from some fact. In this case we have intuitive knowledge. (6) BonJour, however, appears to object to the very possibility of (B). He rejects the ‘doctrine of the given’ according to which ‘basic empirical beliefs are justified, not by appeal to further beliefs ... but rather by appeal to states of “immediate experience”, or ”direct apprehension”, or “intuition” - states which allegedly can confer justification without themselves requiring 44 justification’. And what, according to the doctrine, is the nature of the experience, apprehension, or intuition? ‘The main account’, says BonJour, ‘is well suggested by the terms usually employed in describing such experience: terms like ‘immediate’, ‘direct’, ‘intuitive’ and ‘presentation’. The underlying idea is that of confrontation: in an immediate experience mind or consciousness is directly confronted by its object without intervention of any kind of intermediary. It is in this sense that the object is simply given to or 45 thrust upon the mind’. BonJour’s criticism of the idea of the given, goes like this: 44 45

BonJour 1985: 59. BonJour 1985: 60.

33 The proponent of the given is caught in a fundamental and inescapable dilemma: if his intuitions or direct awarenesses or immediate apprehensions are construed as cognitive, at least quasi-judgmental (as seems clearly the more natural interpretation), then they will be both capable of providing justification for other cognitive states and in need of it themselves; but if they are construed as noncognitive, nonjudgmental, then while they will not themselves need justification, they will also be incapable of giving it. In either case, such states will be incapable of serving as an adequate foundation for knowledge. This, at bottom, is why epistemological givenness is a myth.46

BonJour, then, speaks of a dilemma: either intuitions have cognitive value, or they don’t. I think we can understand BonJour here as saying that intuitions are either instances of intuitive knowledge or instances of knowledge by acquaintance (as I have spelled out these terms in section 2). Both of these alternative ways of understanding intuitions, says BonJour, have unacceptable consequences. Let us see and first consider intuitions construed as instances of intuitive knowledge. If intuitions are thought of as instances of intuitive knowledge, says BonJour, ‘then they will be both capable of giving justification and in need of it themselves’. But, stated thus, this amount to no more then simply contradicting the friend of intuitive knowledge. For the intuitionist is committed to the thesis that an intuition is a belief that is immediately justified, which means that that belief does not derive its justification from other beliefs the subject has and that constitute evidence or reasons for it. At the same time, the intuitionist holds, the immediately justified belief itself may constitute evidence in favour of other beliefs. What makes BonJour think there are no beliefs of this sort? He says the whole idea ‘seems hopelessly contrived and ad hoc’, and that if one allows this idea, then ‘any sort of regress could be solved in similar fashion’. This suggests that it is purely arbitrary to hold that there is intuitive knowledge. But is it? Intuitionists have good reasons for considering some beliefs to be immediately justified, because those beliefs have some unique property, such as being self-evident, or not justified by other beliefs one holds, or being produced in some unique way. To be sure, it would be a task of some magnitude to give a full-blown account of what it is that confers immediate justification on putative cases of intuitive knowledge. But nothing BonJour has said rules out that possibility. When intuitions are construed as instances of knowledge by 46

BonJour 1985: 69.

34 acquaintance, then, says BonJour ‘they will not themselves need justification, [but] they will also be incapable of giving it’. But why should one think this? For, consider the following two examples. (1) S experiences a severe headache, S is, we might say, directly aware of headache. Having headache, for S, is not something that need or even can be justified (‘justified’ taken as the activity concept). Still, this direct awareness may be used to immediately justify S’s belief that he is having a headache. (2) S perceives a blue roundshaped object. The immediate awareness of the blue round-shaped object itself is not the kind of thing that can be justified. Still, it can be used by S to justify his belief that there is a blue ball in front of him. These two examples suggest that it is entirely possible that there are instances of direct awareness, such as awareness of headache, or of coloured objects, that can play a role in justifying a belief, although they themselves cannot be shown to be 47 justified. (7) The next objection I would like to consider is levelled against one segment of the domain of alleged intuitive knowledge. According to the traditional understanding, intuitive knowledge includes our knowledge that two plus two equals four, our knowledge that every event has a cause, as well as our knowledge of the elementary principles of logic. Intuitive knowledge includes, it is widely held, knowledge of apriori truths. Those who separated knowledge and belief held that the truth of these apriori propositions is intuitively grasped. And those who conceive of knowledge as a form of belief have hold that a priori knowledge derives its justification from the believers ability to intuitively grasp (or ‘see’, or apprehend) that the proposition under consideration is true (and, furthermore, is necessarily true). Descartes, Locke and many others have held that reason is operative in the acquisition of apriori truths. Many contemporary philosophers, however, have argued that apriori 48 49 knowledge is not what the tradition has thought it was. Sellars and Rorty , for example, have argued for a linguistic explanation of apriori knowledge, according to which such knowledge is really knowledge of analytic truths and the process of acquiring such knowledge is identical with the process of learning the conventions of one’s language. To know, for example, that all events are caused, it is claimed, simply is to know the meaning of the words 47

What I have argued here is closely akin to Paul K. Moser who holds that there are nonpropositional justifiers which themselves do not need justification and that the given is not a myth. Cf. Moser 1989: 80-88, 182, 186-194. 48 Sellars 1963: 164-170. 49 Rorty 1979; Rorty 1980.

35 ‘event’ and ‘cause’; and to know that two plus two equals four, it is claimed, simply is to know the meaning of the words ‘two’, ‘plus’, and ‘four’. In order to find out whether someone really has knowledge of apriori truths, then, the only criterion we have to apply is whether that person knows his language. If someone claims that every event has a cause, we cannot show that his belief is unjustified, unless we discover that his use of the words ‘event’ and/or ‘cause’ is idiosyncratic. According to the linguistic explanation, then, the truth of apriori propositions depends on the language one happens to speak. Apriori truths are true by convention. And since many apriori truths are necessary truths, necessity is a matter of convention as well. It has, I believe, been cogently argued that the linguistic explanation of 50 apriori knowledge is deeply flawed. For one thing, it can be questioned whether the assumption that all apriori truths are analytic is tenable. One oft cited counter example is the proposition ‘nothing can be red and green all over at one time’, which is clearly apriori but not (obviously) analytic. For another thing, the thesis that apriori truths owe their necessity to linguistic convention (or the meaning of the terms involved) fails to come to grips with the distinction between sentences (ordered sequences of words) and propositions. It is a matter of convention indeed that the sentence ‘Two plus two equals four’ expresses the necessarily true proposition that two plus two equals four. But this, of course, in no way shows that the proposition that is expressed by the English sentence ‘Two plus two equals four’ owes its property of being necessarily true to such conventions. The ‘linguistic account of apriori knowledge’ seems to be put forward in order to explain apriori knowledge, i.e. in order to show that such propositions owe their necessity to linguistic conventions. Since, as indicated, this account is by no means convincing, we are free to reconsider the traditional understanding of this important area of intuitive knowledge. (8) Another objection that might be levelled against the existence of intuitive knowledge derives from the fact that many propositions that traditionally are claimed to be intuitively knowable, need to be reflected on for some time before they are believed or accepted as true, or seen to be true. It may take some time and reflection to believe such proposition as ‘if a implies b, and a is the case, then b is the case’, ‘no proposition can be both true and false at the same time’and ‘wat is in no degree voluntary, can neither deserve moral 50

BonJour 1985: 191-211. A much more elaborate argument is to be found in BonJour 1998: chap. 2.

36 approbation nor blame’ and to see they are true. And, one might think that this constitutes an objection against such propositions being intuitively knowable. But I don’t think it does. For, to repeat myself once more, the hallmark of intuitive knowledge is that the justification for the belief involved is immediate, i.e. does not rest on other of one’s beliefs. But the fact that it takes time and reflection before a proposition is believed or seen to be true, does not exclude the possibility that the proposition at hand is immediately justified, i.e. justified by other things than other beliefs the subject has. Maybe Karl-Otto Apel’s ‘transcendental’ reflection that aims to establish that some propositions are indubitable (especially propositions regarding the necessary preconditions of argumentation) is best understood as an illustration of just this. point. Reflection on what is presupposed whenever we enter argumentative discourse, Apel claims, results in, or leads to, intuitive knowledge. It leads one to believe various propositions (propositions that I need not specify here) and furthermore leads one to see that they have immediate justification. Two points Apel stresses are of importance here. First, reflection is not deduction – the truth of the propositions at hand is not proved, or inferentially justified in the way many mathematical theorems can be. Second, the propositions at hand are believed, or seen to be true, as soon as they are understood. But reflection and understanding may take time. Still, once the proposition is understood, it may well turn out to be self-evident. And according to the traditional view, self-evidence is an immediate 51 justifier. The present point, then, is that the fact that time and reflection are required before one believes a certain proposition, does not imply that the proposition cannot be immediately justified and hence does not imply that it 52 cannot be intuitively known. (9) A final objection against the very existence of intuitive knowledge I want to consider is this: friends of intuitive knowledge have lost sight of history, they ignore the obvious fact that what we know is in many ways determined by the cultures and times we live in. This objection can be met in a way similar to the objection stated under (3): intuitive knowledge is indeed quite independent of such historical and cultural determinations, with respect to the justification requirement. But it is not thus independent with respect to its existence. 51 52

See Apel 1986: 79-99; van Woudenberg 1995: 170-188. This point has recently been made by Aune 1994.

37 (And we have, of course, to bear in mind that not all the knowledge we claim to have is intuitive in character. With respect to such non-intuitive knowledge we are free to allow determinations of the indicated sort even 53 with respect to the justification requirement).

53

Many thanks to John Greco, Tim O’Connor, Bas Jongeling, Alvin Plantinga, Terence Cuneo, Sabine Roeser, Kevin Meeker and Nicholas Wolterstorff for commenting on earlier drafts of this paper.

38

REFERENCES

Albert, Hans. 1985. Treatise of Critical Reason, Princeton (N.J.): Princeton University Press. Alston, William P. 1971. “Varieties of Priviledged Access” in Alston 1989: 249-285. ——— .1983. “What is Wrong with Immediate Knowledge?” in Alston 1989: 57-78. ——— .1989. Epistemic Justification. Essays in the Theory of Knowledge. Ithaca: Cornell University Press. Apel, Karl-Otto. 1986. “Das Problem der phänomenologischen Evidenz im Lichte einer transzendentalen Semiotik” in M. Benedikt / R. Burger (Hrsg.), Die Krise der Phänomenologie und die Pragmatik des Wissenschaftsfortschrittes. Wien: Verlag der Österreichischen Staatsdruckerei. Audi, Robert. 1993. “Ethical Reflectionism”, The Monist 76. BonJour, Laurence. 1985. The Structure of Empirical Knowledge, Cambridge: Harvard University Press. ——— .1998. In Defense of Pure Reason. Cambridge: Cambridge University Press. Delaney, C.F. 1993. Science, Knowledge and Mind. A Study in the Philosophy of C.S. Peirce. Notre Dame: University of Notre Dame Press. Descartes, René.1984. Rules for the Direction of the Mind. The Philosophical Writings of Descartes, vol. 1. Cambridge: Cambridge University Press. Gadamer, H.G. 1976. “Man and Language” in Philosophical Hermeneutics, translated & edited by David Linge, Berkeley: University of California Press. Hartshorne, Charles and Paul Weis (eds.) 1931-1935,. Peirce’s Collected Papers. Cambridge (Mass.): Harvard University Press. Houser, Nathan and Christian Kloesel (eds.). 1992. The Essential Peirce. Selected Philosophical Writings, Vol.I. Bloomington/Indianapolis: Indiana University Press. Locke, John. 1975. An Essay Concerning Human Understanding. Edited with an

39 introduction by P.H. Nidditch. Oxford: Clarendon Press. Moore, G.E. 1902. Principia Ethica. Buffalo: Prometheus. Moser, Paul K. 1989. Knowledge and Evidence. Cambridge: CUP. Reid, Thomas. 1969. Essays on the active powers of man. ed. Baruch Brody. Cambridge (Mass.): MIT Press. Rorty, Richard. 1979. Philosophy and the Mirror of Nature. Princeton: Princeton University Press. ——— . 1980. “Intuitionism” in The Encyclopedia of Philosophy. Russell, Bertrand. 1948. The Problems of Philosophy. Oxford: Oxford University Press. Sellars, Wilfrid. 1963. Science, Perception, and Reality. London: Routledge and Kegan Paul. Shope, Robert. 1983. The Analysis of Knowing. A Decade of Research. Princeton (N.J.): Princeton University Press. Quine, Willard V.O. 1963. From a logical point of view. New York: Harper & Row. Woudenberg, René van. 1995. "On Ultimate Epistemic Foundations", Ratio 8.

BENCE NANAY

Foundationalism Strikes Back? In Search of Epistemically Basic Mental States I. Foundationalism: Two Distinctions I will defend a version of foundationalism in this paper – a view that is not overwhelmingly popular these days.1 Here is a characterization of foundationalism. A belief is justified if and only if it is either epistemically basic or is justified by an epistemically basic mental state. The question is of course what these basic mental states are. Before addressing this question, I need to make two important distinctions within the category of foundationalism. First, structural and substantive structuralism needs to be contrasted:2 Substantive foundationalism: Epistemically basic states are epistemically basic in virtue of their content. Just on the basis of the content of a mental state one can tell whether it is epistemically basic or not. Structural foundationalism: Epistemically basic states are not necessarily epistemically basic in virtue of their content.

1

See for example Audi 1993a, 1993b, 1998, 2001, Alston 1976, 1989, Kornblith 1980, Triplett 1990, Lehrer 1990, BonJour 1985, Sosa 1980. 2 This distinction comes from Williams 2001.

42 Most foundationalist theories want to endorse the former view. I will also argue for a version of substantive foundationalism. The second distinction is even more crucial: Beat-the-sceptic foundationalism: Epistemically basic states are not susceptible to sceptical worries: they are infallible. Never-mind-the-sceptic foundationalism: Epistemically basic states are not necessarily infallible. The main idea here is that solving the problem of scepticism is not the only possible motivation for foundationalism. We can be foundationalist even if we have solved the problem of scepticism in some other ways (by accepting a reliabilist account of justification, for example). If one no longer worries about scepticism, then it is open to one to conceive of epistemically basic mental states as not necessarily infallible. The former version of foundationalism is sometimes called infallibilist foundationalism, whereas the latter is often referred to as fallibilist foundationalism.3 I will defend a form of never-mind-the-sceptic foundationalism. II. What May Epistemically Basic Mental States Be? I will briefly examine the beat-the-sceptic version of foundationalism and consider some famous arguments that are supposed to show that it cannot work. From examining these arguments, some important constraints can be derived that every account of foundationalism should take into consideration. According to the beat-the-sceptic version of foundationalism, epistemically basic mental states are infallible: we cannot be wrong about them. This is a premise the beat-the-sceptic foundationalist needs in order to block the sceptic worries. The main candidates for such infallible mental states are experiences. After all, we cannot be wrong about our experiences. I can be wrong about whether there is a chair in front of me, but I can't be wrong about whether I have the experience that there is a chair in front of me. The beat-the-sceptic foundationalist encounters a very simple problem at this point. If I cannot be wrong about whether I have the experience 3

Audi 1993a, 1993b, 1998, 2001.

43 that there is a chair in front of me, then experiences cannot misrepresent. But one of the uncontroversial claims of philosophy of mind is that any theory of content must allow for the possibility of misrepresentation: I can have a mental state about a papaya even if I encounter a huge yellow pear that looks very much like a papaya. If I mistake a pear for a papaya, my mental state will still be about papayas, in spite of the fact that I am staring at a pear and there are no papayas around. In other words, any theory of mental content must be able to account for the fact that a mental state can be about something (a papaya) and be triggered (or caused) by something else (a pear). Therefore, if a type of mental state cannot be wrong, then this type of mental state does not have content. And it is difficult to see how a state without content could justify anything. Even if it does, it cannot justify anything in virtue of its content, therefore, they cannot provide the basis of a substantive foundationalist theory. It is important to note that the same argument applies not only to experiences, but to any mental states that are infallible. Thus, epistemically basic mental states cannot be infallible, otherwise they would not be contentful mental states at all. This is the first constraint on the notion of epistemically basic mental state that every foundationalist account needs to take into consideration. The second constraint follows from the notion of epistemically basic mental state itself. If we accept the consequences of the impossibility of beat-the-sceptic foundationalism, then epistemically basic mental states need to be construed as fallible mental states. If, however, the basic states are as fallible as the nonbasic ones, what makes them epistemically basic? There must be some epistemic difference between them and the nonbasic states, on the basis of which we can claim that some states are basic and others are not. If all our states are equally fallible, then the asymmetry inherent to any version of foundationalism (that is, the asymmetry between epistemically basic and epistemically nonbasic states) is lost. Thus, we have seen that the difference between epistemically basic and epistemically nonbasic states cannot be constituted by the fact that basic states are infallible. On the other hand, there must be some epistemic difference between them. I will argue that this difference is that while basic mental states are incorrigible, nonbasic ones are not. A note on the notion of incorrigibility. Incorrigibility is often confused with infallibility. In fact it is a very different, and much weaker, notion. All that is required for the incorrigibility of a mental state is that the

44 agent cannot correct it – the agent has no control over whether she has this mental state or not. This mental state may very well be incorrect, however. Incorrigibility is an important epistemic feature of our mental states. If we managed to point at a set of mental states that are incorrigible and that somehow serve as foundations for other, not incorrigible, mental states, then we would satisfy both conditions I outlined above: the difference between basic and nonbasic states is not infallibility, since both basic and nonbasic mental states are fallible. However, there is an important epistemic difference between basic and nonbasic states: basic states are incorrigible, whereas nonbasic ones are not. Unfortunately, incorrigible mental states are few and far between. Our perceptual experiences (or perceptual states), as we shall see shortly, are not normally incorrigible. There are various visual riddles that are based on this feature of our perceptual system. When we are asked to spot the differences between two seemingly very similar drawings or photographs, then even if we cannot spot some of them, when these differences are pointed out to us, we do see them. In other words, we are very much in the position to correct our perceptual states. In this paper, I will argue that there is a special subcategory of perceptual states that are indeed incorrigible. I will call them ‘action-oriented perceptual states’. Before I define them, I will need to introduce the notion of perceptually guided action. III. Perceptually Guided Action Some actions are perceptually guided, others are not. An agent A's action of type Q is perceptually guided if and only if there is a perceptual state of type P such that the reliable successful performance of tokens of Q is not possible unless A is (or at least has been) in a veridical perceptual state of type P. To put it simply, an agent's action is perceptually guided if its reliable successful performance is not possible unless this agent is (or has been) in a perceptual state of a certain type. First of all, a brief note about the notion of perception in the above definition. Some philosophers argued that proprioception is a form of perception. Whether it is or not, I would like to focus on the perception of dis-

45 tal objects. What is necessary for the successful reliable performance of a perceptually guided action is the perception of a distal object.4 Thus, the definition of perceptually guided actions can be rephrased in the following manner: An agent's action-type Q is perceptually guided if and only if there is an object x (or an object-type X) such that the reliable successful performance of Q is not possible unless she perceives (or has perceived) x (or a token of X). Raising one's arm is not a perceptually guided action. I can perform this action reliably even if I do not perceive any distal object at all.5 The reliable successful performance of this action does not presuppose having or having had any perceptual state. Blinking is not a perceptually guided action either. Playing golf, on the other hand, is a perceptually guided action: the reliable successful performance of the action of playing golf does presuppose perceiving certain objects (presumably the ball and the hole). It is important to emphasize that the distinction between perceptually guided and non-perceptually guided actions is a distinction between actiontypes and not between action-tokens. A certain token action may belong to a perceptually guided action-type even if when the agent performs a token of this action-type, she does not perceive anything. An example may be useful to clarify this point. The action-type of scoring a goal is also perceptually guided: One cannot perform this action reliably unless one perceives (or at least has recently perceived) the goal. If I am blindfolded, and hence I cannot perceive the goal, I may also happen to kick the ball in such a way that the ball ends up in the goal. In other words, I may get lucky. All the same, this token action is a token of a perceptually guided action-type, scoring a goal, the re4

Examples where actions are performed on one's own body (such as scratching one's elbow or touching one's nose) constitute an interesting case. The reliable successful performance of these actions does not presuppose the perception of any distal object, but it does depend on proprioception. Hence, these actions are non-perceptually guided (they may be called proprioceptically guided though). 5 The patient Ian Waterman, described in Cole 1991, provides an interesting case. Because of a viral infection, he is thought to lack the ability of proprioception, and as a result, he can move his arm only if he is looking at it. In a dark room, he cannot move (see also Noe 2004, chapter 1). His action of raising his hand may be described as perceptually guided, but this is so only because he uses distal perception in place of proprioseption. See also the previous footnote.

46 liable performance of which does presuppose a perceptual state of a certain kind (perceiving the goal).6 The perceptual states without which the reliable performance of perceptually guided actions is not possible I call 'action-oriented perceptual states'. In order to score a goal, one needs to perceive the goal. This perceptual state without which the successful reliable performance of the action of scoring a goal is impossible is an action-oriented perceptual state. I will argue that these perceptual states are epistemically basic. First, however, I need to say a bit more about the distinction between perceptually guided actions and actions that are not perceptually guided. Some actions are not that easy to sort into one of these two categories. There are actions that have become so automatic that their reliable successful performance does not seem to presuppose any perceptual state. Touch-typing would be a possible example. One can touch-type reliably without even looking. All the same, the reliable successful performance of these actions is not possible without the agent’s having certain perceptual states. Even if I do not need to look at my fingers when I am typing, the reason why I tend to hit the middle of the keys is that my action is guided by some tactile states. How about the following example? Suppose that the dustbin in my office is a couple of meters behind me so that I cannot see it (nor can I touch it), but I spend so much time in the office that I can throw my teabags into my dustbin quite reliably. Does the reliable successful performance of this action presuppose that I now perceive the dustbin? No. Does the reliable successful performance of this action presuppose that I have perceived the dustbin in the past? Certainly. I would not be able to perform this action so successfully and reliably unless I have perceived the dustbin and its whereabouts in the office. The distinction between perceptually guided actions and actions that are not perceptually guided is an interesting one in itself and probably a lot more could be written about it. The only role this distinction plays in this paper, however, is to help introducing the concept of action-oriented perceptual states.7 This is the concept I now turn to. 6

For simplicity, in what follows I will focus on vision rather than perception in general. However, my argument can be generalized to all the sense modalities. 7 The term 'action-oriented representation' was used by Andy Clark (Clark 1997, pp. 49-51) He defines action-oriented representations as "representations that simultaneously describe aspects of the world and prescribe possible actions, and are poised between pure control structures and passive representations of external reality" (Clark

47

IV. Action-oriented perceptual states Action-oriented perceptual states are perceptual states without which the reliable successful performance of perceptually guided actions is impossible. Suppose that I am looking at object x (or an object of type X). My perceptual state is action-oriented if and only if I am performing an action, the reliable performance of which is not possible without perceiving x (or an X) veridically. More precisely, Agent A's perceptual state R at time t is action-oriented if and only if there is an object x (or object type X) and a perceptually guided action Q such that (1) A performs a token of Q at time t (2) A perceives object x (or a token of X) at time t (3) A's reliable successful performance of Q is not possible unless A perceives x (or an X) veridically. Some of our perceptual states are action-oriented, some are not. If I am running to catch my bus, then I see the lamppost in my way in an action-oriented manner. If, on the other hand, I am sitting on a bench in front of the same lamppost admiring it without any particular urge to perform any action, then I am likely to see it in a non-action-oriented manner. Why should we be interested in action-oriented perceptual states? First, because very many of our perceptual states are action-oriented: we perform perceptually guided actions all the time and most of these actions require that we are in some action-oriented perceptual state. More importantly, action-oriented perceptual states are in some sense more basic than perceptual states that are not action-oriented. The perceptual states of animals are likely to be action-oriented. Some animals do perform perceptually guided actions. They escape from predators, approach food, approach their potential mate. The reliable successful performance of these actions is not possible unless the animal is in some veridical perceptual state (presumably, unless it perceives the predator, the food or the potential mate). Hence, it seems uncontroversial that some animals can be in 1997, p. 49). This notion is not really explained, but on the basis of Clark's definition it at least does not contradict my notion of 'action-oriented perceptual state'. See also Clark 1995, Clark 2001, p. 85. Millikan may also mean something similar by her notion of 'pushmi-pullyu' representation (Millikan 1996b).

48 action-oriented perceptual states, whereas it is not at all obvious that these animals can be in perceptual states that are not action-oriented. The same may be true for the perceptual states of small children. Small children do perform perceptually guided actions, therefore, they can be in action-oriented perceptual states. Whether they can be in perceptual states that are not action-oriented is unclear. Thus, it appears that actionoriented perceptual states are in some sense more basic than ones that are not action-oriented from both a phylogenetic and an ontogenetic point of view. I will argue that action-oriented perceptual states are more basic even from an epistemic point of view: they are incorrigible. Before arguing for this claim, however, I need to show that the distinction between actionoriented perceptual states and the rest of our mental states is a real one: the content of action-oriented perceptual states and the content of the rest of our mental states is structurally different. V. The Content of Action-Oriented Perceptual States At the beginning of the paper I argued that most foundationalists want to endorse a version of substantive foundationalism: an account whereby epistemically basic states are epistemically basic in virtue of their content. I will argue in this section that there is a major difference between the content of action-oriented and non-action-oriented perceptual states: the content of action-oriented perceptual states depends counterfactually on the action one performs, whereas the content of the rest of our mental states does not necessarily do so. In other words, whereas the action I perform is part of what individuates the content of my action-oriented perceptual state, it is not necessarily part of what individuates the content of the rest of my mental states. First, what does this claim mean exactly? Under some interpretation, this claim is obviously true. Everyone would agree that the action I perform at t1 does influence my perception at t2, if t2 follows t1. For example, the action of turning my head at t1 does influence my perception in the next moment. There are some other fairly obvious examples of action-perception dependence that everyone would accept: when one is perceiving one's own action, of course, the content of one's perceptual state depends on the action one performs. Also, if I am driving a car, I will see the two sides of the road passing by - something I would not see were I not driving a car.

49 In all of these three examples, the action the agent performs influences her sensory stimulation, thus, her perceptual state. When I am looking at my hand while reaching out to take a sip from a glass and when I am looking at my hand while ringing the doorbell, my sensory stimulation will be different in the two cases, therefore, it is not surprising that the content of my perceptual state will also be different. Also, when I am driving the car, my sensory stimulation depends on this action, therefore, the content of my perceptual state does too. My claim is that the content of one's action-oriented perceptual state depends on the action one performs even if the sensory stimulation is the same. When I am looking at a glass of wine while reaching out to drink it, my perceptual content will be different from what it would be if I were looking at the same glass of wine while reaching out to pour it under the table even if my sensory stimulation is the same. If the action I perform were different, the content of my perceptual state would be different, even if my sensory stimulation were the same. We have seen that this is not an obviously true claim, but it is not so radical either. The content of our perceptual states depends counterfactually on lots of things. For example, when I look at the duck-rabbit drawing and I see it as a duck-picture and when I look at the same drawing and I see it as a rabbit-picture, then the content of my perceptual state is different in the two situations, in spite of the fact that everything else, including my sensory stimulation, is the same. My claim is structurally similar to this: the content of one’s perceptual state also depends counterfactually on the action one performs. One could ask what the motivation for this claim is. Why would anyone be tempted to say that the content of some of one's perceptual states depends on the action one performs? The simple reason is the following. When I am climbing a tree and when I am hiding behind it, the way I see the tree is different, even if my sensory stimulation is the same: in the first case I see it as climbable or suitable for climbing (or affording climbing), whereas in the second case, I see it as suitable for hiding behind (or affording hiding behind). In other words, when I perform action Q, I see the object on or with the help of which I am performing the action as affording action Q. When, on the other hand, I perform another action, Q*, I see the same object as affording action Q*. When I do not perform any action, then I do not necessarily see the object I am looking at as affording an action. This may

50 sound like an intuitively appealing way of rephrasing my claim, but it also raises some worries. First, I can see an object as affording a certain action even if I do not perform this action, I am only tempted to do so. Thus, being in an actionoriented perceptual state is not to be identified with seeing an object as affording an action. So probably describing action-oriented perceptual states with the help of the intuitively appealing notion of seeing an object as affording an action is not so helpful after all. My answer is the following: I aim to show that the content of actionoriented perceptual states depends counterfactually on the action the agent performs. I do not aim to show that action-oriented perceptual states are the only mental states that are such that their content depends counterfactually on a certain action of the agent. It may be the case that if I see a cake as edible without actually eating it, the content of my perceptual state depends on the action I am inclined to perform. To put it simply, I will argue that if one is in an action-oriented perceptual state, then one sees an object as affording a certain action, but one may be able to see an object as affording a certain action even if one is not in an action-oriented perceptual state. The second, and much more serious, problem with using the term 'seeing something as affording certain actions' is that one could deny that there is such a thing. One could argue that when we say that one perceives an object as affording a certain action, this only means that one perceives a certain object and knows that objects of this kind afford a certain action. In other words, one could argue that the action an object affords is not part of the content of our perceptual state, but rather of the belief our perceptual state triggers. This way of describing what it means that we see an object as affording certain actions would be compatible with the thesis that the action one performs does not have any substantial influence on one’s perception. In other words, it would be compatible with the classical picture of perception and action, which I am arguing against. In the next section, I will argue that the action an object affords is part of the content of one’s perceptual state (and not one’s beliefs). In other words, the content of one’s perceptual state depends counterfactually on the action one performs (other things being equal). Now let us see my argument for this claim that when I am in an action-oriented perceptual state, the content of my perceptual state depends counterfactually on the action I perform. David Milner and Melvyn Goodale describe a patient, D. F., who suffered carbon-monoxide-induced visual agnosia (Milner and Goodale

51 1995). D. F. cannot recognize objects or shapes; if she is asked whether an elongated rectangular slot is horizontal or vertical, she cannot tell. She cannot even indicate the orientation of the slot with her hand. If, however, she needs to 'post' an envelope through this slot, she can do so quite reliably (she rotates her hand into the position that is required for the successful performance of this action). Importantly, she cannot do this if the light is turned off or if she is blindfolded. How can we describe D. F.'s action in the conceptual framework I outlined above? The action D. F. is performing is certainly a perceptually guided action: its reliable performance is not possible unless she sees a certain object (the slot). She could perform the action of 'posting' a letter reliably only if the light was not switched off. The perceptual state D. F. is in while attempting to perform this action is an action-oriented one, since (1) she performs the action of 'posting the envelope' through the slot, (2) she perceives the slot and (3) the reliable performance of posting the envelope is not possible unless she sees the slot. Thus, D. F. is in an action-oriented perceptual state. Further, this action-oriented perceptual state must represent the orientation of the slot, otherwise D. F. could not perform the action of posting an envelope through this slot reliably. Now what happens if D. F. does not perform this action? As the experiments show, in this case, she has no way of telling or even indicating with her hand what the orientation of the slot is, even if she is eyeing it keenly. In other words, her perceptual state does not represent the orientation of the slot. Thus, the orientation of the slot is a property D. F.'s perceptual state represents if she performs the action of 'posting an envelope' through this slot (otherwise she could not perform this action reliably), but if D. F. does not perform this action, then her perceptual state does not represent this property. The content of her perceptual state is different if the action she performs is different, even if her sensory stimulation is the same. To put it differently, the content of her action-oriented perceptual state depends counterfactually on what action she performs. This is exactly what we wanted to show. VI. The Incorrigibility of Action-Oriented Perceptual States. Thus, we singled out an interesting subset of mental states: action-oriented perceptual states. We have seen that the content of action-oriented percep-

52 tual states depend counterfactually on the action one performs. Now I only need to show that these mental states are incorrigible. First, what does it mean that an agent's mental state is incorrigible? We have seen that an agent's mental state is incorrigible if she has no control over whether to be in this mental state or not. Beliefs are no incorrigible: if I believe that Arsenal is the best soccer team in the world, and I have enough evidence against this belief, then I can give up this belief of mine. It is important that incorrigibility implies that the agent has no control whatsoever over whether to be in a certain mental state. In the case of some of our mental states, we do have at least some partial control over whether to be in a certain mental state, but this does mean that these states are incorrigible. Perceptual states are possible examples. Our perceptual experiences (or perceptual states) are not normally incorrigible. There are various visual riddles that are based on this feature of our perceptual system. When we are asked to spot the differences between two seemingly very similar drawings or photographs, then even if we cannot spot some of them, when these differences are pointed out to us, we do see them. In other words, we are very much in the position to correct our perceptual states. Also, if I do not see the duck in the famous duckrabbit representation, someone can point out to me where to look for the beak of the duck is and where to look for its eye, and as a result, I may be able to see the duck in the duck-rabbit picture. Therefore, I was in the position to correct and modify my perceptual state. Thus, perceptual states in general are not incorrigible. The main aim of this paper is that a certain subset of perceptual states, namely, action-oriented perceptual states, are indeed incorrigible. Even if we know that an object does not afford action Q, we still cannot help seeing it as affording Q. This seems easy: if I see the depths under my feet as threatening (affording falling down), then even if I know that there is no way I could fall because there is a railing between me and the depths, I still see the depths as threatening (and affording falling down). Similarly, if I see a ball being thrown towards me with great speed, I see it as affording ducking: even if I know it very well that there is a plexiglass between the ball and me, and therefore the ball can never actually hit me, I do see the ball as affording a certain action to me. What these examples suggest is that no matter how hard we try, we just cannot alter our action-oriented perceptual states. They are incorrigible.

53 VII. Conclusion If the argument I presented in the previous sections is correct, then we have found a set of epistemically basic mental states. These mental states are not infallible – hence, we avoid the problems beat-the-sceptic foundationalists face. They are incorrigible though – constituting a major epistemic difference between basic and nonbasic mental states. Also, their content is structurally different from the content of nonbasic mental states – they are epistemically basic in virtue of their content. In other words, if we accept that action-oriented perceptual states are epistemically basic, then we end up with a version of substantive nevermind-the-scpetic foundationalism.

REFERENCES Alston, William P. 1976. “Two types of Foundationalism”, Journal of Philosophy, 73: 165-85. ——— .1989. Epistemic Justification. Essays in the Theory of Knowledge. Ithaca: Cornell University Press. Audi, Robert 1993a. “Contemporary Foundationalism" in Louis Pojman (ed.): The Theory of Knowledge: Classic and Contemporary Readings. Belmont: Wadsworth. ——— .1993b. The Structure of Justification. New York: Cambridge University Press. ——— .1998. Epistemology, London: Routledge. ——— . 2001. The Architecture of Reason. New York: Oxford University Press. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge: Harvard University Press. Campbell, John. 1993. “The Role of Physical Objects in Spatial Thinking” in Naomi Eilan, Rosaleed McCarthy and Bill Brewer (eds.): Spatial Representation. Oxford: Oxford University Press.

54 ——— .1994 Past, Space, and Self. Cambridge, MA: The MIT Press. ——— . 2002 Reference and Consciousness, Oxford: Oxford University Press. Clark, Andy. 1995. “Moving Minds: Re-thinking Representation in the heat of situated action” in J. Tomberlin (ed.) Philosophical Perspectives 9: AI, Connectionism and Philosophical Psychology. Atascadero, CA: Ridgeview. ——— . 1997. Being There: Putting Brain, Body and World Together Again. Cambridge, MA: The MIT Press. ——— . 2001. Mindware. Oxford: Oxford University Press. Edelman, G. M. 1987. Neural Darwinism: The Theory of Neuronal Group Selection. New York: Basic Books. Evans, Gareth. 1982. The Varieties of Reference, Oxford: Clarendon Press. Fodor, Jerry A., Zenon Pylyshyn.1981. “How Direct is Visual Perception? Some Reflections on Gibson's "Ecological Approach"”, Cognition 9: 139–196. Gibson, James J. 1966. The Senses Considered as Perceptul Systems. Boston: Houghton Mifflin. ——— .1979. An Ecological Approach to Visual Perception. Boston: Houghton Mifflin. Hommel, Bernhard, Müsseler, Jochen, Aschersleben, Gisa, and Prinz, Wolfgang. 2001. “The Theory of Event Coding: A Framework for Perception and Action Planning”, Behavioral and Brain Sciences 24: 849-931. Hurley, S. L. 1998. Consciousness in Action. Cambridge, MA: Harvard University Press. Jeannerod, M. 1997. The Cognitive Neuroscience of Action. Oxford: Blackwell. Kornblith, H. 1980. “Beyond Foundationalism and the Coherence Theory” , Journal of Philosophy, 77: 597-612. Lehrer, Keith. 1990. Theory of Knowledge. Boulder: Westview Press. Leibniz, G. W. 1704/1981. New Essays on Human Understanding. Cambridge: Cambridge University Press.

55 Lewis, C. I. 1929. Mind and the World-Order: Outline of a Theory of Knowledge. New York: Charles Scribner’s Sons. Lewis, C.I. 1946. An Analysis of Knowledge and Valuation. La Salle, IL: Open Court. Merleau-Ponty, Maurice. 1945. La phénoménologie de la perception. Paris: Gallimard. Millikan, Ruth G. 1996. “Pushmi-Pullyu Representations” in L. May, M. Friedman and A. Clark (eds.) Minds and Morals. Cambridge, MA: The MIT Press. Milner, A. D., and M. A. Goodale. 1995. The Visual Brain in Action. Oxford: Oxford University Press. Noë, Alva. 2002. “Causation and Perception: The Puzzle Unravelled”, Analysis 63: 93-100. ——— . 2002. “On What We See” in Pacific Philosophical Quarterly 83: 57-80. ——— .2004. Action in Perception. Cambridge, MA: The MIT Press. Peacocke, Christopher. 1986b. “Analogue Content”, Proceedings of the Aristotelian Society Supplementary Volume 60: 1-17. ——— . 1989. “Perceptual Content” in Themes from Kaplan, edited by J. Almong, J. Perry and H. Wettstein. Oxford: Oxford University Press. ——— . 1992a. A Study of Concepts. Cambridge, MA: The MIT Press. ——— . 1992b. “Scenarios, Concepts and Perception” in Tim Crane (ed.): The Contents of Experience. Cambridge: Cambridge University Press. Sosa, Ernest. 1980. “The Raft and the Pyramid: Coherence Versus Foundations in the Theory of Knowledge” in French, Peter A. (ed.): Midwest Studies in Philosophy, Vol. 5: Studies in Epistemology. Minneapolis: University of Minnesota. Triplett, T. 1990. “Recent Work on Foundationalism” in American Philosophical Quarterly 27: 93-116. Ullman, S. 1980. Against Direct Perception. New York: MacMillan. Williams, Michael. 2001. The Problem of Knowledge. Oxford: Oxford University Press.

IGOR DOUVEN

Basic Beliefs, Coherence, and Bootstrap Confirmation Foundationalists and coherentists disagree about the structure of justification. While the former maintain that justification has a layered structure, with basic, self-justifying beliefs ‘at the bottom’ and other justified beliefs being derived from them (in some sense of ‘derived’), coherentists think that there are no self-justifying beliefs and that the justificational status of any belief depends on how well it hangs together with other beliefs. It is no exaggeration to say that, during the 1970s and 1980s, coherentism was the majority position in the field of epistemology. However, several recent publications indicate that foundationalism is making a comeback and that coherentism is losing ground (see, e.g., Plantinga 1993, 2000; Alston 1999; BonJour 1999a, 1999b). At least part of the explanation for this shift in popularity may be that coherentists have never been able to unpack the notion of coherence in an even remotely satisfactory fashion, and that philosophers have become increasingly convinced that it cannot be done. Whether this is a good reason to abandon coherentism in favor of foundationalism is doubtful, however. For it appears that foundationalists must invoke a notion of coherence among beliefs, too. As BonJour (1999a, 124) puts it: [T]he concept of coherence … is also an indispensable ingredient in virtually all foundationalist theories: coherence must seemingly be invoked to account for the relation between the basic or foundational beliefs and other nonfoundational or ‘superstructure’ beliefs, in virtue of which the latter are justified in relation to the former.

The reason for this, BonJour (1999a, 140n) explains, is that, “strictly deductive or even enumerative inductive inference from the foundational beliefs does not suffice to justify most of the superstructure beliefs that the foundationalist typically wants to claim to be justified”.1 And so, as Bon1

It might be suggested that supplementing standard deductive logic (plus, perhaps, enumerative induction, depending on what BonJour means by that) by the so called

58 Jour (1999a, 124) also notes, if the concept of coherence still needs spelling out, that should be no less troubling for the foundationalist than it is for the coherentist. Of course it may be misleading to speak simply of the concept of coherence. Though it cannot be excluded that the concept of coherence foundationalists must invoke and the concept of coherence coherentists must invoke will in the final analysis turn out to be one and the same, there is prima facie reason to believe that, although close cousins, they are not identical twins: the foundationalist concept of coherence must pertain to a relation that holds between a body of non-basic beliefs (or a theory, in a vocabulary that will sometimes be more natural in the context of the discussion to follow) and a body of basic beliefs (the evidence, in that other vocabulary). That is, foundationalist coherence involves beliefs that do not, or at least not initially, all have the same justificational status. For the coherentist, who denies that there is any kind of justification short of that provided by relations of coherence, coherence obtains or fails to obtain between beliefs that all have, or at least may all have, the same justificational status. In this paper I will be concerned only with the foundationalist notion of coherence – in case that differs from the coherentist notion – that is, the notion of a theory cohering with the evidence available at a given point in time (henceforth, the term ‘coherence’ will be understood as referring to this notion). More specifically, I will attempt to render this notion precise by building on the theory of bootstrap confirmation as proposed in Glymour 1980. This theory has received some attention in the philosophy of science in the 1980s but has since practically gone into oblivion. In the epistemological literature, Glymour’s theory even seems to have been overlooked altogether. In any event, epistemologists hitherto have failed to appreciate the theory’s potential to serve as departure point for explicating the concept of coherence. Inference to the Best Explanation (cf. Harman 1965) may be sufficient for justifying all the non-basic beliefs the foundationalist pre-theoretically regards as being justified. However, as for instance McMullin 1996 makes plain, what counts as the best of a number of potential explanations depends, among other things, on how well it coheres with already accepted theory as well as metaphysical presuppositions. In other terms, the notion of coherence seems to be more basic than that of explanation (or at least than that of ‘best explanation’). Apart from that, the Inference to the Best Explanation is a highly controversial rule of inference (cf., e.g., van Fraassen 1989, Ch. 7; but also see Douven 1999, 2002a, 2004, 2005).

59 The paper is in two parts. The first part starts by outlining Glymour’s theory. As will be seen, although the theory as originally proposed can be argued to capture the notion of coherence to some extent, it only permits categorical judgments and therefore is not capable of fully explicating coherence (which clearly comes in degrees). We therefore consider a recent extension of Glymour’s theory that encompasses a measure of degree of bootstrap confirmation and thereby does make graded judgments possible. The second part proposes a rather straightforward analysis of coherence in terms of the extended version of Glymour’s theory. Furthermore, it considers the question of how to formulate a theory of justification on the basis of the notion of coherence defined. Various theories of justification are suggested that all seem worthy of the epithet ‘coherentist’.2 The question which of these theories provides the correct account of justification, if any, I will not endeavor to answer in this paper, though I will make some general remarks on how in my view we should proceed in order to obtain an answer to that question. I. What philosophers of science call a confirmation theory is, roughly put, a theory that purports to specify, for any given evidence statement E and any given hypothesis H, whether or not E is evidence for H, i.e., whether or not E supports H, or, again in different, more colloquial terms, whether coming to know that E should increase or decrease our confidence in H (or perhaps should not affect our confidence in that hypothesis at all). Such a theory may or may not also specify to what extent E should affect our confidence in H. Theories that do are called quantitative confirmation theories, those that do not are called qualitative confirmation theories. While the foregoing formulations suggest that confirmation is a twoplace relation – viz., a relation between an evidence statement and a hypothesis – Duhem (1906/1954) famously pointed out that confirmation is really a three-place relation: evidence generally accrues to a hypothesis only relative to one or more auxiliary hypotheses. To appreciate this point, 2

To forestall misunderstanding, note that theories of justification that are coherentist in the present sense are still foundationalist; it is just that they are foundationalist theories that allow a role to coherence in empirical justification. The name ‘foundherentist’ were perhaps more appropriate for these theories, had that name not already been occupied by Haack 1993 to denote a theory of justification that at best bears some vague similarities to the theories to be presented in this paper.

60 notice that even in forming ‘simple’ perceptual beliefs, we are typically relying on auxiliary hypotheses (such as that we are not dreaming or hallucinating, that lighting conditions are normal, etc.). And to give an example that is less trite: Physicists universally take the deflection of cathode rays toward a metal body charged with positive electricity to constitute evidence for the hypothesis that cathode rays are negatively charged. But of course in doing so they are relying, inter alia, on the background assumption that opposite electric charges attract. Many have taken Duhem’s insight to imply that all confirmation must be ‘relative confirmation’ (some have even taken it to open the door to epistemological relativism). However, according to Glymour (1980), the indispensability of auxiliaries in the testing of single hypotheses is no impediment to absolute confirmation of complexes of such hypotheses or, as we may call them, theories. More exactly, Glymour presents a confirmation theory on which piecemeal confirmation of each of the axioms of some theory relative to one or more other axioms of the same theory under certain conditions adds up to unrelativized confirmation of that theory as a whole. In a capsule formulation, Glymour’s theory comes to the following definition:3 Definition 1 (Bootstrap Confirmation) Let theory T have axioms H1 ,…, Hn. Then evidence E bootstrap confirms T exactly if T ∪{E} is consistent and for each Hi (1 ≤ i ≤ n) hold the following conditions: 1. there is a set S ⊂ {H1,…, Hn} such that Hi ∉ S and (a) E confirms Hi with respect to &S and (b) there is potential – but non-actual – evidence E’ such that E’ disconfirms Hi with respect to &S; 2. there is no S’ ⊆ {H1,…, Hn} such that E disconfirms Hi with respect to &S’. While it will be clear why clauses 1(a) and 2 are part of the definition, it may be less clear what clause 1(b) is needed for. The point of this ‘nontriviality clause’, as it is sometimes called, is that the auxiliaries in each of the particular tests should not guard the axiom being tested from disconfirmation, or, in other words, they should not trivialize the test. The rationale for this requirement is that if a positive test result for whatever axiom 3

In the following I will use a big ampersand to denote the generalized conjunction operator.

61 happens to be tested is guaranteed, whatever the evidence, then we are not really performing an empirical test anymore.4 As Glymour points out, definition 1 does not assume a particular theory of simple, non-bootstrap confirmation. Bootstrapping can, for instance, be combined with a Hempelian positive instance account of confirmation or with a probabilistic analysis of confirmation. In Douven and Meijs 2005 it is explained that, if probabilistic confirmation is assumed – as is most suited for my concerns in this paper, and as seems justified given the current popularity of Bayesian confirmation theory (cf. Earman 1992) – then clauses 1(b) and 2 of definition 1 can be conjoined, resulting in the following definition: Definition 2 (Probabilistic Bootstrap Confirmation) Evidence E probabilistically bootstrap confirms theory T with axioms H1 ,…, Hn precisely if T ∪{E} is consistent and for each Hi (1 ≤ i ≤ n) it holds that 1. there is a S ⊂ {H1,…, Hn} such that Hi ∉ S and p(Hi | (&S)&E) > p(Hi | &S); and 2. there is no S’ ⊆ {H1,…, Hn} such that p(Hi | (&S’)&E) < p(Hi | &S’). It should be noted that the set referred to as S in the first clauses of both definitions is not allowed to contain the axiom under scrutiny itself. This is a later addition to Glymour’s theory. In his first presentation evidence could confirm a hypothesis relative to a set of auxiliaries including that hypothesis itself (this is sometimes called ‘macho-bootstrapping’). However, it was shown by several authors that this had some highly unintuitive consequences (see, e.g., Christensen 1983; Edidin 1983; van Fraassen 1983a). (There is no similar restriction in the second clauses of the definitions. And indeed, it seems that if a hypothesis is disconfirmed by the evidence, it is no less – but rather more – damaging to that hypothesis, and hence also to any theory that includes it, when the hypothesis served itself as an auxiliary in that test than when only other hypotheses did.) How can any of these definitions help with our project? There are two intuitions about coherence that everybody has (and there seems to be no third that is equally generally held), namely, first, that coherent beliefs ‘hang together’, and second, that coherence comes in degrees. Bootstrap confirmation, on any of the foregoing definitions, definitely seems to capture the first of these intuitions. For consider that a positive bootstrap test 4

See Douven and Meijs 2005 for a more extensive discussion of definition 1.

62 from evidence E of a theory T with axioms H1 ,…, Hn is an indication that E and the axioms of T hang together in a very clear sense: the positive result from E for each Hi in T separately has been obtained with the help of other axioms of T, which in turn have also been confirmed by E with the help of axioms in T – including, perhaps, Hi. That is, the axioms and the evidence hang together in the precise sense that the evidence empirically supports each of the axioms if the axioms are allowed to help each other in obtaining support from the evidence. Still, if we now were simply to equate ‘E bootstrap confirms T’ with ‘The elements of the set {H1 ,…, Hn, E} cohere with each other’, and then leave it at that, the second intuition – i.e., that coherence is a matter of degree – would get lost; bootstrap confirmation, whether understood probabilistically or non-probabilistically, is a matter of yes or no. So, while we would like to be able to say that a particular set of beliefs is coherent to a certain degree – depending on how tightly the beliefs in that set hang together – coherence-as-bootstrap-confirmation would only allow us to say that the set is coherent (or is not coherent). And where we would want to express that one set of beliefs is more coherent than a second set of beliefs, coherence-as-bootstrap-confirmation would only allow us to say that both are coherent, or that neither is, or that one is but the other is not. The way around this difficulty seems obvious: Provide a quantitative account of bootstrapping, one that allows us to express that a theory is bootstrap confirmed to a given extent, and that one theory is bootstrap confirmed to a larger extent than another. Recently a start with this project has been made in Douven and Meijs 2005. For present purposes, we do not need to go into all the details of the quantitative account offered there, nor in effect do we need that account as a whole. We do need what is at the heart of it, though, to wit a function measuring the degree of bootstrap confirmation received by a theory from the evidence. Before describing this function, I shall say a few words about the desiderata of this function. What relations between the evidence and the theory and/or between the axioms of the theory themselves do intuitively matter to the degree of bootstrap confirmation that theory receives from the evidence (provided it is bootstrap confirmed by the evidence)? In Douven and Meijs 2005 it is argued that the degree of bootstrap confirmation of a theory T with axioms H1 ,…, Hn by evidence E should be a function of: 1. The degree to which, in the separate tests involved in a bootstrap test of T, the Hi are (non-bootstrap) confirmed by E. Suppose that axiom

63 H1 of T is confirmed by E relative to axiom H2 of T in the measure x (given some measure of confirmation) while axiom H1* of theory T* is confirmed by E* relative to axiom H2* of T* in the measure y, with y > x. Then we would certainly want that, ceteris paribus (!), the measure of bootstrap confirmation accords a higher value to T* given evidence E* than it accords to T given evidence E. 2. The ‘variety of testing’ of T allowed by E. Glymour (1980, 76f) stresses the importance of testing each axiom of a given theory in as much ways as possible. It cannot be excluded that, in testing one axiom relative to another, an error in one compensates for an error in the other, thus leading to spurious confirmation of the axiom that is being tested. But the more ways in which we can test an axiom, the better we will be guarded against such spurious confirmation. So, if, for instance, the axioms of some theory T are all (non-bootstrap) confirmed relative to at least two different other axioms of T, then, ceteris paribus, T should get assigned a higher degree of bootstrap confirmation than T*, a theory all of whose axioms can only be confirmed relative to one other axiom of T*. 3. The degree of non-triviality of the separate tests of the Hi that together constitute the bootstrap test of T. Glymour’s original theory demands, as we saw, that the auxiliaries do not shield the hypothesis under scrutiny from disconfirmation. However, it would seem unrealistic to assume that shielding against disconfirmation is a yes or no affair; auxiliaries can shield a hypothesis from disconfirmation to a greater or lesser degree. And it seems clear that a test is more telling the less the auxiliaries shield the hypothesis from disconfirmation. To accommodate this intuition, the quantitative bootstrap theory should take account of the degree of triviality of each of the tests passed by the various hypotheses of a theory. Though the possibility that there are further desiderata must be left open, the foregoing do seem to me to be the main ones. I shall now first describe the function devised in Douven and Meijs 2005 on the basis of the foregoing desiderata and then briefly explain how it satisfies them. Let T have axioms H1 ,…, Hn. Then for each Hi of T there are exactly 2 n −1 sets of auxiliary hypotheses that are also axioms of T and relative to which Hi can be tested, namely, all the elements of the power set of {H1,…, Hi – 1, Hi + 1, …, Hn} (recall that it is assumed that machobootstrapping is not allowed). Given an ordering 〈 H iT ,K, H iT 〉 of those ele1

2n −1

64 ments, ‘ H iT ’ denotes the jth member of that ordering. ‘ &H iT ’ denotes the conjunction of the axioms in H iT . Now the measure of bootstrap confirmation, B, can be defined as follows: j

j

j

Definition 3 (Measure of Bootstrap Confirmation) B (T , E ) =df

n

2 n−1

∑∑ d * ( H , &H i

i =1 j =1

T ij

& E) .

Here ‘ d * ( H i , &H iT & E ) ’ stands for the function measuring the (nonbootstrap) confirmation Hi receives from E relative to &H iT . It is an adapted version of the familiar difference measure of confirmation – according to which the degree of support a hypothesis H receives from E equals p ( H | E ) − p ( H ) – and is short for p ( H i | & H iT & E ) − p ( H i | &H iT ) . It is clear from this definition that, the higher the degree of nonbootstrap confirmation an axiom receives from the evidence in a given test, the more that test contributes to the value of the degree of bootstrap confirmation of the theory as a whole. Thereby the function satisfies the first desideratum. As to the second desideratum: that is realized in virtue of the fact that the function B adds up the support the evidence provides to each of the hypotheses in the theory relative to each of the possible sets of auxiliaries taken from that same theory. Finally, in Douven and Meijs 2005 it is argued that the conditional probability of an axiom given one or more other axioms can be regarded as measuring the degree of triviality for a test of the former involving exactly the latter as auxiliaries. Given this, and given the definition of d*(-,-) it will also be clear that our measure of bootstrap support also satisfies the third desideratum. To end this part I briefly want to illustrate, by means of a very simple example, how definition 3 ‘works’. Theory T has axioms H1, H2 and H3, each of which has an initial probability of .25, i.e., p(Hi) = .25 for i = 1, 2, 3; the axioms are mutually probabilistically independent, i.e., the probability of each given any of the others alone as well the probability of each given the conjunction of the others equal .25, too. Evidence E has an initial probability of .5. Further we have: p(Hi & E) = .125, p(Hi & Hj & E) = .05 (i, j = 1, 2, 3; i ≠ j ) and p(H1 & H2 & H3 & E) = .015. Each axiom is confirmed by E, both with respect to each of the other two alone and with respect to the conjunction of the others, as can be seen as follows: j

j

j

j

65 p(Hi | Hj & E) = p(Hi & Hj & E)/p(Hj & E) = .1/.125 = .8 > .25, for all i, j = 1, 2, 3; i ≠ j . p(Hi | Hj & Hk & E) = p(H1 & H2 & H3 & E)/p(Hj & Hk & E) = .015/.05 = .3 > .25, for all i, j, k = 1, 2, 3; i ≠ j ≠ k . Thus, E bootstrap confirms T (cf. definition 2). But to what extent does E bootstrap confirm T? To answer this question in accordance with definition 3, we must sum the amount of probabilistic support each Hi receives relative to the other two axioms separately as well as to their conjunction. Spelled out for H1, this yields the following value: d*(H1, H2 & E) + d*(H1, H3 & E) + d*(H1, H2 & H3 & E) = [p(H1 | H2 & E) – p(H1 | H2)] + [p(H1 | H3 & E) – p(H1 | H3)] + [p(H1 | H2 & H3 & E) – p(H1 | H2 & H3)] = (.8 – .25) + (.8 – .25) + (.3 – .25) = 1.15 Since, as is easily seen, we get the same value for each of the other two axioms, the degree of bootstrap support T receives from E equals 3 × 1.15 = 3.45. II With the measure of bootstrap confirmation in hand, it is only a small step to an analysis of coherence. In the previous section, it was said that equating bootstrap confirmation with coherence would fail to entirely accommodate our intuitions concerning coherence. In particular, doing so would not accommodate our intuition that coherence comes in degrees. But now that we can express the degree of bootstrap support a theory receives from the evidence we can easily solve that problem: Just equate coherence with bootstrap confirmation and degree of coherence with degree of bootstrap confirmation. More exactly, given that we are concerned with foundationalist coherence, i.e., with the coherence of pairs consisting of a theory and a body of evidence, the proposal is to equate the coherence of a given theory T together with evidence E with T’s being bootstrap confirmed by E and the degree of coherence of T together with E with the degree of bootstrap confirmation T receives from E. One might object that, while this provides an analysis of the notion of (foundationalist) coherence, it is far from clear that it provides the (true) analysis of that notion. Note, however, that what could be meant by ‘the analysis of the notion of (foundationalist) coherence’ is itself far from clear. Our intuitions concerning this notion

66 simply seem to be too vague to fix a unique, precise concept. It seems that analyses of notions such as this are bound to contain an element of stipulation (cf. Carnap 1962, Ch. 1). The philosophical value of such analyses therefore not only depends on how well they match our pre-theoretical convictions but also, for instance, on how fruitful they are in suggesting further theoretical work. Let us see, then, how far we can get with formulating a (foundationalist) theory of justification by means of the hereproposed concept of coherence. Clearly, equating coherence with bootstrap confirmation and degree of coherence with degree of bootstrap confirmation does not amount to a theory of justification specifying when a belief, or a body of beliefs (a theory, as I call it), that is/are neither basic nor deductively derived from basic beliefs is/are justified. It is in fact not quite a straightforward task to turn the analysis of coherence as proposed here into a genuine theory of justification. For instance – to dismiss right away what may seem the most obvious theory of justification in terms of coherence – we will not simply want to say that a theory T is justified, given evidence E, if T coheres with (i.e., is bootstrap confirmed by) E, irrespective of the degree of coherence, that is (nor, derivatively, that an individual belief is justified by the evidence if it is part of or implied by a theory that coheres with the evidence). And – to dismiss an only slightly less straightforward attempt – it does not make sense to say that E justifies T if T and E maximally (or nearly maximally) cohere, for there is no upper bound on the range of B. But then how are we to define justification in terms of coherence? The measure of coherence we obtain from definition 3 quantifies properties of a given theory in the light of the available evidence that intuitively all seem to be ‘good’ properties of that theory given the evidence; the more coherent the theory with the evidence, the better it seems to be, epistemically speaking. One thus might be tempted to think that coherence, whatever its exact relation to justification, in any case adds to the justificational status of a theory in the following, direct sense: COH If one theory is more coherent with the evidence than a second theory, then we are more justified in believing the former than we are in believing the latter. However, plausible though it may at first seem, the principle is, as it stands, problematic. (Note that, even if it were not problematic, COH would not be a theory of justification, at least not if we want such a theory

67 to inform us about when we are justified in believing some theory – instead of just about when we are more justified in believing this theory than that. And for reasons to be found in, e.g., Maher 1993, Mayo 1996, and Douven 2002b, I believe that a categorical theory of justification cannot be dispensed with.) To see why, we consider a problem that seems to be faced by any attempt to formulate a theory of justification that lets coherence among beliefs add to the justificational status of those beliefs, and that therefore we may simply term the coherence problem. The problem is this: Often, adding a belief to a set of beliefs will result in a more coherent set (cf., for instance, Klein and Warfield 1994). However, unless a belief is absolutely certain, or is implied by some belief(s) already in the set, adding it to the set will result in a set with lower total probability than that of the set minus the belief. Thus, often the more coherent of two sets of beliefs is less likely to be true. Now it is also the case that many philosophers subscribe to the following principle:5 PROB

If one theory is more likely to be true than a second theory, we cannot be more justified in believing the latter than we are in believing the former.

Given this principle, the coherence problem makes one wonder how justification can be a matter of coherence. The coherence problem is mostly presented in terms of an informal notion of coherence. One might perhaps hope that it does not arise for our formal notion of coherence understood as degree of bootstrap support. Unfortunately, however, it does, as is brought out by the following theorem (for a proof see Douven and Meijs 2005): Theorem. For all T, T*, and E, if the set of axioms of T* is a proper subset of the set of axioms of T, and E bootstrap confirms T (according to definition 2), then B(T , E ) > B(T *, E ) . So, if a theory T is bootstrap confirmed by E, it has a higher degree of coherence in light of the evidence than any of (what we might call) its proper subtheories. However, we also have for any proper subtheory T* of T that, unless E in conjunction with the axiom or axioms in T not in T* implies T, 5

For instance, in his 1983b van Fraassen presents it as a truism.

68 p (T | E ) < p (T * | E ) .

Thus it is not generally possible to jointly maximize coherence (as degree of bootstrap support) and probability. But principle PROB can be, and famously has been, denied. Levi (1967), Kaplan (1981a), (1981b), Lehrer (1990), Maher (1993), and others, have argued that justification has the structure of a decision-making problem. In such a problem, one does not just heed the probabilities of the possible outcomes of a certain decision, but also their utilities. More exactly, in decision making one chooses the option that has the greatest expected utility, where this is defined as the sum of the utilities of the option’s various possible outcomes weighted by the probabilities of those outcomes. According to the aforementioned authors, there is nothing in the way decision theory is set up that would prevent applying it to matters epistemological; we can perfectly well assign cognitive or epistemic utilities to accepting particular hypotheses or theories under particular circumstances. For instance, let table 1 represent the cognitive utilities a given agent assigns to the acts of accepting some theory T, rejecting T (i.e., accepting ¬T), and suspending judgment on T, given that T is true and given that T is false, respectively.

Accept T Reject T Suspend judgment

T is T is true false 3 −1 1 −2 0 0

Table 1: Assignment of cognitive utilities Suppose further that according to the agent p(T) = 1/3 (and thus p(¬T) = 2/3). Then the expected cognitive utility of accepting T, relative to the agent’s utilities and probabilities, equals (1/3)(3) + (2/3)( −1 ) = 1/3; the expected cognitive utility of rejecting T equals (1/3)( −2 ) + (2/3)(1) = 0; and, finally, the expected cognitive utility of remaining agnostic with respect to T equals (1/3)(0) + (2/3)(0) = 0. Hence, accepting T has greatest expected cognitive utility for the agent, even though the agent finds T less likely than ¬T. And so we may conclude (according to the aforementioned authors) that the agent is more justified in accepting T than in accepting ¬T, notwithstanding principle PROB. (Since the options are exhaustive in this case, on most approaches to cognitive decision theory the agent would also

69 be absolutely justified in accepting T.6) Put more generally, we may on this approach to justification well be more justified in believing one theory than we are in believing a second even if the former is less likely than the latter, because the former may well have a greater expected cognitive utility than the latter. Applied to our coherence problem, this decision-theoretic approach yields something like the following: We are more justified in believing the set plus the additional belief than in believing just the set (without the additional belief, that is) exactly if doing the former has greater expected cognitive utility than doing the latter (whether we are also absolutely justified in believing the former would depend on the expected cognitive utilities of whatever other options may exist – in this case the options mentioned are not, or at least need not be, exhaustive). So far I have not said anything specific about what utilities are. And doing so is no easy matter. Synonyms are not hard to come by: Utilities are desirabilities, intensities of desire, values. There is little agreement, however, about what these terms are supposed to refer to. Unsurprisingly, some authors have expressed what have come to seem very legitimate doubts about the meaningfulness of the notion of utility (see, e.g., Gillies 2000; Howson 2000). Cognitive utilities even appear to be more problematic (cf. Goosens 1976). They are often said to depend on the informativeness of a hypothesis or theory (cf. Lehrer 1990; Maher 1993). But the notion of informativeness itself is still very much in need of clarification. It therefore seems, at least at the time of this writing, that the cognitive decision theory just described still leaves something to be desired. It is not my intention to defend this theory. The reason that I am bringing it up here is that, in my opinion, it is an important insight of cognitive decision theory that justified belief is not merely a matter of probability, and that to determine the justificational status of a hypothesis or theory we must consider other properties of that hypothesis/theory, too, properties that it may not generally be possible to maximize in tandem with probability. My own proposal is a different attempt to exploit the same insight. The chief difference between my proposal and the foregoing account 6

On some presentations of decision theory we are always to assume that the options facing the agent are exhaustive (even if they are not logically exhaustive, that is). That is a very unrealistic assumption, however, and if we were required to make it standardly – which I think we are not – that would much limit the applicability, and thereby the interest, of decision theory. See for a discussion of this issue Douven 2002c.

70 is that the former replaces the ill-defined concept of cognitive utility by that of coherence, a concept that I have tried to give a rigorous definition. More exactly, where the foregoing decision-theoretic approach to justification weighs probability and cognitive utility against each other, I propose to weigh probability and degree of coherence against each other. There is more than one way in which this can be realized, but a very direct approach begins by defining justification for theories as follows: Definition 4. Theory T is justified, given that evidence E has occurred, precisely if: 1. E bootstrap confirms T; and 2. there is no T* such that: (i) T and T* are jointly inconsistent; (ii) E bootstrap confirms T*; and (iii) p(T * | E ) B(T *, E ) ≥ p(T | E ) B(T , E ) .7 It then defines justification for single hypotheses derivatively, thus: Definition 5. Hypothesis H is justified, given evidence E, precisely if: 1. there is some T such that T is justified by E according to definition 4; and 2. T logically implies H. It is worth mentioning that, on this account of justification, both principle COH and principle PROB hold conditionally: Principle COH holds if all theories competing for acceptance have the same probability, principle PROB if all have the same degree of coherence. Furthermore, it should be noted that given these definitions it can never occur that one and the same body of evidence justifies both a hypothesis and its negation. For suppose some hypothesis H and its negation are both justified by the evidence according to definition 5. Then there must be some theory T such that T logically implies H and T is justified by the evidence according to definition 4, and also some theory T* such that T* logi7

Notice that for some theories T, B(T, E) may not be defined – e.g., take the negation of theory T of the example given at the end of the first part of this paper, ¬T ≡ ¬H1 ∨ ¬H 2 ∨ ¬H 3 . This theory consists of only one hypothesis; if machobootstrapping is disallowed, then obviously it is not possible to bootstrap test this theory. However, in case degree of bootstrap support is not defined for a theory it trivially cannot constitute a theory T* such that, for some other theory T, p(T * | E ) B(T *, E ) ≥ p(T | E ) B (T , E ) .

71 cally implies ¬H and T* is justified by the evidence according to the same definition. Now it cannot be that T = T*, for then T would be inconsistent and thus could not be bootstrap confirmed by any evidence at all (as is immediate from definition 2) so that H could not be justified in virtue of T’s being justified (and T’s entailing H), contrary to what we are assuming. However, nor can it be that case that T ≠ T*, for given that T logically implies H and T* logically implies ¬H, it must be that T and T* are jointly inconsistent. And it follows immediately from definition 4 that in that case at most one of them can be justified by the evidence so that also at most one of H and ¬H can be justified, again contradicting what we are assuming. Hence, our assumption that both a hypothesis and its negation can be justified by the same evidence in either case leads to a contradiction and thus must be false. I said that this is one straightforward manner in which to define justification in terms of coherence and probability, and that there are others. Instead of definition 4, we might for instance adopt one of the following definitions (and alter definition 5 accordingly): Definition 6. T is justified, given that E has occurred, precisely if: 1. E bootstrap confirms T; 2. there is no T* such that: (i) T and T* are jointly inconsistent; (ii) E bootstrap confirms T*; and (iii) p(T* | E) > p(T | E); 3. for all T’ such that T and T’ are jointly inconsistent, if E bootstrap confirms T’ and p(T | E) = p(T’ | E), then B(T, E) > B(T’, E). This definition lets probability dominate bootstrap support in the sense that it only considers the bootstrap support rival (bootstrap confirmed) theories receive from the evidence if they have the same probability conditional on the evidence. Of course this can be reversed, resulting in Definition 7. T is justified, given that E has occurred, exactly if: 1. E bootstrap confirms T; 2. there is no T* such that: (i) T and T* are jointly inconsistent; (ii) E bootstrap confirms T*; and (iii) B(T*, E) > B(T, E); 3. for all T’ such that T and T’ are jointly inconsistent, if E bootstrap confirms T’ and B(T, E) = B(T’, E), then p(T | E) > p(T’ | E).

72 And this evidently still is not an exhaustive list of alternatives for definition 4. But even if it were, how could we determine which definition provides the true theory of justification (if any)? As intimated at the outset, it is not my intention to defend in this paper a particular theory of justification. Nevertheless, in closing I want to say a few words about how I believe we should go about adjudicating between different theories of justification (whether the ones here formulated or any others). So far as I can see, the matter cannot be adjudicated on a priori grounds. In effect, that the history of epistemology is littered with unsuccessful attempts to argue for this or that theory of justification in an a priori fashion should make us suspicious of any new attempt to accomplish the same. I therefore believe that, if it can be accomplished at all, theories of justification should be evaluated empirically. At least in rough outline the procedure for doing so is clear, and is best explained by means of an abstract example in which we assume, for simplicity, that we have it down to two theories of justification, to wit, the ones defined by definitions 6 and 7; all other possible theories of justification have already been rejected. Let Et be our total available evidence at time t. Further suppose that theories T1 and T2 are rivals, are both bootstrap confirmed by Et, and are the only extant theories that are bootstrap confirmed by Et. As is easy to show, it is logically possible for the following conditions to obtain together: (1) (2)

p(T1 | Et) > p(T2 | Et); B(T2, Et) > B(T1, Et).

If these conditions hold, then definition 6 justifies believing T1 while definition 7 justifies believing T2. Now the fact that Et bootstrap confirms both theories does of course not guarantee that an extension Et’ of Et will also bootstrap confirm both. In fact, it may well happen that our evidence at time t’ falsifies one but not the other of these theories. Suppose that is what happens. Suppose, more specifically, that Et’ falsifies T2 but not T1. Then this seems to constitute evidence – however minimal – for definition 6; if we have or had accepted T1 at t we have or would have made a choice that, evaluated at t’, must from a purely epistemic perspective appear to be better than the choice that was suggested by the other definition. We might then start looking for further evidence by evaluating similar choices we made or could have made

73 in the past. And it just might turn out that our further findings also favor definition 6. If that happens, then we would have something of a case for definition 6 and against definition 7, even to the extent that we may be led to embrace the first of these theories of justification. Needless to say, even if things turn out this way, there is no guarantee that future evidence will not point in a different direction. That only means, however, that a verdict concerning the correctness of a particular theory of justification can only have a provisional character – as have many other verdicts about empirical matters. Plainly, the example greatly simplifies. The empirical findings could very easily be less univocal than it supposes. And even if they are not, there are many subtle difficulties associated with the kind of meta-level testing procedure the example suggests. However, I do not believe that such testing procedures are beset by difficulties that are insurmountable in principle. For instance, in Douven 2002a, 2004, 2005 I show how empirical evaluation of the rule of Inference to the Best Explanation is possible in a way that circumvents problems that many believed would render such an evaluation of little significance at best. And I do not see how adjudicating between theories of justification involving the notion of coherence as defined in this paper is essentially different from testing Inference to the Best Explanation.8 Only future work can tell, however, whether I am not being too optimistic about the prospects for the empirical project that here could only be presented in outline.

8

One difference might be that, while in testing Inference to the Best Explanation no use has to made of that rule at the meta-level (as I argued in the papers referred to), in order to determine whether a theory of justification is itself justified, it appears that a theory of justification need already be in place at the meta-level. And thus, it might seem, the empirical project proposed in the text must beg the issue. But note that the fact that a particular theory of justification must be assumed at the meta-level does not ensure—provided the empirical project is set up properly—that that theory (or any other theory) comes out best in an empirical assessment of theories of justification.

74

REFERENCES

Alston, W. 1999. “Perceptual Knowledge” in Greco and Sosa (eds.) 1999: 223-242. BonJour, L. 1999a. “The Dialectic of Foundationalism and Coherentism” in Greco and Sosa (eds.) 1999: 117-142. ——— . 1999b. “Foundationalism and the External World” in J. Tomberlin (ed.) Philosophical Perspectives Vol. 13, Oxford: Blackwell: 229-249. Carnap, R. 1962. Logical Foundations of Probability (2nd edition). Chicago: The University of Chicago Press. Christensen, D. 1983. “Glymour on Evidential Relevance”, Philosophy of Science 50: 471-481. Douven, I. 1999. “Inference to the Best Explanation Made Coherent”, Philosophy of Science (Proceedings) 66: S424-S435. ——— . 2002a. “Testing Inference to the Best Explanation”, Synthese 130:355-377. ——— . 2002b. “A New Solution to the Paradoxes of Rational Acceptability”, British Journal for the Philosophy of Science 53:391-410. ——— . 2002c. “Decision Theory and the Rationality of Further Deliberation”, Economics and Philosophy 18: 303-328. ——— . 2004. “Empirical Equivalence, Explanatory Force, and the Inference to the Best Theory” in R. Festa, A. Aliseda, and J. Peijnenburg (eds.) Logics of Scientific Cognition: Essays in Debate with Theo Kuipers, Amsterdam: Rodopi, in press. ——— . 2005. “Evidence, Explanation, and the Empirical Status of the Realism Debate”, Erkenntnis, in press. Douven, I. and W. Meijs 2005. “Bootstrap Confirmation Made Quantitative”, Synthese, in press. Duhem, P. 1906/1954. The Aim and Structure of Physical Theory (translated by P. Wiener). Princeton: Princeton University Press. Earman, J. (ed.) 1983. Testing Scientific Theories. Minneapolis: University of Minnesota Press.

75 ——— . 1992. Bayes or Bust?, Cambridge MA: MIT Press. Edidin, A. 1983. “Bootstrapping without Bootstraps” in Earman (ed.) 1983: 43-54. Gillies, D. 2000. Philosophical Theories of Probability. London: Routledge. Glymour, C. 1980. Theory and Evidence, Princeton: Princeton University Press. Goosens, W. 1976. “A Critique of Epistemic Utilities” in R. Bogdan (ed.) Local Induction. Dordrecht: Reidel. Greco, J. and E. Sosa (eds.). 1999. The Blackwell Guide to Epistemology. Oxford: Blackwell. Haack, S. 1993. Evidence and Inquiry. Oxford: Blackwell. Harman, G. 1965. “The Inference to the Best Explanation”, Philosophical Review 74: 88-95. Howson, C. 2000. Hume’s Problem: Induction and the Justification of Belief. Oxford: Clarendon Press. Kaplan, M. 1981a. “Rational Acceptance”, Philosophical Studies 40: 129-145. ——— . 1981b. “A Bayesian Theory of Rational Acceptance”, Journal of Philosophy 78: 305-330. Klein, P. and T. Warfield. 1994. “What Price Coherence?”, Analysis 54:129-132. Lehrer, K. 1990. Theory of Knowledge. London: Routledge. Levi, I. 1967. Gambling with Truth. Cambridge MA: MIT Press. Maher, P. 1993. Betting on Theories. Cambridge: Cambridge University Press. Mayo, D. 1996. Error and the Growth of Experimental Knowledge. Chicago: University of Chicago Press. McMullin, E. 1996. “Epistemic Virtue and Theory Appraisal” in I. Douven and L. Horsten (eds.) Realism in the Sciences, Leuven: Leuven University Press, 1334. Plantinga, A. 1993. Warrant and Proper Function. Oxford: Oxford University Press.

76 ——— . 2000. Warranted Christian Belief. Oxford: Oxford University Press. van Fraassen, B. 1983a. “Theory Comparison and Relevant Evidence” in Earman (ed.) 1983, 27-42. ——— . 1983b. “Glymour on Evidence and Explanation” in Earman (ed.) 1983. ——— . 1989. Laws and Symmetry. Oxford: Clarendon Press.

Part II Areas of Basic Belief and Basic Knowledge A: Mathematics and Philosophy

RON ROOD

On the status of axioms in mathematics 1. Introduction. The status of axioms in mathematics, and those of geometry in particular, has always been considered problematic. The aim of this paper is to show that, in reflecting on these axioms, more philosophical as well as more mathematical considerations often are been closely interrelated. We proceed as follows. In § 2, we shall address a few issues related to Leibniz’ views on axioms. Next, in § 3, we shall bring out what may be taken as forming an ongoing concern for reflections on the status of axioms since the time of Kant: Kant’s symmetry thesis. § 4 sets forth a few prerequisites that we shall need in the sections that follow, namely, the structure of the geometry as it is presented in Euclid’s Elements. In § 5, we address the views of what seems to be an important figure in the debate on the foundations of geometry, namely, those of the German mathematician Carl Friedrich Gauss. We proceed to discuss certain elements of Riemann’s more methodological views in § 6. Finally, in § 7, we turn our attention to Hilbert, whose more philosophical views appear to have shaped much of the way modern mathematics is being done. We end up this paper with a concluding section (§ 8). 2. Leibniz on axioms and proofs. For a start, one may propose that an axiom is a true proposition that, in some sense, is based solely on itself. In other words, except for themselves, axioms need no other propositions for their justification. Wherein this supposed self-justifying nature of axioms exactly lies is a difficult question to answer. For example, one may suggest that an axiom is a self-evident proposition. However, though the idea of self-evidence may have a certain appeal, it is very hard to make it more precise. Without pursuing the point into any dept, it suffices to say that a proposition is self-evident when one immediately assents to its truth, as soon one has understood the concepts involved in it.

80 Perhaps a simple arithmetic equation such as 2 + 2 = 4 is self-evident in this sense. However, a proposition like this is typically not considered to be an axiom. Already in the earlier days of modern philosophy, Leibniz thought that this proposition can be proved.1 Though it should be added that there is no evidence indicating that he denied it to be self-evident. Nowadays, a proposition such as 2 + 2 = 4, too, is typically not considered to be an axiom. Indeed, from the point of view of axiomatic systems such as Peano Arithmetic, there is a proof for generally every determinate arithmetic equality. However, in order to make our discussion not unnecessarily complicated, let us stick to Leibniz’s views, and consider these somewhat more closely. In order to prove the aforementioned proposition, Leibniz stated the following three definitions: Definition 1. 2 =def 1 + 1; Definition 2. 3 =def 2 + 1; Definition 3. 4 =def 3 + 1. Furthermore, he added the following axiom: Axiom 1. If equals be substituted for equals, the equality remains. Leibniz’ proof, now, more or less proceeds as follows. Proof. We have: 2 + 2 = 2 + 2. By definition 1 and axiom 1 it follows that 2 + 2 = 2 + 1 + 1. By definition 2 and axiom 1, Leibniz goes on, it follows that 2 + 2 = 3 + 1. By definition 3 and axiom 1 it follows that 2 + 2 = 4, which is what had to be proved.2 Accordingly, Leibniz thinks that he has provided a proof of the proposition 2 + 2 = 4 on the basis of a single axiom and three definitions. Let us 1 2

Cf. Leibniz 1996: 413-4. Cf. Leibniz 1996: 414.

81 note that it has been objected that Leibniz’ reasoning tacitly appeals to at least one further axiom which he did not make explicit, namely, that the operation of addition is associative: Axiom 2. For every number x, y and z, (x + y) + z = x + (y + z). This axiom was left implicit in Leibniz’ proof as presented above. Making it explicit, the reasoning would more strictly proceed as follows. We have 2 + 2 = 2 + 2. By definition 1 and axiom 1 it strictly speaking follows that 2 + 2 = 2 + (1 + 1). (Note the parentheses, which were not used by Leibniz.) By axiom 2, it follows that 2 + 2 = (2 + 1) + 1, which is a step Leibniz did not make explicit. Only now we can infer 2 + 2 = 3 + 1 on the basis of definition 2 and axiom 1. The reader will be able to complete the proof, this time also using axiom 2. One may even add that Leibniz left implicit yet another axiom: Axiom 3. For every number x, x = x. Indeed, Leibniz opened his proof by saying that 2 + 2 = 2 + 2, which is axiom 3 applied to a specific case. However, it may seem that axiom 3 is somewhat exceptional in certain respects. For it appears that axiom 3 holds for every object, and not merely for numbers. The latter does not seem to hold in case of axiom 1 and 2. These axioms, it seems, do not apply to every object, but possibly only to a restricted class such as, e.g., numbers, although especially axiom 1 may appear somewhat doubtful in this respect. We lay this matter to rest. Leibniz’ example of a supposed self-evident proposition that, in his view, still admits of proof does of course not form a counterexample against the idea that all axioms are self-evident (which is what we began this section with). Generalizing upon the example, the upshot thus far is that we have found a distinction between true propositions that admit of proof and true propositions that do not admit of proof. The propositions of the former kind may be called theorems, and propositions of the latter kind may be called axioms. In Leibniz view, then, self-evidence, whatever it precisely is, does not constitute a mark distinguishing the axioms from the theorems. In particular, the above example makes clear that, in Leibniz view, there may be self-evident propositions that are not axioms. It is of some interest to wonder why it would have occurred to Leibniz that a supposedly self-evident proposition such as 2 + 2 = 4 admits of

82 proof, and that he accordingly undertook an attempt to prove it. Since it does not seem to be a lack of self-evidence that motivated him to do so, his reasons must accordingly lie somewhere else. We think that a part of the answer must turn on the following. As said, theorems are propositions that admit of proof. This implies, among other things, that theorems are based on other propositions in turn. Among the latter are axioms and definitions (as in the example above) but generally also other theorems. These theorems, in turn, are themselves based on axioms, definitions and again other theorems, and so on. It is typically assumed, however, and surely also by Leibniz, that this does not lead to a regress. Thus, a kind of “well-foundedness” principle seems to be presupposed, so that, after a finite number of steps, one can find that every theorem is based on axioms and definitions alone. Collectively, then, we may suggestively say that axioms are “basic” in this respect. We may also, and in an equally suggestive manner, say that theorems are accordingly “non-basic.” Ultimately, only axioms (and definitions) lie at the basis of theorems. Now, there are at least two ways in which the notion of “being basic” can be understood. Consequently, given that axioms are basic, there are at least two ways in which the notion of an axiom can be understood. Thus, Leibniz assumed a distinction between what may be called the structure of truths (i.e., true propositions) on the one hand and the structure of knowledge (i.e., known propositions) on the other.3 A proposition in the structure of knowledge, Leibniz would be willing to admit, is in the structure of truths, but not necessarily vice versa. At any rate, a proposition may be basic either in the structure of truths or basic in the structure of knowledge. To say that axioms are self-evident seems to imply that axioms are basic in the structure of knowledge. However, a proposition that is basic in the structure of knowledge need not be basic in the structure of truths and vice versa. The structure of knowledge and the structure of truths may diverge in this structural respect (and perhaps also in other respects). This, we think, is one of the things that bothered Leibniz. An axiom, Leibniz would say, is a proposition that is basic in the structure of truths in the first place.4 The proposition 2 + 2 = 4, Leibniz would hold, is a proposition that is not basic in the structure of truth, and accordingly it is not an axiom (though it may be supposed to be basic in the structure of knowledge). Accordingly, Leibniz’ notion of “being basic” (as it applies to axi3 4

Cf. Leibniz 1996: 412. Cf. Leibniz 1996: 406-7, 414.

83 oms) is more or less divorced from more epistemic notions as “selfevidence.” What makes a proposition basic (i.e., an axiom) does not primarily turn on the specific way it is known, but much more on its position in the global order of truths. But other considerations may also play a role. Why, in Leibniz’ view, a proposition such as 2 + 2 = 4 is basic in the structure of truths but not in the structure of knowledge seems not very clear.5 One consideration that seems relevant in this respect is that Leibniz held that axioms (say, the axioms of elementary arithmetic, or Euclidean geometry6) must be little in number.7 This may be an indication for thinking why 2 + 2 = 4 is not an axiom in Leibniz’ sense. Indeed, there is a myriad of similar arithmetic equalities, which may all considered to be selfevident. However, precisely because they are so plentiful, Leibniz may have thought they are not basic in his preferred sense (i.e., in the structure of truths)—there must be a fewer number of axioms lying at their basis. The idea sketched in the previous paragraph remains somewhat vague unless one makes more precise what it means to say that axioms are “little in number.” On would surely expect that axioms are fewer in number than theorems. But does this mean that they are always finite in number? However, any finite number that one could mention in this respect appears arbitrary. Perhaps, however, to say that axioms are “few in number” has to be understood not merely in the sense of cardinality but also in another way.8 We lay this matter to rest. A condition along the lines of the previous paragraph (i.e., that axioms must be little in number) will very likely not determine a criterion for being an axiom, so that at least other conditions to this end are required. However, it may seem hard to see what these conditions should be, or even whether they would exist. This point becomes even more pressing in the light of our earlier observation that Leibniz manifests a tendency to filter out various more epistemic aspects from his notion of axiom, e.g., aspects such as self-evidence. It is especially epistemic aspects of axioms, what-

5

A similar point seems to hold for “basicness” in the structure of knowledge. The difficulties particularly appear to concern self-evidence. 6 See § 4. 7 Leibniz 1996: 407. 8 For example, it may be required that there is an effective means for deciding whether or not a proposition is an axiom. Cf. Boolos and Jeffrey 1980: 177.

84 ever they precisely are, that a criterion for being an axiom would seem to lean on.9 In the light of these considerations it is interesting to make a jump ahead in time, and take a brief look at Frege. He, too, held that axioms should be little in number, though he did not make clear how, exactly, this is to be understood.10 At any rate, Frege also said that axioms are chosen, and, up to a degree, one’s choice in this respect is arbitrary. Frege’s point, we may take it, is that axioms do not stand on themselves. In contrast, axioms are always axioms in a system of propositions. To illustrate this, consider (using standard logical notation) a system consisting of the propositions p → q, p ∧ q, p and q.11 Within this system, Frege would say, one may consider p → q and p ∧ q as axioms and p and q as theorems. Alternatively, however, it may be possible also consider p → q and p as axioms and p ∧ q and q as theorems.12 There need to be nothing about the propositions involved per se that would force one to reject one option in favor of the other. This point does not constitute an objection against Leibniz. Rather, it reveals some of the issues involved in a view such Leibniz’. What we see in case of a view as endorsed by Leibniz, is that what counts as an axiom in mathematics, does not stand independently from more global methodological issues. However, this point even holds in case one accepts selfevidence as a feature of axioms. For even in that case, not every selfevident proposition can be accepted as an axiom. There must be further conditions on what it is to be an axiom. 3. Kant’s symmetry thesis. Mathematics can be partitioned into geometry on the one hand and arithmetic on the other. In what follows, both arithmetic and geometry should not be too narrowly understood, however. In particular, the suggested partitioning should not be exclusively understood in terms of a division between elementary Euclidean geometry on the one hand and the elementary 9

We do not intend to say that self-evidence, could it be made into a satisfactory notion, would constitute a criterion for being an axiom. Leibniz, for example, would still disagree. 10 Frege 1964: vi. 11 We assume modus ponens and the introduction and elimination rules for the conjunction as admissible principles of inference. 12 Cf. Frege 1979: 205-6.

85 arithmetic of integers on the other. This would certainly give an accurate picture of the state of mathematics in early stages of its development. However, given an eye also on later developments, it is better and safer to be considerably less precise and say that mathematics can be partitioned into geometrically oriented parts on the one hand and arithmetic oriented parts on the other. Among the former, we should at least allow ourselves to include what is broadly known as non-Euclidean geometry and Riemannian geometry.13 Among the latter, we should at least allow ourselves to include what is broadly known as algebra. Kant has put the division of mathematics into arithmetic and geometry in terms of an articulate philosophical theory. The details of this theory are unimportant. It suffices to note that, in Kant’s view, arithmetic is based on an a priori intuition of time, while geometry is based on an a priori intuition of space. The division of mathematics into arithmetic and geometry, as backed up by Kant’s philosophical theory, has exerted considerable influence after Kant. This will become clear at the end of this paper. In Kant’s view, space, on the one hand, is the form of all outer appearances, that is, the form of outer sense.14 Time, on the other hand, is the form of inner sense.15 Given this, it is understandable that, on a view inspired by that of Kant, arithmetic has a more deeply subjective grounding than geometry. While the justification of arithmetic relates to subjective inner experience, the justification of geometry, in contrast, relates much more to experience of the external world. Kant accepted a division of propositions along two dimensions. Along the first dimension, one finds propositions that are a priori and propositions that are a posteriori. Along the second dimension, one finds propositions that are analytic and propositions that are synthetic. Roughly, a proposition is a priori if it is justified on the basis of reason alone. A proposition is a posteriori if its justification involves an appeal to (sensory) experience. The distinction between analytic and synthetic furthermore, turns much more on the methodological aspects of propositions. Thus, in Kant’s view, a proposition is analytic, if its methodological basis lies in (general) logic alone.16 A proposition is synthetic (whether a priori or a posteriori), in con-

13

See § 6. Kant 1998: B42. 15 Kant 1998: B49. 16 Cf. Kant 1998: 151-2/B191. 14

86

analytic

synthetic

trast, if its methodological basis involves the use of what Kant calls “intuition” (Anschauung).17 In Kant’s view, the two aforementioned dimensions are to a high degree independent, but not entirely. In particular, Kant held that an analytic proposition is not a posteriori and vice versa. In other words, for Kant, there are no analytic a posteriori propositions. As a result, Kant knew a tripartite division of propositions as indicated in the figure below:

synthetic a priori

synthetic a posteriori

analytic a priori

a priori

a posteriori

Kant famously held that all the propositions of geometry as well as those of arithmetic are (1) synthetic, and (2) a priori. This point may be taken as implying that geometry on the one hand and arithmetic on the other are symmetric in a certain respect. Let us state this as Kant’s symmetry thesis: Kant’s symmetry thesis. Geometry and arithmetic are symmetric in the sense that (a) the propositions of geometry are (1) synthetic and (2) a priori, and (b) the propositions of arithmetic are (1) synthetic and (2) a priori. Kant’s symmetry thesis implies that, in Kant’s view, the definitions, the axioms, as well as the theorems of mathematics (i.e., geometry and arithmetic) are synthetic and a priori. Many developments in the philosophy of mathematics, but often also mathematics itself, can be understood in terms of a struggle with Kant’s symmetry thesis. In this respect, Kant’s symmetry thesis formed a rather ongoing concern. For example, people have found difficulties with Kant’s 17

As to our reading of Kant’s distinction between analytic and synthetic propositions, we to a considerable extent rely on De Jong 1997.

87 claim that the propositions of arithmetic are analytic. Frege is a notable example in this respect. His logicist program aimed at showing that every theorem of arithmetic could be proved on the basis of nothing but logic.18 Despite the failure of Frege’s own logicist program—Russell discovered an irreparable inconsistency in Frege’s system—there recently has come a revival of Frege’s ideas. Nowadays, various scholars aim at establishing that the propositions of arithmetic are analytic but in some apparently weaker sense than Frege’s. This program is known as neo-logicism or neoFregeanism.19 In what follows, we shall not be primarily concerned with Kant’s symmetry thesis in the light of developments in the foundations of arithmetic, but rather with respect to geometry. In particular, we shall consider how a continuing struggle with Euclid’s axiom of parallels can be understood in terms of a concern with Kant’s symmetry thesis. It will turn out that people have found difficulties with Kant’s belief that the propositions of geometry are a priori. The status of Euclid’s axiom of parallels plays a key role in this respect. 4. Euclidean geometry and the axiom of parallels. The aim of this section is to present the necessary prerequisites to elementary Euclidean geometry as presented by Euclid in his time-honored Elements.20 We begin with presenting the structure of the Elements. Subsequently, we highlight the axiom of parallels. We indicate that this axiom has formed an ongoing concern for mathematicians ever since. Euclid’s Elements opens with 23 definitions, among which are the definition of point, line, surface, circle, equilateral triangle, isosceles triangle, and parallel (i.e. the relation of parallelism that may or may not hold between two lines). In particular, parallel straight lines “are straight lines which, being in the same plane and being produced indefinitely in both directions, do not meet one another in either direction.”21

18

Frege 1964. See Wright and Hale 2001 for comprehensive treatment of the issues involved. Attempts have been undertaken to extend the ideas of neo-Fregeanism to other parts of mathematics such as, for example, set theory and analysis; see Shapiro 2000, Shapiro 2002. 20 Euclid 1956. 21 Euclid 1956: I, 154. 19

88 Next, we have the following five axioms (traditionally called postulates):22 1. 2. 3. 4.

Two points determine one straight line. Any finite straight line can be continuously extended. A point and a finite straight line determine a circle.23 All right angles are equal to one another.

We quote the fifth axiom: 5. That, if a straight line falling on two straight lines makes the interior angles on the same side less than two right angles, the two straight lines, if produced indefinitely, meet on that side on which are the angles less than the two right angles.24

Gray says that these axioms “permit certain geometrical constructions to be made.”25 This seems plausible in case of the first three axioms. However, it appears considerably harder to understand in case of the final two. On the face of it, the final two axioms do not permit a construction. After the above axioms, five so-called common notions are given.26 Let us quote them: 1.Things which are equal to the same thing are also equal to one another. 2.If equals be added to equals, the wholes are equal. 3.If equals be subtracted from equals, the remainders are equal. 4.Things which coincide with one another are equal to one another. 5.The whole is greater than the part. Gray says that the above common notions can be viewed as general principles of inference.27 They are general in the sense that they do not only apply to geometry, but in other fields as well. In other words, the common notions have a highly non-local nature. In this, they contrast with the axi22

Cf. Euclid 1956: I, 154-5. Namely, the circle having the given point as its centre, and the line as its radius. 24 Euclid 1956: I, 155. Two right angles are equal to 180°. 25 Gray 1979: 30. 26 Euclid 1956: I, 155. 27 Gray 1979: 30. 23

89 oms, which have a more local nature in the rough sense that they only apply to geometry. From the point of view of modern logic, the distinction between axioms and common notions corresponds to the distinction between non-logical axioms and logical axioms respectively.28 After the common notions are presented, various theorems are stated and proved. Euclid presents further definitions at other places in the Elements, and proves other theorems subsequently. Euclid’s fifth axiom is often called by a special name: the axiom of parallels. We shall henceforth refer to it by that name. The following picture illustrates the axiom of parallels:

a

ℓ

P

b m k

Comment: if both ℓ and m intersect k, and if a + b < 180°, then ℓ and m intersect at a point P that is on the same side as a and b. The name axiom of parallels might seem remarkable, for on the face of it, the axiom doesn’t say anything about parallel lines. However, it turns out that the axiom is intimately related to various properties of parallel lines. For example, the axiom of parallels is equivalent to Playfair’s axiom:

28

Monk 1976.

90 Playfair’s axiom. Given a line ℓ and a point P. Then there exists exactly one line through P parallel with ℓ. Alternatively, the axiom of parallels is equivalent with the following: if a straight line intersects one of two parallel lines, it intersects also the other. Many more equivalent formulations of the axiom of parallels could be added.29 The axiom of parallels has traditionally always been considered as particularly problematic. Many people have, for example, found the axiom to lack an appeal of self-evidence. However, the various exact reasons underlying the attempts to deal with the problem of the axiom of parallels need not concern us here.30 We think that the difficulties people have expressed with respect to the axiom of parallels have a common core, which may be put as follows: Problem of parallels. Is the axiom of parallels a genuine axiom, or is it (despite its name) a theorem? In view of what was said in § 2, the term axiom can be understood in at least two different ways, namely, as an axiom in the order of truths, or an axiom in the order of knowledge. Thus, it may be objected that the problem of parallels is not an unambiguous question. However, we shall see that part of the problem of parallels is at what level axioms are supposed to live. The problem of parallels has been of enormous importance for intellectual history. For example, it has in part driven important developments in mathematics such as non-Euclidean forms of geometry. This section offers only a glimpse at these developments. On a broader level, philosophical views accompanying the actual development mathematics have gradually changed too. The import of these changes in the philosophy of mathematics must not be underestimated. Often they have drastically changed the outlook on mathematics. For example, investigations on the axiom of parallels have led to the view that our knowledge of the propositions of geometry is not justified entirely a priori, and that experience is likewise an essential ingredient of the justification of geometry. As is well known, any of the investigations with the aim of dealing with the problem of the axiom of parallels had failed. This ultimately led to in29 30

See Euclid 1956: I, 220. See, for example, Heath’s discussion in Euclid 1956: I, 202-20.

91 vestigations into non-Euclidean forms of geometry, based on a denial of the axiom of parallels. The actual history that has lead to non-Euclidean geometry, including the investigations undertaken by the various people involved, is long and complex. It is of no relevance to repeat the story about it here in any great detail. The interested reader is referred to Bolona’s classic text where it is readily available.31 Greenberg and Gray, among others, have provided more modern and up to date treatments.32 Initially, there were two ways of dealing with the problem of the axiom of parallels. First, mathematicians have sought for another but equivalent formulation of the axiom (see, for example, Playfair’s axiom above). It was hoped that such an equivalent formulation was in some sense less problematic than the original. As said, there are several equivalent formulations of the axiom of parallels, none of which need to concern us here.33 Second, people of have tried to prove the axiom of parallels from the other axioms. In other words, people have undertaken attempts to prove the axiom of parallels as a theorem. As it turns out, none of these attempts was were successful. This may lead one to surmise that the axiom of parallels is independent of the other axioms, which indeed turns out to be the case. However, before mathematicians came to this insight, they have also drawn conclusions of a quite different nature. Namely, that the axiom of parallels does not admit of a priori proof, but needs to be justified by experience instead.34 The methods that have been used to prove the axiom of parallels are various. Saccheri, for instance, used the following strategy: infer to the truth of the axiom of parallels from the assumption that it is false and a derivation of a contradiction from that assumption.35 He thought that his attempt at proving the axiom of parallels was successful. However, his proof turned out to be fallacious.36 Beltrami used another method. He provided a model of a denial of the axiom of parallels, including the remaining axioms, within Euclidean geometry. This shows that a denial of the axiom of parallels is consistent with the other axioms, provided that Euclidean geometry is consistent. If the lat31

Bolona 1955. Greenberg 1974, Gray 1979. 33 Euclid 1956: I, 220. 34 See §§ 5-6. 35 To this end, one needs classical logic. 36 For a more elaborate discussion of Saccheri’s attempted proof, see Gray 1979: 5462. 32

92 ter is indeed the case, then it becomes understandable why Saccheri’s attempt must have failed. For, if Euclidean geometry is consistent, then Beltrami’s model shows that the assumption that the axiom of parallels is false is consistent with the remaining axioms of Euclid’s geometry. Therefore, it is not possible to derive a contradiction in the way Saccheri envisaged.37 Let us now turn to Gauss’ views on the axiom of parallels. 5. Gauss on the epistemological status of the axiom of parallels. Gauss did not publish any mathematical work on the axiom of parallels. Known to us are two unpublished memoranda, a few private notes, and various letters to friends. In what follows, we shall mainly consider Gauss’ letters.38 As we said earlier, repeated efforts to deal with the problem of the axiom of parallels all turned out to be unsuccessful. Gauss appears to have been well aware of this. That Gauss occasionally thought about the foundations of geometry is shown by the following quote. It is taken from a letter to the astronomer and mathematician Friedrich Wilhelm Bessel (dated January 27, 1829): I have also occasionally in my spare time thought about another subject that has concerned me for nearly forty years, namely, the first principles of geometry.39

In view of the repeated failures to solve the problem of parallels to date, Gauss appears to have been reflecting on the epistemic status of the axioms of geometry and the axiom of parallels in particular. For Gauss, the repeated failures of providing a satisfactory solution to the problem of parallels formed increasingly convincing evidence that the axiom of parallels

37

See Gray 1979: 135-9. As they are in part translated in Ewald 1996: volume I. 39 Gauss 1829: 301. Incidentally, this remark, if it is to be trusted, seems to indicate that Gauss started to have his convictions since, roughly, the end of the 18th century. That is, approximately a decade after the first edition of Kant’s Critique of Pure Reason was published (an event that happened in 1781), and before Bolyai and Lobachevsky published their results about their investigations in non-Euclidean geometry (which happened in the early 1830s; cf. Ewald 1996: I, 298). The specific dating of Gauss first thoughts on non-Euclidean geometry in is discussed briefly in Detlefsen 1998: 311, note 7. Cf. also Ewald 1996: I, 297. 38

93 cannot be justified a priori. In the same letter from which the above quotation is taken, Gauss said in retrospective that […] my conviction that we cannot establish geometry entirely a priori has, if possible, become even firmer.40

Gauss seems to come close to denying that our knowledge of the propositions of geometry can be justified a priori. In this respect, Gauss comes close to a partial denial of Kant’s symmetry thesis. For Gauss, his reflections on the epistemic status of the axiom of parallels directly influenced his overall outlook on geometry. In a letter to Olbers from 1817, he formulates his views as follows: I have become ever more convinced that the necessity of our geometry cannot be proven, at least not from the vantage of human understanding and in a way that is accessible to human understanding. Perhaps in another life we’ll gain insights into the nature of space that are currently unattainable. Until then, however, we must not classify geometry with arithmetic, which is purely a priori, but rather assign it the same status as mechanics.41

Let us make two observations here. First, Gauss says that: “the necessity of our geometry [i.e., Euclidean geometry] cannot be proven, at least not from the vantage of human understanding and in a way that is accessible to human understanding.” This suggest that Gauss still hopes for (a priori) justification for the axiom of parallels, although this seems currently beyond our reach. Indeed, the next sentence clearly suggests that Gauss considers the “nature of space” as yet an open issue. In the light of this, the axiom of parallels might as well turn out to be false. We shall come back to this point shortly. Second, observe that Gauss tends to compare geometry with mechanics—or at least, he thinks that we must do so for the time being. In Gauss’ view, mechanics is a form of applied science. Accordingly, our knowledge of the axioms of mechanics is justified by experience (i.e., is a posteriori). However, not everyone in those early days of modern science held such a view. Kant, for one, was a notable exception in this respect. Indeed, Kant held that besides the propositions of (pure) mathematics (and hence those of geometry), those of pure physics are likewise (synthetic) a priori: 40 41

Gauss 1829, 301. Quoted from Detlefsen 1998: 312.

94 We can say with confidence that certain pure a priori synthetical cognitions, pure mathematics and pure physics are given; for both contain propositions which are thoroughly recognized as absolutely certain […] and yet as independent of experience.42

(Instead of propositions, Kant speaks in terms of cognitions (Erkenntnisse).) For Kant, pure physics means: Newtonian physics. Rather than being justified by experience, Kant held that the axioms of pure physics are (synthetic) a priori. In the first Critique, Kant says: Natural science (physica) contains synthetic a priori judgments as principles. Let me cite as examples just a few propositions: the proposition that in all alterations of the corporeal world the quantity of matter remains unaltered, or that in all communication of motion effect and counter-effect must always be equal.43

In the former proposition, we recognize the law of conservation of mass (now superseded by the general theory of relativity); in the latter we recognize Newton’s third law of motion. Kant’s belief that the principles of Newtonian physics are justified a priori is remarkable. Especially given the fact that Newton himself thought they were justified a posteriori. In the light of this, it is of some interest to observe that Kant was in a way a traditional mind when he said that pure physics is, like pure mathematics, justified a priori. In this respect, Kant can be seen as taking sides with, for example, Pascal, who, in his Pensées said: Our knowledge of the first principles, such as space, time, motion, number, is as certain as any knowledge we obtain by reasoning. As a matter of fact, this knowledge provided by our hearts and instinct is necessarily the basis on which our reason has to build its conclusions.44

In Pascal’s view, our knowledge of the axioms of geometry, mechanics and arithmetic—“the first principles of space, time, motion, number”—is not established by providing reasons for them. Rather, this knowledge is immediate (i.e., intuitive) and is “provided by our hearts and instinct”. From this it seems to follow that, in Pascal’s view, our knowledge of the 42

Kant 2002: 72. Kant 1998: B17. 44 Quoted from Kline 1980: 45. 43

95 axioms of geometry, mechanics and arithmetic is not justified by experience, i.e., is not a posteriori. Earlier we said that Gauss apparently considers the “nature of space” an open issue. This implies that Gauss considers it an option that the structure of actual space might be non-Euclidean. Underlying Gauss’ views is the presupposition that any of the axioms of geometry is either true or false, and that their truth or falsity is a matter of the structure of physical space. That Gauss indeed seriously entertained the thought that the structure of actual space may very well be non-Euclidean is confirmed by the following quotation from an earlier letter to Wolfgang von Bolyai from 1799:45 […] if one could prove that [for any given surface] a straight lined triangle is possible whose area is greater than the given surface, then we are in a position to prove the whole of geometry with full rigour. Most people would let this stand as an axiom; not me; for it might be that, no matter how far from each other in space one assumes the angles of a triangle to be, nevertheless the area always remains under a given bound.46

In the first paragraph of this quotation, Gauss considers the following proposition: (*) Given any surface with area A. Then there exists a triangle with an area greater than A. It turns out that (*) is equivalent with the axiom of parallels. Hence, (*) could count as a substitute for this axiom.47 Accordingly, given (*), Euclidean geometry could be fully developed, as Gauss says in the first paragraph. In the second paragraph of the above citation, Gauss considers the following proposition (or so we take it): (**) There exists a number B such that the area of every triangle is less than B, no matter how far its angles lie apart.

45

Wolfgang Bolyai (his full name was Farkas Wolfgang Bolyai) was a friend and former fellow student of Gauss’. He was also the father of János Bolyai, who is wellknown for his published results on non-Euclidean geometry in 1823 and later. (See below.) 46 Gauss 1799: 299. 47 Cf. Euclid 1956: I, 220.

96 A space in which (**) holds has non-Euclidean structure. Consider, for example, a two-dimensional sphere. In this space, the surface of any triangle always remains under a given bound (e.g., area that is equal to the surface of the sphere). The geometry of the sphere is non-Euclidean.48 However, note that, in the latter quotation, Gauss is not merely considering whether or not there exists a model (such as e.g. a sphere) such that, in it, there always exists a triangle whose area is greater than any given area. Gauss is pondering over something stronger instead: is it true in space per se (that is, actual space) that there always exists a triangle whose area is greater than any given area? Gauss is not readily prepared to answer this question affirmatively. Therefore, let us conclude that Gauss thinks that non-Euclidean geometry is possible in the sense that the structure of actual space might turn out to be non-Euclidean. However, there is also a sense in which we think Gauss that does not consider non-Euclidean geometry to be possible. Let us explain. Gray says that “[i]t does seem that Gauss was the first to believe that a non-Euclidean geometry was possible” and “the attempts to find a contradiction in other systems were therefore [sic] in vain.”49 We take it that what Gray tries to say is something like the following: for Gauss, the possibility of non-Euclidean geometry forms a reason for the absence of inconsistency in non-Euclidean systems of geometry. Supposing that this is an adequate reading of what Gray says, this is misleading in that it may be tempting to hold that possible here means: absence of inconsistency. Thus, Gauss would have entertained the thought of merely consistent systems of nonEuclidean geometry besides the Euclidean system. However, we do see no reason to believe this. Moreover, we do not think that this captures Gauss’ thought on the matter entirely right. It is the possible that bothers us.50 This should be clear from what we have said earlier. But let us repeat the point in different words. We think that Gauss would have read the possible in a different and perhaps stronger sense. Namely, what Gauss would have meant, we think, is that the structure of actual space is possibly nonEuclidean.

48

If Gauss, in the second paragraph of the quotation given, means to suggest that there is no lower bound on the distance between the angles of a triangle, then the example presented in the main text is obviously not a very good one. For on a given sphere, there is a maximum distance between any pair of angles of a triangle. 49 Gray 1979: 76. 50 We do not dispute Gray’s claim of priority on behalf of Gauss.

97 The difference should not be underestimated. For it is truth that matters. If a system of geometry is merely consistent (in the sense that no contradiction can be derived from it), then it does not follow that its axioms are true. But the latter, we think, is exactly one of the things that bothered Gauss: are the axioms of geometry true and on what grounds do we know? As said, in Gauss’ view, it is the structure of actual space that decides the truth or falsity of the axioms of geometry. The point is, we think, that, in a way, Gray may have read Gauss as a too modern a thinker. It is undeniable that Gauss was far ahead of his time when he, presumably as one of the first, seriously entertained the thought that the structure of actual space is possibly non-Euclidean.51 This is no doubt a mark of Gauss’ groundbreaking and entrepreneurial genius.52 Moreover, he was definitely beginning to think in the spirit of modern science at the moment he started thinking (if only for the time being) that this question can perhaps not be decided solely by a priori means, but that we must perhaps also recourse to experience. However, the idea to study geometry—or indeed, mathematics in general—in terms of mere consistent systems, without bothering about the truth of the axioms, is of much later origin. For this, we had to wait until David Hilbert entered the scene.53 It appears to us, however, that Gauss had not dropped the requirement that axioms are always true. Instead, he appears to have uncritically accepted this requirement from the tradition that preceded him. One thing that makes Gauss’ views remarkable is that he developed them before the early 1830s, the period in which the Hungarian mathematician János Bolyai and the Russian mathematician Nikolai Lobachevsky independently published their results about what is nowadays called nonEuclidean geometry. As soon as Gauss knew about these results, he concluded that they constituted conclusive argument against Kant’s belief that 51

The non-Euclidean structure of space forms an important part of the general theory of relativity. Apparently Gauss speculated about the non-Euclidean structure of space a long time before this theory became established. 52 That Gauss’ ideas were far from widely accepted may be indicated by the following sentence, quoted from a letter from Gauss to Bessel dated 9 April 1830 (Gauss 1830: 302). I we was delighted by the ease with which you entered into my views on geometry, particularly because so few have an open mind to the subject. 53

See § 7.

98 geometry is a priori. In a letter to Wolfgang von Bolyai dated 6 march 1832 he says. […] it is precisely the impossibility of deciding a priori between Σ and S that we have the clearest proof that Kant was wrong to claim that space is only the form of our intuition.54

(Σ denotes “the System of Geometry resting on the hypothesis of the truth of Euclid’s Axiom [i.e., the axiom of parallels], and S denotes “the system founded on the contrary hypothesis.”)”. In this citation, Gauss says “Kant was wrong to claim that space is only the form of our intuition.” Therefore, our knowledge of geometry is not entirely a priori, and hence in part a posteriori. Gauss does not come to his conclusion on the basis of the mere existence of Σ and S. A crucial premise is that, as Gauss’ eventually came to realize, it is impossible to decide between Σ and S solely by a priori means. Incidentally, note that Gauss now speaks in terms of an hypothesis when he talks about the axiom of parallels and of its denial. Gauss expresses effectively the same conclusion in a letter to Bessel: It is my deepest conviction that the theory of space [i.e., geometry] has a completely different position in our a priori knowledge than the pure theory of quantity [i.e., arithmetic]. Our knowledge of the former utterly lacks the complete conviction of necessity (and also of absolute truth) that belongs to the latter; we must humbly admit that, if number is merely the product of our mind, space also possesses a reality outside our mind, and that we cannot prescribe its laws a priori.55

The thought that geometry is, on the whole, not an a priori science but is in part based in experience eventually became the prevailing view. In order to complete the picture, we should add that the idea that the axiom of parallels was based on experience seems to have been entertained earlier. For example, Morris Kline reports that that Georg S. Klügel (a student of Johann Heinrich Lambert’s56), in his doctoral dissertation from 1763, “made the remarkable observation that the certainty with which men accepted the truth of the Euclidean parallel axiom was based on experi54

Quoted from Detlefsen 1998: 313. Gauss 1830: 302. 56 Remarkably, Lambert happened to be an acquaintance of Kant. As far as we can see, Kant himself never addressed the problem of parallels. 55

99 ence.” Kline proceeds by saying that “[t]his observation introduced for the first time the thought that experience rather than self-evidence substantiated the axioms.”57 Again, the conviction that geometry is a posteriori eventually became the prevailing one. This may be very well in part due to Gauss’ influence, especially in the mathematical world.58 For example, Leopold Kronecker, in his only philosophical writing titled On the concept of number (1887), gives the following expression to this view: The difference in principle between geometry and mechanics on the one hand and the remaining mathematical disciplines (here gathered under the term ‘arithmetic’) on the other is, according to Gauss, that the object of the latter, number, is merely our mind’s product, while space as well as time also have outside our mind a reality, whose laws we cannot prescribe completely a priori.59

Kronecker adds a footnote in which he quotes exactly the same passage from Gauss letter to Bessel dated 9 April 1830 as we have done earlier. Kronecker’s remark that the object of arithmetic is grounded in the subject (is “merely our mind’s product”) no doubt reveals the influence of Kant’s thought with respect to arithmetic.60 Let us conclude. A certain asymmetry between arithmetic on the one hand and geometry on the other became eventually became the dominant view: geometry is an posteriori science, while arithmetic is a priori. As both Gauss and later Kronecker suggest, arithmetic has, in a sense, a deeper subjective grounding than geometry. For, as they both say, arithmetic is merely a product of our mind or intellect. In this respect, Kant’s influence seems still apparent. Geometry, in contrast, would bear on a reality outside us. Consequently, the justification of its axioms cannot be entirely a priori; besides reason, we must also recourse to experience. This may be considered a criticism of Kant’s views on geometry.

57

Kline 1980: 81. Kline adds a footnote he which he says that “Newton had also made this assertion but he did not stress it and it was ignored.” 58 Gauss stature as a mathematician was enormous at the time, and presumably it still is. 59 Kronecker 1887: 947. 60 See § 3.

100 6. Riemann’s conceptual revolution in geometry. While Gauss’ views can be seen against the background of the geometry as presented by Euclid, they are by no means intrinsically related to this way of presenting the subject. To this we shall turn now. Bernhard Riemann, one of the pioneers in the development of modern geometry, effectively held the same traditional views with respect to the status of axioms. For Riemann, too, the axioms of geometry are true. What makes the axioms true (or false) is the structure of actual space (see below). Nevertheless, Riemann had a very different approach to the problem of parallels. Rather than proving or refuting the axiom of parallels head on, he changed the fundamental concepts of geometry. Thus, he articulated a view according to which geometry is studied from a different and more general perspective. Riemann’s approach to geometry, however, preserves Euclid’s in a reinterpreted form. As a result, in Riemann’s view, the problem of parallels came to stand in an entirely new perspective. Let us explain. We can take Riemann’s approach towards geometry as proceeding in three successive stages. Let us briefly and roughly explain this without going into the details. First, instead of considering the concept of actual three-dimensional physical space, geometry in Riemann’s approach directs itself to the more general concept of an “extended magnitude” of arbitrary dimension. That is, to the general notion of an n-dimensional extended magnitude, for any nonnegative integer n. Nowadays, these are referred to as n-dimensional manifolds, or simply as n-manifolds.61 From this general perspective, space comes out as a specific type of manifold, namely, one of three dimensions. In the second step, Riemann noticed that, given any n, any ndimensional manifold (and particularly any 3-dimensional manifold) generally admits of different types of metrics, that is, different types of relations that determine the metric properties of that manifold. The different metrics that can be imposed on a manifold (of arbitrary dimension) gives rise to different “geometries.” Riemann now holds that one of the metrics on a three-dimensional manifold gives rise to the geometry of actual space. It is possible that this geometry matches the geometry Euclid has laid down in The Elements. However, it is likewise possible that the geometry of ac61

More specifically, Riemann considered “continuous” manifolds, nowadays called topological manifolds.

101 tual space is of a different nature. For Riemann, this was as yet an open issue. In the third stage, the problem is to determine the metric of a three dimensional manifold that gives rise to the geometry of actual space. Only after this task is successfully fulfilled, one is able to declare the true geometry. However, in Riemann’s view, to determine the metric properties of actual space is an empirical matter. According to Riemann, we will not be able to know those properties on the basis of reason alone. The above views of Riemann’s are compactly formulated in the following passage. It is quoted from his famous habilitations lecture delivered at Göttingen in 1854, titled On the hypotheses [sic] which lie at the foundation of geometry: […] the propositions of geometry cannot be derived from the general notion of magnitude […] the propositions which distinguish space from other conceivable triply extended magnitudes are only to be deduced from experience. Thus arises the problem to discover the simplest matter of fact from which the metric relations of space may be determined; a problem which from the nature of the case is not completely determinate since there may be several systems of matters of fact which suffice to determine the metric relations of space—the most important system for our present purposes being that which Euclid has laid down as a foundation. These matters of fact are—like all matters of fact—not necessary, but only of empirical certainty; they are hypotheses.62

As it turns out, Riemann’s lecture was not widely known until 1868, when it was published for the first time. Nevertheless, the initial reactions where very positive. For example, Dedekind recalls that Gauss, who attended Riemann’s lecture, got unusually exited by Riemann’s ideas at the time.63 The mathematical ideas Riemann presented in his lecture mark a new era in the development of geometry. As explained, Riemann articulated an approach to the subject from a new and much more general perspective. Nevertheless, on some philosophical points, Riemann’s views are, like those of Gauss, fairly traditional. For in Riemann’s view, the propositions of geometry are true, and the basic propositions in particular. In Riemann’s case, their truth is not so much determined by the properties of parallel lines which forms the subject of the axioms of parallels. Rather, they are determined by the metric structure of space. The fundamental truths about 62 63

Riemann 1868: 652-3. Cf. Ewald 1996: II, 650.

102 space are, Riemann says, “deduced from experience”. Therefore, they are not self-evident.64 Moreover, our knowledge of them is not qualified by the kind of certainty self-evident truths of mathematics normally posses. Rather, our knowledge of them is qualified in terms of “empirical certainty”. And thus, Riemann proceeds, they are “hypotheses”. 7. Hilbert on the axioms of geometry. Roughly until the days of Hilbert and his followers, it appears that mathematicians and philosophers generally considered geometry as the theory of physical space. For a long time, it was generally held that the fundamental truths of space can be determined by reason alone (or, as Pascal would say, by our hearts and instincts). Nevertheless, research into the foundations of geometry as initially presented by Euclid (and the axiom of parallels in particular) ultimately led to a radical new perspective on our knowledge of the fundamental truths of geometry. Eventually, people began to adopt the view that our knowledge of geometry is a posteriori. This means that our knowledge of the fundamental truths of geometry can, on the whole, only be decided on a posteriori grounds. Presumably, Gauss played a leading role in this respect. As can be see from Riemann’s views, this conclusion holds fairly independently of what one takes the fundamental concepts of geometry to be. While Riemann’s mathematical views where highly novel in that he changed the fundamental concepts of geometry, his background philosophy was in certain respects constituted by ideas that where endorsed for centuries. For Riemann, too, the axioms of geometry are true, and what makes them true is the structure of three-dimensional physical space. It is precisely this last point that was dropped by Hilbert. Hilbert denied that geometry was about a fixed reality such as physical space, a reality which was traditionally supposed to be given beforehand. In line with this, Hilbert also denied that the axioms (and theorems) of geometry are true with reference to such a fixed reality.65 It may be said that for Hilbert the (truth) of the axioms in not so much dependent on a reality that is supposed to be given beforehand. In contrast, in Hilbert’s view, the structures of objects one studies depend on the axioms posed.66 Thus, in a sense, the axi-

64

Note that, for a modern reader, this is a remarkable use of the verb to deduce. In contrast, the modern reader would certainly rather speak in terms of induction here. 65 Hilbert 1918. 66 Hilbert 1899.

103 oms come first, and the structures considered have to conform to these axioms.67 It may be said that Hilbert’s ideas have become characteristic for modern mathematics. In a nutshell, mathematicians, both in geometry as well as algebra, are prone to pose an interesting collection of axioms. Subsequently, theorems are proved on the basis of these axioms. Accordingly, mathematics now becomes autonomous in a way it was never before. In particular, in a view such as endorsed by Hilbert, we need not to resort to empirical methods (nor, for that matter, to any other method) in order to decide the truth of the axioms. However, we now see that this way of doing mathematics is not without precedent. 8. Conclusion. Actual mathematical work and foundational research often go hand in hand and closely interact with one another. The investigations into the axiom of parallels illustrate this point. Thus, critical investigations into the axiom of parallels have led to important developments in mathematics itself, though these investigations themselves where primarily of a foundational nature. Attempts at proving the axiom of parallels ultimately have led to the development of systems of non-Euclidean geometry. In the other direction, primarily mathematical investigations such as those of Riemann’s, have led to a radical different view on the foundations of geometry. This time the problem of the axiom of parallels was, as it where, not attacked by investigating the properties of parallel lines head on. Riemann, in contrast, articulated a view by means of which the problem of the nature of space is approached from a different and more general perspective. Finally, Hilbert entered the scene, simply denying that the axioms of geometry are true with respect to an antecedently given reality. Nevertheless, it is possible that certain developments in mathematics and its foundations proceed at a different speed than the evolution of certain more general philosophical ideas. In particular, we have seen that mathematicians and philosophers alike have for a long time kept to the thought that axioms are always true. Gradually, however, people eventually came to adopt the view that our knowledge of the axioms is justified a posteriori instead of a priori. As Riemann’s case shows, these philosophical ideas cross-sect important developments in mathematics itself. 67

Hilbert required the axioms to be consistent and independent.

104

REFERENCES Bolona, Robert. 1955. Non-Euclidean Geometry: A Critical and Historical Study of its Development. Translated by H.S. Carslaw, New York: Dover. Boolos, George, Jeffrey, Richard C. 1980. Computability and Logic, 3rd edition. Cambridge: Cambridge University Press. De Jong, Willem R. 1997. “Kant’s Theory of Geometrical Reasoning and the Analytic-Synthetic Distinction. On Hintikka’s Interpretation of Kant’s Philosophy of Mathematics.” Studies in the History and Philosophy of Science 28: 141-66. Detlefsen, Michael. 1998. “Constructive Existence Claims.” In: Philosophy of Mathematics Today, Matthias Schirn (ed.) 1998: 307-35. Euclid. 1956. The thirteen books of The Elements, 2nd ed, 3 vols. Translated by Thomas L. Heath, New York: Dover. Ewald, William B. (ed.). 1996. From Kant to Hilbert: A Source Book in the Foundations of Mathematics, 2 vols. Oxford: Clarendon Press. Frege, Gottlob. 1964. The Basic Laws of Arithmetic: Exposition of the System. Translated by Montgomery Furth, Berkeley: The University of California Press. ———.1979. “Logic in mathematics.” Hans Hermes, Friedrich Kambartel, Friedrich Kaulbach (eds.). Gottlob Frege. Posthumous writings, Oxford: Basil Blackwell: 203-50. Gauss, Carl F. 1799. “Letter to Wolfgang von Bolyai, end of 1799.” In Ewald 1996: 299. ———.1829. “Letter to Bessel, 27 January 1829.” In: Ewald 1996: I, 301 ———.1830. “Letter to Bessel, 9 April 1830.” In: Ewald 1996: I, 302. Gray, Jeremy. 1979. Ideas of space: Euclidean, Non-Euclidean, and Relativistic. Oxford: Clarendon Press. Greenberg, Marvin J. 1974. Euclidean and Non-Euclidean Geometries: Development and History. San Francisco: Freeman. Hilbert, David. 1899. Grundlagen der Geometrie. Leipzig: Teubner.

105 ———.1918. “Axiomatisches Denken.” Mathematische Annalen 78: 405-15. Kant, Immanuel. 1998. Critique of pure reason. Translated Paul Guyer, A. W. Wood, Cambridge: Cambridge University Press. ———. 2002. Prolegomena to Any Future Metaphysics That Will be Able to Come Forward as Science. In: Theoretical Philosophy after 1781, Translated and edited by H. Allison et al, Cambridge: Cambridge University Press: 49-169. Kline, Morris. 1980. Mathematics. The Loss of Certainty. Oxford: Oxford University Press. Kronecker, Leopold. 1887. “On the Concept of Number.” In: Ewald 1996: II, 947-55. Leibniz, Gottfried W. 1996. New Essays on Human Understanding. Translated and edited by P. Remnant and J. Bennett, Cambridge: Cambridge University Press. Monk, James D. 1976. Mathematical Logic. New York: Springer. Proclus. 1970. A Commentary on the First Book of Euclid’s Elements. Transl. by G.R. Morrow, Princeton: Princeton University Press. Riemann, Bernhard. 1868. “On the Hypotheses which lie at the Foundation of Geometry.” In: Ewald 1996: II, 652-61. Shapiro, Stewart. 2000. “Frege Meets Dedekind: A Neo-Logicist Treatment of Real Analysis.” Notre Dame Journal of Formal Logic 41: 335-64. ———. 2003. “Prolegomenon to any Future Neo-Logicist Set Theory: Abstraction and Indefinite Extensibility.” The British Journal for the Philosophy of Science 54: 59-91. Wright, Crispin, Hale, Bob. 2001. The Reason’s Proper Study. Essays Towards a NeoFregean Philosophy of Mathematics. Oxford: Oxford University Press.

BOB HALE

Mathematical Knowledge. A Defence of Modest and Sober Platonism The fundamental questions which any adequate philosophy of mathematics must answer can, in my view, be very easily and simply stated: What is mathematics about? and: How do we know about it? It is answering them—and upholding one’s answers in the face of the difficulties and objections which clamour for attention—that is the hard part. And one very central problem with which one must grapple is the problem of seeing how to tackle these questions in a such a way that one’s answer to one of them does not effectively scupper one’s chances of providing a credible answer to the other. We need, in other words, an account of what mathematical statements are about (and so of what is required for their truth) which meshes with a believable explanation of how we can know them to be true. Since this problem—of seeing how the ontology and epistemology of mathematics can peacefully co-exist—is, arguably, the central and basic philosophical problem in this area, I hope you will agree that it is equally central to our concerns in this conference. In what follows, I shall first give a fuller and more careful statement of the problem, drawing on the work of the philosopher who has done most to make it a primary focus of recent and contemporary philosophical attention; then, after some very brief remarks about other approaches, I shall try to explain—and do as much as space permits to defend—what I take to be the best way of trying to solve it. 1. Benacerraf’s Dilemma Paul Benacerraf begins one of the (deservedly) most influential articles in the last half century of philosophical work on mathematics by suggesting that accounts of mathematical truth have been motivated by two quite distinct concerns: (1) the concern for having a homogeneous semantical theory in which semantics for the propositions of mathematics parallel the semantics for the rest of the language

108 (2) the concern that the account of mathematical truth mesh with a reasonable epistemology1 He goes on to claim that: almost all accounts … [serve] one of these masters at the expense of the other … accounts of truth that treat mathematical and non-mathematical discourse in relevantly similar ways do so at the cost of leaving it unintelligible how we can have any mathematical knowledge whatsoever; whereas those which attribute to mathematical propositions the kinds of truth conditions we can clearly know to obtain, do so at the expense of failing to connect these conditions with any analysis of the sentences which shows how the assigned conditions are conditions of their truth. 2

Developing this thesis, he proposes two conditions which he thinks an acceptable account of mathematical truth should satisfy. One—which I shall call the semantic constraint—he formulates thus: any theory of mathematical truth [should] be in conformity with a general theory of truth … which certifies that the property of sentences that the account calls ‘truth’ is indeed truth … Perhaps the applicability of this requirement in the present case amounts only to a plea that the semantical apparatus of mathematics be seen as part and parcel of that of the natural language in which it is done, and thus that whatever semantical account we are inclined to give of names or, more generally, of singular terms, predicates, and quantifiers in the mother tongue include those parts of the mother tongue we classify as mathematese.3

The other—the epistemological constraint—is that: a satisfactory account of mathematical truth must be consistent with the possibility that some such truths be knowable. To put it more strongly, the concept of mathematical truth, as explicated, must fit into an over-all account of knowledge in a way that makes it intelligible how we have the mathematical knowledge that we have. An acceptable semantics for mathematics must fit an acceptable epistemology.4

1

Benacerraf 1973: 403. Benacerraf 1973: 403-4. 3 Benacerraf 1973: 408. 4 Benacerraf 1973: 409. 2

109 Whilst some further explanation—of what Benacerraf takes to be required for an acceptable epistemology—is obviously needed before the exact force of the epistemological constraint can be clear, its general drift and point is unmistakeable. The semantic constraint, however, is more loosely formulated and calls for some elucidatory comment. Part of what Benacerraf is getting at can be gathered from what he says about an example he discusses5, involving the statements: (1) (2)

There are at least three large cities older than New York There are at least three perfect numbers greater than 17

On the face of it, as Benacerraf observes, these statements share the same ‘logicogrammatical’ form: (3)

There are at least three FGs that bear R to a

They appear both to involve (numerically definite) quantification over objects, one- and two-place predicates of first-level (i.e. true or false of objects) and singular terms for objects, all put together in the same way. Immediately after stating the semantic constraint, Benacerraf says: … if we are to meet this requirement, we shouldn’t be satisfied with an account that fails to treat (1) and (2) in parallel fashion, on the model of (3). There may well be differences, but I expect these to emerge at the level of the analysis of the reference of the singular terms and predicates.6

Taken on its own, this suggests a very exacting reading of the semantic constraint, under which it can be satisfied only by an account of the truthconditions of (2) which respects its surface syntax exactly as just described. I am by no means sure that Benacerraf can have intended quite such a demanding interpretation of his first condition. Certainly, so understood, it would go well beyond the somewhat vague demand that “whatever semantical account we are inclined to give of … singular terms, predicates, and quantifiers in the mother tongue include those parts of [it] we classify as mathematese”. That would seem to come to no more than the weaker requirement—which Benacerraf certainly does endorse—that an account of the truth-conditions of mathematical statements should 5 6

Benacerraf 1973: 405. Benacerraf 1973: 408.

110 accord with a broadly referential (i.e. Tarskian) semantics for the language as a whole. This would leave room for accounts of mathematics which view the surface grammatical form of mathematical statements as more or less misleading as to their logical form, provided that their (alleged) logical form involves only devices amenable to Tarskian treatment. Although Benacerraf does not expressly present himself as posing a dilemma for accounts of mathematical truth, he is widely—and very plausibly—taken to have done so. Prominent among the accounts of mathematical truth Benacerraf discusses is what he terms the ‘standard’ or ‘platonistic’ account, which analyses (2) as being of the form (3)—and, more generally, takes the surface syntax of mathematical statements at face value as an accurate guide to their truth-conditions. This account, Benacerraf argues, complies with the first—semantic—constraint but is in deep trouble when it comes to the second—epistemological—constraint. A satisfactory general epistemology will, he holds, require, if a subject X is to know that p, that there should be a suitable causal connection between p’s truth-conditions on the one side and X’s grounds for believing that p on the other. Since, according to the platonistic account, the truth of a true mathematical proposition consists in the obtaining of suitable relations among numbers or other abstracta—not located in space or time and so outside the causal swim—it is quite unclear how any such suitable causal relationship could hold, and so quite unclear how mathematical knowledge can be so much as possible. If we focus especially on this part of Benacerraf’s argument, it is very natural to state the dilemma as Charles Parsons does: According to Benacerraf, our best theory of mathematical truth (Tarski’s) involves postulating mathematical objects, while our best account of knowledge requires causal relations of the objects of knowledge to us; but mathematical objects are acausal.7

I would express the dilemma a little differently. One requirement on a satisfactory account of the semantics of mathematical language is that it should associate mathematical statements with what can properly be taken to be conditions for their truth. Another is that it should not have the consequence that no true mathematical statements can be known to be true. Neither requirement taken by itself is unsatisfiable. The problem is to give an account which simultaneously satisfies both. 7

Parsons 1979. The quotation is from footnote 12.

111 However we formulate it, it deserves emphasis that it is a dilemma that Benacerraf is posing. Some commentators give the impression that Benacerraf’s primary purpose—or at least the principal effect of his paper—was to present a serious, and perhaps lethal, objection to platonism. Orthodox platonism takes the surface syntax of ordinary mathematical statements at face value. That is, it accepts the appearance such statements present—of involving terms having reference to distinctively mathematical objects (numbers of various kinds, sets, etc.) and quantifers ranging over domains of such objects—as an accurate guide to their logical form and so to their truth-conditions, and so endorses the natural or naïve view that mathematical statements convey truths (or falsehoods) about a distinctive range of mathematical objects. Further, since what mathematical singular terms designate, according to the platonist, are abstract objects, lying—as we figuratively put it—outside space and time, it would seem that the states of affairs depicted by purely mathematical statements (at least) can stand in no causal relations whatever, and so, in particular, can stand in no suitable causal relations to ordinary embodied and so spatio-temporally located knowing subjects. Orthodox—objectual—platonism is thus seems immediately in trouble over Benacerraf’s dilemma. Understandable as it may be, this view of the matter seems to me to involve a quite serious distortion of the issues. Of course, Benacerraf’s argument does raise a serious problem for platonists, and my primary aim in what follows will be to explain how I think a certain kind of platonism can circumvent it. But his problem is not just a problem for would-be platonists—it is a problem, if not quite for everyone, then at least for nearly all positions in the philosophy of mathematics. This is obvious enough in the case of at least one of the principal alternatives to objectual platonism advocated during the period across which Benacerraf’s Dilemma casts its shadow—namely the pure or ante rem8 structuralism which takes the subject matter of mathematical theories to be, not (or not primarily) numbers or other objects but pure structures such as the natural number structure, the real number structure, etc. Just because these structures are themselves every bit as abstract as the orthodox platonist’s objects, there is no evident reason to suppose this kind of structuralism any less subject to the epistemological access problem than orthodox objectual platonism. And whilst the very different—eliminative—variety of structuralism suggested by Benacerraf himself in an earlier and equally celebrated 8

As one of its leading proponents calls it—cf. Shapiro 1997.

112 article9 and more recently advocated in modal form by Geoffrey Hellman10 is precisely designed to avoid the ontological commitments incurred by objectual and structuralist versions of platonism, it is arguable that it cannot avoid epistemologically problematic commitments altogether.11 There are, of course, positions which, by rejecting altogether the conception of mathematics as a body of truths knowable a priori which platonism and structuralism of both kinds seek to uphold, can sidestep the epistemological problems which beset these more conservative philosophies of mathematics. Prominent among the more radical such positions are Quine’s global empiricism, Hartry Field’s highly unorthodox brand of nominalism, and Penelope Maddy’s version of mathematical naturalism. Since I lack the space required for a proper discussion of these and other alternatives which might seem to afford ways past Benacerraf’s Dilemma, I must simply record my view that to the extent to which they escape its epistemological horn, they do so only by either impaling themselves on its other horn, or by incurring unacceptable costs elsewhere. 2. Modest and Sober Platonism Explained Even if it is granted that Benacerraf’s dilemma poses a serious challenge, not just to orthodox objectual platonism, but to other leading approaches in the philosophy of mathematics, it is likely to be felt that it presents a peculiarly telling objection to platonism—that is especially difficult, if not impossible, to see how one can combine a standard platonist account of the truth-conditions of mathematical statements with a credible account of how we might know such statements to be true. In the remainder of this paper, I shall try to explain why I think that the belief that this is so rests upon an inadequate and impoverished understanding of how we may think and talk about abstract objects, and to set forth the essentials of a modest and sober kind of platonism which offers, in my view, the best prospect of a satisfactory positive resolution of Benacerraf’s dilemma. Why do we—or might we—find the idea that we may have knowledge about abstract objects so baffling? Not just, I think, because we subscribe to a causal theory of knowledge and cannot see how abstract objects could possibly conform its demands, however loosely formulated. No doubt—and no doubt largely as a result of Benacerraf’s 1973 paper— this has been the primary focus of anxiety about the abstract in recent 9

Benacerraf 1965. Hellman 1989. 11 For argument to this effect, see Hale 1996a. 10

113 decades. But, as Field and others have pointed out12, doubts about the capacity of platonism to deliver a credible epistemology do not have to be grounded in the adoption of a specifically causal analysis of knowledge13. For even if a causal constraint is not written into the analysis of knowledge, it ought to be possible to provide an explanation of our—or at least mathematicians’—capacity to get the mathematical facts straight, i.e. of how we can be relied upon to form true mathematical beliefs. But it is— or so it may be claimed—quite unclear how this is to be explained, on the assumption that their truth is constituted in relations among abstract objects. This is an important generalisation of Benacerraf’s problem14. However, epistemological perplexity about, and consequent suspicion of, abstract entities has a much longer history, far antedating the intrusion of specifically causalist or, more generally, reliabilist thinking in epistemology. It has, I think, other and more general sources, albeit ones that are less sharply defined and seldom unmistakeably articulated. Without pretending to an exhaustive diagnosis, I would identify two, or perhaps three, contributing factors. The first is that we tend to operate with a wholly negative conception of abstract objects as ‘outside’ space and time. This characterisation is obviously metaphorical, as well as negative—there is, literally, nowhere outside space and time. But this in itself need not be particularly damaging, so long as we remind ourselves, when necessary—that is, when we feel tempted to think of abstract objects as ‘in’ some queer sort of limbo—of the literal content of the metaphor: roughly, that it makes no sense to ask where an abstract object is, or when it came into existence, or how long it will last. It is, rather, the negative aspect of the characterisation that impedes constructive thought. Of course, it is true that abstract objects aren’t located in space or time. And it may be said that since that is enough to ensure that there is an apparently intractable problem about how spatiotemporally located knowers could know of their existence or know anything about them, it is pointless exercising ourselves over what more 12

See the title essay of Field 1989, especially pp.230-39, and its Introduction, pp.25-7. See also Hart 1977 and Maddy 1990: 42-5. 13 Just as well, since it is arguable that any analysis of knowledge that incorporates a causal condition strong enough to make trouble for platonism will be independently objectionable, as ruling out knowledge in other areas in which we should certainly not expect its possibility to be excluded simply by an analysis of what it required for us to know. See Steiner 1975, Wright 1983, section xi, and Hale 1987, ch.4. 14 See the paper by Field cited in footnote 4. Hale 1994a responds to Field’s generalised version of Benacerraf’s challenge.

114 positive characterisation, if any, they can be given. But that just misses my point. If we focus exclusively on what abstract objects are not, with no thought about what they are or might be supposed to be, we can scarcely expect anything but intellectual paralysis when we try to consider how we might get to know about them. The second factor is the idea that knowledge of truths about objects of any kind must involve ‘contact’ with those objects. If ‘contact’ is taken literally, so as to require some sort of physical connection or interaction— perhaps of the sort that occurs in normal sense perception, or even something more indirect—the idea is obviously inimical to platonism, but equally not obviously one that must be accepted. Of course, if it is given a sufficiently broad (and perhaps unavoidably metaphorical) construal, so that possession of any sort of identifying knowledge of an object suffices, the idea reduces, near enough, to a truism—one can hardly be credited with knowledge of truths about objects unless one knows which objects are in question—and it need then cause the platonist no trouble, unless it is coupled with the further idea that such ‘contact’ is presupposed by and must be already in place before any knowledge of truths about objects can be had15. Once we become locked into thinking about the access problem within this straitjacket, we can hardly avoid the further thought—that the access problem is not just a problem about how we can know anything about abstract objects, but goes wider and deeper: how can we even so much as think about them at all. Indeed, it really is incredible that we should be capable of identifying thought/talk about objects of some (any) kind, and yet be somehow incapable, even in principle, of acquiring knowledge about them. Critics of platonism may—and often do16—see this as just so much more grist to their mill: platonism is in trouble on two counts, not just one, 15

Something like this idea can be found in Russell’s writings around 1910-12. Russell distinguishes first between knowledge of truths and knowledge of things (cf. Russell 1912, ch.4), and then, within the latter, between knowledge by acquaintance and knowledge by description (op. cit., ch.5). He holds that knowledge of things, when it is got by acquaintance, is essentially simpler than knowledge of truths. Whilst stopping short of claiming either that we do in fact have such knowledge without having knowledge of any truths or that we must be able to, he does go on to propound his famous principle that “Every proposition which we can understand must be composed wholly of constituents with which we are acquainted” Russell 1912: 58. 16 Benacerraf himself hints (see his 1973: 412) that the problem for the platonist is as much a problem about reference as about knowledge. Field is more explicit on the

115 because it obstructs both a satisfactory epistemology and a workable theory of reference. But this way of putting the difficulty obscures an important connection. The right way to put the objection—if one believed it to be well-taken—would be rather this: even if one could give a platonist account of the truth conditions for mathematical statements (or any other class of statements supposed about abstract objects), we would be unable to explain how such statements, so understood, could be known or reasonably believed; but in fact one cannot even give such a semantical account, since one cannot even so much as make reference to ‘objects’ of the sort such an account takes them to be about—and if one cannot do that much, one cannot so much as state platonistic truth-conditions. I think it is helpful to recast the objection in this way, because it gives us a clearer view of the other side of the coin—of the structure of the task that must be addressed if a sober and responsible form of platonism is to be uncovered. The fundamental problem is not how, given that mathematical statements are about abstract objects, we could know them to be true, but how they could be about such objects in the first place—and that is the problem the platonist should address first. Of course, we have no right to suppose that a positive solution to the epistemological problem would simply fall out of a solution to the problem of reference. On the face of it, solving the latter problem—the reference problem—is merely a necessary, and by no means a sufficient, condition for a solution to the former—the epistemological problem. But I believe that if the platonist approaches, and solves, the reference problem in the right way, he will in fact be well on the way to solving the epistemological problem—at least in so far as it is fuelled by the belief that the abstractness of numbers puts them, or truths about them, beyond our ken17. The essential steps towards a satisfactory explanation how we may attain to concepts of kinds of abstract object and engage in identifying thought and talk about them are taken—or so I believe—in the central sections of Frege’s Grundlagen18. At §62, Frege explicitly confronts the access problem—as a problem about how we can refer, in thought or speech, to abstract objects—in what is, for him and for us, the crucial case: matter, suggesting that the problem of reference is the more fundamental—see, for example, his remarks on the matter in Field 1989: 68. 17 The point of the qualification is that there are other issues—such as how one could be assured of the infinity of the natural numbers, and similar questions about the real numbers and sets—which aren’t resolved just by getting assurance that the abstractness of the objects in question need not be an obstacle. I return to this below. 18 Frege 1884.

116 “How, then, is a number to be given to us, if we can have no idea or intuition of it?” His answer is well-known and has been much discussed, but I hope you will bear with me while I go through its key moves once again. Frege’s answer begins with, and is shaped by, his famous context principle, restated here in the form19: “Only in the context of a proposition does a word mean anything”. From this he directly infers that “what matters [i.e. so far as answering his opening question is concerned] is to explain the sense of a proposition in which a number word occurs”. This so far leaves us, as Frege observes, a pretty free hand in choosing what kind of proposition, containing a number word, to explain. But, he continues, “we have already settled that number words are to be understood as standing for self-subsistent objects”, and that determines a kind of propositions which must have sense—namely ones which “express our recognition [of numerical identity]”. That is, what must be explained is the sense of identity propositions connecting terms for numbers. Since, for Frege, (cardinal) numbers are essentially numbers of concepts, such identity propositions will, in the fundamental case, be of the form: The number of Fs = the number of Gs. And since, for Frege, an explanation of the sense of a proposition in general works by stating its truth-conditions, what is needed is to explain the truth-conditions of propositions of that form. The significance of these first, revolutionary, steps can scarcely be over-emphasised—since if they are well-taken, they contain the essential ingredient—or, more accurately, one of the essential ingredients—of an antidote to aporia about abstract reference. For if Frege is right, it suffices to establish a use for terms purporting reference to numbers, or other abstract objects—that is, objects which are not ‘external’ (located in space), and of which we can have no ‘idea’ or ‘intuition’, but which are nonetheless objective—to explain the truth-conditions of propositions incorporating such terms. And if, under a suitable explanation of the truthconditions of an appropriate range of such propositions, those propositions are true, then those of their ingredient terms which purport reference to 19

The principle is identified in the Introduction (p.x) to Grundlagen as one of the fundamental principles guiding his enquiry, where it is stated: “We must ask after the meaning of a word in the context of a proposition, not in isolation”. On the face of it, this is weaker than the version to which Frege appeals in §62, since it does not strictly imply that a word has no meaning in isolation. But I do not think we need exercise ourselves over this, since the ostensibly weaker principle would suffice to underpin Frege’s answer.

117 numbers or other abstract objects will in fact refer to such objects. To suppose, to the contrary, that before we can entertain true or false thoughts about objects of a certain kind—abstract or concrete—we must have already prepared the ground by first somehow fixing or explaining the use of terms standing for those objects independently of any occurrence of those terms, functioning as names or singular terms, in complete sentences apt for the expression of true or false thoughts, is simply to fail to appreciate the point and force of the context principle. Before I move on, I want to emphasise the scope of the point about meaning in general and singular reference in particular which I take Frege to be making. The context principle is not just—as is sometimes uncharitably supposed—a convenient device Frege reaches for as a means of evading special difficulties about talk of abstract objects. If the context principle is right at all, it is completely general in application, and applies to the meaning and reference of words for concrete objects just as much as to talk of abstracta. It might be supposed otherwise—that in the case of terms for concrete objects, we may somehow fix their intended reference ostensively, simply by pointing to the object(s) for which they are to stand and pronouncing the word. I shall not belabour the points Wittgenstein makes in the early sections of Philosophical Investigations to disabuse us of illusions about the capacities of ostensive definition. Of course it is possible—in a suitable setting—to introduce a singular term for a concrete object by pointing to the object intended. But crucially, it needs to be understood (i) that a term for a particular object of some kind is being introduced and (ii) what kind of object is in question. In both cases, understanding requires a grasp of the rôle of expressions of the kind in question in complete sentences; and possession of a concept of a kind of objects—a sortal concept—calls for understanding of what makes for identity or distinctness among its instances, and so for a grasp of the truthconditions of identity statements involving terms of the kind being introduced. Brute sensory, or otherwise causal, encounters with objects do not suffice to establish an understanding of terms as referring to them— and when we see what else is required, it is at least not clear that such encounters are quite generally necessary, even if they may have an indispensable part to play in establishing a referential use for some terms standing for certain kinds of concrete objects.

118 To return to our main business, Frege next proposes to explain the truth-conditions of numerical identity propositions by means of what is now widely referred to as: Hume’s Principle:

The number of Fs = the number of Gs ↔ there is a one-one correspondence between the Fs and the Gs

He then proceeds, in response to a doubt about the propriety of his proposed explanation, to offer an elucidation of how his contextual explanation of number—as a sortal concept whose instances are the referents of numerical terms of the kind on display—is intended to work. The objection is that it is a mistake to define identity specially for the case of numbers, as he appears to do. His reply is that this is precisely what he is not doing. His aim, rather, is “to construct the content of a judgement which can be taken as an identity, each side of which is a number”, so that rather than defining identity specially for this case, he is “using the concept of identity, taken as already understood, as a means for arriving at that which is to be regarded as identical”. The remainder of his elucidation is given at one remove, in terms of what Frege takes to be the simpler but sufficiently analogous explanation of the concept of direction obtained by laying down the Direction Equivalence:

The direction of line a = the direction of line b ↔ lines a and b are parallel

Of this, he writes: The judgement “line a is parallel to line b”, or, using symbols, a//b, can be taken as an identity. If we do this, we obtain the concept of direction, and say: “the direction of line a is identical with the direction of line b”. Thus we replace the symbol // by the more generic symbol =, through removing what is specific in the content of the former and dividing it between a and b. We carve up the content in a way different from the original way, and this yields us a new concept.20 Frege’s talk of carving up the same content in a new way is, of course, a metaphor—albeit a highly suggestive and potentially helpful one—and it is therefore essential, if we are to make anything of it, to spell out in plainer 20

Frege 1884 §64.

119 terms the process of conceptual surgery he may reasonably be supposed to have had in mind. In particular, we need to understand in what way or sense instances of the left and right sides of the Direction Equivalence (or Hume’s principle, or, more generally, any similar abstraction principle) may be held to share the same content. As a matter of Fregean exegesis, this task is rendered the more difficult by at least two historical facts. One is that Frege’s use of the term ‘content’ in Grundlagen and earlier writings is anything but uniform—he speaks both of the content of whole sentences and of the content of subsentential expressions, including names; sometimes by the content of a symbol he seems to mean what it stands for, but he also speaks of conceptual content in ways that suggest something more like what he later calls sense. The other is that soon after Grundlagen21, he expressly abandons the notion of judgeable content, replacing it by the contrasted notions of thought and truth-value, as a special case of his general distinction between sense and reference. There is one passage in Begriffsschrift22 in which Frege draws attention to the possibility of ‘splitting up’ one and the same sentence in different ways into function and argument(s). His example is ‘Cato killed Cato’, where we may with equal propriety view the first occurrence of ‘Cato’ as supplying the argument to the function ‘… killed Cato’, or the second occurrence of ‘Cato’ as supplying the argument to the function ‘Cato killed …’ (or ‘being killed by Cato’), or, finally, we may view both occurences of ‘Cato’ as supplying a single argument to the function ‘… killed …’, where it is understood that both gaps are to be filled by the same name—so that we have the single argument function ‘ξ killed ξ’, as distinct from the twoargument function ‘ξ killed ζ’. Although this passage has been taken to provide the model for Frege’s subsequent talk of carving content, I do not think it can do so—what is involved here is the possibility of alternative syntactic decompositions of one and the same sentence, and as such it does not begin to explain what he might have meant by speaking of two completely different sentences as sharing their content. That seems rather to involve a semantic, rather than syntactic, relationship.

21

Frege 1892: 198 (cf. Geach & Black 1952: 47). Frege 1879, §9. A partial translation which includes this section appears in Geach & Black 1952.

22

120 Since I do not now have time to pursue this issue in the detail it deserves23, I shall instead simply state, without supporting argument, how I think Frege is best understood. We may view the Direction Equivalence as an implicit definition of the direction operator—‘the direction of …’—and thereby of the sortal concept of direction. The import of the implicitly definitional stipulation is simply that corresponding instances of the left and right sides—matching sentences of the shapes ‘the direction of line a = the direction of line b’ and ‘lines a and b are parallel’—are to be alike in truth-value.24 Thus what a recipient of the explanation immediately learns is that whatever suffices for the truth of a statement of line-parallelism is equally sufficient for the truth of the corresponding statement of directionidentity. However, she also understands that she is to take the surface syntax of direction-identity statements at face value. She already possesses the general concept of identity, and so is able to recognise that the expressions flanking the identity sign must be singular terms. Further, she already understands terms for lines, and so can recognise ‘the direction of …’ must be being introduced as a functional expression, denoting a function from lines to objects. From this, she is able to gather that the objects in question simply are objects for whose identity it is necessary and sufficient just that the relevant lines be parallel. She learns, in other words, that directions just are objects with those identity-conditions, and thus acquires the concept of direction—it is the sortal concept under which fall exactly those objects having those identity-conditions. On this account, the force of Frege’s metaphor is that a statement of direction-identity may be viewed as involving a reconceptualization of the very same state of affairs as suffices for the truth of the corresponding statement of line-parallelism. Instead of thinking of that state of affairs as consisting in two lines bearing to one another the relation of parallelism, we may equally conceive of it as consisting in the relation of identity holding among objects of a ‘new’ kind—the directions of those lines. 23

The suggestion that the Begriffsschrift passage must be the model for Frege’s subsequent talk of carving up content is made by Michael Dummett—see Dummett 1991: 173. For a dissenting view, see Hale 1996b. 24 In other words, what is stipulated is only the truth of the material biconditional ‘the direction of line a = the direction of line b iff lines a and b are parallel’—there is no stipulation, or consequential claim, that the left and right sides of the biconditional have the same sense. Thus Kit Fine’s main objection (see Fine 2002: 35-41) to what he terms ‘definition by reconceptualization’—which assumes that the left and right sides of an abstraction principle are supposed to have the same sense—is simply misdirected, if taken as an objection to the neo-Fregean proposal presented here.

121 Hume’s Principle may likewise be seen as implicitly defining a concept of cardinal number in such a way that statements of numerical identity involve reconceptualizations of the very same states of affairs as are reported by means of statements of one-one correspondence between concepts. And, as Frege undoubtedly saw, the procedure has potentially a much wider application. The Direction Equivalence and Hume’s Principle are two instances of a general type of abstraction principle—that is, a principle of the shape: (Abs) ∀α∀β ( ∑(α) = ∑(β) ↔ α ≈ β) where ≈ is an equivalence relation on entities of the type of α,β, and ∑ is a function from entities of that type to objects. Given any suitable equivalence relation, we may abstract over it to introduce an abstract sortal concept—a concept under which fall abstract objects of a certain type, the identity or distinctness of which is constituted by the obtaining of that equivalence relation among entities in its field. There is a great deal more to be said in explanation, qualification and defence of this procedure of implicit definition by Fregean abstraction. However, it should already be evident that, if it constitutes a viable means of introducing abstract sortal concepts, such as that of direction, and establishing a use for terms potentially denoting their instances, it furnishes also at least a partial means of resolving the problem of epistemological access standardly taken to beset platonism. The problem, in general terms, was to see how statements of a given kind can be understood as involving reference to abstract objects and can yet remain, at least in principle, humanly knowable, given that the objects they concern are outside space and time and in consequence can stand in no sort of intelligible, epistemologically relevant, causal relations to human knowers. And the solution—in the case of directions—is that, provided that certain kinds of statement about lines can be known to be true, there can be no further problem about our capacity for knowledge of the truth of members of a certain basic class of statements about directions. For if the concept of direction is introduced by Fregean abstraction, statements of directionidentity are true, if true at all, in virtue the very same states of affairs as statements asserting relevant lines to be parallel. And so it is with a basic class of statements about numbers. A statement of numerical identity—in the fundamental case, a statement of the kind: the number of Fs = the number of Gs—is true, if true, in virtue of the very same state of affairs

122 which ensures the truth of the matching statement of one-one correspondence among concepts. The way past Benacerraf’s dilemma I am commending is, then, in essence breathtakingly simple. So long as we can ascertain that lines are parallel, or that concepts are one-one correspondent, there need be no further problem about our knowledge of certain basic kinds of truths about directions and numbers, for all their abstractness—provided that the concepts of direction and number can be implicitly defined by Fregean abstraction, we can know statements of direction- and numerical-identity to be true just by knowing the truth of the appropriate statements of parallelism among lines and one-one correspondence among concepts, because if true at all, they are true in virtue of the very same states of affairs. To conclude, I would like to say something—not enough, I’m sure—about some likely sources of dissatisfaction with it, and to note and comment upon some important limitations. 3. Modest and Sober Platonism Defended The Fregean abstractionist may appear vulnerable to attack from two quite different—and indeed doctrinally diametrically opposed—positions. On the one hand, philosophers of a nominalistic persuasion will challenge his entitlement to endorse or stipulate the requisite abstraction principles, when their left hand side syntax is construed—as the Fregean wishes it to be construed—as semantically significant. On the other, philosophers of a more pronounced platonistic tendency will decry his explanation for almost the opposite reason—in their view, the Fregean abstractionist has not rescued anything worth calling platonism at all, but has merely sold out in favour of an Ersatz so dilute in content as to be nugatory. Let me try briefly to indicate how I think the Fregean may avoid being caught in the ensuing crossfire. Nominalists—or at least nominalists of a more orthodox variety— will insist that we may accept abstraction principles as explanatory of the truth-conditions of their left hand sides only if we treat the latter as devoid of semantically significant syntax, beyond their involving genuine occurrences of the term or predicate variables α and β. That is, they will admit as legitimate only an austere reading of the equivalences, according to which their instances involve, on their left hand sides, just occurrences of terms or predicates replacing the variables α and β, embedded in the argument places of a single unbreakable predicate whose surface syntactical complexity betokens no genuine semantic structure. So what,

123 on their only acceptable reading, the equivalences accomplish is the introduction of a new, unstructured two-place predicate—for example: ‘the-direction-of-…-=-the-direction-of-__’ as no more than alternative notation for the equivalence relation that figures on the right hand side—in this case, the equivalence relation expressed by: ‘… is parallel to __’. The crucial question for a nominalist of this stripe is how he justifies his insistence upon this austere reading. His likely answer will be that since the statement ‘lines a and b are parallel’ plainly involves no reference to directions, conceived as a species of abstract object distinct from lines themselves, it can be taken as explaining the truth-conditions of ‘the direction of line a = the direction of line b’ only if the latter is likewise understood as devoid of reference to any such objects. But this seemingly simple answer relies upon a questionable assumption. It is certainly true that statements of line-parallelism involve no terms purporting reference to directions. And it is equally clearly true that two statements cannot have the same truth-conditions if one of them carries a commitment to the existence of entities which is absent from the other. But the premiss the nominalist requires—if he is to be justified in inferring that statements of parallelism can be equivalent to statements of direction-identity only if the latter are austerely construed—is that statements of parallelism do not demand the existence of directions. This does not follow from the acknowledged absence, in such statements, of any explicit reference to directions. The statement that a man is an uncle involves no explicit reference to his siblings or their offspring, but it cannot be true unless he has a non-childless brother or sister. The nominalist may—indeed, must— accept that there can be existential commitment in the absence of explicit reference, but may object that whilst no one would count as understanding the statement that Edward is an uncle if she could not be brought to agree that its truth requires the existence of someone who is brother or sister to Edward and father or mother of someone else, it is quite otherwise with statements of parallelism. Someone may perfectly well understand the statement that one line is parallel to another without being ready to acknowledge—what the nominalist denies—that its truth requires the existence of directions. The abstractionist reply is that so much is perfectly correct, but insufficient for the nominalist’s argument. It is correct precisely because—just as the abstractionist requires—understanding talk of lines and parallelism does not demand possession of the concept of direction. But it is insufficient, because the question that matters is, rather, whether one who is fully cognisant with all the relevant concepts—that is,

124 not only the the concepts of line and parallelism but also with the concept of direction—could count as fully understanding a statement of parallelism without being ready to agree that its truth called for the existence of directions. If the concept of direction is implicitly defined, as the abstractionist proposes, by means of the Direction Equivalence, she could not. Insisting at this point that no such explanation can be admitted, because only an austere reading of the Direction Equivalence is permissible, amounts to no more than an unargued—and obviously question-begging—refusal to entertain the kind of explanation on offer. This kind of nominalist wants to make out that in laying down an abstraction principle, non-austerely construed, one must be illegitmately assuming the existence of new objects, just as one would be if one were to stipulate that the eldest nephew of Edward is the eldest nephew of Bill if and only if Edward and Bill are brothers. But in his attempt to do so, he simply refuses to take seriously the idea that statements of different types may be seen as alternative modes of conceptualizing a single common type of states of affairs—as is manifestly not so, in the case of our statements about nephews and brothers. The accusation that Fregean abstractionist platonism falls so far short of the genuine article as to be unworthy of the name ‘platonism’ at all is harder to come to grips with, partly because it is hard to find a clear, articulate and non-metaphorical account of what ‘genuine’ platonism is supposed to involve, partly because the accusation is consequently somewhat vague, and partly because any of several distinct things may lie behind it. One point around which the vague feeling that Fregean platonism is insufficiently ‘robust’ may crystallise connects directly with the charge— itself emanating from a quite different direction—that the proposal to define number implicitly by means of Hume’s Principle amounts to an illegitimate attempt to secure the existence of numbers by stipulation—to define them into existence25. The would-be more robust platonist may sympathise with the charge, but see it as highlighting the abstractionist’s failure to justify the mind-independence of numbers—instead, they turn out, for the abstractionist, to be mere creatures of our understanding, faint and insubstantial shadows cast by mathematical language as reconstructed on abstractionist lines. The charge itself—and with it, the invidious gloss the robust platonist puts upon it—overlooks at least two vital points. First, 25

For one recent expression of this charge, see Potter & Smiley 2001, to which Hale 2001 responds.

125 what is stipulated, when an abstraction principle is advanced as an implicit definition, is not the existence of certain objects—referents for the terms featured on its left hand side—but the truth of (indefinitely many) biconditionals co-ordinating identity-statements linking such terms with statements involving the relevant equivalence relation over the underlying domain. The truth of any given one of those identity statements, and hence the existence of objects to which its ingredient terms refer, is not stipulated, but follows only given the truth of the co-ordinated statement to the effect that the abstraction’s equivalence relation holds among the relevant objects (or concepts, in case of a higher-order abstraction such as Hume’s Principle). And the truth of that latter statement will be always a matter of independently constituted fact (about parallelism of certain lines, or one-one correspondence between certain concepts, etc.). What is brought into existence by the stipulation—if anything is—is not objects, but a certain sortal concept. What objects, if any, fall under it is—as I’ve said—entirely dependent upon the truth of instances of the abstractive biconditional’s right hand side. Second, whilst it is a contingent matter of human psychology whether the possibility of forming a sortal concept by means of such a stipulation happens to occur to us, the success of the stipulation is an entirely objective matter—essentially, a matter of its compliance with the constraints—including, centrally, consistency and a certain kind of conservativeness, but probably others besides—which govern definition and concept-formation in general. Once it is seen that the existence of abstract objects is thus essentially dependent upon the objective truth of instances of the right hand side of an abstraction principle, whose acceptability is itself dependent upon its conformity to objective constraints, it is hard to see what can remain of the charge that those objects are somehow deficient in point of mind-independence. I want, finally, to emphasise some important limitations of this defence of the Fregean response to Benacerraf’s dilemma, so far as I have been able to develop it here. I have been exclusively concerned with one central aspect of the dilemma, as it confronts the platonist—the thought that the very abstractness of the objects of which, as the platonist construes them, mathematical propositions speak must, in and of itself, render mathematical knowledge impossible. I have tried to explain how, if talk and thought of abstract objects is grounded in abstraction principles such as Hume’s Principle, taken as implicitly definitional of relevant sortal concepts such as cardinal number, this apparent epistemological impasse may be circumvented.

126 There are undoubtedly other aspects of the problem of mathematical knowledge, even in the most elementary case of the arithmetic of natural numbers. Many arithmetic truths, including most mathematically interesting ones, depend essentially on the fact that the natural numbers form a progression, i.e. come in an infinite sequence with a first element, and for each element a unique next element distinct from each of its predecessors and successors. Part of the problem of mathematical knowledge is therefore to explain how we can know that there is such a sequence and know truths about it. This problem does not have anything especially to do with the abstractness of the natural numbers—save, perhaps, in the minimal sense that it is probably only in the case of abstract objects of some kind that we can expect to know, or at least be able to prove, that there exists an infinity of them—so it is not a problem that distinctively afflicts platonism, but a problem for everyone. But it demands a solution, and the Fregean platonist is well-placed—indeed, in my view, uniquely well-placed—to provide one. For, as is well-known, the adjunction of Hume’s Principle to a suitable underlying second-order logic enables us to prove that every natural number has a unique successor. The proof26 depends crucially on a feature of Hume’s Principle which sets it apart from simpler abstractions such as the Direction Equivalence, namely its impredicativity. To show that for a given natural number n, there is another natural number which immediately succeeds it, we take the concept: natural number which ancestrally precedes or is identical with n. Since 0,1, …,n all fall under this concept, its number is n+1. This requires us to take the range of the second-order variables F and G in Hume’s Principle to include concepts under which the numbers themselves fall, and this in turn requires that the first-order quantifiers in the definitional expansion of the principle’s right hand side be understood, impredicatively, as binding variables that range over the numbers themselves. The Fregean must therefore defend himself against the charge that this impredicativity amounts to vicious circularity or is somehow otherwise objectionable27.

26

Sketched by Frege himself (Frege 1884, §§82,83) and reconstructed by Crispin Wright (see Wright 1983, §xix). Other detailed accounts of the proof are given by George Boolos (in Boolos 1987 and in an appendix to Boolos 1990), and by George Boolos and Richard Heck (Boolos & Heck 1998). 27 See Dummett 1991, ch.18; Hale 1994b, Wright 1998a, Dummett 1998, and Wright 1998b.

127 Further, whilst there is less than universal consensus—even among classically minded mathematicians and philosophers of mathematics— about the full extent of our mathematical knowledge, almost no one—if indeed anyone—supposes it stops short at elementary arithmetic. A viable epistemology for mathematics ought to encompass, at a minimum, the foundations of the theory of real numbers, the various branches of analysis, and, arguably, at least a significant portion of set theory, even if not the whole of it. From the Fregean perspective, the question is whether we can formulate otherwise acceptable abstraction principles of sufficient power. If that can be done, extending the programme beyond arithmetic will, so far as I can see, raise no new distinctively epistemological problem. I do not mean to appear unduly sanguine—some steps in the required direction have been taken28, but much work remains to be done, and it is a largely open question, at present, how much can be accomplished. There I must end. As anyone at all familiar with the territory will know, there are many difficulties for and objections to the Fregean approach which I have not been able to mention, much less address29. But I hope to have done enough to persuade you that that approach offers a promising route between the horns of Benacerraf’s famous dilemma30.

28

On the reals and real analysis, see Hale 2000a, Wright 2000 and Shapiro 2000. For some preliminary discussion of the prospects for an abstractionist theory of sets, see Wright 1997, and Hale 2000b. For further details of recent work in this area, see the bibliographies accessible on the Arché website at http://www.st-and.ac.uk/~arche/ 29 For a survey of the difficulties and outstanding problems, see the introduction and postscript to Hale & Wright 2001. 30 This paper was originally presented at the Conference on the Epistemology of Basic Belief, held in the Free University of Amsterdam, June 20th –22nd 2001. I should like to express my gratitude to the organisers for inviting me to participate, to all those who took part in the discussion—and to Ron Rood especially—for their comments, and to Adam Rieger and Crispin Wright for helpful reactions to earlier and later drafts.

128 REFERENCES Benacerraf, Paul. 1965. “What Numbers Could Not Be”, Philosophical Review 74: 4773. ——— .1973. “Mathematical Truth”, Journal of Philosophy 70: 661-80, reprinted in Paul Benacerraf & Hilary Putnam, eds. Philosophy of Mathematics: Selected Readings 2nd edn. Cambridge: Cambridge University Press 1983: 40320. All page references to this article are to this reprint. Boolos, George. 1987. “The Consistency of Frege’s Foundations of Arithmetic” in J.Thomson, ed. On Being and Saying: Essays in Honor of Richard Cartwright. Cambridge, Mass: MIT Press: 3-20; reprinted in Boolos 1998. ——— .1990. “The Standard of Equality of Numbers”, in George Boolos, ed. Meaning and Method: Essays in Honor of Hilary Putnam. Cambridge: Cambridge University Press 1990: 261-77, reprinted in Boolos 1998. ——— .1998. Logic, Logic and Logic. Cambridge, Mass: Harvard University Press. Boolos, George & Heck, Richard Jnr. 1998. “Die Grundlagen der Arithmetik §§82-3” in Schirn 1998: 407-28, reprinted in George Boolos, 1998. Dummett, Michael. 1991. Frege: Philosophy of Mathematics. London: Duckworth ——— . 1998 “Neo-Fregeans: in Bad Company?” in Schirn 1998: 369-87. Field, Hartry. 1989. Realism, Mathematics and Modality. New York: Blackwell. Fine, Kit. 2002. The Limits of Abstraction. Clarendon Press, Oxford. Frege, Gottlob. 1879. Begriffsschrift, eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. Halle:L.Nebert ——— .1884. Die Grundlagen der Arithmetik Breslau: Wilhelm Koebner, translated into English by J.L.Austin as The Foundations of Arithmetic. Oxford: Blackwell 1959. ——— 1892. “Über Begriff und Gegenstand”, originally published in the Vierteljahrsschrift für wissenschaftliche Philosophie 16: 192-205, and translated as “On Concept and Object” in Geach & Black 1952: 42-55. ——— .1893. Die Grundgesetze der Arithmetik vol.1 Jena: H.Pohle, part translated into English by Montgomery Furth in The Basic Laws of Arithmetic. Berkeley: University of California Press 1964.

129 Geach, Peter & Black, Max. 1952. Translations from the Philosophical Writings of Gottlob Frege. Oxford: Blackwell. Hale, Bob. 1987. Abstract Objects. Oxford: Blackwell. ——— .1994a. “Is platonism epistemologically bankrupt?”, Philosophical Review Vol 103, No 2: 299-325, reprinted in A.Casullo, ed. A Priori Knowledge, The International Research Library of Philosophy 24. Aldershot: Ashgate 1999. ——— . 1994b. “Dummett’s critique of Wright’s attempt to resuscitate Frege”, Philosophia Mathematica (3) Vol 2: 122-47, reprinted in Hale & Wright 2001: 189-213. ——— .1996a. “Structuralism’s unpaid epistemological debts”, Philosophia Mathematica (3) Vol 4: 124-47. ——— .1996b. “Grundlagen §64”, Proceedings of the Aristotelian Society 97: 24361, reprinted with a new appendix in Hale & Wright 2001: 91-116. ——— . 2000a. “Reals by abstraction” , Philosophia Mathematica (3) Vol.8, 2000: 100-123, reprinted in Hale & Wright 2001: 399-420. ——— .2000b. “Abstraction and Set Theory”, Notre Dame Journal of Formal Logic Vol 41, No4: 379-98. ——— .2001. “Reply to Potter and Smiley”, Proceedings of the Aristotelian Society 101: 339-358. Hale, Bob & Wright, Crispin. 2001. The Reason’s Proper Study: Essays towards a Neo-Fregean Philosophy of Mathematics. Oxford: Clarendon Press. Hart, W.D. 1977. “Review of Steiner 1975”, Journal of Philosophy 74: 118-29. Hellman, Geoffrey. 1989. Mathematics without Numbers. Oxford: Clarendon Press. Maddy, Penelope. 1990. Realism in Mathematics. Oxford: Clarendon Press. Parsons, Charles. 1979. “Mathematical Intuition”, originally published in the Proceedings of the Aristotelian Society 80 1979-80:145-68 and reprinted in W.D.Hart, ed. The Philosophy of Mathematics. Oxford:Oxford University Press 1996: 95-113. Potter, Michael & Smiley, Timothy. 2001. “Abstraction by Recarving”, Proceedings of the Aristotelian Society 101: 327-38.

130 Russell, Bertrand. 1962 [1912]. The Problems of Philosophy. London: Oxford University Press. Schirn, Matthias (ed.). 1998. Philosophy of Mathematics Today. Oxford: Clarendon Press Shapiro, Stewart. 1997. Philosophy of Mathematics: Structure and Ontology. Oxford and New York, Oxford University Press. ——— .2000. “Frege Meets Dedekind: A Neologicist Treatment of Real Analysis”, Notre Dame Journal of Formal Logic Vol 41, No4: 335-64. Steiner, Mark. 1975. Mathematical Knowledge. Ithaca, New York: Cornell University Press. Wright, Crispin. 1983. Frege’s Conception of Numbers as Objects. Aberdeen: Aberdeen University Press. ——— .1997. “On the Philosophical Significance of Frege’s Theorem” in Richard Heck, Jnr., ed. Language, Thought and Logic: Essays in Honour of Michael Dummett. Oxford: Clarendon Press: 201-44; reprinted in Hale & Wright 2001: 272-306. ——— .1998a. “On the Harmless Impredicativity of N= (Hume’s Principle)” in Schirn 1998: 339-68, reprinted in Hale & Wright 2001: 229-55. ——— .1998b “Response to Dummett” in Schirn 1998: 389-405, reprinted in Hale & Wright 2001: 256-71. ——— .2000. “Neo-Fregean Foundations for Real Analysis: Some Reflections on Frege’s Constraint” , Notre Dame Journal of Formal Logic Vol 41, No4: 317334.

STEVEN D. HALES

A Trilemma for Philosophical Knowledge Philosophers rely on intuition all the time, intuition that does not depend on the senses or deliver hypotheses about the physical world. This sort of intuition, that I will call rational or philosophical intuition, is widely considered to be the way that we come to know necessary truths about morality, metaphysics, modality, and the mind. You can’t pick up a philosophical journal without seeing appeals to intuition or intuitiveness, and it is explicitly endorsed by plenty of our contemporaries such as Ernest Sosa, David Chalmers, Daniel Dennett, Saul Kripke, and Laurence Bonjour. Despite widespread enthusiasm for rational intuition, it is not the only way we can acquire basic beliefs about philosophical propositions. Is it the best way? I will argue that there are no good reasons (so far in the literature) for thinking that beliefs delivered by the method of intuition are likelier to be true or are more epistemically virtuous than those beliefs about philosophical propositions produced by other methods. Since we philosophers have no good reason to prefer intuition to these other methods, we are left with three unpalatable choices. 1. Nihilism. There are no philosophical propositions truths after all. The naturalizers are right and all truths are empirical and are to be discovered by the methods of the natural sciences. Philosophy has no special autonomy or significance. 2. Skepticism. Since we have no defensible reason to prefer one basic method of acquiring beliefs about philosophical propositions over another basic method that gives different results, any true beliefs we have about philosophical propositions is accidental. In other words, we cannot have knowledge of philosophical propositions. 3. Relativism. What philosophical propositions get sanctified as true depends entirely on the basic method one selects to acquire them. If there are knowable philosophical truths, then what they are is deeply relative to method. This is a radical conclusion: either all of intuition-driven philosophy is bogus (option 1), a waste of time (option 2), or relativism is true. Well, you might ask, if we don’t use intuition to get at philosophical truths, what else could we use? The real challenge to the superiority of intuition comes when we examine other belief-acquiring methods that are

132 taken very seriously by serious people. There are such methods, and I will examine one of them: the ritual use of hallucinogens. In his history of drugs, Antonio Eschohotado writes, Before the supernatural became concentrated into written dogmas, when priestly classes interpreted the will of a sole and omnipotent god, what was perceived in altered states was the core of innumerable cults, precisely under the heading of revealed knowledge. The first hosts or holy sacraments were psychoactive substances, such as peyote, wine, or certain fungi.1

There is a complex interconnection between the use of psychoactive substances and religion. In ancient Greece, the use of wine was associated with the god Bacchus, and was seen as a gift of the god. Wine was not the only psychoactive substance the Greeks used in worship. Perhaps the most important of the ancient Greek religions was the worship of the fertility goddess Demeter through the Eleusinian Mysteries. The Eleusinian Mysteries attracted thousands of celebrants from pre-Homeric Greece to the height of Imperial Rome. It was a religion that lasted 2000 years, and was treated with great respect; one hierophant (the high priest or priestess of the cult) bragged that she had crowned three Roman emperors, including Marcus Aurelius.2 The central teachings of the Mysteries are lost to our time. They were an oral tradition, passed down from hierophant to hierophant, and never written down. Furthermore, there were severe civil penalties if initiates into the religion ever spoke about or revealed what they witnessed at the Mysteries. That said, there is a good bit known about the religion. One is that an important part of the initiation ceremony was the drinking of the kykeón, a potion composed of water, a grain (probably barley), and mint. This ritual was connected with the myth of Demeter and her breaking of a fast with the kykeón. Several recent writers have argued that the kykeón was a critical component of the Mysteries because it contained a grain parasite, the ergot mushroom (Calviceps purpurea), which is an hallucinogenic fungus that grows wild near Athens. Eschohotado writes that this was quite psychoactive; to obtain its effects one needs only to pass the cereal sheaves through water and then discard the cereal, because the lysergic acid amides are soluble in water while the poisonous components are 1 2

Eschohotado 1999: 2. Mylonas 1961: 231.

133 not. Considering that water was the medium utilized by the administrators of the sanctuary, we can therefore explain—without resorting to miracles or to the simple credulity of the devotees—the deep and infallible effect of the initiation.3

Indeed, the use of hallucinogenic substances in religious and quasireligious ceremonies is as world-wide as religion itself. The Rig-Veda, the sacred text of the ancient Aryans of Northern India, makes numerous references to the intoxicating plant divinity Soma. It is now widely considered that Soma is none other than the fly agaric mushroom (Amanita muscaria), a tall red mushroom with white spots whose physical properties and psychological effects closely match the descriptions of Soma in the Rig-Veda.4 In the New World, there are numerous hallucinogenic plants that are at the center of indigenous religions, including the mushroom Psilocybe mexicana, and its kin, known to the Indians of Mexico as teonanacatl, a Nahuatl word meaning “flesh of the gods.”5 An Aztec religion involving the consumption of teonanacatl is nearly 3000 years old, despite the fervent attempts of the Spanish conquistadors to stamp it out. In northern Mexico are religions where the ingestion of the peyote cactus (Lophophora williamsii) is a crucial component. Peyote ceremonies have migrated into the United States and today are part of the Native American Church. In the Upper Amazon basin the vine Banisteriopsis caapi is boiled to produce an hallucinogenic drink. The resulting “vine drunkenness,” known to Peruvians as ayahuasca, is the portal to the spirit world.6 There is no real historical tradition of Western logic and philosophical reasoning interacting with shamanic hallucinogen religions in the same way that one finds an Anselm or an Aquinas attempting to wed philosophy and the Christian religion. The result is that the language one finds in the ethnographic reports of anthropologists is not as sensitive to 3

Eschohotado 1999: 17. The seminal study on the role of the ergot mushroom and the Eleusianian Mysteries is Wasson, Hofmann, and Ruck 1978. Ergot poisoning was a common event in medieval Europe. See Sidky 1997: 167-183 and Schultes and Hofmann 1979: 102-105. 4 The locus classicus is Wasson 1968. For criticism of the Soma = fly agaric view, see Rudgley 1993, ch. 3. 5 An excellent overview can be found in Schultes 1972, the Harvard botanist who was responsible for identifying the teonanacatl of the Aztecs as the psilocybe mushroom. A more detailed study is Schultes and Hofmann 1979. 6 Harner 1973.

134 commonplace philosophical distinctions as that of virtually any contemporary Christian theologian. So it is somewhat more difficult to make explicit precisely which propositions are believed in as the result of hallucinogenic experiences. It is clear, however, that hallucinogens are used in many cultures not for mere recreational purposes, or even as a component of a non-doxastic ritual analogous to, say, Christian genuflection, but have a definite epistemic aspect. For example, the ritual use of hallucinogens generates beliefs about the nature of the mind. The anthropologist Michael Harner writes, “there can be little doubt that the use of the more powerful hallucinogens tends to strongly reinforce a belief in the reality of a supernatural world and in the existence of a disembodied soul or souls.”7 For example, the Jívaro Indians believe that the effect of ayahuasca is that part of the soul leaves the body. The Conibo-Shipibo Indians of the Ucayali region of eastern Peru report that ayahuasca can cause the souls of shamans to leave their bodies and assume the form of predatory birds. The Amahuaca, neighbors of the Conibo, also report that souls leave bodies when one drinks ayahuasca, and many of the Desana branch of the Tukano Indians in eastern Columbia hold similar beliefs.8 This is quite easy to understand as hallucinations caused by ayahuasca frequently include the feeling of flying and soaring over great distances. Beliefs about the existence and nature of the divine are also produced by hallucinogen rituals. The Tukano Indians of the territory of Vaupés in the northwest Amazon region of Columbia drink ayahuasca, or yajé. The purpose of doing so is, among other things, to see the tribal divinities. Upon waking from the yajé trance, the Tukanos are convinced of the truth of their religious system, having experienced firsthand, courtesy of the Banisteriopsis caapi vine, their gods and the seminal events of the tribe’s mythology.9 Experience of the gods is also related in the Rig-Veda: We have drunk the Soma, we are become Immortals, We are arrived at the Light, we have found the Gods. What now can hostility do to us, what malice of mortal, O Immortal Soma!10

7

Harner 1973: xiv. See Harner 1973. 9 Reichel-Dolmatoff 1972: 102 10 Cited in Wasson 1972: 210. 8

135 Here the Vedic poet relates the direct experience of the divine through the ingestion of the sacred hallucinogen. The connection is quite explicit: drink the Soma and find the Gods. In the Bwiti religion of sub-Saharan Africa, the Eboka plant is used as a sacramental hallucinogen. Takers of the drug do so for a wide variety of reasons, but one is in order to know God, and to know things of the dead and the land beyond.11 In fact, there is a quite extensive literature on the historical use of entheogenic, or god-spawning, substances and their role in the formation of religions. Some scholars argue that entheogens are the root of all shamanic ecstasy—the mysterium tremendum itself—and ultimately the ancient basis of all religion.12 These more extravagant claims aside, there can be little doubt that for many religions, the ritual use of psychoactive substances had played and continues to play an important role in forming beliefs about the nature of the divine. Finally, hallucinogens have been used to acquire beliefs about moral propositions. In De Legibus (II.14.36), Cicero writes the following concerning the Eleusinian Mysteries: For among the many excellent and indeed divine institutions which… Athens has brought forth and contributed to human life, none, in my opinion, is better than those mysteries. For by their means we have been brought out of our barbarous and savage mode of life and educated and refined to a state of civilization; and as the rites are called “initiations,” so in very truth we have learned from them the beginnings of life, and have gained the power not only to live happily, but also to die with a better hope.13

This is clearly moral language. Cicero is not praising the Mysteries because of the practical knowledge that they communicate, or the metaphysics they endorse, but because they taught the Romans a moral lesson. Cicero claims that it is the sense of community and the knowledge of how to live a good life as conveyed by the Mysteries that civilized the Romans. If Wasson, Escohotado, Schultes and others are right that the key to the Eleusinian Mysteries was the consumption of the hallucinogenic barley ergot, then the initiate Cicero is proposing that the ritualized use of psychoactive agents is a means of gaining moral knowledge.14 11

Fernandez 1972: 251. See for example LaBarre 1972 and especially Wasson et al. 1986. 13 Cicero 45 B.C.E. [1928]. 14 Compare Isocrates, who, in his Panegyricus §28, praises Demeter for “two gifts, the greatest in the world—the fruits of the earth, which have enabled us to rise above the life of the beasts, and the holy rite which inspires in those who partake of it 12

136 The experiences induced by peyote, fly agaric, barley ergot, and the rest of the garden did not on their own give rise to a coherent system of thought. It took the teachings of the hierophants, the shamans, and the tribal elders to make sense of the hallucinogenic visions and give them a context in which they could be understood. It is one thing to have a sense of flying, seeing bird-headed people, and having conversations with dragons who declare themselves to be gods. It is another to have an opinion about what that all meant, and how it is relevant to everyday life. In the so-called Miracle of Marsh Chapel of 1962, 20 Christian theology students participated in a placebo-controlled double-blind psilocybin experiment at Boston University. Walter Pahnke, a Harvard graduate student experimenter, administered 30 mg. of psilocybin to ten students. A control group of ten students received an active placebo of 200 mg. of nicotinic acid, a vasodilator which causes no psychic effects, only warmth and tingling of the skin. Nine out of the ten who had received psilocybin reported having what they considered to be an authentic religious experience. In a chapel, on Good Friday, with theology students as test subjects, psilocybin led to an experience of the Christian God. In 1991, a researcher who did a long-term follow-up of Pahnke’s original test subjects reported “all psilocybin subjects participating in the long-term follow-up, but none of the controls, still considered their original experience to have had genuinely mystical elements and to have made a uniquely valuable contribution to their spiritual lives”.15 Many of these subjects are now Christian ministers and have, like shamans, integrated their psilocybin experience into their religious belief systems. Their hallucinations were not rejected as perceptual outliers but taken as authentic perceptions of the divine. The communities discussed above did not dismiss the experiences caused by psychoactive drugs as recreational cinema, but incorporated them into the ongoing life of the society. That is, the beliefs produced by the hallucinogens didn’t stand on their own, but had to be fitted into a coherent body of beliefs. As in the case of philosophical intuition, reason must be employed to take the foundational propositions about morality, the mind, and the divine, and systematize them into at least a minimally coherent way of understanding those topics.

sweeter hopes regarding both the end of life and all eternity.” Isocrates 380 B.C.E. [1928]. 15 Doblin 1991.

137 Harner argues that the beliefs generated by the use of hallucinogens are “sooner or later bound to come into conflict with the orthodox dogma fundamental to the ideological structure of state religions”.16 He takes this fact to explain why the use of hallucinogens tends to be eventually forbidden in Western societies. Even under the paternalistic obsession with safety currently prevalent in America, it is very difficult to make out an argument for legally prohibiting the personal use of psychotropic plants on the basis of harm to self. So Harner may well have a point. In any case, it does seem that many hallucinogen-produced beliefs are inconsistent with the beliefs of a majority of philosophers who rely on rational intuition. This inconsistency is most clearly seen in the philosophy of mind. An overwhelming number of individuals believe that they possess (at least one) spiritual, non-physical soul as the result of ingesting psychotropic drugs. Yet practically no contemporary philosopher accepts substance dualism. Furthermore, beliefs in the tribal gods are also a common result of the ritual use of hallucinogens. It is a minority of philosophers who rely solely on rational intuition to gain beliefs about philosophical propositions who believe that there is a supernatural realm of spirits and gods. In the case of ethics, it is much more difficult to codify the moral propositions believed by preliterate societies. Ergo it is that much harder to show how their hallucinogen-based moral beliefs are directly in conflict with what most philosophers think follows from rational intuition. However, I will suggest this: it would be astonishing if further ethnography showed the moral beliefs of a Huichol shaman to be isomorphic with those of most rationalist philosophers. In sum, there are four important ways in which the rational intuition and the ritual use of hallucinogens are analogous: (1) both are noninferential belief-acquiring mechanisms that supposedly yield basic beliefs. (2) in both cases, these basic beliefs concern philosophical propositions about the divine, morality and the nature of the mind. (3) in both cases, the basic beliefs generated are not the final word, but must be rationally evaluated for consistency, explanatory cohesiveness, and the like in order to achieve reflective equilibrium.

16

Harner 1973: 7.

138 (4) both approaches are trusted and believed to be legitimate by serious people who have devoted their lives to acquiring knowledge by these methods. However, here is the problem: rational intuition and the ritual use of hallucinogens produce different output beliefs. Rational intuition yields one set of beliefs about the divine, morality, and the nature of the mind, and hallucinogen use yields a different, incompatible, set. So which beliefacquiring methods should we adopt? And on what grounds? Here is an attempt to vindicate rational intuition. Let us call this The Priority of Reason argument. Everyone agrees that we need to use reason to develop epistemically virtuous belief-sets. Reason is used to uncover and resolve inconsistencies, eradicate fallacious inferences, promote explanatory unity, etc. in other words, rational reflection is the basic method by which we come to have justified beliefs about philosophical propositions. Other methods, like ingesting fly agaric or peyote, are add-ons. Rational intuition is the basic model and the others are optional features. So if push comes to shove, and there are inconsistencies between the basic beliefs provided by intuition and those provided by revelation or peyote, we’d better stick with the basic package. Since everyone agrees that reason is essential and not all agree that we need these “bonus” methods, in the case of trouble, the bonus methods are the first to go. Rational intuition remains the best way of acquiring beliefs about philosophical propositions. I have argued that both traditional shamans and traditional philosophers rely upon reason to systematize their beliefs and attain reflective equilibrium. However, there are three problems with the Priority of Reason argument. First, there is no reason to suppose that both groups recognize the same set of properties as being epistemically virtuous. There may be a wide discrepancy as to what qualities they want their belief-sets to exemplify. The fact that both Greek heirophants and Laurence Bonjour count certain beliefs as good inferences from logically prior beliefs does not imply that they are using the same concept of “good inference.” Even among philosophers, some regard justification to be a normative concept and some think it is a causal-nomological one. Some philosophers defend an internalist notion of justification and some an externalist one. We should be cautious not to overstate agreement about exactly how reason is to be used in developing inferential beliefs.

139 Second, there is a difference between the use of reason to evaluate basic beliefs and draw inferences (inductively or deductively) from these basic beliefs, and the use of rational intuition to generate the basic beliefs themselves. That is, once one has a set of basic beliefs, arrived at by whatever method, reason is then employed to operate upon this set and produce inferential beliefs. This procedure is a different activity from the use of a priori philosophical intuition to arrive at basic beliefs. The use of reason as an inferential procedure is logically separate from how one acquires noninferential beliefs. An independent argument is required to demonstrate that intuition is the best way to get noninferential beliefs; the value of reason in making good inferences from the foundations tells us nothing about how to get the right foundations. There is an acronym in computer programming: GIGO— garbage in, garbage out. A computer is an extensional logic machine, but its output is only as good as the data it is fed. The present concern is not over how to process the data, but how to get it in the first place. The Priority of Reason argument is that since reason is a good inference procedure, therefore rational intuition is a good way of gaining noninferential beliefs. This is like arguing that since a computer’s inferences are all logical deductions, therefore its input must be logical truths— a manifest non sequitur. The final difficulty with the Priority of Reason defense is brought out clearly by analogy. In 1610 Galileo published an account of his recent astronomical discoveries in Sidereus Nuncius, revealing among other things that the moon is mountainous, craggy and in general not perfectly spherical. Galileo faced some legitimate criticisms, for example that his telescopes were still fairly crude, imprecise, and not entirely to be trusted. Purely rationalist arguments were offered as well; the Aristotelians of the day opposed Galileo on the grounds that empirical science relies upon the use of reason to adjudicate competing hypotheses, evaluate data, draw correct inductive inferences, and so on. Thus reason, rational reflection, a priori intuition is the primary method by which we gain empirical knowledge. Should there be any discrepancies between what pure reason tells us about the nature of the physical world and the observations of the experimentalists, we must discard experimentation and rely solely on reason, our fundamental method. Since reason dictates that the universe conform to the a priori beauty of the spheres, argued the pedants, we are entirely justified in discarding Signore Galileo’s telescopic findings to the

140 contrary. In fact, Giulio Libri, one of the foremost philosophers at Pisa, refused even to look through Galileo’s telescope.17 Obviously, this mulishness is precisely the sort of scholastic hubris that fomented resistance to experimental science in the Renaissance. It is clear in the case of empirical knowledge that reason alone is inadequate. We must also use the method of sense perception to acquire basic beliefs. Simply because reason (in the form of science) operates upon these basic beliefs to produce intricate theories is no evidence that sense perception is not also needed as a method of belief-acquisition. Just as reason operates upon the beliefs produced by sense perception to develop a systematic scientific understanding of the world, so too reason operates upon the beliefs produced by psychoactive drugs to develop a coherent view about various philosophical propositions. In short, the Priority of Reason argument is precisely the same argument that Aristotelian dogmatists used to dismiss Galileo’s experimental method of acquiring empirical beliefs. It is employed in the present context to dismiss the ritual use of hallucinogens as methods of acquiring beliefs about philosophical propositions. If the Priority of Reason argument wasn’t a good argument then, it isn’t a good one now. At this point we have considered an argument designed to show the relative superiority of intuition: the Priority of Reason argument. It turns out to be inadequate, lending no more credibility to intuition than to peyote. Thus we have a bit of a quandary; the methods of intuition and hallucinogen use have inconsistent results, and we can’t seem to show that one is better than the other. So what do we do? At this point we might well be tempted by nihilism. As naturalists will be eager to argue, the preceding reflections constitute a reductio ad absurdum on the very idea of non-naturalistic philosophical knowledge. Our mistake all along was to think that there are any purely philosophical truths to be had. Once we give that up, along with the suspicious beliefacquiring methods we have been considering, we will be able to see that the only truths to be had are either contingent or nomologically necessary. In other words, all truths are fundamentally scientific ones, and Michael Devitt is right when he claims, “[we should] reject a priori knowledge and embrace ‘naturalism,’ the view that there is only one way of knowing, the empirical way that is the basis of science.”18 Traditional intuition-driven philosophy is empty foolishness. 17 18

Langford 1966 : 41. Devitt 1999 : 96.

141 If we resist the nihilistic-cum-naturalistic response, perhaps we must reluctantly accept skepticism. Skeptical arguments are generally based on the notion that S doesn’t know P because S’s true belief that P is improperly dependent on good luck. Sara doesn’t know that it is 3:30 because it is no more than luck that she glanced at the stopped clock at exactly 3:30. Smith doesn’t know that either Jones owns a Ford or Brown is in Barcelona because he is just fortunate that logical addition yielded a true disjunction from the justified yet false premise that Jones owns a Ford. Henry doesn’t know that he is seeing a barn, even though it is a barn and his true belief was caused by the state of affairs that it is a barn because it is simply luck that he is not fooled by one of the many papier-mâché barns in that area. More global skepticism proceeds from the contention that we can’t perceptually discriminate among veridical sense perception and dreams, evil demons, or the neural inputs of alien scientists. Since we can’t tell the difference, if we have true beliefs based on sense perception then it is the sheerest fortune that we are not dreaming, deceived, etc. Thus our true empirical beliefs never rise up to knowledge. The same argumentative strategy can be deployed against knowledge of putative philosophical truths. We have no defensible reason to prefer one basic method of acquiring beliefs about philosophical propositions over another basic method that gives different results. Any true beliefs we have about philosophical propositions are accidental—it is just good fortune if we pick the right method. If my intuition-based belief that “necessarily, it is wrong to use persons merely as means and never as endsin-themselves” is true, then it is no more than good luck that I chose intuition as my method instead of one of the competitors. Hence I don’t know the categorical imperative or mutatis mutandis any other intuitionbased philosophical proposition to be true. In short, our failure to vindicate intuition means that we don’t know any philosophical truths. There may be truths about morality, the divine, metaphysics, and the rest, but we’ll never know them. Intuition-driven philosophy is a waste of time; as far as we know, we’re just as well off chewing peyote buttons. These are frightening alternatives. Fortunately there is one more choice: relativism. When it comes to philosophical propositions, there is no way to rationally decide among basic belief-acquiring methods. Dogmatic faith in intuition is no better than faith that peyote will reveal the right way to live, or faith in...well... faith. If we select rational intuition as our method, certain propositions will come out necessarily true. If we pick ritualistic peyote consumption, a different set of propositions will be

142 deemed true. According to the relativist, what propositions are true is therefore dependent on, and relative to, method. There is more than one true philosophical story to be told about morality, epistemology, metaphysics, and the divine. Relativism has a bad reputation among most analytic philosophers, and development of an adequate theory of relativism and an account of how it will elude the traditional objections made against relativistic theories are beyond the scope of this paper.19 Nevertheless, philosophers had better work hard to overcome these objections, since unless relativism is true, traditional philosophy is either worthless or pointless. It is difficult to imagine how a more powerful argument for relativism about philosophical truths could be constructed. Another way to view the results of this paper is this: philosophers are well advised to be either naturalists or relativists. If you resist naturalism then relativism beats the remaining alternatives.20

19

20

Although see Hales 1997 for a solution to the problem of self-referential inconsistency. Earlier versions of this paper received helpful criticisms from audiences at Bloomsburg University, The College of William and Mary, William Patterson University, and the Conference on Basic Belief at the Vrije Universiteit in Amsterdam.

143 REFERENCES Cicero, Marcus Tullius. 1928 [45 B.C.E.]. De Legibus. Translated by C. W. Keyes. Edited by T. E. Page, E. Capps, W. H. D. Rouse, L. A. Post and E. H. Warmington, The Loeb Classical Library. Cambridge, MA: Harvard University Press. Devitt, Michael. 1999. “A Naturalistic Defense of Realism” in Metaphysics: Contemporary Readings, edited by S. D. Hales. Belmont, CA: Wadsworth Publishing Co. Doblin, Rick. 1991. “Pahnke’s Good Friday Experiment: A Long-Term Follow-Up and Methodological Critique”, The Journal of Transpersonal Psychology 23 (1). Eschohotado, Antonio. 1999. A Brief History of Drugs: From the Stone Age to the Stoned Age. Translated by K. A. Symington. Rochester, Vermont: Park Street Press. Fernandez, James W. 1972. “Tabernanthe Iboga: Narcotic Ecstasis and the Work of the Ancestors” in The Flesh of the Gods: The Ritual Use of Hallucinogens, edited by P. T. Furst. New York: Praeger Publishers. Gordon, Stella Kramrisch, Jonathan Ott, and Carl A. P. Ruck. 1986. Persephone's Quest: Entheogens and the Origins of Religion. New Haven: Yale University Press. Hales, Steven D. 1997. “A Consistent Relativism”, Mind 106: 33-52. Harner, Michael J. 1973. “Common Themes in South American Indian Yagé Experiences” in Hallucinogens and Shamanism, edited by M. J. Harner. Oxford: Oxford University Press. ———, ed. 1973. Hallucinogens and Shamanism. Oxford: Oxford University Press. Isocrates. 1928 [380 B.C.E.] Panegyricus. Translated by G. Norlin. Edited by T. E. Page, E. Capps, W. H. D. Rouse, L. A. Post and E. H. Warmington, The Loeb Classical Library. Cambridge, MA: Harvard University Press. LaBarre, Weston. 1972. “Hallucinogens and the Shamanic Origins of Religion” in Flesh of the Gods: The Ritual Use of Hallucinogens, edited by P. T. Furst. New York: Praeger Publishers. Langford, Jerome J. 1966. Galileo, Science, and the Church. Ann Arbor: University of Michigan Press.

144 Mylonas, George E. 1961. Eleusis and the Eleusinian Mysteries. Princeton: Princeton University Press. Reichel-Dolmatoff, Gerardo 1972. “The Cultural Context of an Aboriginal Hallucinogen: Banisteriopsis Caapi” in Flesh of the Gods: The Ritual Use of Hallucinogens, edited by P. T. Furst. New York: Praeger Publishers. Rudgley, Richard. 1993. Essential Substances: A Cultural History of Intoxicants in Society. New York: Kodansha International. Schultes, Richard Evans. 1972. “An Overview of Hallucinogens in the Western Hemisphere” in Flesh of the Gods: The Ritual Use of Hallucinogens, edited by P. T. Furst. New York: Praeger Publishers. Schultes, Richard Evans, and Albert Hofmann. 1979. Plants of the Gods: Origins of Hallucinogenic Use. New York: McGraw-Hill Book Company. Sidky, H. 1997. Witchcraft, Lycanthropy, Drugs, and Disease: An Anthropological Study of European Witch-Hunts. New York: Peter Lang. Wasson, R. Gordon 1968. Soma: Divine Mushroom of Immortality. New York: Harcourt, Brace & World, Inc. ———.1972. “What Was the Soma of the Aryans?” in Flesh of the Gods: The Ritual Use of Hallucinogens, edited by P. T. Furst. New York: Praeger Publishers. ———. Gordon, Albert Hofmann, and Carl A. P. Ruck. 1978. The Road to Eleusis: Unveiling the Secret of the Mysteries. New York: Harcourt Brace Jovanovich, Inc.

B: Religious Belief

CHRISTIAN B. MILLER

Defeaters and the Basicality of Theistic Belief Recently much work has been done in developing the reformed approach to religious epistemology in new and interesting ways.1 This paper hopes to continue that trend by investigating the relationship between basic belief in the existence of God and the impact of what purport to be various epistemic defeaters for that belief. Ultimately, I hope to show that while theistic beliefs might arise and even be warranted in the manner described by reformed epistemologists, they quickly will become non-basic in the presence of putative defeaters.2 Such a result may serve not only to elucidate the noetic structures of religious believers, but also to help find common ground between the warring parties of evidentialism and reformed epistemology. My plan is as follows. First a brief description will be given of one of the most prominent positions in reformed epistemology. With the necessary background in place, we can then begin to examine the role that undefeated-defeaters and defeater-defeaters will play in such a view. Along the way, it also will be important to provide a preliminary sketch of the workings of a defeater system and of those conditions necessary for a belief’s being epistemically basic. 1. Plantinga’s Approach Given limitations of space, we will be able to examine the implications of defeaters for only one view of religious belief, namely that of Alvin Plantinga. In a series of papers in the early 1980s, Plantinga outlined many of 1

See in particular McLeod 1993, Zagzebski 1993, and Sudduth 1999a, 1999b, 1999c. Discussion of defeaters for properly basic theistic beliefs is hardly new in the literature. In particular, Alvin Plantinga and Philip Quinn have been engaged in an important debate over how theists should respond to various intellectual challenges to their religious beliefs. See Quinn 1985, 1993 and Plantinga 1986. For an interesting evaluation of this exchange, see Hasker 1998.

2

148 the central elements of his reformed approach to theistic belief.3 In the intervening years, he has continued to develop the view, culminating in the recent publication of Warranted Christian Belief.4 It is that work in particular which will serve as the focus of our discussion below. Plantinga presents both a descriptive and a normative model for theistic belief, and it will be important for our purposes to keep these two facets of his project separate. Taking the descriptive model first, suppose that human beings have as part of their cognitive endowment what Calvin called a sensus divinitatis (hereafter SD).5 Such a mental faculty takes certain experiential stimuli as inputs, and outputs beliefs about God. More rigorously, (1) S forms theistic beliefs about God as a result of the workings of S’s sensus divinitatis SD at time t if: (a) S’s cognitive faculties at t include a SD. (b) S at t is in circumstances C*, where C* is a member of the set of SD-eliciting circumstances C. (c) S’s SD at t is functioning properly.6 What are these SD-eliciting circumstances? It would be hard to give a precise characterization, but according to Plantinga they might include observing great beauty in nature, feeling tremendous guilt over a past wrongdoing, experiencing grave danger, and so on.7 So the inputs to the sensus divinitatis typically include certain powerful experiential stimuli. What about the outputs? According to Plantinga, the SD produces beliefs about God which are basic for the individual in question, where: (2) S’s belief B at time t is basic iff:

3

See for example Plantinga 1979, 1981, 1982, 1983. See Plantinga 2000, as well as the first two volumes of his warrant trilogy 1993a, 1993b. 5 For Plantinga’s use of Calvin, see his 2000: 170-174. 6 Compare Plantinga: “The sensus divinitatis is a disposition or set of dispositions to form theistic beliefs in various circumstances, in response to the sorts of conditions or stimuli that trigger the working of this sense of divinity” (2000: 173). 7 Plantinga 2000: 174. 4

149 (a) (b)

S at t has B as part of S’s noetic structure.8 B at t has property P in common with all the other basic beliefs in S’s noetic structure.

(2) can be left intentionally circular for the time being until we need to specify property P in section three. For now, we tentatively can think of P as the property of not being accepted on the basis of any other beliefs in S’s noetic structure. Thus on Plantinga’s model, while theistic beliefs which result from the operation of the sensus divinitatis might be occasioned by certain events in S’s experience, they are not arrived at as a result of arguments from those experiences to the conclusion that God exists.9 In this regard, reformed epistemologists think that there is a clear analogy with belief formation in the case of perception, remembrance, and a priori intuition.10 The details of exactly how this analogy is meant to work are complex,11 but something like the following thesis seems to be at work: (3) Descriptive Parity Thesis: The beliefs which result from the operations of S’s sensus divinitatis can play roughly the same foundational role in S’s noetic structure as do those which result from S’s cognitive faculties responsible for perception, memory, and a priori belief.12 For example, just as the perceptual belief that there is a computer in front of me spontaneously arises in my noetic structure as I type this paper, so too may a basic belief in God be formed in a person when he or she is in SD-eliciting circumstances. Having provided at least a preliminary sketch of the workings of the sensus divinitatis, the natural question to ask next is whether human beings actually possess such a cognitive faculty. The answer, perhaps not surprisingly, depends on the de facto truth of theism. If God exists and has created us in his image, then according to Plantinga it is quite probable that he has

8

A noetic structure is, “the set of propositions [S] believes, together with certain epistemic relations that hold among [S] and them” (Plantinga 1993a: 72). 9 Plantinga 2000: 175. 10 Plantinga 2000: 175-176. 11 As the book length treatment of parity principles in McLeod 1993 suggests. 12 Plantinga 1983: 82ff, 2000: 343-4 and Sudduth 1999a: 307.

150 endowed us with such a faculty aimed at producing true beliefs about himself.13 Of course, whether anyone is epistemically justified or warranted in forming beliefs in this way is another question altogether, and here we turn from the descriptive to the normative model. Should epistemic agents hold basic beliefs in the existence of God as a result of the proper functioning of the SD? In other words, are such beliefs properly basic, where: (4) S’s belief B at time t is properly basic iff: (a) B at t is basic. (b) B at t satisfies certain normative epistemic criteria.14 The criteria in (b) depend on one’s preferred epistemological theory. Plantinga himself considers four primary candidates – deontological justification, internal rationality, external rationality, and warrant – and argues in some detail that it is possible for theistic beliefs formed according to his descriptive model to satisfy the normative standards articulated by each of these views.15 Hence if God exists, it is possible on Plantinga’s view for agents to arrive at properly basic theistic beliefs as a result of the proper functioning of their sensus divinitatis.16 As far as this paper is concerned, we need not bother with the details of these various epistemic theories, nor with the question of whether Plantinga is in fact right in thinking that his model does allow for properly basic theistic beliefs. Since I am primarily interested in the contribution made by various putative defeaters which arise only after S has formed a basic belief in God and is no longer in SD-eliciting circumstances, we can grant Plantinga both his descriptive and normative claims for our purposes here.

13

Plantinga 2000: 186-190. (4) resembles Plantinga’s construal of proper basicality in his 1993a: 70. 15 Plantinga 2000: chapter six. 16 Part of Plantinga’s strategy always has been to rely on the normative equivalent of the parity thesis: Normative Parity Thesis: If S’s sensus divinitatis is functioning properly and produces a basic theistic belief B, then B will enjoy at least the same degree of justification or warrant as that had by beliefs which result from the proper functioning of S’s faculties of perception, memory, and a priori belief. See for example Plantinga 1983: 80 and McLeod 1993: chapters six and nine. 14

151

2. Undefeated-Defeaters Most ‘intellectually sophisticated adult theists’ in our society will face certain serious epistemic challenges to their religious beliefs.17 Whether they involve the existence and magnitude of evil, the plurality of world religions, the results of historical-critical scripture scholarship, or the findings of projection theories in psychology, these challenges take the form of potential defeaters for theistic belief. In general, a defeater is a belief D18 such that if S has belief B in S’s noetic structure and S comes to acquire D, then other things being equal S should no longer continue to hold B as strongly (if at all) given D.19 What bearing do defeaters (so defined) have on the epistemic status of properly basic theistic beliefs? The answer is likely to be both long and complicated, and so we shall consider only a restricted version of this question as it pertains to the bearing of what from the first person perspective are taken to be defeaters for the beliefs in question. First we will consider the descriptive case of how religious believers in fact typically react to what purports to be a defeater, and then we shall move on to the normative question of how they ought to have reacted.

17

The phrase is taken from Quinn 1985: 481. For the sake of simplifying what follows, defeaters will be restricted to beliefs and hence will not also include experiences. In general, however, I have no reservations about this richer understanding. For relevant discussion, see Bergmann 1997a: 87-90. 19 For helpful discussion of defeaters, see Pollock 1986: 37, Bergmann 1997a, and Sudduth 1999a: 299. According to Sudduth, “S acquires a defeater D against some belief B just if S acquires reasons R (of which S is conscious) that are such that, given the relevant rest of S’s noetic structure, R is an appropriate ground for S’s revising his noetic structure in a particular way” (1999c: 218). And according to Plantinga: (*) D is a defeater of B for S at t if and only if: (a) S’s noetic structure N at t includes B. (b) S comes to believe D at t. (c) Any person: (i) Whose cognitive faculties are functioning properly in the relevant respects. (ii) Whose noetic structure is N and includes B. (iii) Who comes at t to believe D but nothing else independent of or stronger than D. would withhold B (or believe it less strongly). See Plantinga 2000: 362. He also develops the notion of a purely epistemic defeater to handle cognitive faculties aimed at non-alethic ends, but such refinements need not concern us here. 18

152 But before we begin, it is important to make another distinction. Since defeaters come in many shapes and sizes, we will not get very far simply by examining their general epistemic properties. Instead let us distinguish between defeater-defeaters and undefeated-defeaters, where: (5) DD is a defeater-defeater for S’s defeater D and belief B at time t iff: (a) S’s noetic structure at t includes B, D, and DD. (b) D is a defeater for B prior to t. (c) At t and as a result of possessing DD, S should longer regard D as a defeater of B, or at least not to the same extent. (6) UD is a undefeated-defeater for S’s belief B at time t iff: (a) S’s noetic structure at t includes B and UD. (b) UD is a defeater for B at t. (c) The other beliefs in S’s noetic structure at t are not defeaterdefeaters for UD. Less formally, an undefeated-defeater is simply a defeater for a belief which is such that it is not at present defeated by any countervailing beliefs. Yet if an agent does come to acquire a relevant defeater-defeater, then her initial belief typically will enjoy an increase in epistemic status from the status it had solely in the presence of the previously undefeateddefeater. In the remainder of this section we shall focus on undefeateddefeaters, saving the more complex case of defeater-defeaters for section three. With these distinctions in mind, our first question is this – as a matter of descriptive fact, what results when a person possesses a basic theistic belief acquired via the proper functioning of the sensus divinitatis, and later comes to acquire what he or she takes to be an undefeated-defeater for that belief? In many cases, the outcome seems clear enough – the agent in question will form a new, non-basic belief. Perhaps an example will help. Suppose that Jim came to believe in God along the lines of Plantinga’s descriptive model. Yet one day he discovers Mackie’s formulation of the problem of evil, and is deeply impressed by the force of the argument.20 Even after much thought and reflection, Jim still finds the argument convincing and feels compelled to accept its conclusion. So he comes to believe: 20

Mackie 1955.

153 (7) God does not exist. But (7) is not basic in Jim’s noetic structure; it clearly is based on his prior belief that: (8) There is a sound argument from evil for the claim that (7). Furthermore, the following counterfactual seems to be true of Jim: (9) If Jim had not discovered Mackie’s argument, then he probably would have continued to possess a basic belief in the existence of God as part of his noetic structure. So Jim’s construal of the problem of evil as an undefeated-defeater has undermined whatever positive epistemic status belief in God originally had for him, and has contributed directly to the formation of a new non-basic belief. As a descriptive account, our example can be generalized to cover many cases of putative undefeated-defeat. Thus there seems to be good reason to accept the following: Thesis 1: As a matter of fact, when S takes it to be the case that S has an undefeated-defeater D for S’s basic theistic belief B which requires revision in S’s attitude towards B, then if S forms a new belief B* in response to D, B* typically will be non-basic in virtue of being at least partially based on D.21 But ought S do so? In other words, when S carries out such revisions upon acquiring what S takes to be an undefeated-defeater, is S satisfying the appropriate epistemic norms?

21

If the Descriptive Parity Thesis is correct, then the analog of Thesis 1 should hold true for other kinds of beliefs as well. While it would take us too far afield to investigate the matter to the degree that it deserves, it does seem as if basic perceptual beliefs, for example, are typically rendered non-basic by an putative undefeated-defeater. Standard perceptual defeaters such as the acquisition of evidence regarding the influence of mirages or hallucinogenic drugs can illustrate this point nicely.

154 The answer may depend on the details of the particular epistemic theory under consideration. But rather than engaging in the arduous task of surveying a wide variety of views in contemporary epistemology, we hopefully can reach a substantive conclusion by a more direct route. To begin with, notice that whatever normative standards there are for the defeater activity in question, they will have to arise from an internalist theory of justification or rationality. Since we are only concerned with an agent’s first person understanding of the makeup of her noetic structure, we are only concerned with the special epistemic access that is typical of internalism. So the epistemic norms in question are such that they require an agent to adjust her relevant beliefs accordingly in light of what reflection tells her about whether certain types of defeaters are present in her noetic structure.22 Externalists themselves even recognize such norms, and thus typically concede that at least one of the conditions individually necessary and jointly sufficient for warrant is an internalist no-defeater condition.23 In light of the above, then, it seems obvious enough that the epistemic standards for responding to putative defeaters would be met if an agent like Jim revises his theistic belief in the light of what purports to be an undefeated-defeater such as the problem of evil. Even if this alleged defeater is arrived at in an irrational manner or as a result of cognitive malfunctioning, it is still a defeater by the agent’s own lights, and thus such that unless he has a defeater-defeater, he is epistemically required given his noetic situation to either abandon his basic belief or hold it less firmly.24 We can arrive at this conclusion in another way given that reformed epistemologists also typically endorse the normative version of the parity thesis.25 Whatever response to a putative defeater is required in the case of religious belief will be structurally analogous to what should be done in the

22

See Pollock 1984 and Plantinga 2000: chapter 11, esp. page 361. The interpretation of internal access and internalism follows Kim 1993 and especially Bergmann 1997b. Such a broad characterization allows room both for what Plantinga calls ‘internal rationality’ and for the more traditional notion of deontological epistemic justification. See Plantinga 2000: 365. 23 See Nozick 1981: 196, Goldman 1986: 62-3, 111-112, Plantinga 1993a: 40-42, Bergmann 1997b, and Sudduth 1999b: 171, 1999c. 24 Relevant here is Sudduth 1999a: 303-4 and Plantinga 2000: 364-365. 25 For the thesis, see footnote 16. Michael Sudduth uses a similar strategy in his 1999a: 306-309 and 1999c.

155 case of supposed defeaters for perception, memory, and a priori beliefs.26 And certainly for those beliefs, we are epistemically required to revise our initial basic beliefs upon taking there to be strong countervailing and undefeated evidence concerning the influence of hallucinogenic drugs, mirages, cosmic rays, and the like. Not only should agents revise their basic theistic beliefs in the presence of what purport to be undefeated-defeaters, but they also should revise them in the appropriate way. What is appropriate will depend on the kind of undefeated-defeater at issue, and so we need some further distinctions: (10) PD is a partial-defeater for S’s belief B at time t iff: (a) S’s noetic structure at t includes B and PD. (b) PD is a defeater for B at t such that S ought to hold B with less firmness. (d) None of the other beliefs in S’s noetic structure at t are defeater-defeaters for PD. (11) CD is a complete-defeater for S’s belief B at time t iff: (a), (c) Same as in (10) for CD rather than PD. (b) CD is a defeater for B at t such that S ought to cease holding B. (12) RD is a reversing-defeater for S’s belief B at time t iff: (a), (c) Same as in (10) for RD rather than PD. (b) RD is a defeater for B at t such that S ought to cease holding B and instead believe the negation of the proposition affirmed by B. In the theistic case, some defeaters might be such that an agent is required merely to believe in the existence of God with less of a degree of commitment, others such that she should suspend belief in both the existence and

26

In his 1993b, Plantinga seems to agree. After discussing a case of perceptual defeat, he goes on to note that, “[b]y parity of reason, the same goes, I should think, for the believer in God of a couple of paragraphs back. She too has an undercutting defeater for belief in God; if that defeater remains itself undefeated and if she has no other source of evidence, then the rational course would be to reject belief in God” (231).

156 non-existence of God, and still others such that she ought to believe that God does not exist. With these distinctions in place, we can now state our normative thesis about the relationship between basic theistic belief and putative undefeated-defeaters: Thesis 2: For any theistic belief B arrived at by S in the basic way as a result of the proper functioning of the sensus divinitatis in circumstances C, if S is not in C and S takes there to be an undefeated-defeater D for B, then S ought to either: (a) Hold B less firmly (and as a non-basic belief partially based on D). (b) Cease holding B and form a new (non-basic) belief that S ought to suspend belief in both the existence and non-existence of God.27 (c) Cease holding B and form a new (non-basic) belief that God does not exist.28 Having considered both the descriptive and normative impact of what purport to be undefeated-defeaters, we can now move on to the more complex case of defeater-defeat. 3. Defeater-Defeaters Recall that defeater-defeaters are beliefs which are such that they can either partially or totally undermine the epistemic role played by a prior defeater in S’s noetic structure. In order to evaluate the bearing of putative defeaterdefeaters on basic theistic beliefs, we again shall proceed by separating out the relevant descriptive and normative considerations. To help fix intuitions about the impact of defeater-defeaters, let us return to the case of Jim and the problem of evil. For a time, Jim possessed a non-basic belief in his noetic structure as a result of sustained reflection on Mackie’s argument. However, one day he comes across a presentation of the free-will defense, and realizes after much consideration of the defense that Mackie’s argument can be met. Suppose he again comes to believe that: 27 28

This understanding of agnosticism follows Benn 1999: 173. Sudduth seems to defend a similar thesis in his 1999a: 300-1 and 1999b: 172.

157 (13) God does exist. But unlike the case when Jim was in SD-eliciting circumstances and initially came to believe (13), his present affirmation is no longer basic; instead, it seems to rest in part on the following: (14) Mackie’s problem of evil is not a sound argument for the nonexistence of God. Without the positive contribution of (14), Jim may not feel that he is epistemically entitled to reaffirm (13). In other words, the following counterfactual very well could accurately describe Jim’s epistemic situation: (15) If Jim had not discovered a response to Mackie’s argument, then he probably would have continued to hold a non-basic belief in the nonexistence of God. So it seems that Jim’s believing that (14) is true can make both a counterfactual and a justificatory contribution to Jim’s believing (13). Our Jim example can be generalized once again, this time to cover many cases where a purported defeater-defeater is taken by the agent in question to help support an initial theistic belief: Thesis 3: As a matter of fact, if S initially possessed what purported to be an undefeated-defeater D for S’s basic theistic belief B which S utilized in the formation of a new non-basic belief B*, then upon obtaining what S takes to be a defeater-defeater DD for D, if S forms a new belief B** adjusted to the level of defeat taken to be provided by DD, B** typically will be non-basic.29 So even if a theist succeeds in neutralizing the epistemic threat posed from the first person perspective by her initial undefeated-defeater, something very important will have changed in the structural arrangement of her noetic structure.

29

Given the Descriptive Parity Thesis, we should expect a similar thesis to hold in the case of perception, memory, and a priori belief as well.

158 At this point, it is tempting to simply move on and consider the normative requirements which pertain to defeater-defeaters. However, it is worth noting that the only real argument that has been offered for the claim that defeater-defeaters contribute to the generation of non-basic beliefs, has relied on the use of specific examples and our pre-philosophical experience of epistemic defeat. This approach may be sufficient in order to satisfactorily motivate Thesis 3, and it does nicely avoid the formidable task of having to deal with all the extant formulations of basicality in the literature. Nevertheless, more should be said about exactly how such defeaterdefeaters can contribute to the generation of non-basic beliefs. Not surprisingly, this proves to be a difficult task given that there is very little consensus about what conditions are individually necessary and jointly sufficient for a belief’s being basic. In the first section of the paper, we said that: (2) S’s belief B at time t is basic iff: (a) S at t has B as part of S’s noetic structure. (b) B at t has property P in common with all the other basic beliefs in S’s noetic structure. It would take another paper altogether in order to properly survey the various construals of P that have been offered. Instead we shall compromise by briefly testing the impact of defeater-defeaters given some of the most influential understandings of the basing relation.30 The first three views can be passed over rather quickly: (16) Argument View. S’s belief B at time t is basic only if B at t is such that it is not arrived at by way of a prior argument which S is disposed to appeal to in either explaining or justifying why S believes B at t.31 (17) Causal View. S’s belief B at time t is basic only if B at t is such that it is not causally sustained by some other belief(s) in S’s noetic structure that S has associated with B.32

30

What follows can be applied straightforwardly to the case of undefeated-defeaters as well. 31 See Audi 1986: 31-32, 1998, Korcz 1997: 175, and Greco 1998: 36-37. 32 See Armstrong 1973: 85, Pappas 1979a: 56-57, Audi 1986: 41, Pollock 1986: 37, Moser 1989: 157, Fumerton 1995: 92, 103, and Koehl 1998: 89.

159 (18) Counterfactual View. S’s belief B at time t is basic only if B at t is such that there are no other beliefs in S’s noetic structure such that if S did not possess them at t, then other things being equal, S (probably) would not believe B at t.33 On the surface at least, little needs to be said to show how most purported theistic defeater-defeaters would render S’s belief in the existence of God non-basic on all three of these views. And the same holds for the following: (19) Explanatory View. S’s belief B at time t is basic only if B at t is such that there are no other reasons in S’s noetic structure which S has associated with B at t and which (partially) explain why S has come to believe B at t.34 In our example, Jim believes once again in the existence of God in part because he has found important reasons which enable him to do so. Furthermore, these reasons play a fundamental role in the explanation of his doxastic activity such that if they had not been present, external observers could not adequately account for the change in Jim’s belief about the existence of God. Perhaps the most popular construal of the relevant necessary condition for basicality focuses on the role of justification: (20) Justification View. S’s belief B at time t is basic only if B at t is such that it is self-justifying, i.e., B does not derive its justification from any other justified beliefs in S’s noetic structure at t.35

33

See Pappas 1979a: 56-7 and Moser 1989: 157. Of course (16) through (18) can be combined in various ways – Swain, for example, defends a combined causal and counterfactual view in his 1979 and 1981. 34 See Audi 1986: 33, 45, 53, 63. 35 See Pollock 1974: 25-29, 1986: 26-7, 93-4, Rescher 1974: 702, BonJour 1978: 1, 5, 1985: 17, 30-33, 2001: 22-3, Pappas 1979a: 60, Lehrer 1990: 13, 41-2, Triplett 1990: 93, Haack 1993: 14-16, Pollock and Cruz 1999: 29-31, and Fumerton 2001: 3. Thus Pollock writes that, “[b]asic beliefs must be justified independently of reasoning – if a belief can only be justified through reasoning, its justification is dependent on the justification of the beliefs from which the reasoning proceeds, and hence, by definition, it is not a basic belief” (1986: 26).

160 While not uncontroversial, one could think that a putative defeater-defeater can prevent S’s belief B from being basic since it could contribute a great deal to the degree of justification enjoyed by B. Here the amount bestowed may vary with the strength of the defeater-defeater – partial defeaterdefeaters would confer less by way of justification than complete defeaterdefeaters, while reversing defeater-defeaters might contribute a great deal towards the positive epistemic status of a new non-basic theistic belief. Finally we should consider Plantinga’s own account of basicality: (21) Evidential View. S’s belief B at time t is basic only if B at t is such that it is not accepted by S on the evidential basis of any other beliefs in S’s noetic structure at t.36 At first it might seem as if the defeat of an influential argument from evil or the vindication of certain historical evidence for the claims of the New Testament would count as evidence in favor of S’s belief that God exists.37 While this may well be so, we need to be careful here. For what exactly is it for something to count as evidence?38 According to Plantinga, the paradigm case of an agent S having evidence for the truth of some proposition is the presence in S’s noetic structure of either deductive, inductive, or abductive arguments which are epistemically relevant to the truth of one or more of S’s beliefs.39 Given this understanding, the evidential view can be restated as follows:

36

See Plantinga 1983: 54, 1993a: 68-7, 1993b: 177-8, 185, 2000: 83, 175-6, 178, as well as Pappas 1979a: 53 and Grigg 1990: 389. According to Plantinga, “a person S accepts a belief A on the basis of a belief B only if (roughly) S believes both A and B and could correctly claim (on reflection) that B is part of his evidence for A” (1986: 306, emphasis his). 37 For a similar point, see Quinn 1985: 484. 38 According to Richard Feldman, evidence can be understood quite generally as all the information a person has which is relevant to a proposition. On this view, previously defeated theistic beliefs would rest in part on the evidential basis of their appropriate defeater-defeaters. See Feldman 1988: 89 and 1995: 252. 39 Plantinga has made this claim in conversation. It is also worth noting that even if the paradigm case of evidence for Plantinga is argument, he also allows for certain forms of non-propositional evidence. See in particular his 1993b: 177-8, 192-3.

161 (22) Evidential View*. S’s belief B at time t is basic only if B at t is such that it is not accepted by S on the basis of any other arguments in S’s noetic structure at t. Plantinga’s construal of the basing relation thus commits him to a form of what might be more generally called the inferential view: (23) Inferential View. S’s belief B at time t is basic only if B at t is such that it is not arrived at by way of any inferences from S’s other beliefs.40 If we interpret (22) and (23) loosely as claiming that in order for B to be a basic belief, S must not have accepted any prior arguments or undertaken any chains of inference which are (taken to be) relevant to the epistemic status of B, then clearly it follows on these two views that a previously defeated basic theistic belief will remain non-basic in the presence of the relevant defeater-defeaters. But such an implication does not follow so clearly on a strict interpretation of the evidential and inferential views according to which B is basic only if the proposition it affirms is itself not the conclusion of any prior chains of argument or inference. Thus the proposition: (13) God does exist. would be basic so long as S does not accept it as a result of any arguments with (13) as their conclusion. For purported partial and reversing theistic defeater-defeaters, this strict interpretation of (22) and (23) will have the same results as the other accounts of the basing relation which were considered above did. Partial defeater-defeaters are not sufficient by themselves for overcoming an agent’s non-basic atheistic belief. And reversing defeater-defeaters have (13) as their conclusion, thereby failing to satisfy the strict interpretation of (22) and (23). But the case of complete defeat is unique. For complete defeater-defeaters can undermine a defeater without in any way providing additional argumentative or inferential support for the agent’s initial belief. In Jim’s case, the free will defense does not have 40

See Pappas 1979a: 51, Winters 1980: 10, 12-4, Goetz 1983: 473, Audi 1986: 31-35, and Appleby 1988: 31.

162 (13) as its conclusion; rather it serves the merely negative task of undermining Mackie’s argument from evil. Thus we seem to have found an exception to our general descriptive claim about the impact of putative defeater-defeaters on the basicality of theistic belief.41 Much more could be said in precisely spelling out how defeaterdefeaters impact the structural makeup of a theist’s noetic structure, but hopefully Thesis 3 has at least some initial plausibility. So let us consider the normative considerations which arise with respect to this class of defeaters. Fortunately much of the groundwork already has been laid by our treatment of undefeated-defeaters in the previous section. For example, the relevant evaluative standards must arise from an internalist theory of defeater justification, and be such that they prescribe what the appropriate response is to S’s doxastic experience of a supposed defeater-defeater. In general terms, the epistemic prescriptions seem clear enough. When confronted on the one hand with a non-basic religious belief which arose as a result of the epistemic impact of a putative defeater, and on the other hand with what the agent takes to be a defeater-defeater for the defeater in question, the correct thing for her to do is to revise her noetic structure in such a way that she has a (non-basic) belief which accurately reflects the strength of the defeater-defeater.42 Once again, this requirement holds even if the defeater-defeater was produced by an irrational process or malfunctioning faculty. If this result is correct, then we can state our fourth and final thesis as follows: Thesis 4 For any theistic basic belief B arrived at by S at time t1 as a result of the proper functioning of the sensus divinitatis, and non-basic belief B or B* held by S at time t2 as a result of a putative undefeated-defeater D for B,43 if at time t3 S acquires what S takes to be a defeater-defeater DD for D, then: (1) If D is partial-defeater, then: 41

A similar exception might hold for the case of complete defeaters and a strict interpretation of the justification view (20). 42 We also can hope to establish this result by employing the Normative Parity Thesis and examining analogous epistemic standards for defeater-defeater activity. 43 As we saw in section two, the epistemic adjustments that should be made in light of a purported undefeated-defeater could be simply adjustments in the firmness that S holds B, rather than adjustments which require the formation of a new belief altogether.

163 (a) If DD is a partial defeater-defeater, then S ought to hold B less firmly than B was held at t1. (b) If DD is a complete defeater-defeater, then S ought to hold B with the same (or greater)44 strength that B was held at t1. (2) If D is a complete-defeater, then: (a) If DD is a partial defeater-defeater, then S ought to hold B less firmly than B was held at t1. (b) If DD is a complete defeater-defeater, then S ought to hold B with the same (or greater) strength that B was held at t1. (c) If DD is a reversing defeater-defeater, then S ought to hold B with greater strength than B was held at t1. (3) If D is a reversing-defeater, then: (a) If DD is a partial defeater-defeater, then depending on DD’s purported degree of defeat, S either ought to: (i) Continue to believe in the non-existence of God but less firmly than at t2. (ii) Suspend belief in both the existence and nonexistence of God. (iii) Believe in the existence of God but less firmly than at t1. (b) If DD is a complete defeater-defeater, then S ought to hold B with the same (or greater) strength that B was held at t1. (c) If DD is a reversing defeater-defeater, then S ought to hold B with greater strength than B was held at t1. This thesis is certainly unwieldy, but it does serve to helpfully articulate a theist’s epistemic obligations when it comes to appropriately responding to what purport to be defeater-defeaters by her own lights. 4. Further Complications Unfortunately we cannot leave matters as they stand. For there are two important complications which arise when considering the epistemic impact 44

The defeat of an influential defeater for theistic belief might be taken to merit even greater confidence in the truth of theism.

164 of putative defeater-defeaters – the role of so-called intrinsic defeaterdefeaters, and the influence of various kinds of defeater-defeaters which can be distinguished based upon the way in which they undermine the positive epistemic status of prior undefeated-defeaters. Let us take each of these in turn. Recall that we initially characterized defeater-defeaters as follows: (5)

DD is a defeater-defeater for S’s defeater D and belief B at time t iff: (a) S’s noetic structure at t includes B, D, and DD. (b) D is a defeater for B prior to t. (d) At t and as a result of possessing DD, S should longer regard D as a defeater of B, or at least not to the same extent.

We then proceeded in section three to treat such defeaters as beliefs typically formed by S at some time later than both B and D initially were formed. Thus it was assumed that all defeater-defeaters are extrinsic, where: (24) DD is an extrinsic defeater-defeater for S’s defeater D and belief B at time t iff: (a) DD at t is a defeater-defeater. (b) The content of DD at t is not identical to the content of B. In the case of Jim and the problem of evil, Jim later acquired an extrinsic defeater-defeater in the form of the free-will defense which allowed him to justifiably preserve his belief in the existence of God by his own lights. But as Plantinga observed in his well-known exchange with Philip Quinn, there may also be cases of intrinsic defeat in which S regards S’s original belief B as having enough by way of justification or warrant to itself defeat a potential defeater.45 Let us grant Plantinga the existence of such defeaters. At first it is tempting to emulate (24) and try to characterize them as follows: (25) DD is an intrinsic defeater-defeater for S’s defeater D and belief B at time t iff: 45

For Plantinga’s initial characterization of intrinsic defeater-defeaters, see his 1986: 311-312. For additional discussion, see also Quinn 1993: 37-44, Hasker 1998: 61-64, and Sudduth 1999b: 180-2.

165 (a) (b)

DD at t is a defeater-defeater. DD at t just is B.

But this would be a mistake. For despite its name, an intrinsic defeaterdefeater is not even a defeater to begin with. The reason is simple – condition (b) in our account of defeater-defeaters, the condition whereby D is a defeater for B at some time prior to t, fails to be met in cases of intrinsic defeat. The belief in question never rises to the status of being a defeater in the first place. Fortunately we can easily remedy (25) to avoid this problem: (26) Belief B is an intrinsic defeater-defeater at time t for some belief D whose content conflicts with the content of B iff: (a) S’s noetic structure at t includes B and D. (b) At t and solely as a result of possessing B, S should not take D to be a defeater for B. What implications does the possibility of intrinsic defeat have for the basicality of theistic belief? We have been concerned throughout this paper with the impact that various alleged defeaters have on theistic beliefs arrived at in the basic way as a result of the proper functioning of an agent’s sensus divinitatis. But in the case of intrinsic defeat, there is no such impact by defeaters to investigate – from the perspective of the agent in question, intrinsic defeater-defeaters preclude the obtaining of any relevant defeaters. So strictly speaking, the existence of intrinsic defeater-defeaters has no bearing on the issues in question in this paper. The second complication arises from the various ways in which putative defeater-defeaters might go about undermining the positive epistemic status of a defeater. Recall that we earlier distinguished between three types of defeaters – partial, complete, and reversing – and observed in section three that the same distinctions could be made for defeater-defeaters. Here it is important to note that what motivated such a taxonomy were the differences in epistemic outcomes which result from an agent’s taking there to be a defeater or defeater-defeater in her noetic structure. Thus partial defeater-defeaters are such that the agent no longer regards her prior defeater as having quite the epistemic bearing it formerly did on her original belief, whereas complete defeater-defeaters are such that the agent takes them to entirely offset the previous impact of a defeater.

166 But while this tri-fold distinction helps to clarify the possible results of putative defeater-defeater possession, it does little by way of illustrating the various means by which these defeater-defeaters actually go about undermining the positive epistemic status of their corresponding defeaters. For example, if a defeater comes in the form of an argument against the truth of the content of a basic belief, then by an agent’s own lights a defeater-defeater might show how the argument rests on a faulty premise or is formally invalid. In general, we can distinguish three means of defeaterdefeat:46 (27) DD is a soundness defeater-defeater for S’s defeater D and belief B at time t iff: (a) DD at t is a defeater-defeater. (b) At t and as a result of possessing DD, S should think that ~D is true.47 (28) DD is a validity defeater-defeater for S’s defeater D and belief B at time t iff: (a) DD at t is a defeater-defeater. (b) At t and as a result of possessing DD, S should think that š(D & B) is true.48 (29) DD is a conclusion defeater-defeater for S’s defeater D and belief B at time t iff: (a) DD at t is a defeater-defeater. (b) At t and as a result of possessing DD, S should think that B is true.49 50

46

An analogous distinction also applies to undefeated-defeaters, but there the distinction does not raise the same important issues about basicality as it does for the case of defeater-defeaters. 47 If D comes in the form of an argument, then S should think that one or more of D’s premises is false. 48 Again if D is an argument, then S should think that D is formally invalid such that it is possible to affirm its premises but deny its conclusion. Validity defeaters are similar to what Pollock has called undercutting defeaters. See his 1986: 39. 49 If D is an argument with premises P1 . . . Pn and conclusion C, then S should think that ~C is true. Here the resemblance is to Pollock’s notion of rebutting defeaters in his 1986: 38.

167 Thus if Jim had taken himself to have discovered a sound deductive argument for the existence of God, then he might have treated that as a conclusion defeater-defeater for his Mackie-style defeater. But instead he was introduced to the free will defense, and thereby came to acquire what looked to be a soundness defeater-defeater which fully restored his confidence in the existence of God, and hence also functioned as a complete defeaterdefeater.51 With this admittedly brief discussion in mind, we can examine what implications these three kinds of defeater-defeaters have for the basicality of an agent’s initial theistic belief. The case of conclusion defeat is rather straightforward – if S has what purports to be a conclusion defeaterdefeater DD for some prior defeater D, then regardless of whether DD is a partial, complete, or reversing defeater-defeater, the resulting belief in S’s noetic structure will be non-basic.52 DD would provide S with evidence for the truth of S’s initial belief, evidence upon which S would base that belief. Unfortunately, the case of validity and soundness defeat is more complex. Either demonstrating the falsity of a premise in a defeater argument or showing how the argument is formally invalid might be a necessary condition for continuing to be justified in holding the initial belief. But perhaps it would be a mistake to go on and say that the original belief would be based on the defeater-defeater; after all, the validity or soundness defeater-defeater simply might remove an obstacle to continuing to hold the initial belief in the same manner and with the same degree of confidence.53 But we must be careful here. For such a conclusion only can be established once we have gotten clear about what conditions are necessary for a belief’s being basic. And when we examine the accounts of basicality that were briefly outlined in section three, it is not at all obvious that a belief would be basic as the result of an agent possessing what he or she takes to be either a soundness or validity defeater-defeater. In fact, it seems clear 50

For similar distinctions, see Pollock 1986: 37-39, Plantinga 1993b: 40-42, 2000: chapter 11, esp. 359-362, and Sudduth 1999b, 1999c. 51 As this example is meant to bring out, there are partial, complete, and reversing defeater-defeaters which are either soundness defeating, validity defeating, or conclusion defeating. 52 Sudduth seems to agree with this conclusion in his 1999b and 1999c. 53 Plantinga (1993b, 185) and Sudduth (1999b, 1999c) both seem tempted by this line of thought.

168 that the first four views – Argument, Causal, Counterfactual, and Explanatory – all imply the opposite result. Nor is the Justification view much help; the justificatory status of S’s theistic belief would depend on that of the arguments employed by S as soundness and validity defeaterdefeaters.54 That leaves Plantinga’s own Evidential view. Yet even here basicality may fail since S could base belief in God in part on the arguments S employs as soundness and validity defeater-defeaters – arguments which S takes to be sufficient in some cases for allowing S to justifiably return to belief in the truth of theism.55 So while the existence of intrinsic, soundness, validity, and conclusion defeater-defeaters significantly complicates our model of an agent’s defeater system, it seems to do little by way of challenging our original four theses. 5. The Parallel Scenario for the Atheist Interestingly enough, we can run the same line of argument about the role of defeaters in the opposite direction. In other words, starting with an atheist who has a basic belief in the non-existence of God, we can show how he or she may come to acquire a non-basic belief when challenged by a putative theistic defeater. Given that the epistemic situation of the atheist would structurally parallel that of theistic belief discussed in the previous four sections, we can proceed relatively quickly to some interesting results. But first it is important to consider how a person might come to have such a basic belief in the non-existence of God. While it would take us too far afield to discuss the subject at any length, we can get an intuitive understanding by looking at the case of a person’s having an immediate experience of horrendous evil. For some, such an experience is the occasion for them to believe, without the aid of any arguments or evidence, that God

54

Sudduth seems to concede this point in both his 1999b and 1999c. Thus he writes in the latter article that, “[d]efeater-defeaters, if they do not all provide evidence for the truth of theism, do provide evidence for [the] truth of higher-level beliefs about the epistemic rationality of theistic belief. This kind of evidence is sometimes necessary if a person is to remain warranted in holding her theistic belief” (227). 55 As we saw in section three, Plantinga’s evidential view can be given both a loose and strict interpretation. If we adopt the strict interpretation, then again we seem to get an exception to our general thesis about non-basicality for the special case of what are taken to be complete soundness and validity defeater-defeaters.

169 does not exist.56 In a way analogous to the outputs of the sensus divinitatis, such atheistic basic beliefs also should be susceptible to what the agent takes to be various theistic defeaters and atheistic defeater-defeaters. If our earlier treatment of descriptive and normative epistemic defeat for theistic basic beliefs is correct, then similar conclusions should hold for the cases at issue in this section as well. For example, a putative undefeated-defeater typically would render a basic atheistic belief non-basic. Similarly, an atheist ought to revise his original basic belief in the appropriate way when confronted with what looks to be a defeater in his noetic structure. It is important to stress that even if theism is in fact true, we need not make any claims about the rationality or warrant of the initial atheistic belief or of its various defeaters. For our purposes, we are only interested in what typically is and ought to be done given a certain foundationalist noetic structure. With these considerations in mind, we can reformulate our first two theses as follows: Thesis 1* As a matter of fact, when S takes it to be the case that S has an undefeated-defeater D for S’s basic atheistic belief B which requires revision in S’s attitude towards B, then if S forms a new belief B* in response to D, B* typically will be non-basic in virtue of being at least partially based on D. Thesis 2* For any atheistic belief B arrived at by S in the basic way in circumstances C, if S is not in C and S takes there to be an undefeateddefeater D for B, then S ought to either: (a) (b) (c)

56 57

Hold B less firmly (and as a non-basic belief partially based on D). Cease holding B and form a new (non-basic) belief that S ought to suspend belief in both the existence and non-existence of God.57 Cease holding B and form a new (non-basic) belief that God does exist.

Plantinga 2000: 484. This understanding of agnosticism follows Benn 1999: 173.

170 (d)

Similar revisions would be required for Theses 3 and 4, but they are straightforward enough that we can pass over them for now.

6. Unresolved Issues With Theses 1 through 4 before us, the main goal of this paper has now been met. Yet in the course of investigating the relationship between epistemic defeaters and basic religious beliefs, a number of issues have arisen which deserve further attention. (a) The Scope of the Argument The focus of the paper has been on basic religious beliefs, but it is quite natural to ask whether our four theses could be generalized to encompass the bearing that purported defeaters have on basic beliefs as such. The outcome of such an investigation would be of significant interest to reformed epistemologists given their commitment to the Descriptive and Normative Parity Theses. Thus if it turned out that, for example, basic perceptual beliefs do not bear a similar relationship to putative defeaters, then this would provide some impetus towards either revising the parity theses or reconsidering the arguments given above. (b) The Special Case of Basic Defeat Throughout this paper, putative defeaters typically have been construed as non-basic beliefs whose contents serve as arguments for or against the truth of theism. But such an approach overlooks the special case of undefeated- and defeater-defeaters which are themselves basic beliefs. Thus, for example, when in SD-eliciting circumstances an agent might form a basic theistic belief but later on when in circumstances of horrendous evil come to believe that God does not exist. Finally at some third time the agent might reflect on both of these experiences and try to decide what to believe about God’s existence. Regardless of the content of the resulting final belief, would it be basic or non-basic? (c) Agnosticism The discussion above focused primarily on theistic and, to a lesser extent, atheistic belief. But what bearing do defeaters have on agnosticism? This is a complex issue due in large part to various difficulties which arise when

171 trying to give a precise characterization of the view.58 But without worrying about the technical details here, it seems clear enough that an agnostic has similar epistemic obligations which must be satisfied when he or she is faced with what purports to be a defeater. The real question as far as this paper is concerned, however, is whether there is such a thing as basic agnostic belief. If not, then the issues raised above will be of little relevance to the agnostic. (d) Evidentialism One important issue that merits further consideration is the extent to which the results argued for above have any bearing on the debate between evidentialists and reformed epistemologists in contemporary philosophy of religion. While my claims here are speculative, it seems that we can carve out a moderate position which draws on the advantages of both approaches. This view, what we might call Moderate Evidentialism, would on the one hand preserve the possibility of basic theistic beliefs which satisfy those conditions individually necessary and jointly sufficient for justification and warrant. Yet the view also would pay serious attention to the important role played by various kinds of propositional evidence which might serve as putative defeater-defeaters for some of the alleged defeaters that most reflective religious believers face today.59 Giving substance to this Moderate Evidentialist position strikes me as a fruitful and important research project.60

58

For a brief survey of various construals of agnosticism, see Benn 1999: 171-174. Moderate Evidentialism may be similar to what Michael Sudduth has called defeater-based evidentialism in his 1999c: 228. 60 Many thanks to Leopold Stubenburg, David Hemp, Kevin Meeker, René van Woudenberg, and especially Alvin Plantinga for helpful comments on earlier drafts. Thanks also to the University of Notre Dame for a Presidential Fellowship which supported my graduate studies and thereby my work on this paper. 59

172

REFERENCES Appleby, Peter C. 1988. “Reformed Epistemology, Rationality, and Belief in God”, International Journal for Philosophy of Religion 24: 129-141. Armstrong, D. M. 1973. Belief, Truth and Knowledge. Cambridge: Cambridge University Press. Audi, Robert. 1986. “Belief, Reason and Inference”, Philosophical Topics 14: 27-65. ———. 1998. Epistemology: A Contemporary Introduction to the Theory of Knowledge. London: Routledge. Benn, Piers. 1999. “Some Uncertainties about Agnosticism”, International Journal for Philosophy of Religion 46: 171-188. Bergmann, Michael. 1997a. Internalism, Externalism, and Epistemic Defeat. Notre Dame: University of Notre Dame Ph. D. Dissertation. ———. 1997b. “Internalism, Externalism and the No-Defeater Condition”, Synthese 110: 399-417. BonJour, Laurence. 1978. “Can Empirical Knowledge Have a Foundation?”, American Philosophical Quarterly 15: 1-13. ———. 1985. The Structure of Empirical Knowledge. Cambridge: Harvard University Press. ———. 2001. “Toward a Defense of Empirical Foundationalism” in Resurrecting Old-Fashioned Foundationalism. Ed. Michael DePaul. Lanham: Rowman & Littlefield, 21-38. Feldman, Richard. 1988. “Having Evidence”, in Philosophical Analysis: A Defense by Example. Dordrecht: Kluwer Academic, 83-104. ———. 1995. “Evidence” in The Cambridge Dictionary of Philosophy. Cambridge: Cambridge University Press, 252. Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham: Rowman and Littlefield Publishers. ———. 2001. “Classical Foundationalism” in Resurrecting Old-Fashioned Foundationalism. Ed. Michael DePaul. Lanham: Rowman & Littlefield, 3-20.

173 Goldman, Alvin. 1986. Epistemology and Cognition. Cambridge: Harvard University Press. Goetz, Stewart C. 1983. “Belief in God is Not Properly Basic”, Religious Studies 19: 475-484. Greco, John. 1998. “Foundationalism and Philosophy of Religion” in Philosophy of Religion: A Guide to the Subject. Ed. Brian Davies. Washington D.C.: Georgetown University Press, 34-41. Grigg, Richard. 1990. “The Crucial Disanalogies between Properly Basic Belief and Belief in God”, Religious Studies: 389-401. Haack, Susan. 1993. Evidence and Inquiry: Towards Reconstruction in Epistemology. Oxford: Blackwell. Hasker, William. 1998. “The Foundations of Theism: Scoring the Quinn-Plantinga Debate”, Faith and Philosophy 15: 52-67. Kim, Kihyeon. 1993. “Internalism and Externalism in Epistemology”, American Philosophical Quarterly 30: 303-316. Koehl, Andrew. 1998. Implicitly Grounded Beliefs. Notre Dame: University of Notre Dame Ph.D. Dissertation. Korcz, Keith. 1997. “Recent Work on the Basing Relation”, American Philosophical Quarterly 34: 171-191. Lehrer, Keith. 1990. Theory of Knowledge. Boulder: Westview Press. Mackie, J. L. 1955. “Evil and Omnipotence”, Mind 64: 200-12. McLeod, Mark S. 1993. Rationality and Theistic Belief: An Essay on Reformed Epistemology. Ithaca: Cornell University Press. Moser, Paul. 1989. Knowledge and Evidence. Cambridge: Cambridge University Press. Nozick, Robert. 1981. Philosophical Explanations. Cambridge: Belknap Press. Pappas, George. 1979a. “Basing Relations” in Pappas 1979b, 51-64. ———. 1979b. Justification and Knowledge: New Studies in Epistemology. Ed. George Pappas. Dordrecht: D. Reidel Publishing.

174 Plantinga, Alvin. 1979. “Is Belief in God Rational?”, Rationality and Religious Belief. Ed. C. F. Delaney. Notre Dame: University of Notre Dame Press. ———.1981. “Is Belief in God Properly Basic?”, Nous 15: 41-51. ———.1982. “Rationality and Religious Belief”, Contemporary Philosophy of Religion. Ed. Steven Cahn and David Shatz. New York: Oxford University Press. ———. 1983. “Reason and Belief in God,” in Faith and Rationality. Ed. Alvin Plantinga and Nicholas Wolterstorff. Notre Dame: University of Notre Dame Press, 16-93. ———.1986. “The Foundations of Theism: A Reply”, Faith and Philosophy 3: 298313. ———.1993a. Warrant: The Current Debate. New York: Oxford University Press. ———. 1993b. Warrant and Proper Function. New York: Oxford University Press. ———. 2000. Warranted Christian Belief. New York: Oxford University Press. Pollock, John. 1974. Knowledge and Justification. Princeton: Princeton University Press. ———. 1984. “Reliability and Justified Belief”, Canadian Journal of Philosophy 14: 103-114. ———. 1986. Contemporary Theories of Knowledge. Totowa: Rowman Littlefield. Pollock, John and Joseph Cruz. 1999. Contemporary Theories of Knowledge. Second Edition. Lanham: Rowman & Littlefield. Quinn, Philip. 1985. “In Search of the Foundations of Theism”, Faith and Philosophy 2: 468-486. ———.1993. “The Foundations of Theism Again,” in Zagzebski 1993, 14-47. Rescher, Nicholas. 1974. “Foundationalism, Coherentism, and the Idea of Cognitive Systematization”, Journal of Philosophy 71: 695-708. Sudduth, Michael. 1999a. “Can Religious Unbelief be Proper Function Rational?”, Faith and Philosophy 16:3: 297-314. ———. 1999b. “The Internalist Character and Evidentialist Implications of Plantingian Defeaters”, International Journal for Philosophy of Religion 45: 167-187.

175 ———. 1999c. “Proper Basicality and the Evidential Significance of Internalist Defeat: A Proposal for Revising Classical Evidentialism,” in The Rationality of Theism. Ed. Brüntrup and Tacelli. Dordrecht: Kluwer Academic Press, 215236. Swain, Marshall. 1979. “Justification and the Basis of Belief” in Pappas 1979b, 25-50. ———. 1981. Reasons and Knowledge. Ithaca: Cornell University Press. Triplett, Timm. 1990. “Recent Work on Foundationalism”, American Philosophical Quarterly 27: 93-116. Winters, Barbara. 1980. “Reasonable Believing”, Dialectica 34: 3-16. Zagzebski, Linda (ed.). 1993. Rational Faith: Catholic Responses to Reformed Epistemology. Notre Dame: University of Notre Dame Press.

DUNCAN PRITCHARD

Reforming Reformed Epistemology* 0. Introduction There has been a renaissance of interest in the epistemology of religious belief over the last twenty years which has been largely inspired by the work conducted by Alvin Plantinga, William Alston, Nicholas Wolterstorff and others on the so-called “reformed” defence of the rationality of religious belief. The starting-point for this reformed conception of religious epistemology is a rejection of the supposedly evidentialist assumptions which drive standard sceptical arguments regarding religious belief.1 I think that this general negative claim is, in its bare essentials, correct, although canvassing support for an argument for this contention will not be my primary concern here. Instead, I will be outlining one way in which the reformed epistemological stance can be modified to make it resistant to a certain sort of attack. I suggest that the manner in which the reformed conception of the epistemology of religious belief is often motivated with respect to a supposed analogy with perceptual belief has tended to overemphasize certain features of religious belief, and that recognising this fact enables one to offer a more fine-grained account of the epistemic status of religious belief. In particular, I argue that making this point clear draws out the sense in which reformed epistemology, properly understood, should be allied to a specific form of virtue epistemology. In §1, I give an overview of the standard sceptical attack on the rationality of religious belief and briefly outline the sense in which such scepticism is held to rest upon an evidentialist thesis. In §2, I elucidate the manner in which the theses of evidentialism, classical foundationalism and epistemological internalism are intertwined within the debate between the reformed epistemologist and the sceptic about religious belief. In §§3-4, I explain how reformed epistemology is typically motivated with respect to a *

A slightly different version of this article was published (with the same title) in International Philosophical Quarterly 43, 2003: 43-66. I am grateful to the editors of that journal and to the publishers, the Philosophy Documentation Center, for permission to reprint this article here. 1 For an excellent overview of the debate, see Plantinga and Wolterstorff 1983.

178 “parity” argument concerning certain relevant analogies between religious and perceptual belief, and I go on to consider some problems that this parity argument faces. In §5, I consider Keith DeRose’s intriguing suggestion that one can resolve some of these problems by reconfiguring reformed epistemology in terms of the sort of “foundherentist” epistemological model advocated by Susan Haack. I argue that this proposal should be endorsed provided that it is understood at a very general schematic level. So construed, however, it fails to offer us the fuller account that we are after. Accordingly, in §6, I argue that the best way of adding content to this foundherentist structural model of religious epistemology is by conceiving of reformed epistemology in terms of a specific type of virtue epistemology. Finally, in §7, I offer some concluding remarks. 1. Scepticism about Religious Belief Perhaps the most common form of scepticism about religious belief has an ontological rather than (at least directly) an epistemological form. That is, the focus of the attack is not directly on the epistemic status of the belief in question but rather concerns the ontology that is thought to underlie that belief. One might naturally think that any dispute over ontology would always spill-over into a dispute about epistemic status. If I think that you are wrong to commit yourself to a certain ontology then do I not thereby regard your beliefs in such an ontology as lacking the requisite epistemic status? Interestingly, the answer to this question is, despite first appearances, “no”. Think, for example, of two eminent scientists who, whilst sharing many beliefs, have differing beliefs regarding the ontological status of a certain ‘entity’ that is sometimes postulated by scientists working within a particular area of scientific inquiry, the one thinking that it exists and the other thinking that it doesn’t. Although each is committed to regarding the other as being in error in some way, both of them could recognise that the other’s belief had some substantial degree of positive epistemic status. Why, then, is it the case that when it comes to disputes between religious believers and their detractors, the latter tends to formulate the ontological challenge such that it has direct, and devastating, epistemological ramifications for the epistemic status of the religious beliefs in question? The standard answer is that the difference between the scientists in our imagined example and the religious believer is that the scientists can in principle offer appropriate evidence to support their belief whereas the religious believer cannot.

179 Consider the contrast between the two cases. In the case of the scientists, each of them could, conceivably, offer the other evidence which they both accept as being good evidence for the belief in question (even though, of course, they are both obliged, as long as they maintain their positions, to disagree about the extent of the relative evidential support in each case). The situation facing the religious believer is very different. The only “evidence” that he could propose which might potentially be adequate for the task will be contentious in this context. He might base his religious beliefs upon certain sorts of religious experiences, revelation for example, but this sort of evidence will only be apt in this context provided that one is already willing to grant the ontology that is presupposed by counting these experiences as “religious” in the relevant sense.2 Indeed, the only sort of evidence that would seem to be appropriate for this purpose would be an a priori argument of some description, or at least some sort of incorrigible empirical evidence, and few people would now hold that there are grounds of this sort available to license religious belief. There is thus a sense in which the religious believer is unable to adduce non-question-begging evidence in support of his religious belief.3 In both cases the dispute is being characterised in terms of an evidentialist requirement to the effect that one should proportion one’s belief to the evidence available to support that belief, but the thought is that whereas certain sorts of beliefs can, in principle, meet this requirement, there is an a 2

At least in the standard case. Of course, there may be certain sorts of events which everyone might be willing to recognise as providing corrigible empirical evidence in support of religious belief that is non-question-begging (the sky opens up across the world and a being comes down proclaiming certain verses from a key religious text, and so on). But even if one is willing to grant that such cases might be able to offer the requisite non-question-begging evidence, this concession will offer little comfort to the apologist for religious belief. After all, the epistemic status of one’s religious belief would then just be hostage to the occurrence of such events and, in the meantime, completely lacking. 3 In this respect scepticism about religious belief mirrors more general forms of scepticism. Consider external world scepticism, for example, where the claim is that one is unable to adduce the kind of non-question-begging evidence required which would support the belief without already assuming (as do most empirical beliefs) the existence of an external world in the first place. Accordingly, any anti-evidentialist strategy as regards religious belief will have application to scepticism in general. For more discussion of this question-begging aspect of radical sceptical arguments, see Wright 1985; 2000 and Pritchard 2002a. For a discussion of the general relationship between scepticism about religious belief and radical scepticism, see Pritchard 2000. For a survey of recent work on radical scepticism, see Pritchard 2002b.

180 priori difficulty with religious belief which means that it can never adequately fulfil it. We shall explore the thesis of evidentialism more fully below, but it is worth noting that, at first glance at any rate, the evidentialist requirement seems an entirely uncontentious hurdle for religious belief to clear. 2. Evidentialism, Classical Foundationalism and Epistemological Internalism Historically, one common response to this sort of sceptical argument has been to try to supply the very sort of a priori grounds that are being demanded. In contrast, another type of response has been to embrace the sceptical conclusion of that argument by claiming that religious belief is not the sort of belief that requires epistemic support.4 What both of these approaches to scepticism about religious belief share is an implicit acceptance of the nature of the sceptical argument such that any response to that argument must consist in either some sort of rapprochement with the conclusion or a rejection of the key premise that religious beliefs lack the evidence in question. In contrast, the reformed epistemological stance under discussion here adopts the completely different approach of calling into question the very evidentialist doctrine that drives the sceptical argument.5 In many of the key accounts of reformed epistemology, the locus for evidentialism has been W. K. Clifford’s provocative claim that “[I]t is wrong always, everywhere, and for anyone, to believe anything upon insufficient evidence”.6 There are, however, eminent proponents of the evidentialist thesis in more recent epistemological debate, so it is perhaps better to focus on one of these contemporary accounts. For example, in two recent papers (one of them co-authored with Earl Conee), Richard Feldman has offered a spirited defence of the evidentialist position. He defines the thesis as follows: For any person S, time t, and proposition p, if S has any doxastic atti4

This style of response has, of course, found a great deal of support amongst those guided by what they consider to be certain insights in Wittgenstein’s later remarks on religious belief, especially as they appear in Wittgenstein 1966. See, for example, Nielsen 1967 and Phillips 1976. 5 Note that these three alternative defences of religious belief are not all in competition. One could, for instance, argue that there are a priori grounds available in support of religious belief whilst also contending that evidentialism is an erroneous doctrine. 6 Clifford 1879: 186.

181 tude at all toward p at t and S’s evidence at t supports p, then S epistemically ought to have the attitude towards p supported by S’s evidence at t.7 Although this formulation may meet some of the objections that are most naturally levelled at the much less precise rendering of the doctrine due to Clifford, it still faces several seemingly insuperable difficulties. To begin with, we need to look a little closer at how we are to understand the notion of evidence in play here. Presumably, we cannot just understand an agent’s evidence to be whatever grounds that are available to the agent in support of his belief regardless of whether or not he attends to (or even can attend to) that evidence. There are many reasons for this, but the most pressing is that we want the evidence to play the required supporting role, and unless that evidence is attended to then it need not perform this function. Consider, for example, the case of a man who has excellent evidence in support of his belief that a certain person is guilty, but who does not attend to this evidence when forming the belief. Instead, he forms his belief as a result of a bigoted ill-feeling towards the person in question. Clearly, this would not be a case of someone meeting the kind of epistemic requirement that is meant to be imposed by evidentialism because this person’s belief would not enjoy any substantial measure of positive epistemic status, despite the existence of evidence in support of that belief that is available to the subject. As Feldman’s characterisation of evidentialism stands, however, it makes no distinction between an agent who believes what he epistemically ought to believe because of bigoted ill-feeling and an agent who believes what he epistemically ought to believe because he is sensitive to the evidence that is available to him in support of that belief.8 It appears, then, that we must restrict the evidentialist thesis so that it is understood as requiring that an agent should proportion his belief to the evidence that he has for that belief and which he is presently attending to.9 Even under this construal, however, problems remain. For unless we offer 7

Feldman 2000: 679; cf. Feldman and Earl Conee 1985. This example is adapted from one used by DeRose 2000, 697-8, in his astute response to Feldman 2000. For some of the other, to my mind, decisive criticisms that can be levelled against evidentialism, one should consult DeRose’s paper. 9 Of course, a more detailed characterisation of the evidentialist account would involve an elucidation of this notion of “attending to”. Since I am not defending the evidentialist position, however, and since I do not think that anything consequential hangs upon our understanding of this notion, I will not be undertaking a detailed discussion of it here. 8

182 some restriction on what can count as evidence in this respect then it is far from clear that evidentialism, so understood, could plausibly drive a sceptical argument regarding religious belief. After all, as it stands, this characterisation of evidentialism would seem to permit one to adduce such evidence as revelation in support of one’s religious belief and, as noted above, the religious believer has (by his own lights) plenty of evidence of this sort. How, then, do we form the connection between a general evidentialist thesis and scepticism about religious belief? The standard answer to this question is to argue, whether explicitly or implicitly, that it is not evidentialism alone that is the motivation behind scepticism about religious belief, but rather the combination of evidentialism and what Plantinga refers to as ‘classical foundationalism’. He characterises the classical foundationalist account in terms of a certain conception of the criteria that a belief needs to meet if it is to be “properly basic”: For any proposition A and person S, A is properly basic for S if and only if A is incorrigible for S or self-evident to S.10, 11 It is certainly true that, historically, classical foundationalism and evidentialism have been closely intertwined. Although there is some debate about the historical roots of the evidentialist requirement on religious belief, it is certainly present in the work of Locke, and, significantly, Locke combines such a thesis with a version of classical foundationalism that demands that the foundational propositions be certain.12 Locke famously wrote in the Essay Concerning Human Understand13 ing that “reason must be our last judge and guide in everything”, and, accordingly, he maintained that religious beliefs should be put before the tribunal of reason just like any other. This line of thinking did not lead to scepticism about religious belief as far as Locke was concerned, however, because he held that the necessary evidential grounds were available in support of religious belief. The conclusion was thus only that we must dis10

Plantinga 1981: 49. In later works, such as Plantinga 1983, Plantinga extends the class of properly basic propositions to include those propositions that are also evident to the senses. 12 More specifically, as Wolterstorff 1991, 81, puts it: “Locke’s official view is that the only things we know immediately are those necessary truths that are self-evident for us, whereas his unofficial view is that we can know immediately whatever we are certain of without inference. In either case, knowledge is grounded in certitude.” 13 Locke 1979 IV, xix: 14. 11

183 tinguish between those believers whose religious beliefs were grounded in reason and those believers (whom Locke called the “enthusiasts”) whose religious beliefs were grounded only in revelation. He describes those who fall into this latter camp as follows, arguing that if they regard what they believe as being true solely [...] because it is a revelation, and have no other reason for its being a revelation but because they are fully persuaded, without any other reason, that it is true, they believe it to be a revelation only because they strongly believe it to be a revelation; which is a very unsafe ground to proceed on, either in our tenets or actions.14

What we find with Locke is the claim that evidence must be available to support the religious belief in question and that the only evidence suited to the purpose, given his prior commitment to (one form of) classical foundationalism, will be that which is derived from propositions which are certain. Although this line of thought is, of course, distinct from scepticism, one need only remove the conviction that the relevant certain foundational propositions are available for the sceptical challenge to emerge.15 Likewise, in contemporary discussion, the thought is that the legitimation of religious belief requires not just evidence but evidence of a certain kind: foundational evidence of an a priori or incorrigible variety. So construed, the connection between evidentialism and scepticism about religious belief becomes transparent. Nevertheless, one should be careful about identifying evidentialism too closely with classical foundationalism because this can tend to obscure the underlying nature of the dialectic in play here. In particular, too close an identification can lead to the impression that merely denying classical foundationalism would suffice to meet this sceptical challenge. This conception of the debate is certainly wrong, however, because one could state the evidentialist challenge without making any mention of classical foundationalism. In order to see this point, it is worthwhile returning to the contrast drawn earlier between the two scientists who were arguing about the ontological status of a certain “entity” and the debate between the religious be14

Locke 1979 IV, xix: 11. There are, of course, subtleties to Locke’s view in this respect that this brief overview cannot do justice to. Nevertheless, this short account should suffice for our purposes here. For an excellent discussion of Locke’s view in this regard that is sympathetic to the line taken here, see Wolterstorff 1996. 15

184 liever and the sceptic about religious belief. Significantly, in both cases one can characterise these debates in terms of an evidentialism that does not incorporate a commitment to classical foundationalism. Take the scientific case first, which was meant to be a situation in which one could, in contrast to the religious case, respond to the evidentialist challenge with the requisite evidence. It seems perfectly acceptable in such a scenario to regard one of the disputants as possessing the greater degree of evidential support for his belief, and thus as being able to convince the other scientist that he should weaken, if not completely change, his opposing belief. Moreover, there need be no assumption in play here that the evidence adduced should be a priori or incorrigible. Indeed, one would expect it to be ordinary corrigible empirical evidence. Of course, one might further demand that, at some point, this chain of support must lead to incorrigible or a priori evidence, but there need be no mention of this claim directly and it is far from obvious that such a further move is entailed by the mere evidentialist thesis alone. Similarly, one can characterise the debate about religious belief without mention of classical foundationalism. For on at least one plausible construal of evidentialism, which demands not just evidence but nonquestion-begging evidence, then it will be true that the religious believer will be unable to offer the requisite grounds. Moreover, we can put this point in terms of evidentialism alone without making any direct recourse to classical foundationalism because there is no obvious sense in which this “non-question-begging” demand must entail a demand for a priori or incorrigible evidence. Of course, as we saw above, the thought that the reformed epistemologist has is that this “non-question-begging” form of evidentialism is simply the result of conjoining a more neutral form of evidentialism with the classical foundationalist thesis, and thus that classical foundationalism is implicit within this purely evidentialist account of religious scepticism after all. The reformed epistemologist will typically claim, for example, that when it comes to religious belief the only way to offer such nonquestion-begging evidence is to adduce a priori or incorrigible grounds, and thus that this variety of evidentialism presupposes classical foundationalism. On this understanding of the debate, it is merely a matter of taste whether one characterises the sceptical argument in terms of a beefed-up evidentialism or in terms of a neutral rendering of the evidentialist doctrine which one then supplements with classical foundationalism. Such a conception of the debate can tend to mislead, however, since

185 what ultimately underlies the move from the neutral form of evidentialism to the “non-question-begging” version is not classical foundationalism at all but rather epistemological internalism. Epistemological internalism is, roughly, the thesis that positive epistemic status demands reflective access on the part of the subject to those facts that determine that epistemic status.16 In the case of evidentialism, for example, an internalist variant of this thesis would not just demand the availability of evidence to the subject, but also that the subject is actually in a position to not only reflectively access that evidence but also reflectively determine that it is evidence. This is the kind of demand that is being made in the more restricted version of evidentialism just considered where the agent has to actually vouch for the evidence as being evidence that is suited to the purpose (rather than question-begging evidence). Evidentialism and epistemological internalism tend to go together because any natural rendering of the evidentialist thesis invokes the internalist doctrine.17 Nevertheless, it is best to keep the two theses apart as much as possible to bring out the role that epistemological internalism plays in the sceptical argument regarding religious belief. In particular, by being clear about the role played by epistemological internalism in this respect, we avoid the temptation of thinking that a simple denial of classical foundationalism would suffice to meet the sceptical problem about religious belief. For even if one granted that corrigible empirical nonquestion-begging evidence could support legitimate belief (as we granted above in the case of the two scientists), it would still remain that we have a sceptical problem regarding religious belief because of the inherently question-begging nature of the evidence involved. Note that this is not to deny that the role that classical foundationalism plays in the sceptical argument about religious belief is closely intertwined with that played by epistemological internalism. After all, the most obvious way of meeting this “non-question-begging” demand is by adduc16

Though an extremely rough characterisation of this doctrine, it should suffice for our purposes here. The locus classicus for epistemologically internalist approaches is, of course, Chisholm 1989. For more on the epistemic externalism/internalism contrast, see the papers contained in the excellent anthology on this subject edited by Kornblith 2001. 17 One reason for this is that evidentialism tends to be closely associated with a certain deontological thesis regarding justification that has explicitly internalist overtones. Indeed, Feldman e.g., 2000 is a good example of someone who holds both an epistemologically internalist thesis characterised along deontic lines and, for related reasons, also endorses a form of evidentialism.

186 ing a priori or incorrigible empirical evidence in the manner that classical foundationalism demands. But this merely reflects the fact that classical foundationalism tends to be a natural consequence of epistemological internalism, and thus that it is epistemological internalism that is the underlying force behind scepticism about religious belief. That scepticism about religious belief should ultimately find its source in epistemological internalism should not surprise us, since, as is well-known, epistemological internalism drives a number of other sceptical arguments as well.18 Moreover, I think we gain a better understanding of the core elements of the reformed epistemology response to scepticism about religious belief by understanding it as, primarily, an epistemologically externalist response to such scepticism, rather than as an antievidentialist or anti-classical-foundationalist thesis. Accordingly, henceforth when I refer to evidentialism it will be the explicitly internalist variant of this thesis that I have in mind. 3. The Parity Argument One of the great benefits of being clear about the nature of the evidentialist demand as regards religious belief is that it highlights just how austere that demand is. The reformed epistemologist can use this observation to his advantage. In particular, reformed epistemologists have pointed out that such a requirement on positive epistemic status would seem to rule-out a great deal of belief as being epistemically lacking. The standard example employed here is that of perception since, if we know anything much about the world at all, then it would seem that we must have some perceptual knowledge. Consider how perceptual knowledge fares with respect to the evidentialist thesis, however. Is it really true that we are able to offer evidence that we can vouch for which is sufficient to support the level of conviction present in our standard perceptual beliefs? In order to make the parallel between the religious case and the perceptual case more direct, consider someone who is sceptical about perceptual belief in general on the grounds that he is sceptical about the external world ontology that it commits one to. How would one convince such a sceptic? Clearly, as with the religious case, it would not do to cite empiri18

Plantinga certainly would not be surprised by this observation, of course, because it informs a great deal of his discussion of warrant in Plantinga 1993a. Nevertheless, as I will be arguing below, Plantinga has not adequately recognised the nuanced manner in which the externalism/internalism distinction impacts on this debate.

187 cal grounds that are gained by perception since this would be questionbegging. This evidence is only apt for the purpose provided that the beliefs at issue are genuine perceptual beliefs (rather than merely beliefs about how things seem), and insofar as they are assumed to be genuine perceptual beliefs then the ontology in question is being illicitly taken for granted. As with the religious case, then, the only way around this requirement would appear to be to offer either a priori or incorrigible support for our perceptual beliefs, and such support is not obviously forthcoming. Allowing evidentialism in the case of religious belief would thus seem to license, by parity of reasoning, radical scepticism as regards perceptual belief. Provided that we grant that if we know anything much at all then we must have some perceptual knowledge, we thus have independent grounds to be sceptical about the evidentialist requirement itself.19 Reformed epistemologists have been keen to exploit this parallel between religious belief and perceptual belief as a means of putting religious belief on the same sort of secure footing that is typically granted to perceptual belief. This line of reasoning has been characterised by Alston and others as a “parity argument”. Alston argues that the point of such an argument is to show that what he calls “Christian Practice” [...] has basically the same epistemic status as [Perceptual Practice] and that noone who subscribes to the latter is in any position to cavil at the former.20, 21

In order to achieve this end, the possibility which Alston explores is that [...] religious experience is basic to religious belief in somewhat the way in which sense experience is basic to our beliefs about the physical world. In both cases [...] we form certain beliefs about the subject matter (God, the physical environment) on the basis of experience [...].22

19

Of course, the religious sceptic could at this point retreat into a general scepticism that was not solely confined to religious belief. The trouble with this manœuvre, however, is that it only serves to undermine the original sceptical argument about religious belief. After all, that argument contended that there was something peculiarly defective about religious belief, not that belief in general was problematic. Indeed, if belief in general is problematic then the religious believer is no worse off than the agnostic or the atheist. Newman is sensitive to this point. See Newman 1844; 1985. 20 Alston 1982: 12. 21 The original quotation has the “former” and “latter” in the reverse order but, as DeRose 1999a notes, this is surely a mistake on Alston’s part. 22 Alston 1986: 2.

188 In particular, reformed epistemologists tend to concentrate on two parallels between religious and perceptual belief. The first parallel, which we have already noted, is that both religious beliefs and perceptual beliefs are prone to an evidentialism-based sceptical argument. Insofar as one regards perceptual belief as legitimate, then, one is obliged to regard religious belief as being, at least prima facie, legitimate as well. That is, if evidentialism is inapplicable as regards perceptual belief, then it cannot simply be assumed to play a role as regards religious belief. This is the “negative” element of the parity argument since it merely contends that there are prima facie grounds for thinking that religious belief is no worse off, epistemically speaking, than perceptual belief.23 The second parallel is more positive in that it draws upon relevant similarities between the nature of perceptual experience and the nature of religious experience which would appear to license the same sort of nonevidentialist epistemology that is often applied in the perceptual case to the religious case.24 More specifically, the thought is that religious belief can sometimes enjoy the very sort of “directness” that is often found in perceptual belief and which, in the perceptual case, licences the adoption of a non-evidentialist epistemology. As Laurence Bonjour25 has put it, ordinary perceptual beliefs tend to arise “spontaneously” out of certain perceptual experiences, and the same might be said regarding how certain religious beliefs arise in response to particular religious experiences. By exploiting this positive analogy between religious and perceptual belief, reformed epistemologists, such as Alston and Plantinga, have argued that the kind of non-evidentialist epistemological model that is applied in the perceptual

23

Another “negative” claim that Plantinga 1983 makes in this respect is that classical foundationalism is self-refuting, because an agent’s belief in the classical foundationalist doctrine will not itself be grounded in either incorrigible or a priori evidence. 24 What is meant here by “religious experience”? The best account that I know of is due to Alston 1986 who argues that it is those experiences that give rise to the “Mbeliefs” (or manifestation beliefs) that are the focus for a great deal of his discussion on this matter. He writes that religious experience concerns: “[...] experiences that would naturally lead the experiencer to formulate that he has experienced something about God’s current relation to himself; that God said “ ” to him, that God was enlightening him, comforting him, guiding him, sustaining him in being, or just being present to him.” Alston 1986: 6. One of the advantages of this characterisation is that the beliefs that it gives rise to will tend to have a clear propositional content. 25 BonJour 1985, passim.

189 case ought to be just as apt in the religious case.26 In particular, the failure of evidentialism as regards perceptual belief points towards the need for a form of externalist epistemology that characterises perceptual knowledge in terms of an appropriate reliable relationship between one’s beliefs and the physical environment which those beliefs are supposed to track. Such an account is externalist in the sense that it does not require that the agent who has epistemically supported perceptual beliefs should be able to reflectively recover the grounds that support such beliefhis “evidence”in the way that evidentialism demands. This kind of approach to perceptual belief takes its cue from a certain construal of the writings of Thomas Reid, and in effect maintains that perceptual belief enjoys a kind of “default” epistemic support such that, in the absence of countervailing evidence and provided that the belief-forming mechanisms in question are as a matter of fact reliable, an agent’s perceptual belief can enjoy sufficient positive epistemic status even when the agent is unable to adduce adequate evidential grounds to support that belief. Prima facie, an externalist epistemology understood along these lines ought to be equally applicable to religious belief. Just so long as the beliefforming mechanisms in question (however they are to be characterised) are appropriately reliable, and just so long as there is no countervailing evidence to take into account which would imply that those mechanisms are not functioning adequately, then such beliefs seem to enjoy a default positive epistemic status. Of course, anyone who does not share the religious belief in question will be sceptical about the putative reliability of the belief-forming mechanisms just as they will be sceptical about the ontology presupposed by such beliefs, but this situation is no different in relevantly similar cases, such as with perceptual belief. Saying that perceptual belief gives us knowledge so long as it is indeed reliable in the appropriate way will not persuade someonethe radical sceptic, saywho doesn’t already accept the ontology that is being presupposed here. Nevertheless, such a manœuvre would support the claim that perceptual knowledge is at least conditionally possible given that certain conditions actually do obtain. Similarly, the religious believer can respond to an evidentialism-based 26

See Alston 1986; 1991, and Plantinga 1993a; 2000. As I discuss below with respect to Alston, however, different reformed epistemologists have differing conceptions of what can be concluded from this “positive” element of the parity argument. The reader should also note that although it is common practice to refer to both Alston and Plantinga as reformed epistemologists, both of these writers have rejected this categorisation.

190 scepticism about religious belief by arguing that religious knowledge is indeed possible after all, just so long as certain factual conditions obtain. In effect, what the religious believer is doing here is distinguishing the epistemological question of whether it is ever possible for one’s religious beliefs to enjoy a sufficient degree of positive epistemic status from the metaphysical question of whether there really does exist the kind of ontology to which the religious believer is committed. Though a negative answer to the metaphysical question would, of course, prejudice the possibility of a positive answer in the epistemological case, in the absence of such a negative metaphysical conclusion the epistemological question can be met. The burden is thus placed back on the sceptic about religious belief to offer the relevant negative metaphysical argument, and it is hard to see how such an argument could be supported. Moreover, since the epistemological model in question is independently plausible in the case of perceptual belief, and since there are relevant similarities between the perceptual and the religious case, hence it seems that the defender of religious belief has made significant headway against the sceptic. The legitimacy of perceptual belief is rescued from the threat posed by an evidentialism-based scepticism and thus, by parity of reasoning, so is religious belief.27 4. Problems with the Parity Argument I think that this general externalist defence of religious belief on these grounds of parity is persuasive. As so often in philosophy, however, the devil is in the detail, since the problems arise once one opts for a specific externalist construal of the epistemology of religious belief. In particular, I want to focus upon the most common account that runs along these lines that is due to Plantinga. His claim is that we should treat religious belief as being “properly basic” in just the same way that perceptual belief is treated as properly basic on the externalist account just considered. This claim of 27

For two interesting discussions, and overviews, of the kind of parity arguments employed by reformed epistemologists, see Penelhum 1986; cf. Penelhum 1983; 2000, and McLeod 1993. In the former work, Penelhum also points out that Plantinga’s early characterisations of the reformed epistemological position e.g. Plantinga 1983 were misleading since they tended to imply that grounds were being offered to support religious belief over other sorts of non-religious belief. This is not the case, however, since all that is achieved is a kind of advantageous impasse with the religious sceptic (though this is achievement enough). I think that in later work Plantinga is more explicit about this point, though see DeRose 1999b for more discussion of this worry in the light of Plantinga’s more recent work, in particular, Plantinga 2000.

191 proper basicality comes down to the contention that, although one might have evidence in support of one’s religious or perceptual beliefs, such beliefs (at least in favourable cases) do not stand in need of an evidential grounding in order to be properly held. Provided that certain external conditions are met, a perceptual or religious belief can enjoy an immediate warrant that arises directly out of the agent having an experience of a certain sort, rather than being a transferred warrant, or partly transferred warrant, which is dependent upon an evidential grounding.28 This line does more than merely rescue religious belief from an evidentialism-based scepticism, however, it goes further to actually allow, as with perceptual belief, that evidence need play no substantive warranting role when it comes to certain religious beliefs. It is this claim that religious beliefs can be properly basic in this way that I think is questionable. The problem with this construal of the epistemology of the religious belief is that it overstates, in relevant respects, the parallels between religious experience and perceptual experience. In particular, there is a worry about the putatively analogous “spontaneity” of religious and perceptual beliefs. It was noted above that religious beliefs, like perceptual beliefs, can sometimes seem to have the same sort of “directness” that one might find in the perceptual case, as if one is directly responding to a religious being in the way that one directly responds to objects in the physical world via perception. It was this spontaneity of perceptual belief that made it apt for a radically non-evidentialist construal since evidence seemed to play no essential warranting role as regards standard perceptual belief. The problem, however, is that whereas this sort of “directness” is the norm in the perceptual case, it is more naturally thought of as the exception to the norm in the religious case. Indeed, whereas perceptual beliefs seem to be, in the main, “forced” upon us, religious beliefs often seem to be formed in a far less immediate and compelling fashion. As Keith DeRose29 has put it, normal religious belief is rarely understood in terms of being “zapped” by a divinity, as Plantinga seems to understand it; instead, the more common way of conceiving of such belief is in terms of being “nudged” or “invited” towards a certain sort of doxastic commitment. In general, there does seem to be a certain ‘voluntary’ element in re-

28

Following Plantinga 1993a I will understand “warrant” to be that epistemic notion that is sufficient, with true belief, for knowledge. 29 DeRose 1999a.

192 ligious belief that is absent in most forms of perceptual belief.30 Relatedly, whereas an agent might gain a warrant for his perceptual belief without engaging his reflective capacities at all, religious beliefs seem to directly implicate such capacities. For example, whereas we are happy to attribute perceptual knowledge to small children, we are unwilling, I think (except in rare cases), to ascribe anything but the most basic of religious knowledge to persons who lack reasonably developed reflective capacitates. This hints towards the fact that well-formed religious beliefs seem to demand more of the subject than well-formed perceptual beliefs. Of course, for the parity argument to go through all that is needed is that some religious beliefs are basic in the same way that some perceptual beliefs are; it is not essential that religious belief be in general analogous to perceptual belief. Indeed, Plantinga’s own discussion of the role that defeaters can play regarding basic religious beliefs would appear to indicate that the religious beliefs held by most reasonably sophisticated religious believers are, for the most part at least, non-basic.31 Nevertheless, the concern about the disanalogies between religious belief and perceptual belief is important because it directs us to look again at the idea that, on grounds of parity, we should consider religious beliefs to be basic. After all, if we concede that properly formed religious belief may be constrained by more imposing demands than properly formed perceptual belief, then it becomes far from clear that it follows from the fact that certain perceptual beliefs are basic that any (or hardly any) religious beliefs are basic as well. It is open to reformed epistemologists to query these putative intuitions, of course, or at least question the epistemological ramifications that they are meant to hold. Rather than engaging in such dialectical warfare, however, a better approach might first be to see whether the general re30

Note that by these remarks I am not committing myself to some sort of “doxastic voluntarism” thesis. Rather, I am merely highlighting the fact that, insofar as we have control over our beliefs at all, then it is part of our ordinary conception of our doxastic capacities that we have more control over the formation of the standard religious belief than we do regarding the formation of the standard perceptual belief. Moreover, it may seem as if I am also endorsing some sort of general thesis to the effect that belief control is a direct result of the belief being formed in a “non-spontaneous” way. This is not the case. My remarks here are solely confined to the cases of religious and perceptual belief. 31 Though not necessarily. A basic belief that has been subject to a defeater which, in turn, has been defeated by a second defeater, will return to being properly basic. For an excellent discussion and overview of the issue of the relationship between basicality and defeaters, see Miller 2004.

193 formed epistemology framework couldn’t simply be modified, in nonessential respects, to enable it to accommodate these disanalogies. For if it can then the dialectical warfare is unnecessary and the claim that there are these disanalogies has constructive, rather than destructive, consequences for the view. One motivation for pursuing the irenic goal of integrating this concern into the reformed epistemology thesis rather than attempting to argue it down, is that the “intuitions” that drive this concern find expression in the work of one of the most prominent reformed epistemologists, Alston. Not only has Alston noted these disparities between perceptual and religious experience but, as a result, he has advocated a far more cautious nonevidentialist epistemology as regards religious belief than one finds in Plantinga. In particular, he argues that religious experience alone cannot suffice to warrant one’s religious beliefs, contending that further transferable epistemic support must be sought from other sources, such as salient historical evidence and other relevant evidence that is socially transmitted from other members of the agent’s religious community.32 That Alston is willing to even consider such a move offers prima facie grounds for thinking that perhaps the basicality thesis does not play quite such a central role in the reformed epistemology stance than is often thought. I think that this is right, and that, properly understood, the driving motivation for reformed epistemologyand which enables it to evade scepticism about religious beliefis its commitment to epistemological externalism. This point is in keeping with a claim made earlierthat, at root, it is internalism rather than evidentialism that drives scepticism about religious belief. Accordingly, the rejection of evidentialism contained within reformed epistemology is not nearly so important as the move away from epistemological internalism and towards epistemological externalism. As a result, I will be arguing that provided this externalist element of the view is retained then one can forge a version of the thesis that can accommodate these disanalogies. Note that in what follows I will allow that there may be some religious beliefs that are basic in the appropriate sense. What I will be claiming is that the reformed epistemological stance is not hostage to the existence of these beliefs since, even if no religious belief ever, in fact, meets the basicality rubric, the general elements of the view can still be regarded as secure and capable of evading the sceptical attack.33 32

See, in particular, Alston 1982; 1986. DeRose 1999a suggests that the reason why Plantinga is not as sensitive to the disanalogies between religious and perceptual belief is that he has a very specific model

33

194

5. DeRose’s Foundherentist Proposal Before we consider how this modified reformed view is to work, it is worthwhile looking at a suggestion made by DeRose that runs along similar lines. DeRose (1999a) argues that we can respond to these disanalogies between perceptual and religious belief by understanding religious belief in terms of the sort of foundherentist model advocated by Susan Haack (1993; 1995). One effect of adopting this line of response is that it leads to the rejection of the kind of foundationalism that is implicit within Plantinga’s version of reformed epistemology as it currently stands. In particular, the foundherentist idea as it applies to Plantinga’s account is precisely not to treat religious beliefs as properly basic at all, except perhaps in rare cases. This may not be quite such a dramatic change as it might at first appear, however, for two reasons. First, because foundherentism does incorporate some key foundationalist insights and therefore can allow a key sense in which religious beliefs enjoy direct epistemic support. Second, because foundherentism can allow that perhaps some religious beliefs are properly basic in unusual circumstances. Accordingly, one could plausibly view this foundherentist proposal as merely restricting the class of religious beliefs that the parity argument has application to without thereby discounting this manœuvre altogether. First, however, we need to get an idea of what is involved in foundherentism. In essence, it is meant to be an epistemological model that can capture insights from both foundationalist and coherentist schools of thought in epistemology. Consider the following passage from Haack: Foundherentism is an intermediate theory which (unlike coherentism) allows the relevance of experience but (unlike experientialist foundationalism) requires neither privileged beliefs justified exclusively by experience nor an essentially onedirectional notion of evidential support.34

of religious experience in mind which is much more akin to perceptual experience, standardly understood. If this is so, then the dialectical moral to be drawn from the modified version of reformed epistemology offered here is not that this model should replace Plantinga’s own (since he is welcome to model his conception of religious belief in any way he pleases), but rather that this model generalises the core insights within Plantinga’s account to make that view applicable to a broader range of religious experience. 34 Haack 1993: 113

195 The thought is thus that foundherentism retains a core thesis from both foundationalism and coherence theory. On the one hand, it retains the foundationalist idea that some beliefs could enjoy sufficient positive epistemic status without that status resting upon the epistemic status of any other beliefs (such that they are properly basic). On the other hand, it retains the coherentist idea that the positive epistemic status of some beliefs can be the result solely of the coherence of this belief within an appropriate set of other beliefs. The way this is achieved is by arguing that whilst some beliefs enjoy a “direct” epistemic support which is not based on other beliefs, other beliefs enjoy a completely indirect ‘transferred’ epistemic support which is the result of its relation to other beliefs, and still other beliefs gain some of their epistemic support in the direct fashion and some of it in the indirect fashion. Haack often describes this position in terms of the metaphor of a crossword puzzle, where the clues stand for experience. It could well be that certain clues directly point to a certain answer (epistemically supported belief), and thus that the gaining of this answer is not dependent upon any other answers that one might already have. This would thus represent the standard perceptual case where our perceptual beliefs directly gain a sufficient positive epistemic status without that status being dependent upon the epistemic status of any other beliefs we hold. In contrast, other clues won’t suggest any particular answer, but one could determine an answer by looking at the way that certain possible answers ‘cohere’ with the answers already gained. A “real-world” example of such a case might be a trial where the jury must form an opinion on the basis of the evidence in front of them. An epistemically supported judgement in this case may be entirely the result of some reflective process whereby one comes to recognise that a certain alternativethe not-guilty verdict, saycoheres with the evidence presented in a far more adequate way than the opposing guilty verdict. These are two extreme cases, however, and, more usually, our belief formation falls between these two poles. Employing the crossword metaphor again, we can say that often a clue directs us towards a small selection of answers, and that our final determination of the correct answer is dependent upon our forming a judgement about which of these possible answers best coheres with the other answers that we already have. This thus corresponds to cases where we form beliefs which enjoy some degree of direct positive epistemic support, but where this support does not take one to the threshold necessary for warrant. In such cases the epistemic support

196 in question needs to be augmented with further epistemic support from one’s other beliefs if it is to meet this threshold. An example of this might be a case where there is some ambiguity present in experience, as when one is unsure whether the person that one sees in the distance is one’s brother or one’s father. Here, we might call on further beliefs that we hold (that he’s too tall to be one’s father for example) to form our final (warranted) judgement. Foundherentism thus offers a spectrum of possibilities, from completely direct non-inferential warrant at one extreme to completely indirect transferred warrant at the other extreme, with various degrees of combination of these two alternatives in between.35 The advantage of employing this model is that it can accommodate the thought that religious belief enjoys some measure of direct epistemic support (perhaps even a sufficient measure in certain rare cases), whilst also allowing that, in general, this support is insufficient by itself to warrant the religious beliefs in question. What must be added is thus further epistemic support from other beliefs to bring the positive epistemic status up to the required threshold for warrant. This account thus allows us to mark the contrast with perceptual belief (which tends to generally enjoy sufficient direct epistemic support) in a way that accords with the intuition that religious belief is formed in some ways that are analogous with perceptual belief.36 In particular, it allows that one’s religious beliefs do enjoy some measure of positive epistemic status that is direct and therefore not dependent upon the epistemic status of any other beliefs that one holds. In this sense the foundherentist modification of the reformed epistemology thesis is just as resistant to the sceptical challenge we witnessed earlier on. For so long as the religious belief is indeed formed in the right kind of circumstances then, even though the agent might lack good reflectively accessible grounds for his belief, it can still have, contra the sceptic, a sig35

The other advantage of this crossword metaphor is that it can offer a vivid description of belief-change, even where those beliefs were previously taken as being very secure. As any crossword enthusiast will tell you, even the most compelling of “answers” to a particular clue can start to look shaky if evidence against it begins to build in the form of several not so compelling answers that will not fit with it. 36 As DeRose has pointed out to me (in correspondence) his point is actually slightly different from this in that his emphasis is on how religious belief is lacking in indirect warrant relative to perceptual belief. The key point about there being a disanalogy here between religious and perceptual belief stands either way, of course, but I retain this particular reading of the foundherentist claim because I think that this understanding of the difference between perceptual and religious belief best captures the disanalogies noted earlier in §4. I am grateful to DeRose for helping me to be clear on this point.

197 nificant degree of positive epistemic status. Again, then, we find an externalist thesis emerging and it is this externalism that is doing the work of undermining the sceptical challenge. Furthermore, this proposal can account for why it is that certain sorts of religious belief seem to be preferable to others. After all, if coherence in one’s religious beliefs can contribute to the epistemic status of those beliefs, then it is little wonder that an inchoate set of religious beliefs would seem to be lacking in rationality. This approach therefore enables us to distinguish between properly held religious belief and the religious belief held by the “enthusiasts” that Locke talked of.37 Evidence thus does have a central role to play in religious epistemology after all; it is just not quite as central as the evidentialist contends. That is, the account sketched here attempts to offer a compromise view between evidentialism and the radical anti-evidentialism espoused by Plantinga. Whereas the evidentialist sees epistemic support as being entirely concerned with evidential considerations and Plantinga views evidential considerations as only relevant once we have moved away from basic religious beliefs (as happens, for example, when the believer is exposed to a defeater for his basic belief), this modified reformed account contends that evidential considerations are nearly always relevant but usually only in concert with other epistemic support that is direct and non-evidential. In this way evidential considerations can be accorded a role in the determination of a religious belief’s epistemic support without this role thereby inviting the usual evidentialism-based sceptical challenge. Finally, this account explains why it is that we rarely ascribe religious knowledge to small children, even though we are often happy to ascribe to them perceptual knowledge. The reason is that the former sort of knowledge requires the agent to reflect on the relevant evidential considerations and therefore implicates certain reflective capacities in a way that perceptual knowledge does not. I think that this is an intriguing proposal, but as it stands it does not offer us quite what we are looking for. The reason for this is that the very schematic foundherentist account under consideration merely presents us with an appealing description of the structural nature of the epistemology that we want rather than going further to distinguish between the various sorts of epistemological analyses that might fit this template and adjudicat-

37

Insofar, of course, as their beliefs really are lacking in coherence.

198 ing between them.38 Accordingly, in the next section I will try to show how a very specific sort of epistemological positiona form of virtue epistemologycan be put into service to account for the epistemology of religious belief. This account will fit the framework offered by foundherentism whilst also being contiguous with the core formulation of reformed epistemology due to Plantinga that is also conceived along (broadly speaking) virtue-theoretic lines. 6. A Virtue-Theoretic Proposal The suggestion that reformed epistemology is best understood along virtue-theoretic lines may not at first seem particularly novel because, at least in the case of Plantinga, reformed epistemology is already regarded by some as being a form of virtue epistemology. I want to argue, however, that reformed epistemology, where it is understood as a virtue-theoretic account, is not conceived of in terms of the right virtue-theoretic account. Moreover, the account that I propose will fit the foundherentist template outlined above. What makes an epistemological account virtue-theoretic is that it is agent-based rather than belief-based. In particular, a belief counts as knowledge only if it is the result of, as John Greco puts it, “an agent’s cognitively virtuous character”. We can, I think, get a better handle on this distinction between belief-based and agent-based epistemology by considering some of the earliest forms of virtue-theoretic proposals in epistemology which were expressed in terms of agent reliabilism. This kind of proposal has been put forward by, amongst others, Ernest Sosa and Alvin Goldman. Moreover, it is the sort of account that Plantinga himself offers and which he explicitly applies to religious belief.39 The basic idea behind agent reliabilism is that we need to amend the key process reliabilist accountas expressed by, for example, (an earlier) 38

Indeed, it is important to remember that the type of foundherentism that we are dealing with here is just Haack’s basic schematic account of the position rather than the particular variant of that position which she goes on to outline. Note that this is no criticism of DeRose since his suggestion was only meant to be structural in the first place. 39 For the main accounts of virtue epistemology, see Sosa 1985; 1991; 1993; Montmarquet 1987; 1993; Plantinga 1988; 1993b; 1993c; 2000; Kvanvig 1992; Goldman 1993; Greco 1993; 1999; 2000; Hookway 1994; and Zagzebski 1996. See also the excellent survey article by Axtell 1997.

199 Goldman40 – along virtue-theoretic lines in order to meet some of the standard challenges to the view. For example, one of the problems that reliabilism faces is that it seems to count certain beliefs as being warranted even when they are formed via processes which, whilst reliable, are clearly not knowledge-conducive. There are three main examples of this ilk. For now I will concentrate on two of them.41 The first type of counterexample to reliabilism is concerned with those reliable processes where the success of the process does not seem to reflect any cognitive achievement on the part of the agent. One could imagine, for example, that an agent reliably forms true beliefs about a certain subject matter solely because some benevolent demon makes it the case that his beliefs in this regard are reliable. In this case we have strong intuitions that knowledge does not result because the agent is not tracking the world in the relevant sense (instead “the world” appears to be tracking the agent’s beliefs). And given that there is no cognitive achievement on the part of the agent, it does not seem right to say that his reliable true beliefs can count as knowledge. The second type of counterexample concerns certain “malfunctions” on the part of the agent which, nonetheless, actually enable the agent to reliably form true beliefs about a certain subject matter.42 Because the reliability is due to a malfunction, however, we have a strong intuition that it cannot count as being knowledge-conducive. Of course, committed reliabilists could respond to both of these examples by modifying the view in subtle respects, but such a move would fail to pay due attention to the heart of the difficulty here. This is that reliabilism goes wrong in only considering certain properties of the belief rather than focussing instead upon properties of the agent who formed that belief. In particular, the agent reliabilist thought is that not just any reliable belief-forming process can produce knowledge, but only those processes 40

e.g., 1986. All three of these examples are discussed, in one form or another, in Plantinga 1993a. For a general discussion of Agent Reliabilism, see Greco 1999. It is important to note that there are a number of other advantages to adopting a virtue-theoretic epistemology that I do not have the space to expand upon here. For example, such proposals seem to be able to adequately respond to both Gettier-type scenarios and sceptical arguments. Moreover, a virtue-theoretic epistemology may also be able to meet the socalled ‘generality’ problem that has bedevilled reliabilist accounts of knowledge. 42 The example that Plantinga 1993a offers of a reliable cognitive malfunction is an epistemically “lucky” brain lesion that enables the agent to form true beliefs about his condition in this respect. 41

200 that perform certain appropriate roles within the cognitive character of the agent. Agent reliabilists therefore argue that instead of defining knowledge purely in terms of properties of the belief in question, one should instead focus upon the stable natural cognitive traits, or faculties, of the agent. Paradigm examples of such traits are our senses which, if they are working correctly and in a stable manner relative to the appropriate environmental conditions, will lead us to true beliefs. Plantinga, for example, characterises his version of this kind of thesis in terms of a “cognitive design plan” as follows: A belief B has warrant for S if and only if the relevant segments (the segments involved in the production of B) are functioning properly in a cognitive environment sufficiently similar to that for which S’s faculties are designed; and the modules of the design plan governing the production of B are (1) aimed at truth, and (2) such that that there is a high objective probability that a belief formed in accordance with those modules (in that sort of cognitive environment) is true; and the more firmly S believes B the more warrant B has for S.43

What is important about such cognitive faculties is that they are more than just (in the ideal case) reliable. In Plantinga’s account just cited, for example, the reliability of these faculties is merely a necessary condition for knowledge gained through them. What is needed to make such reliability knowledge-conducive is the further claim that this reliability arises out of a kind of stable cognitive excellence, or virtue, that the agent exhibits and which he can therefore take credit for.44 This conception of knowledge in terms of natural stable cognitive faculties meets the two objections outlined above by explaining why the agent lacks knowledge despite exhibiting a reliable belief-forming process. In the first example, the agent cannot take any credit for his true beliefs, and thus have beliefs which count as knowledge, because they are not due to any trait of his, let alone a stable natural belief-forming process which exhibits a kind of cognitive excellence. As regards the second example, the agent reliabilist account explains why cognitive malfunctions can never 43

Plantinga 1993b: 19. It is important to note that not everyone agrees that Plantinga’s account is a virtuetheoretic one, including Plantinga himself (who prefers to call his position a “proper function” thesis). I think, however, that Plantinga’s account has enough in common with the views expressed by the key virtue theorists to fall into this camp. For more discussion on this point, see the exchange between Sosa 1993 and Plantinga 1993c, and the commentary by Axtell 1997, 3-4. 44

201 give us knowledge. For not only are such processes usually lacking in the required stability, they are also not natural cognitive faculties either. No wonder, then, that they are unable to provide us with knowledge. An agent reliabilist thesisa thesis which employs an understanding of the notion of a cognitive virtue in terms of natural cognitive facultiesis thus able to meet certain objections to basic process reliabilism by restricting the conception of what is to counts as a knowledge-producing reliable process in ways that focus upon attributes of the agent. Henceforth, I will refer to such early virtue epistemological accounts as faculty virtue theories. In essence, where they differ from reliabilist accounts is merely in their stress on the agent’s cognitive character and their focus on natural cognitive faculties of the agent. Such faculty virtue accounts are ideally suited to capturing perceptual knowledge because it seems entirely uncontentious in the perceptual case to view knowledge as being purely the result of cognitive faculties functioning correctly in the right circumstances. Note that to conceive of perceptual knowledge in this way is to adopt the kind of “pure” externalist epistemology that was discussed above. There is no demand here that the agent need bring his reflective capacities to bear in order to exhibit knowledge; instead, he need only meet purely external conditions in order to know. As one might expect given his use of the parity argument, Plantinga directly employs the kind of faculty virtue account he offers for perceptual beliefwhat he calls a “proper function” accountto religious belief. In line with the Calvinist tradition to which he belongs, Plantinga argues that we have an innate natural cognitive facultya sensus divinitatiswhich enables us to form reliable religious beliefs and that, so long as this cognitive faculty is functioning correctly (so long as, for example, it is not adversely affected by the agent’s sin), then one can gain religious knowledge in the same unmediated fashion that one gains perceptual knowledge. Perceptual beliefs and religious beliefs can thus both be thought of, at least in the standard case, as properly basic. We have already seen, however, that the nature of religious experience, in contrast to perceptual experience, is such that the application of a “pure” externalist epistemology to religious belief is suspect. Accordingly, Plantinga’s agent reliabilist account of knowledge, when applied to religious belief, will have the unfortunate consequence that an agent could come to know certain religious propositions via the exercise of a particular sort of cognitive faculty even though he does not bring any of his reflective capacities to bear upon the formation of his belief in these propositions. As

202 we noted above, however, for the vast majority of religious beliefs (if not all of them) this seems entirely unintuitive. Properly formed religious beliefs are not usually simply formed as a direct response to certain stimuli, as one might naturally think in the case of properly formed perceptual beliefs, but instead seem to standardly invoke certain reflective capacities on the part of the subject.45 Of course, this observation alone does not entail that the virtuetheoretic approach in general is suspect. Instead, all it shows is that those versions of the virtue-theoretic account that cannot leave room for our reflective capacities to play an essential role in the acquisition of certain sorts of knowledge are problematic. If there are other virtue-theoretic accounts configured along similar lines that can allow these capacities to play the required role, then the virtue-theoretic model will be back in business. I think that there are such models available. The kind of accounts that I have in mind are those virtue theories that do not concentrate solely on the faculty virtues but also incorporate a role for reflective virtues. Such virtues may include such cognitive traits as the ability to weigh-up evidence impartially, or the ability to integrate one’s beliefs so as to gain, for example, a greater degree of doxastic consistency. This general line of thought has its roots in the distinction that Sosa46 makes between “brute” or “animalistic” knowledge and “reflective” knowledge, although it expands upon this basic distinction by allowing that between these two extremes there can be various types of knowledge which demand different combinations of “brute” and “reflective” cognitive virtues. The thought is that by combining both faculty and reflective cognitive virtues one attains a more fine-grained account of what is involved in knowledge possession in different cases. That is, certain sorts of knowledge, such as perceptual knowledge, might just require properly functioning cognitive faculties, whereas other sorts of knowledge, such as that which can result from abstract reasoning for example, might solely depend upon the reflective virtues. In between, one will find the vast majority of knowledge that requires a mixture of both properly functioning faculty virtues and reflective virtues.47 45

For an insightful, and more general, critique of Plantinga’s agent reliabilism, see Zagzebski 1996, §3.5. 46 e.g., 1991. 47 One finds virtue-theoretic accounts which emphasise the importance of reflective virtues in recent work by Greco 1993; 1999; 2000 and Zagzebski 1996, although the stress in Greco’s account is more on what he terms “subjective justification” than on

203 Such an account has a number of attractive features. For example, one could use such a theory to explain why certain types of knowledge are regarded as being more “refined” than others. It may be, for example, that two knowers both meet the threshold for knowledge but that one of them surpasses that threshold in important epistemically relevant respects, perhaps because he exhibits a certain sort of understanding of what he knows which is lacking in the case of the other agent (and which is not simply the result of knowing more truths than the other agent). This would thus be a case where the “refinement” in question was due to the activation of a reflective capacity on the part of the subject. A further advantage of this sort of virtue-theoretic approach is that it can meet the third type of counterexample that is often made against process reliabilism that I alluded to earlier. This counterexample concerns situations where one reliably gains true beliefs by forming beliefs in ways which seem, antecedently at least, to be undesirable. For example, one could imagine a scenario in which one forms beliefs about a certain subject matter on the basis of bias and yet, because of some stipulated feature of the circumstances in which the beliefs are acquired, beliefs formed on this basis turn out to be reliable. Seemingly, however, one cannot gain knowledge through bias, no matter how reliable one’s beliefs are. This example should sound familiar because it is a variation of the example offered earlier in our critique of evidentialism. There it was noted that evidentialism (in at least some of its forms at any rate) is unable to make any adequate distinction between those agents who believe what they epistemically ought to believe for all the wrong reasons (such as because of bias), and those who believe what they epistemically ought to believe for all the right reasons (because of a sensitivity to the weight and extent of their evidence, for example). Again, then, we have a situation in which an agent has met the relevant epistemic rubric that has been set, but has nevertheless met it in a way that seems to preclude that agent from possessing knowledge. Agent reliabilism lacks the resources to respond to examples of this sort because there does not seem to be any way in which one can trace the reflective virtues as such. Moreover, although the accounts offered by Greco and Zagzebski are in this respect similar to that sketched here, they do tend to go further to make the activation of such reflective capacities necessary for knowledge. In contrast, my claim is much weaker in that I allow that in certain casessuch as in the perceptual casean agent might have knowledge without exhibiting any reflective capacities at all.

204 cognitive shortcoming in question back to the agent’s cognitive faculties where these are understood in non-reflective terms. The more developed form of virtue epistemology under consideration here has no such difficulties, however, because it can explain the agent’s lack of knowledge in terms of a failure to exhibit the appropriate reflective virtue. In this case, for example, the agent forms his beliefs in terms of bias (a reflective vice) rather than in response to the weight and extent of his evidence (as reflective virtue would dictate). The same goes for the evidentialist variant of this example. If one characterises knowledge in terms of true belief that arises out of a cognitive virtue, then those agents who believe what they epistemically ought to believe for all the wrong reasons lack knowledge precisely because that true belief (if it is true) does not arise out of a reflective virtue. The attraction of applying such a thesis to religious knowledge should be clear. For example, a reflective virtue epistemology of this type can allow that whilst it might be true in the perceptual case that one can gain knowledge without exhibiting any developed reflective capacities, the same need not also be true in the case of religious knowledge. This view can thus do justice to our intuition that, at least in the standard case, the role of the faculty virtues alone as regards religious belief is insufficient to afford us religious knowledge. Instead, a precondition of acquiring religious knowledge (at least in the standard case) will be that the agent has (and brings to bear) the appropriate reflective virtues as well. An account of this sort is therefore able to capture the reformed intuition that religious knowledge is similar to perceptual knowledge in certain respects, in that both forms of knowledge presuppose that the agent has certain properly functioning faculty virtues, whilst also allowing that this analogy is not complete and thus that some epistemological explanation of the disanalogies present here should also be given. These disanalogies are accounted for in terms of the role played by the reflective virtues in the acquisition of religious knowledge.48 7. Concluding Remarks In effect, what we have done here is show how knowledge in general can 48

Indeed, it may be that we need to incorporate not only reflective virtues into this account but also moral ones as well, in that religious knowledge is the sort of knowledge which might directly implicate certain virtuous moral traits. For more on this point, see Zagzebski 1996.

205 be adequately understood in terms of a virtue-theoretic account that employs both faculty and reflective virtues, and then further shown that such an account can accommodate the nature of religious belief, especially as regards the respects in which it differs from perceptual belief. Moreover, since we are retaining the key externalist thought that knowledge need not be dependent upon the agent being able to adduce sufficient non-questionbegging evidence, we are no longer troubled by an evidentialism-based sceptical argument concerning religious belief. It is still true that the legitimacy of an agent’s religious belief is not solely dependent upon that agent being able to adduce evidence appropriate and proportionate to the belief in question. All that is different about this account as opposed to Plantinga’s “properly basic” account is that if an agent’s religious belief is to enjoy a warrant then that agent must exhibit the relevant reflective virtues as well, and this will typically involve the ability to adduce some appropriate degree of evidence. Far from being a sceptical hurdle for religious belief to clear, however, this condition can actually serve to distinguish responsible religious belief from the undisciplined religious belief of Locke’s “enthusiast”. Provided that this reflective condition is met then, ultimately, whether or not an agent’s belief is actually warranted will depend upon whether the appropriate ‘external’ facts obtain, just as in the perceptual case. Note also that this suggestion is entirely in the spirit of the foundherentist proposal made by DeRose, since a natural way of modelling the different types of knowledge here, whether faculty-virtue-based, reflectivevirtue-based, or a mixture of the two, is in terms of a foundherentist structure. What we have done is merely add specific content to the basic foundherentist structure by offering a particular epistemological proposal that is structured along these lines that is both independently plausible and which can accommodate important features of religious experience and the belief that it gives rise to. Moreover, since the virtue-theoretic proposal offered here is merely an extension of the sort of earlier virtue-theoretic account offered by Plantinga, it ought to be in the spirit of the general reformed epistemological approach. Adapting our understanding of the parity argument thus gives rise to a reformed conception of reformed epistemology that is able to meet at least one of the key problems that the unreformed version faces.49 49

An earlier version of this paper was presented at “The Epistemology of Basic Belief” conference, Vrije Universiteit, Amsterdam, The Netherlands, in June 2001. Thanks to the audience that day and, in particular, to Andrew McGonigal, Christian

206

REFERENCES Alston, W. P. 1982. “Religious Experience and Religious Belief”, Noûs 16: 3-12. ———. 1986. “Is Religious Belief Rational?” in The Life of Religion, edited by S. M. Harrison and R. C. Taylor, 1-15. Lanham, Maryland: University Press of America. ———.1991. Perceiving God: The Epistemology of Religious Experience. Cornell, Ithaca: Cornell University Press. Axtell, G. 1997. “Recent Work on Virtue Epistemology”, American Philosophical Quarterly 34: 1-26. BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge, Massachusetts: Harvard University Press. Chisholm, R. 1989. Theory of Knowledge, third edition. Englewood Cliffs, New Jersey: Prentice Hall. Clifford, W. K. 1879. Lectures and Essays, edited by F. Pollock. London: Macmillan. DeRose, K. 1999a. “Are Christian Beliefs Properly Basic?” Unpublished manuscript. ——— .1999b. “Voodoo Epistemology.” Unpublished manuscript. ——— . 2000. “Ought We to Follow Our Evidence?” Philosophy and Phenomenological Research 60: 697-706. Feldman, R. and Conee, E. 1985. “Evidentialism”, Philosophical Studies 48: 15-34. Feldman, R. 2000. “The Ethics of Belief”, Philosophy and Phenomenological Research 60: 667-96. Goldman, A. 1986. Epistemology and Cognition. Cambridge, Massachusetts: Harvard University Press. ——— . 1993. “Epistemic Folkways and Scientific Epistemology”, Philosophical IsMiller, Alvin Plantinga, and to the chair, Hunter Davies. Thanks also to Michael Brady, Keith DeRose and an anonymous referee for this volume who read and commented on an earlier draft of this paper.

207 sues 3:271-84. Greco, J. 1993. “Virtues and Vices of Virtue Epistemology”, Canadian Journal of Philosophy 23: 413-32. ——— . 1999. “Agent Reliabilism”, Philosophical Perspectives 13: 273-96. ——— .2000. Putting Skeptics in Their Place: The Nature of Skeptical Arguments and Their Role in Philosophical Inquiry. Cambridge, UK: Cambridge University Press. Haack, S. 1993. “Double-Aspect Foundherentism: A New Theory of Empirical Justification”, Philosophy and Phenomenological Research 53: 113-28. ——— . 1995. Evidence and Inquiry: Towards Reconstruction in Epistemology. Oxford: Basil Blackwell. Hookway, C. 1994. “Cognitive Virtues and Epistemic Evaluations”, International Journal of Philosophical Studies 2: 211-27. Kornblith, H. 2001. Epistemology: Internalism and Externalism (editor). Oxford: Basil Blackwell. Kvanvig, J. 1992. The Intellectual Virtues and the Life of the Mind. Savage, Maryland: Rowman & Littlefield. Locke, J. 1979. Essay Concerning Human Understanding, edited by P. H. Nidditch. Oxford: Oxford University Press. McLeod, M. S. 1993. Rationality and Theistic Belief: An Essay on Reformed Epistemology. Ithaca, New Jersey: Cornell University Press. Miller, C. B. 2004. “Defeaters and the Basicality of Theistic Belief.” This volume. Mitchell, B. 1973. The Justification of Religious Belief. Philadelphia: Temple University Press. Montmarquet, J. 1987. “Epistemic Virtue”, Mind 96: 487-97. ——— . 1993. Epistemic Virtue and Doxastic Responsibility. Lanham, Maryland: Rowman & Littlefield. Newman, J. H. 1844. Sermons, Chiefly on the Theory of Religious Belief, Preached Before the University of Oxford. London.

208 ——— . 1985. An Essay in Aid of a Grammar of Assent, edited by I. T. Ker. Oxford: Oxford University Press. Nielsen, K. 1967. “Wittgensteinian Fideism”, Philosophy 42: 237-54. Penelhum, T. 1983. God and Skepticism. Dordrecht, Netherlands: Reidel. ——— . 1986. “Do Religious Beliefs Need Grounds?”, Nederlands Theologisch Tijdschrift 40: 227-37. ——— . 2000. “Reflections on Reformed Epistemology.” Unpublished manuscript. Phillips, D. Z. 1976. Religion Without Explanation. Oxford: Oxford University Press. Plantinga, A. 1980. “The Reformed Objection to Natural Theology”, Proceedings of the American Catholic Philosophical Association 49-62. New York. ——— .1981. “Is Belief in God Properly Basic?”, Noûs 15: 41-51. ——— .1983. “Reason and Belief in God” in Faith and Rationality, edited by A. Plantinga and N. Wolterstorff, 16-93. Notre Dame, Indiana: University of Notre Dame. ——— .1988. “Positive Epistemic Status and Proper Function”, Philosophical Perspectives 2: 1-50. ——— . 1993a. Warrant: The Current Debate. New York: Oxford University Press. ——— . 1993b. Warrant and Proper Function. New York: Oxford University Press. ——— . 1993c. “Why We Need Proper Function”, Noûs 27: 66-82. ——— . 2000. Warranted Christian Belief. New York: Oxford University Press. Plantinga, A. and Wolterstorff, N. 1983. Faith and Rationality, editors, Notre Dame, Indiana: University of Notre Dame. Pritchard, D. H. 2000. “Is “God Exists” a “Hinge” Proposition of Religious Belief? International Journal for Philosophy of Religion 47: 129-40. ——— . 2002a. “McKinsey Paradoxes, Radical Scepticism, and the Transmission of Knowledge across Known Entailments”, Synthese 130: 279-302. ——— . 2002b. “Recent Work on Radical Skepticism”, American Philosophical Quarterly 40: 215-57.

209 Sosa, E. 1985. “Knowledge and Intellectual Virtue”, The Monist 68: 224-45. ——— .1991. “Intellectual Virtue in Perspective” in Knowledge in Perspective: Selected Essays in Epistemology, E. Sosa. Cambridge, UK: Cambridge University Press. ——— .1993. “Proper Functionalism and Virtue Epistemology”, Noûs 27: 51-65. Swinburne, R. 1979. The Existence of God. Oxford: Oxford University Press. ——— . 1996. Is There a God?. Oxford: Oxford University Press. Wittgenstein, L. 1966. Wittgenstein’s Lectures and Conversations on Aesthetics, Psychology and Religious Belief, edited by C. Barrett. Oxford: Basil Blackwell. Wolterstorff, N. 1991. “The Migration of the Theistic Arguments: From Natural Theology to Evidentialist Apologetics.” in Contemporary Classics in Philosophy of Religion edited by A. Loades and L. D. Rue, 71-92. La Salle, Illinois: Open Court. ——— . 1995. Divine Discourse: Philosophical Reflections on the Claim that God Speaks. Cambridge, UK: Cambridge University Press. ——— .1996. John Locke and the Ethics of Belief. Cambridge, UK: Cambridge University Press. Wright, C. 1985. “Facts and Certainty”, Proceedings of the British Academy 71: 429-72. ——— . 2000. “Cogency and Question-Begging: Some Reflections on McKinsey’s Paradox and Putnam’s Proof”, Philosophical Issues 10: 140-63. Zagzebski, L. 1996. Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge, UK: University of Cambridge Press.

CHRISTIAN WEIDEMANN

Why basic theistic belief is probably not warranted, even if it is true I Most people who believe in the existence of God do not base their belief on propositional evidence. They do not arrive at their religious convictions by way of deductive or probabilistic argument. The beauty of nature, the experience of guilt or gratitude, mystical practices, or reading the Holy Scriptures induce them to believe that there is a powerful and kind being who has created the world, loves us etc. But there is usually just as little ground for saying that religious people infer the existence of God from experience as for claiming that somebody looking out of the window and seeing a tree infers the existence of the tree from his sense data. In general, religious people simply find themselves with the belief that God exists. In most cases theistic belief comes to mind immediately. So there are without much doubt basic theistic beliefs. But are they also properly basic? According to Alvin Plantinga a belief p can be properly basic for a person S in at least two senses (Plantinga 2000, 175-180). It is properly basic with respect to deontological justification, roughly, if and only if 1) S holds p in the basic way, 2) S has examined carefully and at length all serious objections against p she knows, and 3) S sees no sufficient reason for believing that her cognitive faculties involved in the production and examination of p do not function properly or are not aimed at truth. But the critic’s concession that in this sense theistic belief can be properly basic is of course nothing a theist should count on too much. Unfortunately there are many implausible, pathological, and even incoherent beliefs which can be and often are properly basic with respect to justification, too. For example a child who believes that the Great Pumpkin returns every Halloween may be perfectly justified in doing so. A schizophrenic who suffers from persecution mania is perhaps not flouting any epistemic duty in believing that a big black monster is shadowing him. He simply can’t do better. Although he may be perfectly justified in his beliefs, he

212 needs help. To show that theistic belief is often justified in the above sense is therefore obviously not enough to answer the de iure-objection to theism. However, according to Plantinga, a belief p can also be properly basic for a person S with respect to warrant, roughly, if and only if 1) S holds p in the basic way, and 2) p is produced by cognitive faculties functioning properly in a congenial epistemic environment according to a design plan successfully aimed at truth. It is warrant, not justification, that constitutes knowledge: S knows p, if and only if S believes p, p has a certain degree of warrant for S, and p is true.1 Of course, several objections could be raised to externalistic theories of knowledge in general, and to Plantinga’s conception in particular. However, for the sake of argument I will assume that Plantinga’s externalism is at least close to the truth. (By the way: I actually think it is). In the following I shall focus on the question whether theistic belief is warranted in the Plantingian sense: Is theistic belief produced by cognitive faculties functioning properly in a congenial epistemic environment successfully aimed at truth? Plantinga’s own answer is: It depends. If there is no God, it is very likely that theistic belief has little or no warrant. It would indeed be hard to contradict this claim: a world without a God, in which, nevertheless, a cognitive faculty successfully aimed at truth, that is a cognitive faculty that predominantly produces true beliefs, causes the false belief that God exists, seems quite odd, to put it mildly. But how do things stand, if there is a God? According to Plantinga, in this case the epistemic probability of theistic belief’s being warranted is very high. Therefore the problem whether theistic belief is warranted can be reduced more or less to the problem whether theistic belief is true. The de iure question is inextricably linked with the de facto question. Because the latter cannot be answered philosophically, there is also no philosophical solution to the former. Plantinga’s argument for the high probability of theistic belief’s being warranted, given the existence of God, can be reconstructed as follows (cf. Plantinga 2000, 188-190):

1

Plantinga develops and defends this conception of knowledge in the second part of his warrant-trilogy (Plantinga 1993).

213 (1) (2) (3)

(4)

(5)

If there is a God, he is omnipotent, omniscient and morally perfect. [premise1] If God exists, to know and to love him is the greatest human good. [premise 2] If God exists, it is true that a) he knows, due to his omniscience, that the greatest human good is knowledge and love of God, b) wants, due to his morally perfection, that we are able to achieve that good, and c) makes sure, by means of his omnipotence, that we are able to achieve it. [follows deductively from / is highly probable on (1) and (2)] If God exists, the cognitive processes, which in fact produce belief in God, are aimed by God at producing that belief in order to provide us with knowledge of him. [highly probable on (3)] If God exists, the cognitive processes, which in fact produce belief in God, are functioning properly in a congenial epistemic environment according to a design plan successfully aimed at truth, that is: if God exists, theistic belief is warranted. [highly probable on God’s omnipotence (1) and (4)]

II Before checking whether this argument is valid, let us take a quick glance at its premises. Premise (1) is a definition. It states some of what is involved in the main Western tradition’s understanding of God. There are of course theologians who think this definition is inadequate on biblical grounds, and there are philosophers, for example Spinozists, Whiteheadians or Wittgensteinians, who defend a different understanding of God. With regard to the religious beliefs of these authors and their adherents, nothing follows from Plantinga’s argument, even if it were valid. But without a doubt many and perhaps most people do share premise (1). Only their beliefs are at stake here. Premise (2) seems more controversial. According to Søren Kierke2 gaard , for example, knowledge of God is anything but the greatest human good. Quite the reverse, knowledge of God amounts to the end of faith and the death of religion. Essential for Christian faith is not only the belief in God’s incarnation, but also that the believer holds that this belief is highly improbable, an inexplicable paradox, an obvious absurdity. If anyone knew 2

See especially his Concluding Unscientific Postscript.

214 God, he would have a policy dictated by prudence, not a faith. The value of religious commitment is therefore totally dependent on its irrational character. For reasons that will become clear later, I think a grain of truth is hidden in this objection. Nevertheless it is no serious threat to premise (2), because it is just incomprehensible that someone should believe in a certain fact, orientate his whole life towards this fact, and at the same time defend its alleged absurdity. If someone does not welcome arguments which confirm her belief that God exists, if she is not interested in knowing that God exists, the suspicion seems justified that she also is not really interested in God himself. The act of faith has become more important than its object. This, however, would be nothing but a very subtle form of idolatry. Thus, knowledge of God is, like almost all knowledge, a human good. Is it also the greatest good? If there is a God, he created the immense universe, knows everything that is possible to know, is morally perfect, and the key to an eternal life. It is hardly possible to imagine a good that would be greater than the acquaintance with such a Being and the knowledge of its love. III So the premises stand, I suppose, but what can we say about the conclusions? Consider (3). Does (3) follow deductively from the conjunction of (1) and (2)? That an omniscient God knows the greatest human good is a necessary truth. There is also little doubt that, if an omnipotent God intends that we be able to be aware of his presence and to know something about him, he can bring it about. But is it also a necessary truth that a morally perfect being wants to make possible the greatest human good? I do not think so.3

3

A scholastic distinction may be helpful here. We can discriminate two wills in God. With his antecedent will (voluntas antecedens) he considers only the intrinsic value of things, while with his consequent will (voluntas consequens) he takes all relevant considerations into account. It seems rather unproblematic to claim that God necessarily wants the greatest human good with his antecedent will. But unfortunately only God’s second, consequent will is crucial for his actions. Therefore, whenever I speak of something God wants, I refer to his consequent will. For the two wills in God see for example Thomas Aquinas, Summa Theologiae Ia, 19. 6, ad.1; G.W. Leibniz, Theodicy, I 22-23.

215 For example, it is logically possible that the realization of the greatest human good would destroy higher non-human goods, and that therefore God refuses to realize it.4 Is (3b) at least highly probable on (1) and (2)? That’s a very difficult question. However, it seems clear to me that Plantinga’s treatment of the problem of evil makes it impossible for him to answer that question consistently in the affirmative. Admittedly, we do not have any idea, which higher good could be destroyed by our knowledge of God. But does it really follow from our ignorance that probably there is no such good? In some articles on the socalled evidential problem of evil Alvin Plantinga has stressed an old argument, whose modern version was developed especially by Stephen Wykstra (Wykstra 1984, 1996; Plantinga 2000, 458-481)5. Plantinga points out that from the fact that we often cannot see any reason which would morally justify God in permitting certain evils, it does not follow that probably there is no such justifying reason. In addition to this, Plantinga admits explicitly the realization of higher non-human goods, for example angelic ones, as a possible excuse for God. In comparison with God’s omniscience our knowledge is extremely slight. Therefore, we are not able to infer reliably God’s plans from our limited knowledge. I confess that I never found that kind of argument very convincing (for my reasons see e.g. Gale 1996), but if its consequence is true, and Plantinga clearly believes it is, it follows that the probability of (3b) is simply inscrutable, given the truth of (1) and (2). From the fact that we cannot see any reason which would morally justify God in preventing the greatest human good, it does not follow that there probably is no such reason.

4

5

A Kantian moral philosopher may argue that our moral reality is such that using a person solely as a means, rather than an end, is absolutely forbidden. But if there is such an absolute ban and if we assume that moral perfection is part of God’s nature, so the Kantian might continue, it seems to follow that God necessarily wants the greatest human good, because it would not be morally permissible for him to keep us in ignorance for the sake of some other good, extraterrestrial, celestial or whatever. Kantian ethics may indeed save (3b), but I heavily deny that its adoption is a serious option for traditional theists like Plantinga. If God’s moral obligations really are to be constructed among Kantian lines, traditional Christian apologetics is doomed to failure from the beginning. It is obviously hopeless to find an answer to the problem of evil, if the truth of Kantian ethics is presupposed. Wykstra’s first article (Wykstra 1984) set off an extensive and fruitful discussion, see especially Howard-Snyder 1996, which includes two articles by Plantinga.

216 The only way out of this difficulty, so it seems to me, is the revision of premise (1). The supposition of a minimal theism does not suffice for our purposes; we need a more extended version. Given such a version, we need neither infer deductively nor probabilistically from God’s moral perfection that God wants to reveal himself to us, or that he wants to enable us to know him. According to this extended version of theism, self-revelation is itself part of God’s nature. This moderate extension of premise (1) is not yet very problematic. It is true, in principle, that the more content premise (1) has, the less relevant the argument will become, but at least the eschatological self-revelation of God is not an esoteric addition to theism. Quite the reverse, it belongs, pace some deistic philosophers, to the core of Christianity and Islam. But unfortunately a similar difficulty emerges with regard to the argument’s second step. If we take Wykstra’s objection seriously, the probability of (4) on (3), is again neither high nor low, but inscrutable. It may well seem to us that, if God exists, basic theistic beliefs are warranted, because we can’t see any reason why God should deceive us. But again one might ask: does it really follow from our ignorance that probably there is no such reason? A second extension of premise (1), that remedies this difficulty, will be at any rate less innocent than the first one (see section VII) IV However that may be, let us suppose for the sake of argument that the probabilities of God’s actions are not inscrutable. Let us also assume that (3) is highly probable on the conjunction of (1) and (2). Have we saved Plantinga’s argument? What shall we say about its next step? Is (4) highly probable on (3)? As far as I can see, (4) at least does not follow probabilistically from (3), if we also take into account some background knowledge concerning the widespread atheism and general religious pluralism of today. To begin with, it is important to notice why there is no necessary correlation between (3) and (4). Firstly, it does not follow from the fact that God wants to reveal himself to us, when he will do so. Perhaps it is part of the conditio humana that we are not able to know anything about God, until we die and leave this world. Perhaps God will install a sensus divinitatis only eschatologically, or he will create only eschatologically a congenial epistemic environment in which our innate sensus divinitatis can begin to produce warranted be-

217 liefs. According to such a scenario, during our earthly life we are at best able to discern the possibility of God’s existence, but not whether he really does exist. Secondly, from the fact that God wants to reveal himself to us, it does not follow either in which way he has done or will do so. Perhaps at least during our earthly lives it is only possible to know God’s existence after a process of considerable length, through decades of careful reflection, the painstaking study of theological and philosophical works, the participation in discussions and congresses, but by no means in the basic way. If one of these scenarios is true, the cognitive mechanisms which produce basic theistic beliefs either do not function properly or are not successfully aimed at truth. But isn’t it at least highly probable, given that many people do believe basically in the goodness and power of God, that God, if he exists, has chosen this way to let them know something about him? It is true that anyone who denies the proper basicality of theistic belief will have to explain why there are basic theistic beliefs at all. But it is also true that anyone who claims that theistic beliefs are often properly basic will have to explain something else: why do not all human beings produce properly basic theistic beliefs in this life? Why does not the almighty and morally perfect God enable all of us to know him? Furthermore, it must be explained why religious people often believe contradictory things of God. Plantinga puts the exclusiveness and variability of theistic convictions down to the noetic consequences of original sin (Plantinga 2000, ch.7). The cognitive faculty, the sensus divinitatis, which would enable us to know God’s existence, is seriously disturbed since the human fall. The work of the Holy Spirit compensates or repairs this damage, but only selectively and partly. So the crucial question is this: what is, given the existence of God, more plausible: that the omnipotent and morally perfect Designer holds all of us collectively responsible for the sins of our progenitors without regarding our respective individual guilt? That he withholds the greatest human good from many of us, while he gives it freely to others, again obviously without regarding individual merits or the degree of individual guilt? Or is it more plausible, given God’s existence, that many people falsely pretend to know God’s existence in the basic way? I am strongly inclined to take the latter view: not to be among the elect, to be without the grace of God, is an enormous evil that according to the doctrine of original sin is often suffered without any special amount of personal guilt. The arbi-

218 trariness of election is, at least according to human standards, unjust. Even if the doctrine of original sin were consistent with the existence of an omnipotent and morally perfect being, as far as I can see, nobody could sensibly expect its truth. On the other hand, to overestimate the epistemic status of basic theistic beliefs is a minor evil that often happens through one’s own fault and can be overcome by one’s own efforts. It cannot sensibly be denied that human beings often display forms of wishful thinking, that many people tend to project ideal properties on romantic or political heroes. But if such phenomena in the field of health, love or politics, do not render the existence of God improbable, why should similar phenomena do so in the field of religion? God’s goals in permitting unwarranted theistic beliefs may be quite varied. Perhaps he wants to lead us to the right questions, to an adequate object of thought and hope. And at the same time he may try our humility, modesty, and tolerance. Perhaps some of those who claim to know God’s existence do not pass that test. It might well be true that the latter scenario is not very probable, given the existence of God, but at any rate in my eyes it is far more plausible than the alleged noetic consequences of original sin and the Holy Spirit’s unequal work of repair. Here is a counter-objection that comes quickly to mind: It is true, so one might argue, that according to the doctrine of original sin, only some people are blessed, but according to the second model the situation is even worse: nobody knows God in the basic way. So there seems to be no reason at all for claiming that the latter scenario should be more probable than the doctrine of the noetic consequences of original sin, given the existence of an omnipotent and morally perfect God. There is something in that. However, what makes it so difficult to believe in the noetic consequences of original sin and the selective work of the Holy Spirit is not the idea that only some are blessed in this life, but rather that God treats people, who are equally guilty or not guilty, in quite different ways, that he shows his love and grace so arbitrarily, that at least according to the Augustinian version many are damned forever without any chance to change their fate. It is true, of course, that many human goods, like health, intelligence, beauty or wealth are distributed extremely unequally, but shouldn’t we expect that, if the greatest human good is at stake, everyone’s chances are approximately equal? Only if we accept the view that, given God’s existence, nobody knows God basically in this life, but everyone will know him after he or

219 she will have died, our hope in the infinite and impartial love and justice of God can remain unequivocal. V But let us set aside that line of argument for the moment and suppose instead that the doctrine of original sin offers a somewhat plausible explanation of the fact that in spite of the sensus divinitatis, there is so much scepticism and religious pluralism in the world. Is a defender of Plantinga’s probability thesis now at least on a par with his opponent? Again, I do not think so. That is because until now we have completely ignored God’s possible reasons for choosing a certain time and a certain way of selfrevelation. However, God has a very strong motive for making knowledge of himself not too easily available. Such knowledge would heavily restrict and perhaps even destroy human freedom of choice between the good and the bad, respectively between the good and the wrong. Suppose someone has an overwhelmingly clear vision of God. Can she still freely choose to oppose God, to reject his love and the offered eternal salvation, and instead dedicate her life to cheap pleasures or even terrible vices? As far as I can see, such a thing would not only be psychologically implausible, but even metaphysically impossible. A free and rational being, who knows God’s existence beyond all reasonable doubt, and therefore knows that the loving communion with God is the greatest good, cannot intentionally do anything that would make it impossible to achieve that good. But doesn’t this contradict biblical thought? Didn’t Adam and Eve oppose God’s will in spite of their full knowledge of his existence and nature? No, not at all. On the contrary, the human fall would not have happened without ignorance. Adam and Eve falsely believed that they would become as powerful as God by eating the apple. They falsely believed that they were able to hide themselves from the wrath of God. They mistook the omniscient and omnipotent God for an ordinary king or landlord. They did not really know God. And what shall we say about the devils who believe in God and tremble? It is true that they know God and nevertheless do wrong, but they don’t do so freely. They can’t help it, because they are devils. They are not free to do otherwise. Existentialist writers like Dostojevskij show in a psychologically plausible way that man’s nature is rebellious, that if you demonstrated him

220 with mathematical exactness the way to individual happiness or to the Golden Age of mankind, he nevertheless would necessarily sooner or later not only violate the principles of reason and morals, but deliberately contravene his own benefit. However, this only seemingly indicates that it is possible to act against that which one believes to know as the greatest good. Dostojewskij’s underground man talks of the most advantageous advantage of all advantages: not to be a passive part of the hustle and bustle of the world and to demonstrate instead one’s own capricious will. For him exactly this is the greatest good. However, who really knows God also knows that the communion with God is a good nothing can surpass. Admittedly, in order to know God’s existence, it is not necessary to hold that belief with the maximum degree of firmness or even with Cartesian certainty. The belief in question need not have the same strength as, for example, my belief that 2+2=4, or that here is a desk in front of me, or that the world is older than five minutes. It is also true, unfortunately, that human beings do not always behave perfectly rational. Thus, one might argue that the less than maximum degree of theistic beliefs, in conjunction with irrational influences on the believer, can sometimes lead to selfdeception. Therefore, so it seems, someone who knows God is at least free to resist the tendency towards self-deception or to give way to it (Schellenberg 1993, ch.5). Let me introduce an example. Imagine a person named Paul. Paul knows that life is worth living and amongst other things likes eating fruit. Unfortunately he is allergic to strawberries. Every time that he eats strawberries, he gets terribly sick for some days. Is Paul nevertheless free to eat strawberries? Of course he is. We all know people, I guess, who regularly drink an enormous amount of alcohol, although they know which result their drinking will show the next morning. Obviously those people are prepared to pay a price for their drinking or they successfully repress all memory of the drinking’s after-effect. But now suppose Paul’s allergy has become even worse. Several specialists told him that the next time he ate strawberries he would inevitably die. Paul believes this diagnosis. Now imagine that someone is offering Paul a strawberry: Is Paul free to choose between eating and not eating it? From my point of view it would be at least very odd to say so. Perhaps, due to the closeness to death, Paul is somewhat attracted by the strawberry, perhaps he feels a little tremble, similar to the fascination we feel when we are looking down an abyss. He may even have an appetite for strawberries. But Paul, of course, is not really wondering whether to eat or to refuse the

221 strawberry. He is not fighting an inner struggle. To eat a strawberry is no longer a live option for him.6 Imagine a third situation. Paul suffers from incurable cancer. He knows that he won’t live longer than a few weeks and that the rest of his life will be full of pain. Is Paul now free to choose between eating strawberries and refusing to do so? Certainly he is. The good Paul would loose by eating strawberries is much less than before his cancer, and perhaps Paul even has come to the conclusion that a life with incurable cancer would be no good at all. Consider a final case. Paul is in serious trouble. He is the only survivor of an air crash. The pilot managed to make an emergency landing. It could have been a great deal worse, so one might think. But Paul’s problem is that all this happened in the middle of Sahara Desert. The plane burned out completely, and two days later still no help is in sight. The only thing which has not been destroyed in the fire, is a big box of strawberries. Is Paul free to eat one of these? Paul’s situation looks desperate. He has nothing to eat or drink. He does not know when or even if help will arrive. The more hungry and thirsty he gets, the more attractive the box of strawberries becomes in his eyes. Increasingly he will be in danger of losing his rational judgement. Perhaps he will begin to make himself believe that the allergy specialists were mistaken, or that his hypersensitivity was cured miraculously in the meantime, or that he will die soon anyway with or without eating the strawberries and so on. To resist these temptations Paul undoubtedly will have to fight an hard inner struggle. Now let us return to theistic beliefs. Suppose a person named Carl. Carl believes in God’s existence, and also believes that to humiliate, to torture, or to murder other people would separate us from God. He holds the latter belief with the same strength as Paul holds the belief that eating strawberries would kill him. Is Carl free to choose between humiliating etc. other people and refusing to do so? What we have learned from the strawberry-case is that at least one of the following two requirements must be fulfilled for that: Either Carl believes from the beginning that the good he would lose by humiliating other people is not greater than the good he might gain by doing so. Or Carl feels a very strong, irrational inclination or

6

It is important to emphasize that Paul knows that he would inevitably die after eating strawberries. Why, for example, does not the threat of death penalty prevent murder? Because it is possible that culprits remain unpunished. Why do not statistics concerning lung cancer stop nicotine consumption? They are just statistics!

222 impulse to humiliate or murder other people and therefore begins to deceive himself as to the good he would lose by doing so. The first requirement cannot be fulfilled. Carl knows God’s existence, by assumption, and therefore knows that the communion with God is the greatest human good, a good that might even be incommensurable. But to humiliate your neighbour means to oppose God’s will and therefore to destroy the communion with God, in a perhaps irreversible way. What about the second requirement? Is it logically possible that there is an irrational impulse or inclination which would be strong enough to trigger off a mechanism of self-deception in Carl? It is possible, but such an impulse would have to be enormously strong. Because the good Carl would lose by humiliating or murdering other people is immense, his inclination to humiliate others has to be even stronger than, for example, our inclination to eat strawberries after some days of hunger and thirst in Sahara Desert.7 As far as I can see, in our world there is no such natural inclination, leave alone as a general phenomenon. However, if there were a strong impulse to do evil, that certainly would be a bad thing. A world with such a general inclination would ceteris paribus be worse than a world without it. Here we are confronted with a situation Richard Swinburne called „logical straitjacket“ (Swinburne 1999, 203-212; see also Swinburne 1979, 211-212; Hick 1966, ch.6). It is good that we are able to choose between good and wrong and to mould gradually our character, it is an even greater good to know God, and it is also good, if we do not have an innate or inherited inclination to evil. Unfortunately even for an omnipotent being it is impossible to realize all these three goods at the same time. But he can bring them about successively. The natural thing to think is therefore that God, if he exists, created us without an innate inclination to do evil, but with the faculty of free choice between good and wrong, so that we are able to form our character on our own; and that then, after a long process of person-making, God will enable us to know him. This will happen, either after we will have thought a lot about God and the richness of nature, have studied theological and philosophical works, or simply after we will have died.

7

It is exactly this kind of interrelation that John Schellenberg underestimates in his interesting book on divine hiddenness (1993: 108-130).

223 VI But didn’t we overlook something? If strong belief in God’s existence limits the human freedom of choice, that fact does not depend on whether such a belief constitutes knowledge or not. So, if God is interested in human freedom of choice, the natural thing to expect is that he not only prevents knowledge of him, but also unwarranted belief in him. However, given that he really exists, he obviously does not do so. Therefore, if God exists, either human freedom of choice is not such a great good, or even strong belief in God cannot considerably limit it. The weakness of this objection can easily be seen by a simple analogy. Does the existence of murder show that murder is not a bad thing, that God does not abhor it? Of course not. Most analytic philosophers of religion, including Plantinga, hold that God has to take the incalculable risk (Swinburne) – or even to accept the foreknown existence (Plantinga) – of moral evil, if he wants to create beings capable of free choice. Paradoxically God values human freedom so highly that he even permits the free decision to destroy it. We are able to deprive others of their free will by torture, indoctrination, brainwashing, drugs etc., and we are also capable of destroying or limiting our own freedom of choice. One way of doing so, I have claimed, is strong basic belief in God. The limitation of freedom by basic belief in God is not, of course, irreversible. A believer who grasps the connection between human freedom and knowledge of God, who notices the spreading of religious pluralism and atheism, and recognizes the implausibility of the doctrine of original sin and its alleged cognitive consequences, should no longer believe that theistic convictions are properly basic, and should thus regain her freedom of choice between good and wrong. Let us consider a more promising objection: It belongs to the core of the great theistic religions that at least some persons, – Moses, the Apostles, Mohammed – did have basic knowledge of God. They taught and transmitted it, and the holy texts give an account of its content. Therefore, if we, while we are reading or listening to these texts, accept their testimony in the basic way, we obtain new convictions which, via warrant transfer, are properly basic, too. Thus, even if there is no general sensus divinitatis and no work of the Holy Spirit, there can be, due to religious testimony, properly basic theistic belief. There is not sufficient space here to deal exhaustively with the epistemic status of testimony, but I shall outline my position. For simplicity’s

224 sake I will concede two points to the objector: 1) God enabled some protagonists of the heilsgeschichte to obtain basic knowledge of him. 2) Under suitable circumstances testimony can be a source of warranted basic belief. But I reject the view that with regard to the holy texts the circumstances are suitable. Someone who believes in the assertions of a text that is thousands of years old and genuinely written in a foreign ancient language cannot sensibly claim that she knows these assertions on the basis of testimony alone. She needs information about author, time of origin, philology, tradition, historical background, other archaeological or literary sources etc. Testimony is a source of warranted basic belief only in a congenial epistemic environment. With regard to ancient texts we are obviously not in such an environment. A modern scholar who claims to know, for example, ancient Greek and Roman history exclusively on the basis of Herodotus and Livius, makes a fool of herself. Therefore, if one wants to ascribe a high degree of warrant to basic beliefs formed on the testimony of ancient texts which are supposed to be holy, it is not sufficient to refer to the epistemic value of testimony simpliciter, but one has to bring in the work of the Holy Spirit or something similar that guarantees the authentic tradition, translation, and understanding of the teachings of Moses, Jesus, St. Paul, or Mohammed. Traditional (Christian) theology indeed has done and still does so (cf. Plantinga 2000, 249-252). However, there are not only fierce controversies about the genuine form, the translation, and above all the interpretation of the holy texts, they also obviously contradict each other. The various theistic religions do not agree which texts are holy. And many people do not even believe that there are holy texts at all. Hence, if there is the Holy Spirit or something similar at work, it intervenes only selectively. But, as we have already seen, there is little ground for believing in the selective work of the Holy Spirit or any other similar divine power, even given the truth of theism. Therefore, I conclude, Plantinga’s probability thesis cannot be saved by reference to religious testimony either. VII There still seems to be a last, desperate option open to the defender of Plantinga’s argument: further extension of premise (1). Let us assume additionally not only that, given God’s existence, God wants to reveal himself to us, but also that there is an innate sensus divinitatis, that mankind suffers from transworld depravity, that the first human sin has had far-

225 reaching noetic consequences for all following generations, and that the Holy Spirit repairs some of these consequences in a selective and seemingly arbitrary way. In this case it would indeed be true that, if God exists, basic theistic belief is probably warranted. But at the same time Plantinga’s argument would be devoid of any philosophical interest, because after the above extension of premise (1) only a triviality would remain: if, given that God exists, the omnipotent and omniscient Holy Spirit wants to produce true basic beliefs about God in some of us, while he keeps the rest suffering from the noetic consequences of original sin, it is not very surprising that, given God’s existence, the objective epistemic probability of basic theistic belief’s being warranted is high and cannot be diminished by reference to the widespread atheism and religious scepticism in the world. But such reasoning would not only come close to simple tautology, it would also presuppose a special form of Christian theism, which is, of course, absolutely unacceptable for Jews and Muslims and at least highly problematic for many Christians as well. Especially the doctrine of original sin and its noetic consequences, so it seems to me, is not very popular among Christian believers of today. Hence, the price a defender of Plantinga’s argument has to pay for the extension of premise (1) is twofold: the argument becomes trivial and the conclusion loses its relevance for those people who reject the new assumptions concerning original sin and the Holy Spirit’s unequal work of repair. By the second point a standard objection to Plantinga’s project is reinforced: many critics have complained that Plantinga has little to say to atheists, to adherents of other religions, and to people whose Christian beliefs are rather weak.8 But if my considerations are correct, Plantinga’s main argument in Warranted Christian Belief is of no interest for many devout Christians either, as long as they do not share certain dubious assumptions concerning original sin etc. VIII The unsatisfactory defensive attempt above set aside, and given a notion of theism which is not too special, Plantinga’s probability thesis fails: basic theistic belief is probably not warranted, even if it is true. To make myself clear: this does not imply that theistic belief is probably not warranted at all. There is non-basic propositional evidence which confirms theistic be8

See most recent Swinburne 2001; see also Plantinga’s rejoinder (Plantinga 2001).

226 lief. Whether this evidence can be sufficient for knowledge is an open question. I personally believe that it is at least sufficient to justify hope for God.9

9

I am indebted for stimulating discussion to Andreas Hansberger, Peter Rohs, Hartmut Sitzler and the audience at the Basic Belief Conference, especially Alvin Plantinga. I am grateful to Konrad Jocksch, Sabine Roeser and René Van Woudenberg for helpful comments on earlier versions of this paper.

227 REFERENCES

Gale, Richard. 1996. “Some Difficulties in the Theistic Treatment of Evil” in HowardSnyder 1996: 206-218. Hick, John. 1966. Faith and Knowledge, 2d ed. Ithaca. Howard-Snyder (ed.), Daniel. 1996. The Evidential Argument from Evil. Bloomington . Plantinga, Alvin. 1993. Warrant and Proper Function. New York: Oxford. ——— . 2000. Warranted Christian Belief. New York: Oxford. ——— . 2001. “Rationality and Public Evidence: a Reply to Richard Swinburne”, Religious Studies 37: 215-222. Schellenberg, John L. 1993. Divine Hiddenness and Human Reason. Ithaca, London. Swinburne, Richard. 1979. The Existence of God. Oxford. ——— .1999. Providence and the Problem of Evil. Oxford. ——— .2001. “Plantinga on Warrant”, Religious Studies 37: 203-214. Wykstra, Stephen J. 1984. “The Humean Obstacle to Evidential Arguments from Suffering: On Avoiding the Evils of Appearance”, Journal of Philosophy of Religion 16: 7394. ——— . 1996. “Rowe’s Nosseum Arguments from Evil” in Howard-Snyder 1996: 126150.

C: Morals, Testimony, and Proprioception

SABINE ROESER

Defending Moral Intuition I. Introduction In the last decade or two, ethical intuitionism has experienced a revival after a long period in which moral philosophers would take any occasion to declare that they are ‘of course not intuitionists’. Present-day intuitionists such as Jonathan Dancy are mainly inspired by W.D. Ross. This leads Dancy to define intuitionism as a combination of cognitivism, non-reductive moral realism and pluralism of morally relevant features. However, historically this is not the best definition of intuitionism. As a matter of fact, there have been intuitionists long before the 20th century-intuitionists Ross, A.C. Ewing, H.A. Prichard and G.E. Moore, namely in the 18th and 19th century,1 and these early intuitionists did not all defend a form of pluralism.2 Rather, what connects all intuitionist accounts up to and including Moore, Prichard, Ross and Ewing is cognitivism, non-reductive moral realism and foundationalism.3 Foundationalism says that our moral knowledge is based on basic moral beliefs, or in other words, intuitions. It is this aspect of intuitionism that is responsible for its name. A complete re-establishment of intuitionism has to incorporate a defense of the notion ‘intuition’ in its original sense, as used by the classical intuitionists. In this paper I will defend this aspect of intuitionism. In sections II and III, I will analyze how intuition functions as a foundation for moral reasoning. To this purpose I will mainly discuss the ideas of Thomas Reid. In section IV, I will analyze how the notion intuition can help to explain how we can bridge the is-ought gap and how we can assess morally complex situations. This account will be based on the early twentieth century-intuitionists Ross, Prichard and Ewing. 1

These intuitionists are for example Hutcheson, Clarke, Price, Reid et.al. in the 18th century and Whewell and Sidgwick in the 19th century. For a historical introduction into intuitionism, see Hudson 1967. 2 For example, Henry Sidgwick defends a form of monism: the most basic moral principle is that of utilitarianism; Henry Sidgwick 1901: 442. 3 Cf. Hudson 1976: 1, and Brink 1989: 102.

232 II. Foundationalism and basic moral beliefs To give the reader a flavor of what intuitionists mean by ‘intuition’, let us look at the following passage by Thomas Reid: When men’s faculties are ripe, the first principles of morals, into which all moral reasoning may be resolved, are perceived intuitively, and in a manner more analogous to the perception of sense than to the conclusions of demonstrative reasoning.4

According to Reid, each epistemic justification is based on first principles or, to use contemporary terminology, basic beliefs.5 Basic beliefs are 1. a necessary foundation of our reasoning and 2. self-evident, which means that they are not justificatorily based on other beliefs. We know basic beliefs by intuition and all other beliefs are based on those. Foundationalists distinguish two classes of beliefs: there are basic beliefs and there are non-basic beliefs. Basic beliefs are ‘not accepted on the evidential basis of other beliefs’.6 Rather, a person is justified in believing a basic belief by the content of the belief itself or because it is a belief about our sensory experiences, whether one is in pain and so on. A non-basic belief is justified in so far as one has propositional evidence for such a belief from other beliefs one has. Propositional evidence can be deductive, inductive and abductive.7 Candidates for basic beliefs can be perceptual beliefs and rational beliefs, for example mathematical beliefs (insight in axioms) or other forms of a priori knowledge. As in general foundationalism, intuitionists argue that if there were no basic moral beliefs that we are justified in holding by themselves, then that would mean that all beliefs had to be justifiable by other beliefs which would then lead to an infinite regress or to circular reasoning. In the following subsections I will discuss important aspects of foundationalism in intuitionism.

4

Thomas Reid 1969b: 727. In the following, I will mainly use the contemporary terminology of ‘basic belief’, but when I quote Reid the reader should realize that this is what he means by ‘first principle’. For Reid a particular belief can also be a ‘first principle’, namely by being a foundation for other beliefs. 6 Plantinga 1993: 177. 7 Plantinga 1993: 177, 178. 5

233 A. Intuitive beliefs cannot always be justified by argument A central feature of foundationalism in intuitionism is that basic moral beliefs can very often not be argued for. Says Reid: It is a first principle in morals, that we ought not to do to another, what we should think wrong to be done to us in like circumstances. If a man is not capable of perceiving this in his cool moments, when he reflects seriously, he is not a moral agent, nor is he capable of being convinced of it by reasoning.8

Moore says that basic or self-evident beliefs are the kinds of beliefs that are ‘incapable of proof or disproof’.9 Other beliefs are based on intuitions, but basic beliefs do not need to be based on other beliefs.10 All intuitionists agree that basic moral beliefs are not like theorems that can be proven by reasoning or deduction, rather, other beliefs are based on them. There are propositions that we can apprehend without judging them to be true or false. Selfevident propositions, though, as it were, force a judgment upon us. We cannot but consider them true or false. I call these first principles, because they appear to me to have in themselves an intuitive evidence which I cannot resist. I find I can express them in words. I can illustrate them by examples and authorities, and perhaps can deduce one of them from another; but I am not able to deduce them from other principles that are more evident.11

According to Reid, if compared with mathematical beliefs, basic moral beliefs resemble mathematical axioms more than theorems. Hence, says Reid, just as mathematical axioms cannot be proven, neither can basic moral beliefs.12 According to Reid, a ‘clear and intuitive judgment, resulting from the constitution of human nature, is sufficient to overbalance a train of subtle reasoning on the other side’.13 This means that, according to the intuitionists, basic beliefs are not inferior to reasoning and argumentation. 8

Reid 1969a: 234, 235. Moore 1988: x. 10 But they can be, cf. section IIB. 11 Reid 1969b: 369. 12 Reid 1969a: 721, 722, 727. 13 Reid 1969b: 409. This role of intuitions is similar to the role of Michael DePaul’s notion of ‘formative experiences’, De Paul 1993. 9

234 Instead, moral intuitions are a separate source of evidence that can be even stronger and more convincing than a complex argument. The fact that self-evident beliefs cannot always be proven means that if somebody does not accept a basic belief after having had appropriate experiences,14 we cannot convince him of this belief by reasoning. We can at most try to make him see a situation in a certain light, but we cannot convince him using arguments based on more fundamental premises that would establish the belief at stake as conclusion, because there simply are not any more basic beliefs involved. Somebody who is not able to see the truth of certain basic moral beliefs is in the end not a moral agent, his moral faculty15 is not functioning properly (e.g. in the case of a sociopath). On the other hand, Reid allows for the possibility that somebody who is not able to make real moral judgments can also be motivated to behave morally out of egoism. Just as we can see in Hobbesian theories such as rational-choice theory, the assumption of egoism can lead to the adoption of certain moral norms. Intuitionists would say that, although this can be helpful to convince otherwise morally blind people to behave morally, such convictions are not genuine moral judgments. Compare this with colorblind people who can discern colors by the shades of gray they perceive. Although these people are able to make the appropriate color distinctions, we would hold that they do not perceive colors in the genuine sense.16 B. We can also intuit non-basic propositions Philosophers often assume that self-evident principles can by definition not be proven or argued for. However, Robert Audi argues that it might be the case that different moral principles that are understandable by themselves and are in that sense self-evident might be derivable from an even more fundamental moral principle, namely the Kantian categorical imperative which demands respect for persons and to regard them as ends and never merely as means.17 One could then say that this principle is the most basic principle and hence it cannot be argued for. Some derived beliefs are 14

I will come back to this in section IIC. The notion ‘moral faculty’ is one of the controversial concepts that lead philosophers to reject intuitionism. However, first of all, only intuitionists up to and including Reid spoke in terms of a moral faculty (or conscience or the moral sense). Secondly, they all meant by it our ability to make moral judgments, such as reason, emotion or both, and not some mysterious additional sense organ. 16 This example can be found in different versions in the literature, for example, in Reid 1969b: 232, McNaughton 1988 and Little 1995: 117-137. 17 Audi 1997: 47, 48. 15

235 justification-wise overdetermined, meaning that they are self-evident and also derivable from more basic beliefs. Indeed, according to Reid, intuition serves not only to cognize axioms but also theorems. Reid argues that the ‘system of morals’ is not so much like mathematics but rather like botany and mineralogy: where the subsequent parts depend not for their evidence upon the preceding, and the arrangement is made to facilitate apprehension and memory, and not to give evidence.18

In his discussion of Reid’s moral philosophy, Keith Lehrer puts this as follows: In such a system, there would be no epistemic priority of theorems to axioms. The system would be a collection of propositions, a taxonomy of moral propositions, which, though they stand in various logical relations to each other, and are not independent, do not obtain their evidence from deduction. They posses it immediately.19

Hence, although the different moral principles are connected by logical relations, the theorems can be as much intuited as the axioms. Even if it is possible in principle to derive certain moral beliefs from basic moral beliefs, we can also see the truth of derivable beliefs or theorems directly. According to Reid, this is apparent from the fact that the ability to reason is no guarantee for being virtuous. Reasoning can also be abused to further one’s own interest instead of following one’s duty. Furthermore, somebody who is not good at reasoning can still have correct moral insights. Hence, moral knowledge does not always depend on reasoning but is more analogous to perception.20 According to Reid all human beings can make moral judgments without complicated arguments and deductions. Whether it concerns general moral principles or concrete moral judgments, every human being is in principle as able as any other to see what is right or wrong, provided that their moral faculty is sufficiently developed, which leads me to the next point.

18

Reid 1969b: 376. Lehrer 1989: 238. 20 Reid 1969a: 727. 19

236 C. Our ability to moral insight needs to be developed Reid emphasizes that we must not think that because man has the natural power of discerning what is right, and what is wrong, that he has no need of instruction; that this power has no need of cultivation and improvement...21

To be able to have justified moral intuitions, it is a precondition that our moral faculty is sufficiently developed. According to Reid, all of our cognitive faculties have to be developed to be able to generate beliefs in a reliable way. Reid thinks that where, for example, vision and hearing develop at an early age, our conscience is the most sophisticated of all our cognitive faculties and hence needs the most time to develop. All our cognitive faculties are innate faculties, which only means that we have the potential to use them. To activate conscience, exercise and education are needed.22 By this Reid does not mean merely intellectual education, but moral education, which can be provided by all virtuous people in all social groups. An appropriate education and experience will help us to be able to make intuitive moral judgments. Although Reid thinks that a proper education is a necessary condition to understand moral truths, he does not think that morality is just a matter of conditioning. Reid is a moral realist and not a constructivist. Moral education influences our cognitive development, but not the moral truths that we can learn to understand. D. Does foundationalism imply infallibilism? Many people have the idea that intuitionism, because of its foundationalist aspects, implies infallibilism, and that because of this an intuitionist is naive, dogmatic or even a fundamentalist. However, foundationalism does not necessarily imply infallibilism. Instead, it is open to a foundationalist whether she will adopt an infallibilist or a fallibilist account.23 The infallibilist claims that there can be a sure foundation; basic beliefs are taken to be infallible, incorrigible or indubitable (Descartes). In contrast, the fallibilist says that we do not have external evidence for all our beliefs. The latter is a rather modest claim. Fallibilist foundationalism says that not all of our beliefs can be justified by other beliefs. Says Moore:

21

Reid 1969b: 248. Reid 1969b: 246-250, 369-373. 23 Cf. Pollock 1986: 58. 22

237 The expression “self-evident” means properly that the proposition so called is evident or true, by itself alone; that it is not an inference from some proposition other than itself. The expression does not mean that the proposition is true, because it is evident to you or me or all mankind, because in other words it appears to us to be true.24

Self-evidence does not imply any easy process of insight, rather it means that a certain belief is not derived from another belief. A basic belief is immediate, but not understood as immediate in time. Indeed, acquiring selfevident beliefs can take time, for example, understanding certain mathematical axioms and moral principles. That something is a basic belief does not mean that we cannot be mistaken about it, as Moore emphasizes: …in every way in which it is possible to cognise a true proposition, it is also possible to cognise a false one.25

Intuiting is just a way to cognize a proposition, it does not guarantee that one will acquire a true belief. The result is that our intuitions can be fallible.26 So it is possible to be a foundationalist and a fallibilist at the same time. Infallibilism concerns epistemology, i.e. our human capacities. None of the intuitionists I discuss here thinks that our human capacities are infallible. What might give rise to the idea that intuitionists are infallibilists is that many of them defend the position that moral principles are necessarily true or that they have axiomatic status. However, all these statements do not concern our epistemic access to moral propositions but the status of the propositions themselves. Intuitionists think that, although moral propositions are objectively true independently of our cognizing them, we can always err in our beliefs concerning these propositions. III. Intuition and justification A major concern of critics of intuitionism is that the foundationalist aspect of intuitionism implies that discussion or justification of moral beliefs becomes impossible. In this section I will argue that this is not the case. A. Internal and external justification 24

Moore 1988: 143. Moore 1988: x. 26 Cf. Ewing 1929: 191. 25

238 David Brink thinks that foundationalism is wrong. His argument is that although foundationalists reject coherentism because of its supposed unavoidable circularity or infinite regress, foundationalism (in the case of intuitionism but also in general) itself is circular in that it presupposes selfjustifying beliefs. Says Brink: But can there be such beliefs? Justification is justification in believing true. In order to be justified in holding one’s belief p, one must have reason to hold p true. But p is a first-order belief that such and such is the case and, as such, cannot contain the reason for thinking p is true. Indeed, selfjustification can be regarded as the limiting case of circular reasoning – that is, self-justification is the smallest justificatory circle imaginable. And everyone – even the coherentist – regards such small circles of justification as nonexplanatory and, hence, as nonjustifying.27

Brink’s main objection against intuitionism is that it is circular, a circularity with an even more vicious form than that of coherentism. I take Brink to claim the following: Premise 1: Foundationalism presupposes the existence of self-justifying beliefs. Premise 2: Self-justifying beliefs are a case of circular justification. ———————————— Conclusion: Foundationalism is committed to circular justification (and so is intuitionism, as being a form of foundationalism). Are premises 1 and 2 true? In section II I mentioned a few ways in which self-evident beliefs can be justified through non-propositional evidence: e.g. a sensation on which a belief is based or the occurrence of a belief, or through propositional evidence such as the content of a belief. So the formulation ‘self-justifying beliefs’ is misleading. The belief does not literally justify itself, it is its content, its occurrence or that on what it is based that provides for justification. Hence, premise 1 is false but then premise 2 is irrelevant. This means that the conclusion has not been established. Brink demands that we give second-order beliefs as justification for our first-order beliefs.28 This is a form of epistemic internalism. An internalist says that the occurrence of a belief, the content of a belief or the way it is formed are incapable of providing justification. Instead, justification 27 28

Brink 1989: 116. Brink 1989: 118, 119.

239 should always be internalistic, by for example giving arguments and providing proofs. Present-day Reidians such as William Alston and Alvin Plantinga deny this claim. Instead, they defend externalism. According to William Alston, in order to know it is not needed that I know that I know. On an internalist account, in order to be justified in believing p, I have to know that I am justified in believing that p. According to Alston, this leads to an infinite regress.29 However, on a foundationalist account this infinite regress might be avoided by proposing that there could be a basic belief like ‘I am justified in believing that there is a red ball in front of my eyes because this is selfevident’. Against such a reply Alston would launch an additional argument, which is that we do not always have such a reply available and still would be justified in our belief that, for example, there is a red ball in front of us. A belief can be self-evident without us knowing that it is self-evident. According to Alston, internalism would restrict that what we think to be knowledge to an incredibly small amount of beliefs for which we really have internal justification. We have to distinguish between having justification (requiring internalistic criteria) and being justified (externalistic).30 Reid and the other intuitionists argue that having self-evident beliefs is a state of being justified, even though one need not be aware of having this justification. Reid attacks philosophical approaches according to which we have to prove each of our beliefs. The kind of evidence (namely internally accessible, propositional evidence) these philosophers want to accept is much too limited. Reid advocates that nonpropositional kinds of evidence such as self-evidence and evidence of the senses, the fact that a belief is reliably formed or the content of a belief have to be taken seriously as kinds of evidence or justification. Even if we are not aware of these kinds of evidence, they can give us justification. Reid seems to apply externalism to basic beliefs, because he states that basic beliefs, from their very nature, do not allow for any direct proof. If basic beliefs are generated by a reliable, properly functioning faculty, they are warranted as knowledge. With regard to derived beliefs, though, it can be the case that they involve reasoning from premises to conclusion, so Reid might say that internal justification can be possible in those cases.31

29

This is also pointed out by Prichard 1912: 34, 35. Cf. Alston 1989. 31 However, cf. IIB: we can also intuit derivable beliefs. 30

240 In any case, Reid thinks that we have certain epistemic duties. He emphasizes that we have to do our best to make a good moral judgment.32 Proper functioning of our cognitive faculties is a constitutive and necessary condition for knowledge, but it might not always be sufficient. We also have to have the intention to find out what is right or wrong. The man who neglects the means of improvement in the knowledge of his duty, may do very bad things, while he follows the light of his mind. And though he be not culpable for acting according to his judgment, he may be very culpable for not using the means of having his judgment better informed.33

Reid says that somebody may not think that he may safely rely upon the suggestions of his mind, or upon opinions he has got, he knows not how.34

Hence, Reid thinks that we have to do our best to inform ourselves of the ways in which we form our beliefs, to make sure they are formed by a reliable source of knowledge. In conclusion, Brink’s attack against intuitionism does not work, since intuitionism does not involve circular justification. Demanding second order beliefs leads to an infinite regress and unduly reduces the amount of our knowledge. Intuitionism allows for external justification of our basic beliefs, but secondary beliefs can allow for internal justification. B. Intuition and verification A.J. Ayer thinks that the main problem with intuitionism is that it makes statements of value unverifiable. For it is notorious that what seems intuitively certain to one person may seem doubtful, or even false, to another. So that unless it is possible to provide some criterion by which one may decide between conflicting intuitions, a mere appeal to intuition is worthless as a test of a proposition’s validity.35

Ayer here says that intuitionism cannot provide for a criterion to decide which of two conflicting intuitions is true and that because of this, 32

Reid 1969b: 361. Reid 1969b: 249. 34 Reid 1969b: 248; italics mine. 35 Ayer 1952: 106. 33

241 intuitionism is a useless theory. However, the intuitionist will answer that in this respect ethical beliefs are not completely different from other kinds of beliefs. In (fallibilist) foundationalist fashion, the intuitionist will say that it is the case with many beliefs that we have that if we are confronted with somebody who is convinced of the contrary of what we are convinced of, it can be impossible to establish beyond doubt who is right. Reid says that this is simply a consequence of the fact that we have no new senses with which we can ‘sit in judgment upon the old’.36 We have to take the deliveries of our senses as starting points: Nature has doomed us to believe the testimony of our senses, whether we can give a good reason for doing so or not.37

This is also the case with our moral beliefs. Nevertheless, through discussion, evaluation and reflection we can sometimes reconsider our beliefs. This is even more the case with moral judgments which concern not just a simple impression on our senses but an evaluation with many aspects that are relevant in a situation. The intuitionist would say that in the case of sense perception it can also be problematic to decide between conflicting beliefs, although admittedly it might be even harder in the case of moral perception. Intuitionists think that concerning the difficulty of deciding between conflicting beliefs, moral knowledge is in similar shape as other forms of knowledge. Because of this, Ayer’s demand for an independent criterion might rather lead one to universal skepticism than only to a rejection of intuitionism. And this consequence of Ayer’s account might backfire against itself. C. Intuition and common sense Reid’s epistemological project is directed against philosophical approaches that demand that we give a justification for all our beliefs. According to Reid, this inevitably leads to skepticism. Or as Prichard puts it: …if, as is almost universally the case, by Moral Philosophy is meant the knowledge which would satisfy this demand [of provability], there is no such knowledge, and all attempts to attain it are doomed to failure because

36 37

Reid 1969b: 237. Reid 1969b: 462.

242 they rest on a mistake, the mistake of supposing the possibility of proving what can only be apprehended directly by an act of moral thinking.38

Reid defends foundationalism in all realms of epistemology. His form of foundationalism starts from our common sense beliefs on which we build our knowledge. Ross also takes the moral judgments of the common person as knowledge: ...it seems as self-evident as anything could be, that to make a promise, for instance, is to create a moral claim on us in someone else. Many readers will perhaps say that they do not know this to be true. If so, I certainly cannot prove it to them; I can only ask them to reflect again, in the hope that they will ultimately agree that they also know it to be true. The main moral convictions of the plain man seem to me to be, not opinions which it is for philosophy to prove or disprove, but knowledge from the start; and in my own case I seem to find little difficulty in distinguishing these essential convictions from other moral convictions which I also have, which are merely fallible opinions based on an imperfect study of the working for good or evil of certain institutions or types of action.39

Reid says in a similar vein as Ross that the task of philosophy is not to doubt our common sense judgments but to start from them. Without common sense, philosophy would be impossible.40 This does not mean that we should hold on to our naive prejudices in the face of other evidence, but our common sense judgments are as it were innocent until proven guilty. That means that we can hold on to our initial beliefs until we have good reasons to doubt them. Of course a critical person should always examine her beliefs to see whether they are tenable or not. Reid and Ross are rather confident in our knowledge of general moral principles. Still, they consider concrete moral judgments to be more problematic since contextual information is needed. People might often judge a situation wrongly, simply because they do not get all the factual information needed or because they have to weigh different conflicting duties (Ross’s famous account of prima facie-duties). In any case, according to Reid, we cannot start from scratch, common sense knowledge has to be the starting point. Says Reid: In matters beyond the reach of common understanding, the many are led by the few, and willingly yield to their authority. But, in matters of common 38

Prichard 1912: 21-37, 36, 37. Ross 1967: 20, 21, n. 1. 40 Reid 1997: 19. 39

243 sense, the few must yield to the many, when local and temporary prejudices are removed.41

It is not that Reid thinks that concerning all subject matters, the beliefs of the common person should be the guiding line. He definitely thinks that certain fields of knowledge are the prerogative of philosophers and scientists, e.g. concerning knowledge of abstract and necessary propositions and concerning complex reasoning. But concerning the data we obtain by our sense organs, philosophers, ‘the few’, have no right to tell the common person that their cognitive faculties are mistaken. Instead, the philosopher should consult the common person on those matters. We should be careful to understand exactly how Reid’s argument from common sense runs. Intuitionists and common sense philosophers are often criticized for being naive and dogmatic. However, Reid’s rejection of skepticism can hardly be called dogmatic; to the contrary, its upshot is very liberal. Concerning common sense, the few have no right to tell the many what to believe, but the many instead can tell the few that they are wrong. With the few, Reid refers to the skeptical philosophers, who are the target of his ironic polemic that accompanies all of his philosophical works. Reid is here obviously not thinking of ideological mass movements that can be a form of repression themselves. These movements do not consider all human beings to be on equal footing, so they do not fit in the common sense picture. To take every human being as having in principle an equal right to judge is at the heart of Reid’s common sense-philosophy, which is one of the aspects that make Reid a real enlightenment-philosopher. Says Reid: The judgments grounded upon the evidence of sense, of memory, and of consciousness, put all men upon a level.42

As to the supposed naivety of a common sense-approach, Reid would say that, compared with the enormous demands that skeptics have on knowledge, common sense philosophy might appear naive. Still our common sense is the only source of knowledge we have concerning certain domains of knowledge, and when it comes to practical matters, even the skeptical philosopher will have to rely upon his ‘naive’ common sense beliefs. We are born under a necessity of trusting to our reasoning and judging powers; and a real belief of their being fallacious cannot be maintained for 41 42

Reid 1969a: 605. Reid 1969a: 540.

244 any considerable time by the greatest skeptic, because it is doing violence to our constitution. It is like a man’s walking upon his hands, a feat which some men upon occasion can exhibit; but no man ever made a long journey in this manner. Cease to admire his dexterity, and he will, like other men, betake himself to his legs.43

To refute skepticism in the sense that the radical skeptic could be convinced that we have knowledge, is impossible according to Reid. This is because skepticism starts out by asking unanswerable questions, namely how we can justify our common sense beliefs. These demands on knowledge, though, are too high. The general idea behind Reid’s common sense epistemology is the following: the skeptical project is an impossible project. To demand a justification of our common sense beliefs implies either that we assume that we have certain cognitive faculties, for example reason, that are infallible and through which we can judge our other faculties. But this is an arbitrary, unjustified assumption. Or we prove the reliability of some of our faculties by the fact that they have been successful in a lot of cases. However, this is a circular way of reasoning, because in order to judge that, we already presuppose the reliability of some of our cognitive faculties, otherwise we could not state that some judgments are reliably formed.44 Or, as a third possibility, we conclude that we cannot prove the infallibility of our cognitive faculties and hence become skeptics, but this leads to a pragmatic paradox, as in everyday life we will rely on our faculties. All this indicates, according to Reid, that the skeptical project is a pragmatically impossible project. The faculties which nature has given us, are the only engines we can use to find out the truth. We cannot indeed prove that those faculties are not fallacious, unless God should give us new faculties to sit in judgment upon the old. But we are born under a necessity of trusting them.45

It is impossible to determine how often our moral judgments are true. We do not have neutral standards by which we could judge that the reliability of any of our cognitive faculties is more beyond doubt than that of our other faculties. It is impossible to judge our faculties from the outside. The only way we might proceed is to ask ourselves how often we experience in our daily life that it turns out that our moral judgments were wrong, e.g. 43

Reid 1969a: 632. Cf. Alston 1993 45 Reid 1969b: 237. 44

245 that somebody feels himself mistreated although we thought we were doing our best. However, even then it might always turn out that our initial judgment was right after all. It is as speculative to say that our moral judgments are more fallible than we think as to assume that they are infallible. Our experience might indicate that sometimes we get it pretty right, other times we err. Reid thinks that we can check our moral judgments by calm reflection and by comparing them to certain points of reference. 46 Furthermore, in making our moral judgments we are not only left to ourselves with our subjective opinions or stuck to our tradition, but by interaction with others and by being open for other points of view, we are able to examine our judgments and to develop them further. This can never be an ultimate guarantee that we make the right judgments, but it is a more promising route than to just rely on our own initial judgments. Here the skeptic about intuitionism might get puzzled and wonder whether this idea does not undermine the foundationalist claims that intuitionists are supposed to defend. If we use other beliefs to check our socalled intuitions, surely these beliefs can no longer be called basic. However, I think that this conclusion is mistaken. The above considerations do not undermine foundationalism. This is because no deeper argumentation is given, but the same is said in different words. For example, if we want to defend or examine the claim that all human beings have equal worth, we can reflect on times and places where this moral principle has been violated. We can reflect about for example sexist or racist societies and conclude that these are very cruel societies. However, our judgment that such societies are cruel already presupposes the moral principle at stake. Reflection about concrete examples merely facilitates an insight into a basic moral belief, but it does not provide for more fundamental arguments. IV. Normative judgments as intuitions In the foregoing sections I have elucidated and defended the notion of intuition that the older intuitionists such as Reid have used. However, in the writings of intuitionists from the beginning of the 20th century we can find some further interesting ideas concerning the notion intuition. As I will show in this section, according to early 20th century intuitionists moral intuition fulfils two purposes in our moral assessment of a situation: 1., intuitions 46

Reid 1969b: 261. This is one of Reid’s first principles of morals. Although it refers mainly to our knowledge about our own duties, we can also extend it to our moral judgments in general.

246 bridge the fact-value gap, and 2., intuitions are needed to make judgments concerning morally complex situations (i.e., situations involving more than one moral consideration). According to Ross and Prichard, all moral judgments, even the more general ones, are initially justified on the basis of particular judgments. The following passage from Prichard is the locus classicus of this view: …if we do doubt whether there is really an obligation to originate [an action] A in a situation B, the remedy lies not in any process of general thinking, but in getting face to face with a particular instance of the situation B, and then directly appreciating the obligation to originate A in that situation.47

How exactly do we judge whether we should perform a concrete action or not? Ross says that we should consider all morally relevant properties of a concrete case and based on that we can form a moral judgment about the resulting moral properties.48 The perception of the non-moral base properties requires normal empirical observation. Once we have assessed the base properties, we will be able to make a moral judgment. According to Ross, moral rightness refers to the ‘suitability’ of an action in a situation.49 Moral judgments cannot be reduced to non-moral judgments, as this would mean committing the naturalistic fallacy.50 Hence, the moral judgment is not an inference from the non-moral observations, but it needs these observations as input. Because the final moral judgment is non-inferential, it can best be characterized as intuitive. The procedure might be illustrated in the following schema: Schema 1: 1. is1 2. is2 3. is3 .... n. isn ~~~~ ought 47

Prichard 1912: 37. Ross 1968: 168. 49 Ross 1927: 113-127, 127. 50 Cf. Moore 1988. 48

247 Or more explicitly as follows: Schema 1’: 1. Situation x is a. 2. Situation x is b. 3. Situation x is c. .... n. Situation x is z. ~~~~ In situation x, person p ought to φ. By observing as many of the relevant descriptive facts of a situation as possible we are prepared to form a normative judgment, but the judgment is not a logical inference from the descriptive judgments since we cannot derive ought from is. Hence the ~~-line instead of a straight line in a logical inference. One could argue that one can construe a logical reasoning involving a premise that connects is-statements with ought-statements. However, then the operator that connects the is-antecedent with the ought-consequent would neither be a logical nor an empirical operator but, as it were, a moral operator. In fact, schema 1 as well as schema 2 below can be read as elucidations of exactly such a premise. The descriptive judgments are not premises in an argument. Instead, they are preliminaries that enable us to form a moral judgment. A similar point is also made by Ewing and Prichard.51 Ewing says the following about these preliminaries: they are ‘the arguments and data that help us arriving at, but do not, as we have seen, prove the truth of, our intuition’.52 Ewing points out that there is also another way in which our moral judgments are non-inferential. His argument concerns the relation between the evaluative judgment of a whole to that of its parts. According to Ewing, we first have to perceive the values of the parts to determine the value of a whole. However, since Ewing accepts Moore’s account of organic wholes53 he believes that this determination cannot be a direct inference but must be an intuition. By intuition Ewing means ‘knowledge otherwise than by mediate inference or by mere observation’.54 51

Ewing 1929: 185, 186, and Prichard 1912: 28. Ewing 1929: 187, note 1. 53 The theory of organic wholes means that the value of a whole need not be identical to the sum of the values of its parts, cf. Moore 1988: 28. 54 Ewing 1929: 186. 52

248 We can illustrate this as follows: Schema 2: 1. is1 ~ ought1/value1 2. is2 ~ ought2/value2 3. is3 ~ ought3/value3 .... n. isn ~ oughtn/valuen ~~~~ oughtoverall/valueoverall This means that if one adopts Moore’s account of organic wholes (or a holistic approach such as Dancy’s particularism55), complex moral judgments that concern the balancing of different moral considerations, as opposed to moral judgments based on non-moral observations as illustrated in schema 1, are still best taken to be intuitions.56 I hope these schemas help to shed more light on what is meant by ‘intuition’ and make it more digestible for the reader who is still a skeptic about the legitimacy of intuitions. This manner of understanding intuition does not imply: ‘just sitting down and having an ethical intuition’, what Mackie called a ‘travesty of our actual moral thinking’.57 Ewing was well aware of this skepticism: This word [intuition] is, I admit, associated with some very unphilosophical modes of thinking, since nothing is easier than to fancy you have true “intuitions” when you really have not; and I shall therefore substitute for it the term “non-inferential cognition”, if by so doing I may avoid some scandal (Ewing 1929, p. 186).

V. Conclusion The notion of intuition is much less obscure than philosophers in the past have made it seem to be. It is time to reassess this aspect of intuitionism and give intuitionism the place in moral epistemology that it deserves to have: i.e. not as a strawman that nobody wants to identify with but as a 55

Dancy 2004. Ross 1968: 185 makes the same point. Dancy (2004: 148 ff.) actually rejects foundationalism, but the interpretation of the notion intuition that I give in this section should be compatible with Dancy’s ideas. 57 Mackie 1977: 38. 56

249 common sense-approach which can also provide for interesting insights concerning the relation between moral and non-moral beliefs and the assessment of morally complex situations.

REFERENCES

Alston, William. 1993. The Reliability of Sense Perception. Ithaca and London: Cornell University Press. ——— . 1989. Epistemic Justification. Essays in the Theory of Knowledge. Ithaca & London: Cornell University Press. Audi, Robert. 1997. Moral Knowledge and Ethical Character. Oxford: Oxford University Press. Ayer, A.J. 1952 [1936]. Language, Truth and Logic. New York: Dover Publications Inc. Brink, David. 1989. Moral Realism and the Foundations of Ethics. Cambridge: Cambridge University Press. Dancy, Jonathan. 2004. Ethics Without Principles. Oxford: Oxford University Press. DePaul, Michael. 1993. Balance and Refinement: Beyond CoherenceMethods of Moral Inquiry, London: Routledge. Ewing, A.C.1929. The Morality of Punishment. London: Kegan Paul. Hudson, W.D. 1967. Ethical Intuitionism London: Macmillan. Lehrer, Keith. 1989. Thomas Reid. London, New York: Routledge. Little, Margaret Olivia.1995 “Seeing and Caring: The Role of Affect in Feminist Moral Epistemology” in Hypatia 10: 117-137. Mackie, J.L. 1977. Ethics: Inventing Right and Wrong. Hammondsworth: Penguin. McNaughton, David. 1988. Moral Vision. Oxford: Basil Blackwell. Moore, G.E. 1988 [1903]. Principia Ethica. Buffalo, NY: Prometheus Books.

250 Plantinga, Alvin. 1993. Warrant: The Current Debate. New York: Oxford University Press. Pollock, John. 1986. Contemporary Theories of Knowledge. Totowa: Rowman & Littlefield. Prichard, H.A. 1912. “Does Moral Philosophy Rest on a Mistake?”, Mind 21: 21-37, 36, 37. Reid, Thomas. 1969a [1785]. Essays on the Intellectual Powers of Man, Introduction by Baruch Brody. Cambridge, Massachusetts, and London, England: The M.I.T. Press. ——— . 1969b [1788]. Essays on the Active Powers of the Human Mind, Introduction by Baruch Brody. Cambridge, Massachusetts, and London, England: The M.I.T. Press. ——— .1997 [1764]. Inquiry into the Human Mind, ed. by Derek R. Brookes. Edinburgh: Edinburgh University Press. W.D. Ross. 1967 [1930]. The Right and the Good. Oxford: Clarendon Press. ——— .1968 [1939]. Foundations of Ethics. The Gifford Lectures. Oxford: Clarendon Press. ——— .1927. “The Basis of Objective Judgments in Ethics”, International Journal of Ethics 37: 113-127, 127. Sidgwick, Henry. 1901. The Methods of Ethics. London, New York: Macmillan.

DAVID ENG

Basic Beliefs, Testimony, and Blind-Trust In the epistemological literature on testimony, the trend recently has been to adopt a view that I call the ‘Blind Trust account’.1 According to this account, a hearer can be justified in believing a speaker’s utterance simply in virtue of understanding the utterance. This means that a hearer can form a justified belief on the basis of a speaker’s utterance without possessing any evidence about the reliability of the speaker or the truth of the asserted proposition. Although there are a number of problems facing this view (Fricker 1994), this paper focuses on a particular motivation for this account, one that appeals to the claim that testimonial beliefs are epistemically basic. In particular, I want to consider whether one can defend the Blind-Trust account on the grounds that testimonial beliefs are basic or because testimony is a basic source of justification. In doing so, I will consider two different senses of epistemic basicality: 1) the traditional foundationalist sense that applies to particular beliefs and 2) a sense of epistemic basicality that applies to cognitive practices.2 I will argue that if testimony is epistemically basic in either sense, it does not follow that the Blind-Trust account is plausible. In section I, I begin by clarifying what I mean by testimonial beliefs and how I understand the Blind-Trust account. In section II, I discuss the foundational sense of epistemic basicality. I then consider and reject arguments for why testimony is epistemically basic in this sense and for why if testimony is epistemically basic, it follows that the Blind-Trust account is correct. In section III, I consider a different sense of epistemic basicality that applies to cognitive practices and distinguish it from the foundationalist sense of epistemic basicality. I then consider and reject arguments that 1 For defenses of this view, see Burge 1993; Foley 1994; Kornblith 1987; and Webb1993. It is unclear whether Plantinga, 1993 endorses the Blind-Trust account. He grants that children can form beliefs on the basis of blind-trust, but he claims that our cognitive practices are modified according to a design plan. This allows Plantinga to say that adults do not form justified beliefs on the basis of blind-trust. 2 Although Alston does not explicitly use the term ‘epistemic basicality’, this sense of epistemic basicality is drawn from his work (Alston 1989).

252 attempt to motivate the Blind-Trust account on the basis of the epistemic basicality of testimony in this sense. I Testimonial Beliefs and the Blind-Trust Account In speaking of testimonial beliefs, I have in mind, roughly, cases where a hearer forms a belief that p on the basis of interpreting a speaker as having asserted that p. Since assertions are a kind of linguistic utterance that involve an intention that the utterance be taken as true, a testimonial belief is one that is formed on the basis of interpreting a speaker as having made a linguistic utterance with an intention that the utterance be taken as true. This is a broad way of construing testimonial beliefs for several reasons. First, since a hearer only has to interpret a speaker as having made an assertion, the account includes beliefs that are formed on the basis of interpreting a speaker as having made an utterance with a more specific intention (e.g., the intention that the utterance be taken as true and that it be taken as evidence toward a controversial matter).3 The account only excludes beliefs that are formed on the basis of interpreting a speaker as having made a non-assertive linguistic utterance (e.g., a command, a question, etc.). Second, notice that the account allows hearers to form testimonial beliefs on the basis of interpreting a speaker as having asserted p, and does not require that a speaker actually assert p. Thus, a hearer can form a testimonial belief that p on the basis of mistakenly interpreting a speaker as having asserted p (e.g., the speaker utters p with a sarcastic intention). Finally, note that a hearer can form a testimonial belief when the speaker has or lacks the relevant expertise to assert p, and when the hearer possesses a great deal of evidence or no evidence at all regarding the speaker’s reliability. A desirable reason for construing testimonial beliefs so broadly is that it allows for a discussion of the justificatory status of testimonial beliefs without fear of begging important epistemic questions. In particular, the current dispute between the two dominant accounts of the justification of testimonial beliefs, Humean weak individualism and the Blind-Trust account, has focused on whether a hearer must possess good reasons for believing that the speaker is reliable. For the Humean weak individualist, a 3 For example, Coady 1992, ch.2 argues for a much narrower account of testimony, and if we were to adopt his account, we would get a much narrower account of testimonial beliefs.

253 hearer’s testimonial belief is justified only if the hearer possesses good reasons for believing that the testimonial source is reliable, and these reasons must ultimately be grounded in non-testimonial reasons, in order to be good. In contrast, on the Blind-Trust account, the mere fact that a speaker asserts p is sufficient for a hearer to acquire a justified belief that p. Thus, a hearer does not have to possess any evidence about the reliability of the speaker in forming a justified testimonial belief. Now on a narrower construal of testimonial beliefs, according to which a testimonial belief is one that is produced on the basis of an utterance from a speaker who has the relevant expertise, the Blind-Trust account seems quite plausible. If we restrict testimonial beliefs to only beliefs where the speaker has the relevant expertise, then it seems natural that hearers should not have to possess evidence about the reliability of a speaker as the Blind-Trust theorist suggests. The problem though, as the Humean weak individualist notes, is that by defining testimonial beliefs in this narrow way, we avoid the crucial question that divides Humean weak individualism and the Blind-Trust account: do hearers have to possess reasons for believing that a speaker is reliable if we consider cases when the source is reliable and when it is unreliable? Thus, to be fair to the weak individualist, we should adopt the broad construal of testimonial beliefs that I proposed earlier. Before turning to whether testimonial beliefs are epistemically basic, I should make a few clarificatory remarks regarding the Blind-Trust account. First, recall that on the Blind-Trust account, it is sufficient for a hearer to interpret a speaker as having asserted p in order to acquire a justified belief that p. Hence, a hearer might not possess any evidence about the reliability of the speaker. If a hearer were unable to see or hear any features of the speaker that might indicate the speaker’s reliability (e.g., tone of voice, manner of presentation, etc.), the hearer’s belief would still be justified. Second, we need to clarify what it means to say that a hearer does not have to possess any evidence about the reliability of the speaker. In particular, it is important to note a distinction between the evidence that a hearer might acquire as a result of the speaker having made the utterance, and other ways in which an agent might acquire evidence about the reliability of the speaker. In defending the Blind-Trust account, Burge (1993) has argued that if a hearer is entitled to believe that a source has made an assertive utterance, the hearer is entitled to believe that the source is rational. This in turn suggests that hearers can acquire evidence about the ra-

254 tionality and thus the reliability of the source on the basis of interpreting the speaker as having made a meaningful utterance. So, for Burge, a hearer can acquire evidence about the speaker’s reliability on the basis of merely interpreting the speaker as having made an assertion. This evidence, for Burge, can play a crucial role in the justification of the hearer’s testimonial belief. But, according to the Blind-Trust theorist, this is not the kind of evidence that is unnecessary in forming a justified testimonial belief. Rather, it is evidence that a hearer might acquire prior to the agent having made the utterance or evidence that a hearer might acquire on the basis of how the speaker made the utterance (e.g., tone of voice, structure of the utterance, etc.) that is unnecessary. Third, the claim that a hearer can form a justified belief simply on the basis of interpreting a speaker as having made an utterance that p implies that beliefs formed on the basis of blind-trust are justified. What exactly is blind-trust? A hearer has a habit of forming beliefs on the basis of blind-trust just in case the hearer has the habit of forming beliefs on the basis of believing what anyone says about anything. According to Reid,4 this is the kind of belief forming habit with which we are all born, and that gets modified in light of experience as we mature. Why does the Blind-Trust account entail that beliefs formed on the basis of blind-trust are justified? If understanding an utterance that p is sufficient for a hearer to form a justified belief that p, then when children understand an utterance and form a belief on the basis of blind-trust, they must, according to the Blind-Trust account, form a justified belief. Moreover, defenders of the Blind-Trust account readily seem to endorse the idea that children’s testimonial beliefs, and thus beliefs formed on the basis of blind-trust, are justified. As Burge writes, “As children and often as adults, we lack reasons not to accept what we are told. The justification I develop below is a reflective philosophical account of an epistemic entitlement that comes with being a rational agent…Acceptance underlies language acquisition.”(Burge, 1993, 467-468)

I focus here on cases involving children, because it is clearer that children, rather than adults, form beliefs on the basis of blind-trust. As a final clarification of the Blind-Trust account, note that interpreting a speaker as having asserted p only provides a prima facie justification 4 Thomas Reid 1983, VI, 5: 281-282.

255 for believing p. The prima facie justification could be defeated if there were evidence available to the agent that the speaker should not be trusted or that p is false. With this understanding of testimonial beliefs and the Blind-Trust account, let us turn to how one might argue for the epistemic basicality of testimony and the claim that the Blind-Trust account must be correct if testimony is epistemically basic. II:Basicality, Foundationalism, and Blind-Trust As I noted earlier, there are at least two senses of epistemic basicality, and the first that I will consider here is the foundationalist sense of epistemic basicality. According to foundationalism, our justified beliefs have the following structure. At the foundation are justified beliefs that are basic, justified beliefs that do not depend on and that do not derive their justification from other beliefs. Above this foundation are justified beliefs that are nonbasic, justified beliefs that do depend on or that do derive their justification from other beliefs. In some cases, these non-basic beliefs might derive their justification directly from basic beliefs, but, in other cases, they might derive their justification directly from other non-basic beliefs. In either case, one can ultimately trace the justification of non-basic beliefs to basic beliefs lying at the foundation. Given the foundationalist sense of epistemic basicality, let us consider why one might think that testimonial beliefs are epistemically basic in the foundationalist sense, and moreover, let us consider why one might think that if testimonial beliefs are epistemically basic in the foundationalist sense, it follows that the Blind-Trust account is correct. Consider first the following argument for why one might think that testimonial beliefs are epistemically basic in the foundationalist sense. The majority of our testimonial beliefs are formed without any inference; we immediately believe what we are told. We do not reason through any belief or any evidence about the speaker’s reliability. When I ask strangers for the time, I simply believe what they say. Hotel receptionists immediately believe me when I tell them my name, and I immediately believe what I read when I read the paper. If the majority of our testimonial beliefs are formed in this way and if we were to say that these beliefs are unjustified, then very few of our testimonial beliefs would be justified. Moreover, any nontestimonial beliefs based on these testimonial beliefs would also be unjustified. The supposition that testimonial beliefs that are not formed on the ba-

256 sis of an inference relying on evidence about the speaker’s reliability would lead to an undesirable skepticism. In order to avoid skepticism, we must grant that testimonial beliefs formed without any inference must be justified and thus epistemically basic. How does it follow that beliefs formed on the basis of blind-trust are justified? The argument, I take it, is that if an agent can form a justified belief without any inference through any evidence about the reliability of the speaker, then an agent can form a justified belief on the basis of blind-trust. Therefore, blind-trust beliefs must be justified. Although the above arguments have never been explicitly offered, this way of arguing is, I believe, a common way of motivating why testimonial beliefs are epistemically basic and why blind-trust beliefs are thus justified. However, there are problems with both arguments. To begin, notice that the first argument rests on the assumption that if a justified testimonial belief is not formed on the basis of an inference from any evidence, E, or a belief, B, then its justification does not depend on E or B.5 In other words, if a belief is psychologically basic (i.e., produced by a process that does not have other beliefs as inputs to the process) and justified, it must also be epistemically basic. Not all foundationalists, I suggest, would accept this assumption. The central claim of foundationalism, as I understand the view, is one about the structure of justification, that there are basic versus non-basic beliefs. If this is right, then foundationalism does not take a stand on what constitutes the justification for any belief, whether it is basic or non-basic. Addressing this issue requires addressing the following questions. Why are certain beliefs epistemically basic, but others not? For example, what is it that allows certain perceptual beliefs, for example, to be epistemically justified, but other beliefs, such as those formed on the basis of wishful thinking, not? Similarly, why are certain inferential beliefs justified, but others not? Historically, foundationalists have answered these questions in different ways. On classical versions of foundationalism such as Descartes’, basic and non-basic beliefs are justified in virtue of being indubitable. Moreover, what makes a justified belief, such as the cogito, basic rather than non-basic is the fact that it is indubitable in virtue of its content. In contrast, justified beliefs that are non-basic are those that derive their indu5 I am using the term “evidence” in a broad sense so as to include belief states and non-belief states such as perceptual states.

257 bitability and thus their justification on the basis of deductive inferences. On Descartes’ account, the standards of justification are high. On more moderate versions of foundationalism, basic beliefs only have to be reliably formed, and non-basic beliefs can derive their justification from other beliefs via deductive and inductive inferences. Finally, on less objective and admittedly less traditional versions of foundationalism, such as Annis’ contextualist account (1978), the justification of a belief is determined by whether the agent can respond to the objections of a group. A basic belief is one against which no objection is raised while a non-basic belief is one against which objections are raised, but the agent is able to respond to these objections. The salient point here is that for some of the above foundational accounts what makes a belief basic is not tied to how it is formed. In the case of the cogito, this belief is indubitable and basic in virtue of its content and not how it is formed.6 Similarly on Annis' account, since a basic belief is determined by whether the group raises any challenges, one can form an epistemically basic belief that is not psychologically basic. In these cases, whether a belief is psychologically basic is only contingently connected to whether it is epistemically basic. Thus, the fact that some testimonial beliefs are psychologically basic does not necessarily imply that they are epistemically basic. In response, one might argue that these foundational accounts are implausible, and that on any plausible version of foundationalism, whether a belief is epistemically basic is tied to how it is formed. Consider for example Goldman’s reliabilist account of justification (1979), according to which, roughly, a belief is justified if and only if it is produced by a reliable process. On Goldman’s account, whether a belief is epistemically basic is tied to whether it is psychologically basic. Goldman distinguishes belief-dependent processes from belief-independent ones. Belief-dependent processes take beliefs as inputs, and thus produce output beliefs that are psychologically non-basic. Moreover, the justification of these output beliefs depends on the justification of the input beliefs. So psychologically non-basic beliefs can only be epistemically non-basic. In contrast, belief6I do not mean to imply here that for Descartes the indubitability of all beliefs has nothing to do with how the beliefs are formed. As George Pappas pointed out to me, for Descartes beliefs about one’s current sensations are indubitable because of how they are formed. My point here is that there is only a contingent connection between whether a belief is foundational in the epistemic sense and how the belief is formed.

258 independent processes do not take beliefs as inputs, and thus produce psychologically basic beliefs. Obviously, the justification of these output beliefs does not depend on any input beliefs. Thus, psychologically basic beliefs can only be epistemically basic. If Goldman’s account is correct and if we grant that most of our testimonial beliefs are psychologically basic and justified, then most of our testimonial beliefs must be epistemically basic. Moreover, as the second argument suggests above, it would follow that blind-trust beliefs are justified. Although I agree that if Goldman’s account is correct, and if most of our testimonial beliefs are psychologically basic, then most of our testimonial beliefs are epistemically basic, I do not agree that blind-trust beliefs would therefore be justified. This is because many of our testimonial beliefs that are psychologically basic are not formed on the basis of blindtrust. There are two common ways in which a testimonial belief can be psychologically basic without being formed on the basis of blind-trust. First, numerous studies on persuasion7 indicate that when agents rely on testimonial sources, they are unconsciously influenced by features they perceive in testifiers (e.g., whether the testifier is cognitively similar to them, tone of voice, features in the surrounding environment, the mode of presentation, etc.). Had the agents not perceived these features, they would not have formed the testimonial belief. In cases where the agent is discriminative in this way, the agents do not reason through any belief that the testifier is reliable because of their tone of voice or any other feature. The studies indicate that certain features, such as tone of voice, can affect whether the agent will form the belief, even though the agent is unaware of this fact. The agents can perceive features that are associated with a speaker’s reliability without necessarily forming any beliefs about whether the speaker has or lacks this feature and whether it is associated with reliability. When agents are discriminative in this way, notice that the testimonial beliefs are psychologically basic, since the beliefs are based only on a perceptual state and not on some other belief state. But at the same time, the beliefs are not formed on the basis of blind-trust, since the agents are discriminative. A natural response to my characterization of the above kind of case is to argue that these testimonial beliefs are not psychologically basic. Although the belief states might not be mediated by a perceptual belief state 7 See Alpert and Anderson 1973; Rogers1983; McCroskey 1969; Calder et al, 1974; Festinger and Maccoby 1964; McGuire1964.

259 that a testifier has feature, F, the testimonial belief is still inferential insofar as it is based on a general background belief (e.g., the belief that there is a general correlation between F and trustworthiness or reliability). One obvious response to this challenge is that it seems very implausible that agents would have these background beliefs, given that agents are typically unaware about why they trust the testifier. That is, if agents are unaware that they are affected by the tone of voice of the speaker, it is unlikely that they have a general belief that a certain tone of voice indicates reliability. But a second, more serious problem with the above response is that if it is correct, it makes it even more difficult to show that the Blind-Trust account is plausible. Recall that the central thrust of the original argument was that blind-trust beliefs must be prima facie justified in order to avoid skepticism, because most of our testimonial beliefs are psychologically basic. But if the current objection is correct (i.e., most of our testimonial beliefs are not psychologically basic), then we can avoid skepticism without granting that blind-trust beliefs are prima facie justified.8 A second way in which testimonial beliefs can be psychologically basic without being formed on the basis of blind-trust is illustrated by the following case. Consider an ordinary case of testimonial belief formation such as a belief formed on the basis of a New York Times report. According to blind-trust theorists, these beliefs are psychologically basic and formed on the basis of blind-trust because we immediately believe what we read, and we do not reason through any belief or evidence about the reliability of the New York Times. Does it follow that these beliefs are formed on the basis of blind-trust? No. In most cases, the beliefs are produced by beliefindependent processes that involve relying on a particular source. True, the agents do not reason through any evidence or belief about the reliability of the source, and thus the processes are belief-independent. But in most cases, the agents possess (or at one point possessed) evidence or a belief about the reliability of the source. This evidence plays an important role in developing a belief-forming testimonial habit that involves relying on a particular source, and in some cases, this testimonial belief-forming habit can be a belief-independent process. My general point here is that in many cases, agents reason through evidence about the reliability of a source early 8 My claim here is not that agents have to be aware of when they are trusting blindly, but rather it is to suggest that agents are not trusting blindly in these cases. They cannot be used in support of the Blind-Trust account.

260 on, but over time, agents go on to develop a belief-independent process of relying on the source. Because the evidence now only plays an indirect role and because the beliefs are immediately formed, it is easy to mistakenly think that the beliefs are formed on the basis of blind-trust. But they are not formed on the basis of blind-trust because the agents would not have formed the belief had the testimony come from a different source. So far, I have described two common and general ways in which testimonial beliefs can be psychologically basic without being formed on the basis of blind-trust. In both cases, I assumed that what an agent would do in non-actual circumstances is relevant to determining the process or mechanism the agent uses in forming a testimonial belief. If a hearer forms a testimonial belief on the basis of what a speaker says, but would not have formed the belief had the speaker used a different tone of voice, the testimonial belief forming process responsible for the belief, I claim, is not blind-trust. One might question my assumption that what agents do in counterfactual situations is relevant to determining the process that produces the belief. The blind-trust theorist might argue that when we immediately believe what others say and there is no evidence that the source is unreliable, the open or default position is blind-trust. Our background evidence that certain features indicate unreliability or untrustworthiness acts only as a defeater, and is triggered only if those features are present.9 So if a hearer forms a testimonial belief on the basis of what a speaker says, but would not have formed the belief had the speaker used a different tone of voice, the testimonial belief forming process responsible for the belief is blindtrust. What the agent would do in counter-factual situations simply is not relevant in these cases. The dispute here is whether what an agent would do in counterfactual situations is relevant to determining the process that is responsible for a belief. The blind-trust theorist would argue that they are not, but I would argue that they are, and this dispute could be settled if a general criterion could be given for identifying the counter-factual situations that are relevant to identifying an agent’s belief forming process. Unfortunately, I do not have a clear sense of what such a criterion looks like, and even if I 9 I believe that this way of characterizing how our testimonial beliefs are formed, (i.e., the default position is blind-trust and background evidence only plays a defeater role) has motivated many, such as Burge, Foley, and Coady, to argue for a Blind-Trust account.

261 did, I do not have the space to present it here. Instead, I propose to consider other kinds of belief formation in order to show that we generally take into account what an agent would believe in counter-factual circumstances in determining the process that is responsible for the agent’s belief. Consider an instance of inferential belief formation. Suppose that Melissa forms the belief that q on the basis of an inference from the belief that p and the belief that p ⊃ q. This belief is analogous to someone forming a belief on the basis of a New York Times report. What is the process responsible for her belief? Consider two candidate processes: 1) a process that is embodied by a modus ponens inference or 2) a process that is embodied by any conditional inference, regardless of the direction of the conditional. On my view, but not on the Blind-Trust theorist’s view, whether Melissa would form the same belief if the direction of the conditional had been different (i.e., if she would have concluded q if the propositions were q ⊃ p and p) is relevant to fixing the process that is responsible for her belief. There are clear reasons for preferring my view in this case. If the process that is responsible for her belief is a process embodied by any conditional inference, regardless of the direction of the conditional, the following problem arises. Suppose Betty, unlike Melissa, is insensitive to the direction of the conditional for any conditional inference. Unlike Melissa, Betty commits the fallacy of affirming the consequent. Moreover, suppose that both Betty and Melissa form the belief that it rained on the basis of the beliefs that it is wet and the belief that if it is wet, then it rained. If considerations about what Melissa and Betty would believe in counter-factual circumstances (e.g., if the direction of the conditional were different) are irrelevant to determining the processes that are responsible for their beliefs, then we must say that they use the same process and thus their beliefs are equally justified. But this is counter-intuitive. Analogously, suppose that a child and I were to form a belief on the basis of a New York Times report. If we were to say that the process responsible for both of our beliefs is blind-trust, then we must say that our beliefs are equally justified. Again, this is counter-intuitive. On my view, we can easily explain why Melissa’s belief is more justified than Betty’s by appealing to the fact that the process responsible for Melissa’s belief is one that is embodied by modus ponens while Betty’s is not. Similarly, we can explain why the testimonial beliefs of adults are justified while the testimonial beliefs of children or gullible people are not. There are different processes that are responsible for the beliefs. The proc-

262 ess responsible for their beliefs, but not ours, is actually blind-trust because they, unlike us, would have formed the same belief had it been any other testimonial source. Counter-factual situations involving what agents would believe in different situations must be relevant in order to allow us to explain the difference in the justificatory status of beliefs. If two agents form the same belief under the same circumstances, we need to be able to explain why one agent’s belief is justified while the other’s is not or why one agent’s belief is more justified than the other’s. Admittedly, we do not want to consider all counter-factual situations, but in the case of testimony, we do, I take it, want to draw a distinction between how we form our beliefs and how children form their beliefs. So far, I have argued that if testimonial beliefs are epistemically basic in the foundationalist sense, it does not follow that beliefs formed on the basis of blind-trust are prima facie justified. In this last section, I consider a different sense in which testimony is epistemically basic, again for the purpose of determining whether blind trust beliefs are prima facie justified. III: Source Basicality and Blind-Trust A different way in which testimony or other cognitive practices, such as perception and inference, can be epistemically basic concerns whether agents must possess evidence for believing that the cognitive practice has a certain feature (e.g., being reliable) in order for the practice to produce justified beliefs. The issue here can be put in other ways, and does not have to be framed in terms of practices. In forming a justified belief, do agents have to possess reasons for believing that they have good reasons or is it sufficient for the agent to have good reasons? Do agents have to possess evidence for believing that a process is reliable or is it sufficient for process to be reliable? The issue here is whether a plausible account of justification must include a higher level requirement, what I call ‘an iterative requirement’ on justification. An account that rejects an iterative requirement is one that treats certain cognitive practices as basic sources of justification. Thus, a cognitive practice is a basic source of justification if and only if the agent does not have to possess any evidence about the reliability of the practice (or perhaps some other relevant epistemic feature) in forming a justified belief on the basis of that practice.

263 Two familiar problems motivate rejecting an iterative requirement for justification. The first has to do with circularity. For many, if not all, of our cognitive practices such as perception and induction, it is impossible to possess a non-circular justification for believing that the practice is reliable. As Alston writes, “And to show, against the skeptic, that perception is a source of justified belief (knowledge), we have to show that some mode of forming perceptual beliefs is reliable…The main difficulty is that there seems to be no otherwise effective way of showing this that does not depend on sense perception for some or most of its premises. Take the popular argument that sense perception provides its veridicality by the fact that when we trust our senses and build up systems of belief on that basis we have remarkable success in predicting and controlling the course of events. That sounds like a strong argument until we ask how we know that we have been successful at predicting and control. The answer is, obviously, that we know this only be relying on sense perception. Somebody has to take a look to see whether what we predicted did come to pass and whether our attempts at control were successful.”(Alston, 1989, 2)10

If perception is our only interface with the external world, how else could we verify that our perception is reliable without relying on perception? A similar problem arises for our inductive belief forming practices. As Hume noted, how can we verify induction without relying on induction? So if a circular justification for any belief forming process is all that is available, any reasons for believing that we have good reasons are not worth having. The second problem that motivates rejecting an iterative requirement is the threat of a regress. If agents must possess good reasons for believing that they have good reasons, why not also require that they possess good reasons for believing that they have good reasons for believing that they have good reasons, and so on? Without any reason for thinking that this kind of iterative requirement can be stopped at a certain level, we are led to a regress, and to a requirement that no agent could satisfy. Given these two problems, it would seem that in order for an agent to have a justified belief, the agent should only be required to possess good reasons for the belief, and so some of the agent’s cognitive practices are epistemically basic. As one can see, this latter sense of epistemic basicality (call it ‘source basicality’) is motivated in a similar way as the foundational sense of epistemic basicality. The familiar problem facing the foundationalist is 10 For a more detailed version of the circularity problem, see Alston 1986.

264 that if justified beliefs acquire their justification from other justified beliefs, then three options are available: 1) there is an infinite chain or regress of justified beliefs, 2) there is a circle of justified beliefs, or 3) there are justified beliefs that do not acquire their justification from other beliefs. The first two options seem implausible, and so we are led to foundationalism and to basic beliefs in the foundationalist sense. In a similar way, what motivates source basicality is the fact that we seem to be confronted with three different options: 1) there is an infinite regress of having good reasons for having good reasons for having good reasons, etc., 2) we have a circular chain of justification insofar as our cognitive practices justify one another, or 3) we can simply rely on our cognitive practices without possessing any evidence that they are reliable. Although the problems motivating source basicality and foundationalism are similar, they are not exactly the same. The regress problem facing the foundationalist does not involve the agent having beliefs about their own cognitive states, but this is not true of the iterative regress problem facing the source foundationalist. This difference is important for it is much more plausible that an agent might have a long, but not infinite chain, of beliefs than it is for an agent to have a long iterative belief (i.e., a belief about a belief about a belief, etc.). It is also worth pointing out an important difference between epistemic basicality in the foundationlist sense and source basicality. It is possible for beliefs to be produced by basic sources of justification that are not epistemically basic in the foundationalist sense. Consider, for example, a belief formed on the basis of an inductive inference. If inductive inferences are basic sources of justification, these inductive beliefs are not epistemically basic in the foundationalist sense since the justification of these inductive beliefs will depend on the justificatory status of the input beliefs of the inductive inference. This second difference raises an interesting question: if a belief is epistemically basic in the foundationalist sense and it is produced by a practice P, does it follow that P is a basic source of justification? Yes. If a belief is epistemically basic in the foundationalist sense, its justification does not depend on the justification of any other beliefs, including beliefs for or evidence about the reliability of the practice. So if the justification of a belief does not depend on this evidence, then an agent cannot be required to possess evidence about the reliability of a speaker in forming a justified belief.

265 Let us turn now to issues regarding testimony. There are two questions that concern us: 1) is testimony a basic source of justification? and 2) if testimony is a basic source of justification, are blind-trust beliefs prima facie justified? Of course, if the answer is yes to both questions, then blindtrust beliefs are prima facie justified. But why should we think that both questions have affirmative answers? I suspect the following rationale might underlie such intuitions. First, the reason why we should think that testimony is a basic source of justification is that if we reject a general kind of iterative requirement for justification, then we ought to reject it in the particular case of testimony. Second, consider the following reason why we should think that if testimony is a basic source of justification, then blind-trust beliefs are justified. In any case of testimonial belief formation, an agent forms a belief on the basis of the following kind of process: forming a belief that p on the basis of interpreting someone as having asserted that p. If this is right, then rejecting an iterative requirement on justification is tantamount to rejecting the requirement that an agent has to possess evidence that the speaker is reliable. Finally, if an agent does not have to possess evidence about the reliability of speakers, then blind-trust beliefs are prima facie justified. It would seem therefore that blind-trust beliefs are prima facie justified. This is the most compelling argument I have heard so far for why we should think that blind-trust beliefs are justified. In showing where this argument goes wrong, let me begin by noting an important point about what rejecting an iterative requirement on justification entails. Rejecting an iterative requirement does not entail that agents can use any process in forming a justified belief. This is because rejecting an iterative requirement only tells us what agents don’t have to do in forming a justified belief; however, it does not tell us what agents have to do to form a justified belief. To see this point more clearly, consider a reliabilist account of justification that rejects an iterative requirement. Rejecting an iterative requirement in the case of our inferential and perceptual practices does not mean that any inferential or perceptual practice can produce justified beliefs. Inferential processes such as wishful thinking or hasty generalizations still produce unjustified beliefs. The same is true of testimonial belief forming practices. The fact that there should not be an iterative requirement on our belief forming practices does not mean that any testimonial belief forming practice can produce justified beliefs. In short, what we need to be wary of is the following inference: if we do not have to possess evidence about the

266 reliability of a process, the process is therefore a basic source of justification. Returning to the above rationale for the Blind-Trust account, I see two problems with the argument. Recall first the rationale for the claim that testimony is a basic source of justification. In rejecting the general iterative requirement, we thereby reject the iterative requirement in the particular case of testimony, and so testimony is a basic source of justification. But as I just noted, rejecting an iterative requirement for a practice does not thereby show that the practice is a basic source. For reliabilists such as Goldman (1979) and Alston (1989), we still need to know whether the practice is reliable. There is a related problem concerning the second claim: if testimony is a basic source of justification, then blind-trust beliefs are prima facie justified. The problem with this claim is that the term ‘testimony’ here is ambiguous. If it means ‘blind-trust’, then of course, the inference is trivially true, and it would beg the question since it would assume that blind-trust is a basic source of justification by showing that blind-trust is a basic source of justification. Now if by ‘testimony’ we mean something different (i.e., practice of only forming beliefs on the basis of a particular kind of speaker, experts let us say), then the inference clearly does not follow. Why should we think that testimony is a basic source of justification? If such a practice is basic, in the sense that we do not have to possess evidence for its reliability and it is reliable, then it does not follow that blind-trust beliefs are prima facie justified. The inference makes reference to two different testimonial belief-forming practices. The fact that certain inferential and perceptual practices are basic does not entail that inference and perception, in general, are basic. For it may turn out that certain testimonial belief forming practices are reliable, but blind-trust is not. In this paper, I have focused on a particular way of motivating the blind-trust account, one that involves appealing to considerations involving epistemic basicality. I hope to have to shown that many of our testimonial beliefs are not epistemically basic even though they might be psychologically basic. Moreover, I have tried to show that even if our testimonial beliefs are epistemically basic or if testimony is a basic source of justification, it does not follow that blind-trust beliefs are prima facie justified.11

11 I would like to thank Sondra Bacharach and the participants at the Epistemology of Basic Belief conference in Amsterdam, June 2001 for their helpful comments.

267 REFERENCES

Alpert, M.I. and W.T. Anderson. 1973. "Optimal Heterophily and Communication Effectiveness-Some Empirical Finding", Journal of Communication 23. Alston, William P. 1989. “A ‘Doxastic Practice’ Approach to Epistemology” in Knowledge and Skepticism, edited by Marjorie Clay and Keith Lehrer.Boulder: Westview Press. ——— .1986. “Epistemic Circularity”, Philosophy and Phenomenological Research 47. Annis, David. 1978. "A Contextualist Theory of Epistemic Justification", American Philosophical Quarterly 15. Burge, Tyler. 1993. “Content Preservation”, The Philosophical Review 102. Calder, B.J. et al. 1974. "The Relation of Cognitive and Memorial Processes to Persuasion in a Simulated Jury Trial", Journal of Applied Social Psychology 4. Coady, C. A. J. 1992.Testimony: A Philosophical Study. Oxford: Clarendon Press, Festinger, L. and Maccoby, N. 1964. "On Resistance to Persuasive Communications", Journal of Abnormal and Social Psychology 68. Foley, Richard. 1994. “Egoism in Epistemology” in Socializing Epistemology, edited by Fred Schmitt. Lanham: Rowman & Littlefield Publishers. Fricker, Elizabeth. 1994. “Against Gullibility” in Knowing from Words, edited by Bimal Krishna Matilal and Arindam Chakrabarti. Dordrecht: Kluwer Academic. Goldman, Alvin. 1979. “What is Justified Belief?” in Justification and Knowledge, ed ited by George Pappas. Dordrecht: Reidel. Hume, David. 1992. An Enquiry Concerning Human Understanding. Edited by L.A. Selby-Bigge and P.H. Nidditch. Oxford: Clarendon Press, Kornblith, Hilary. 1987. “Some Social Features of Cognition.”, Synthese 73. McCroskey, J.C. 1969. "A Summary of Experimental Research on the Effects of Evidence on Persuasive Communication.", The Quarterly Journal of Speech 55. McGuire, W. J. 1964. "Inducing Resistance to Persuasion: Some Contemporary

268 Approaches" in Advances in Experimental Social Psychology I, edited by L. Berkowitz. New York: Academic Press. Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press, Reid, Thomas. 1983. Essay on the Intellectual Powers of Man in Thomas Reid’s Inquiry and Essays. Edited by. R. Beanblossom and K.Lehrer. Indianopolis: Hackett. Rogers, E.W. 1983. Diffusion of Innovation. New York: Free Press. Webb, Mark Owen. 1993. “Why I Know About As Much As You: A Reply to Hardwig”, Journal of Philosophy 40.

ANDY HAMILTON

Proprioception as Basic Knowledge of the Body Proprioception is a faculty which is both familiar - of necessity, because it underlies the possibility of action - yet mysterious. It is the capacity which yields ordinary knowledge of bodily position and movement - what is often rather loosely termed “bodily awareness”. In Philosophy it has until relatively recently been neglected; indeed, in my experience it still remains rather unfamiliar to general philosophical audiences. Proprioception has been called a “sixth sense” of bodily awareness and is still sometimes referred to as the “muscle sense”. This article, however, challenges the ubiquity of the sensory model, and tries to pose an alternative to it. It argues that proprioception, like memory, is a kind of direct knowledge, and defends the claim that proprioception constitutes basic knowledge of one’s body 1 immediate and nonperceptual. I refer to „basic knowledge“ rather than “basic belief“ because, as will become clear, there is only a limited possibility of error in judgments based on proprioception. Husserl was probably the first philosopher to recognise the importance of proprioception, and the physical-intentional ambiguity concerning the 2 body which it suggests - although there are interesting precedents in Locke. Husserl’s account is a great advance on that of his post-Cartesian precursors, who almost universally regarded the body as a purely material and epistemically outer entity – one whose states, in contrast to those of the mind, are not known immediately and transparently. In Ideas Book II he argues that the human body is defined by intentional attributes of action and proprioception as well as by spatio-temporal material attributes: ...the Body is originally constituted in a double way: first, it is a physical thing, matter; it has extension, in which are included its real properties, its color, smoothness, hardness, warmth...Secondly, I find on it, and I sense „on“ it and „in“ Thanks for comments and discussion to Roger Squires, Jonathan Lowe and Louise Richardson, and to members of the audience at the Basic Belief Conference at The Free University Amsterdam, 2001. 1 Hamilton 2003 defends the claim that memory is also a kind a direct knowledge. 2 These are discussed in Hamilton (forthcoming).

270 it: warmth on the back of the hand, coldness in the feet, sensations of touch in the fingertips... [The Body is] a material thing which, as localization field for sensations and for stirrings of feelings, as complex of sense organs, and as phenomenal partner and counter-part of all perceptions of things...makes up a fundamental component of the real givenness of the soul and the Ego.3

Husserl distinguishes two senses of “the body”, der Leib, the „animated flesh of an animal or human being“ - the „lived Body“ or what may be referred to as “my body” - and der Körper, „inanimate physical matter“, the „mere“ 4 body or body viewed purely as a physical object. This contrast underlies the discussion in this article. The lived Body is sometimes referred to by commentators as the „mindful body“; a better alternative might be „the body5 in-action“. But talk of the lived Body or body-in-action is useful chiefly as indicating essential features of, simply, my body. Despite the efforts of recent commentators to portray him in a thoroughgoing anti-Cartesian light, the metaphor of animation of my body, and reference to the “real givenness of the soul and the Ego”, shows the residual Cartesian bias in Husserl’s 6 treatment. It was not until Merleau-Ponty’s investigation of the „body7 subject“ that the Cartesian bias was eliminated completely. These writers in the phenomenological tradition influenced Gareth Evans’ focus on the importance of bodily self-ascription in an account of self-consciousness. Their ideas, filtered through Evans’ work, have achieved currency among some analytic philosophers, if not those of a naturalistic 8 persuasion. However, the work of the phenomenological tradition suggests an alternative to the sensory model which has not received much attention, and which this article is concerned to develop. A nonperceptual account undermines the view that proprioception is a capacity for tracking a fundamentally material entity. Indeed, on this view proprioception should not be regarded as a conceptual superimposition on pre-identified and 3

Husserl 1989, sections 36 and 40: 153, 165. Husserl 1989, for instance section 62: 297. 5 It has also been termed the “Living Body” by Cassam 1997: 52, and by Bell 1990: 208; but I avoid this term because of it biological connotations. I am grateful for discussion on these points with Paul McDonald; they are treated in his 2000. 6 It is discussed below. 7 See B. Smith 1995 in B. Smith and D. Smith eds. 1995: 406-7. D. Smith 1995: 324-6 argues that the break with Cartesian dualism in Heidegger and in Merleau-Ponty is a development of insights found in Husserl. 8 Evans 1982. Husserl and Merleau-Ponty do not rate a mention in Bermúdez 1998 or Baker 2000, for instance. 4

271 individuated living Bodies. Rather, I would argue, proprioceptive knowledge is part of the material from which the latter concepts are formed. In this article, however, I am not concerned to vindicate the larger claim, but simply to defend a non-perceptual treatment of proprioception. My present concern with proprioception originates in a broader treatment of the nature of self-consciousness, and in particular of the phenomenon of immunity to error through misidentification (IEM) which Gareth Evans did much to clarify. This concept lies in the background to the present article, but should be briefly outlined. IEM is exhibited by a range of self-ascriptions including those of sensation, perceptual experience, belief, intention, memory and proprioception. I argue elsewhere that proprioceptive judgments of bodily position, posture and movement are immune to error through misidentification of the subject, while judgments about one’s body 9 based on visual perception or touch are not. That is, if I judge immediately that my legs are crossed, and then come reasonably to doubt that they are, it will be senseless for me to cite the original justification as a reason for believing that nonetheless someone’s legs are crossed. It will make no sense for me to say “Well, I at least just know that someone’s legs are crossed”. (Even the term “justification” sounds forced – since in the normal case I just know that my legs are crossed.) In contrast, when in unusual circumstances the subject judges that their legs are crossed on the basis of vision, there is no guarantee of IEM. „I just know“ functions in the same way that the continuous-verb justification “I remember o-ing” does in the case of personal memory; indeed a direct knowledge account of proprioception is structurally analogous in important respects to the direct knowledge account of memory. Thus IEM extends to bodily as well as psychological self-ascription - a significant development of Wittgenstein’s original contrast in The Blue Book 10 between “I”-as-subject and “I”-as-object uses. As in the case of memory, traditional accounts of proprioception militate against a proper understanding of the IEM – hence a further reason for the present discussion, in addition to its intrinsic significance. The preceding considerations, I believe, support the anti-materialist view that “My body” is the body of which, when I am conscious, I have selfconscious knowledge, and which I can move basically sense – as in Danto’s account of basic action. Distinctively self-conscious knowledge - that which necessarily yields knowledge only of the subject - is manifested in IEMexhibiting judgments of posture, orientation, intention and action. “My body 9

Hamilton 1995, Hamilton (forthcoming). Wittgenstein 1969: 66-7. The claims in these paragraph are defended in Hamilton 1995, and (forthcoming). 10

272 is the body of which I have self-conscious knowledge...” functions only as an elucidation of the concept of my body, and not as a criterion of identity, for there is no conceivable use for such a criterion. The phrase „my body“ emphasises the subject’s knowledge of and concern for their own (living) 11 body, and was used in this sense by Carnap. The IEM claim suggests that proprioception is not simply a mode of awareness of events that happen to occur within the body’s boundaries, and this is the position I will defend. My claim is that proprioception does not just happen to yield knowledge only of one subject – like seeing oneself in the mirror or by looking down at one’s body. The elucidation „My body is the one of which I have self-conscious (IEM-exhibiting) proprioceptive knowledge and which I can move basically“ suggests, rather, that proprioception and bodily identity form a circle of interdefined concepts. Just as the concepts of personal identity and selfconscious, IEM-exhibiting capacities notably memory interlock, so do the concepts of bodily identity and self-conscious, IEM-exhibiting bodily 12 awareness and movement. Underlying these claims is scepticism concerning the traditional mental-physical distinction. „I am not that hard up for categories“ was Wittgenstein’s justified riposte to the question of whether there is one 13 substance or two. The dichotomy of mental and physical is a crude one, and the ambiguous status of the body is the crucial illustration of its arbitrariness and inadequacy. It would be some advance on materialism to say, still using the terminology of the traditional distinction between mental and physical, that the body is a psychological as well as a material unity. But this remains an imperfect transitional position. Merleau-Ponty, in his account of bodily intentionality, was more radical and more correct when he wrote that „The experience of our own body...reveals to us an ambiguous mode of existing“ - neither thing nor consciousness. He continues: “If I try to think of it as a cluster of third person processes...[these] are all obscurely drawn together and mutually implied in a single drama. Therefore the body is not 14 It follows, I believe, that materialism concerning selfan object...”. consciousness - the view that I am conscious of myself as a material entity, defended by Cassam in Self and World - is a misguided way of opposing 15 Cartesianism. But these wider considerations concerning selfconsciousness, and the concept of IEM itself, remain largely in the 11

Carnap 1967, sections 129-131 - a discussion which shows the influence of Husserl. This general position was defended in Hamilton 1995. 13 Wittgenstein 1980, II: para. 690. 14 Merleau-Ponty 1962 : 198. 15 Cassam 1997. 12

273 background in the rest of the article, the arguments of which are, I hope, independently intelligible and significant. 1. Proprioception as direct, non-inferential knowledge: rejecting the image theory (i) varieties of proprioception, and proprio-blindness First, a brief outline of the varieties of proprioception is necessary. The core capacity of proprioception yields knowledge of bodily position and movement. Kinaesthesis is knowledge of movement of parts of the body, as opposed to their posture or position. It is one of the main categories of proprioception, though confusingly is sometimes used as an alternative term for proprioception itself – perhaps understandably since knowledge of 16 movement requires knowledge of position. Other varieties are knowledge of fatigue and warmth and cold (as opposed to merely feeling tired, hot or cold); the vestibular system in the inner ear that gives information about balance and posture; interoception (the visceral sense); and „visual proprioception“, the term coined by J.J. Gibson to describe the kinaesthetic function of vision, which enables the subject to differentiate between a change of place by the observer and a change of state of an external object changes that are reversible by movement of the observer (back to the original 17 Interoception yields position of observation) and those which are not. knowledge of the non-muscular organs, blood-vessels and intestines, and is a function of the autonomic nervous system. (Some researchers maintain that one can learn to distinguish visceral events just as well as muscular ones.) Proprioception also embraces self-locating capacities which involve more than mere internal or surface bodily knowledge - such as knowledge of orientation in space, for instance that I am lying down or standing up. Indeed there is a case for arguing that knowledge of orientation relative to objects in the immediate environment - for instance, that I am standing in front of Marble Arch - is distinctively self-conscious because IEM, and thus on a continuum with proprioception. Its many varieties led the psychologist J.J. Gibson to reject the traditional idea of proprioception as a specific “body sense”, referring instead to self-specifying information which cuts across the 18 different sensory modalities. I will return to his view towards the end of 16

Cole 1991: xix-xx, corrects this error. Gibson 1966: 37-8. A more complete list of the varieties of proprioception is found in Bermúdez et al. 1995: 13, and Bermúdez 1998: 132-3. An excellent discussion of Gibson's theoretical perspective is found in Reed 1988. 18 Gibson 1966. Husserl had earlier commented on the interwovenness of perception and 17

274 this article. First I will attempt to refute directly the view that proprioception is sensory, while arguing – as Gibson does not – that it is definable by the direct character of the knowledge that it yields. One further preliminary issue. Considering cases of proprioceptive deficit assists the imaginative effort involved in trying to grasp a familiar yet apparently obscure capacity of proprioception. Anscombe’s Sensory Deprivation Tank thought-experiment postulates radical if temporary 19 proprioceptive deficit – what may be termed proprio-blindness. Actual proprio-blindness is caused by conditions ranging from stroke to viral infection, and is more common, at least in a partial and mixed form, than is 20 often imagined. Devastating cases of proprio-blindness such as Ian 21 Waterman’s are rare. Waterman’s profound proprio-blindness seems to have been caused by an immune reaction, devastating his nervous system from the neck down while leaving the vestibular system and visual proprioception still functioning. Unlike other severely proprio-blind patients, however, Waterman did not remain in a wheelchair but re-acquired mobility through the constant effort - the „daily marathon“ - of a mostly visual tracking of his body. His knowledge of bodily position and posture is sustained by constant looking and checking. He retains a capacity to feel pain, temperature and muscle fatigue - presumably he has some knowledge of the location of pain, at least as expressed through primitive reactions of pain-response such as movement. Without this residual proprioception his extraordinary achievement in regaining bodily control and learning to walk again would have been impossible. General ignorance of the prevalence of partial proprio-blindness is attributable in part to a failure to recognise the distinct neuropathies of the efferent and the afferent systems. Proprio-blind patients suffer from the latter but not necessarily the former. Stroke victims usually suffer from a complex mixture of damage to both efferent and afferent systems; some parts of a limb may be truly paralysed, while others exhibit loss of proprioception, bodily movements, as discussed by B. Smith 1995: 403. 19 Anscombe 1981. 20 The experiment is discussed in Hamilton 1991. Talk of „proprio-blindness“ should not be taken to imply a perceptual model of proprioception. 21 His heroic story is told in Cole 1991, and in a BBC2 TV programme, broadcast on 16.10.97. Cole writes that „damage to his nerves was extraordinarily, perhaps uniquely, specific. It had affected some of the sensory fibres, but none of the motor nerves“ (p. 2; further physiological details are found on pp. 24-34). For earlier discussion of a proprioblind patient see Sacks 1985. I am indebted to CB for discussion of the experience of proprio-blindness.

275 temperature awareness or sensation of pain. Though regarded as paralysed, proprio-blind patients may be only effectively so because, as neurologists put it, the brain cannot tell a muscle what to do if it does know where it is. Ian Waterman found an unorthodox route to this knowledge by visually tracking his body and thereby overcome his effective paralysis. (ii) proprioception as direct, immediate knowledge The position defended here is that proprioception yields direct, immediate and spontaneous knowledge of the body - centrally its position, posture and movement. This knowledge is immediate not just in the sense that it is noninferential but also in the sense that the subject never has to do anything to acquire it. Indeed, I do not really „acquire“ it at all; immediate knowledge is knowledge for which I do not have to do anything in order to have it; I „just know“, for instance, that my legs are crossed when they are. So proprioception differs both from knowledge based on bodily sensation, and from perception of the body by the five senses. Other varieties of proprioceptive knowledge such as awareness of fatigue and temperature must be distinguished from avowals of bodily sensation. However, the focus in the present discussion will be on the central cases of kinaesthetic knowledge, viz. position, posture and movement. The characterisation of proprioceptive knowledge as direct, immediate and spontaneous is preferable to Anscombe’s term “non-observational”, which appears in her influential remarks on bodily awareness. Nonobservational knowledge, she believes, includes knowledge of one’s own intentional actions and of the causes of some involuntary movements, as well as knowledge of the position of one’s limbs - that is, proprioceptive knowledge. Her justification for describing this knowledge as nonobservational is that “nothing shows [someone] the position of his limbs; it is not as if he were going by a tingle in his knee...Where we can speak of separately describable sensations, having which is in some sense our criterion 22 for saying something, then we can speak of observing that thing...”. Anscombe is right to doubt whether there is a separately describable basis for proprioceptive judgment – an issue picked up later. But her terminology is unsatisfactory, because the term „non-observational“ implies a contrast with perceptual knowledge, while Anscombe’s elucidation of the term seems to contrast it with inferential knowledge based on bodily sensation. Anscombe is wrong to say that, in itself, observation implies 22

Anscombe 1963: 13. In common with many earlier writers, she refers to “bodily awareness“ rather than „proprioception”.

276 „separately describable sensations“, since this suggests that observation is inferential. If this were so, knowing that my legs are crossed by looking could also count as non-observational, since it does not seem to involve “separately describable sensations”. This makes nonsense of her criterion. This error results in a conflation of two distinctions: between inferential and non-inferential knowledge, and between (self-) knowledge by the five senses and distinctively self-conscious self-knowledge (knowledge which necessarily concerns only the subject). Both contrasts need to be made explicit and each, I will argue, involves a distinct challenge to the direct knowledge account. There are further problems with Anscombe’s terminology, for it is not even the case that “non-observational” implies “non-perceptual”. Despite the history of so-called observation sentences in the philosophy of science, there is something artificial about equating ordinary perception with observation, since “observation” implies attentive perception, usually visual. The characterisation of proprioceptive knowledge as “direct, immediate and spontaneous” is therefore greatly preferable to Anscombe’s term “non-observational”. (iii) rejecting the image theory My claim is that proprioception is not inferential knowledge, and that it should not be assimilated to knowledge by the five senses. Hence I reject both the image theory and the perceptual model. The image theory of proprioception, analogous to the image theory of memory, maintains that proprioceptive knowledge is inferential. It is a representative theory, and assumes that proprioception may be distinguished phenomenologically, by its feel - by the sensations associated with it. The presence of these sensations is the inferential basis of the proprioceptive judgment of bodily posture or position. Proponents of this view take the concept of a “body-image” rather literally; indeed many writers unsatisfactorily describe the subject’s total proprioceptive knowledge as a body-image, when the term is more appropriately applied to the highly distorted self-image possessed for instance 23 by anorexics. The image theory may be regarded as a particular version of the perceptual model; but I criticise it directly first, before targeting the latter model. It is important for the treatment of self-consciousness that these theories should be discredited. For one reason, they make q-proprioception appear more plausible, just as the image theory of memory makes q-memory more plausible. Q-proprioception is meant to be a peculiar kind of proprioception, involving an information-link between the subject and 23

Cole 1995 in Bermúdez et al eds. 1995 has further discussion of „body-image“.

277 someone else’s body, such that the subject allegedly registers information from it, thus contravening the IEM of proprioception. The image theory makes it seem more imaginable that the proprioceptive „feel“ could arise from the bodily state of a distinct subject. Interestingly, memory and proprioception are linked by Wittgenstein when he rejects the idea that kinaesthetic sensations advise me of the movement of my limbs: “It is the same with the idea that it must be some feature of our pain that advises us of the whereabouts of the pain in the body, and some feature of our memory 24 image that tells us the time to which it belongs”. The picture offered by the image theory is this. If my legs are crossed with one resting on the other, I will experience feelings of pressure, touch and so on. Even when my arm is by my side but not touching anything, there will be feelings of muscle tension or skin stretching, or tension in my shoulder. The image theory claims that such sensations “advise” me of the positions of my limbs, etc., presumably through inductive inference. For this account to be plausible, the sensations would have to be characterised independently of the state of whose occurrence they advise. This is possible when I infer from bodily sensations to the existence of a medical condition, having learned from experience that the feeling is associated with this condition – for instance when I infer, from the recurrence of stomach pains, that I have a gastric ulcer rather then indigestion. As Anscombe and Wittgenstein argue, in the proprioceptive case it is difficult to see how sensations could be characterised independently of the state of whose occurrence they advise. Wittgenstein’s position is presented in the following passage: My lower arm is now lying horizontally and I should like to say I feel that; but not as if I had a feeling that always goes with this position (as one would feel ischaemia or congestion) - rather as if the „bodily feeling“ of the arm were arranged or distributed horizontally, as e.g., a film of damp or of fine dust on the surface of my arm is distributed like that in space. So it isn’t really as if I felt the position of my arm, but rather as if I felt my arm, and the feeling had such and such a position. But that only means: I simply know how it is lying - without knowing it because....As I also know where I feel pain - but don’t know it because.....25 24

A more obscure version of the image theory seems to be defended by O'Shaughnessy 1995, in Bermúdez et al. eds. 1995. The editors of that volume state that Anscombe's view of proprioception “is rejected, implicitly or explicitly, by all the contributors” (p. 19), as if it were an outdated scientific theory rather than an enduring philosophical position. 25 Wittgenstein (1958) Part II, p. 185. Wittgenstein 1980 Vol I, para 786; the idea is also expressed in paras 784-5. See also Wittgenstein's discussion in his 1958 Part II, pp. 185-

278

I will return to these elusive remarks later, because they suggest an interpretation of the sense of “feeling” involved in proprioception which undermines the perceptual model as well as the image theory. Here I will address some considerations which at first sight offer support to the image theory. It should not be assumed that the acquisition of a learned capacity has to involve inference; or that information-processing amounts to inference. Children learn to control their movements, but an inferential model of this learning process is unconvincing; in particular, the suggestion of unconscious 26 inference should be rejected. However, there are adult learning processes where the claim of conscious inference may seem more plausible. In the practice of Alexander Technique, for instance, one learns to attend closely to one’s bodily posture, aiming to release bodily tensions and direct bodily use efficiently, by „thinking“ upwards for instance. The new habits which the Technique encourages one to acquire at first feel strange, but gradually become automatic. For instance, I may learn on the basis of close attention that I have a detrimental habitual use of stooping forwards, wrongly assuming that I was standing bolt upright. Alexander Technique attempts to make this knowledge immediate and habitual, by means of a process that is hard to describe, but which does not seem to involve a conscious inference; indeed the very elusiveness of the technique counts against the idea that inductive inference is involved. There are other cases where knowledge of bodily posture and movement is not immediate and has to be learned. In many of these, visual perception or touch is essential in refining proprioceptive knowledge, which is usually rough and approximate. Figure skating, ballet, playing a musical instrument and singing are examples; for instance, one may learn complicated dance positions with the aid of mirrors. Proprioceptive knowledge is not always accurate enough for such specialised purposes. If I am instructed to close my eyes and put my arms out horizontally, the result will be close enough; knowledge that my arm is horizontal, based on proprioception, means „horizontal as opposed to vertical or at 45 degrees“. In contrast, singing lessons or learning to speak a foreign language with a correct accent may involve more precise bodily knowledge; I may need to know where my tongue is when making certain sounds. Before the knowledge becomes automatic, the learning process involves exploratory 6; and Merleau-Ponty 1962: 93 on the localisation of pain. 26 See Budd 1989: 147-9. For other criticisms of an image theory, see Anscombe, „On Sensations of Position“ in her 1981, and Candlish 1996.

279 touch; „my tongue is touching my upper teeth“ is more like „My fingers are in the jelly“ than „My fingers are curled up“. (Exploratory touch occurs when I move my hand to feel part of my body or another object, passive or proprioceptive touch occurs when I feel the rain on my face; they are contrasted below.) Ian Waterman learned to control his body by a process of constantly looking which never became automatic or unconscious. His bodily knowledge must often involve inferences such as „When my body is at this angle, it’s likely that I’m about to topple over“. 2. Proprioception as direct, immediate knowledge: rejecting the perceptual model (i) The perceptual model While the image theory of proprioception has many proponents, the perceptual model - of which the image theory may be one variety - is almost 27 ubiquitous. As noted earlier, proprioception has been called a „sixth sense“ of bodily awareness or the „muscle sense“, and physiological affinities with touch seem to support the perceptual view. Indeed „proprioception“ is a contraction of „proprioperception“, that is, „self-perception“. The perceptual model is supported by physiological affinities between proprioception and touch. The proprioceptive nerve receptors in the muscles give feedback from joints, tendons and muscle spindles, while cutaneous receptors - those near the skin-surface - respond to touch. These receptors might be regarded as „organs of proprioception“ and „organs of touch“. If there is an organ of exploratory touch, one of the five senses, then it seems that there must be an organ of proprioception. However, I will argue that there are important differences between proprioception and the five senses in the character of the knowledge which they yield. This divergence is sufficient to undermine the standard conception of proprioception as a variety of perceptual knowledge which happens to be limited in its objects to the body and its surface. In assessing the perceptual model, two questions must be considered: how does proprioception differ from the five senses, and is this difference sufficient to show that proprioception is not a variety of perception at all? My conclusion is that proprioception has a status intermediate between, on the one hand, the five senses which yield perceptual knowledge, and on the other hand, sensations such as pain which figure in avowals, and which are not objects of knowledge at all. Thus the epistemic status of „I am cold“ is 27

In fact it is a difficult question how the image theory and the perceptual model are related; Bermúdez et al eds. 1995: 18-19, run the two together in their criticism of Anscombe.

280 intermediate between „The ice is cold“, and „I feel cold“ (or „I have a headache“) – though it is closer in certain respects to sensation. Now there are writers who assimilate proprioception to sensation by subsuming both under the perceptual model, thus neglecting the well-grounded distinction between sensation and observation. Their view updates the Cartesian model of sensations, which regarded them as objects of inner perception, mistakenly postulating a sensational object distinct from the awareness. The contemporary version regards pain not as an object of perception, but as a mode of perceiving one’s body; thus pain has a physical object distinct from the awareness. Martin for instance claims that the object of pain is the bodypart that feels painful, just as the object of visual experience is the physical object that one is perceiving: „...[O]ne perceives one’s body through 28 sensation, just as one perceives other objects through the five senses“. Sensation is surely not, as this view has it, a mode of perception. The most that can be conceded to the perceptual model is that the self-ascription of pain assumes proprioceptive knowledge, and that, as noted earlier, the experience of pain may itself be the basis for inferential knowledge, for instance that I have a gastric ulcer. Aside from its innate implausibility, it is hard to see how the immunity to error exhibited by avowals of pain can be accounted for on a perceptual model. Their authority implies that if X truthfully, attentively and comprehendingly asserts or avows „I have a pain in my leg“, and there is no relevant cognitive defect in the subject, then the truth 29 of „X has a pain in their leg“ is guaranteed. I can be mistaken about the causal origin of the pain, but not about its phenomenal location; and it is the phenomenal location which determines the truth of the avowal. It follows that apparently mistaken self-ascriptions resulting from referred pain and phantom-limb phenomena, often cited as counterexamples to the authority of avowals, are not convincing as objections. An example of referred pain would be a sinus condition which causes a headache, a pain not phenomenally located in the sinus region; but the fact that the headache has its immediate causal origin elsewhere does not make the avowal that I have a headache a mistaken one. A similar point applies against objections that cite phantom-limb phenomena, where the subject is inclined to avow „I have a pain in my left leg“ but has no left leg. Certainly the subject has a pain, and it is phenomenally located where the left leg would have been.

28

Martin 1995: 269. The implications of the authority of avowals are developed in Hamilton 2000 and (forthcoming).

29

281 (ii) the role of sensory orientation Although the distinction between sensation and observation is, I believe, well-grounded, proprioception seems to belong to neither category. If proprioception is treated simply as a mode of knowledge of one’s body, the perceptual model of proprioception seems inevitable. Nonetheless, it is possible to make a principled distinction between proprioception and sensory perception. An initial move is the suggestion that sensory knowledge involves reference to a possible action of looking, listening, tasting, smelling or touching, while no kind of action is required to gain proprioceptive knowledge. Only in the exceptional situation where the subject has to look or touch to find out - when they have partial proprio-blindness due to a stroke for instance - do they have to do something to acquire the knowledge that their legs are crossed. (My knee may also feel as if it has a bump on it - the skin feels stretched and so on, perhaps.) In the normal case, that is, where the knowledge is proprioceptive, it is not something that one gains or acquires. I „just know“. One explanation for the lack of action in the proprioceptive case might be that sense-organs exhibit directionality - they pick up information from a certain direction and may need to be re-oriented. I may need to concentrate my gaze, or strain to catch what someone is whispering. There is no such directionality in the case of proprioception; no equivalent to the moving in or focussing found in exploratory touch and the other senses. For this reason, looking at one’s legs, although it does not involve experiencing a sensation or applying a criterion – except in the sense of ownership perhaps - does not yield immediate knowledge. Immediate knowledge is knowledge which I do not have to do anything in order to have; I „just know“. To develop this contrast it is necessary to distinguish exploratory and proprioceptive touch. Exploratory touch occurs when I move my hand to feel part of my body or another object; passive or proprioceptive touch occurs when I feel the rain on my face, or someone treading on my toes, and is comparable to feeling bodily sensations such as pain. Exploratory touch and visual perception contrast with proprioception in that they are ways in which I can also gain knowledge of someone else’s bodily position, and so the judgments to which they give rise do not exhibit IEM. Exploratory touch, when stationary, proves hard to distinguish from proprioceptive touch, as in what Merleau-Ponty describes as „double sensations“: „When I press my two hands together, it is not a matter of two sensations felt together as one perceives two objects placed side by side, but of an ambiguous set-up in

282 30

which both hands can alternate the roles of ‘touching’ and ‘being touched’„. These double-sensations simultaneously yield knowledge of one’s body and one’s environment. When I feel by touch the stationary sphere in the palm of my hand, without moving my hand across its surface, both proprioceptive touch and exploratory touch - in this case stationary - seem to be involved. By means of proprioceptive touch I feel the pressure, the coldness and perhaps the smoothness or roughness of the sphere. By means of exploratory touch I feel the shape, the hardness, and the roughness of the sphere; but I do not feel these when my hand is quite stationary. (Feeling the shape of the box is not like feeling the edge of the box, because it involves inference.) Thus exploratory touch seems to be essentially active. How significant is this lack of action and directionality? Does it really suggest that proprioception is not a mode of perception? A first response by proponents of the perceptual model might be to argue that gaining knowledge on the basis of proprioception often does involve an action of some sort. Perhaps I sometimes need to move my limbs about in order to activate or alert my proprioceptive capacity and get coordinated. This however is not a correlate of the particular action often required with the five senses, I would argue. It may also be argued that there is an action of paying attention, as there is in the case of pain. But while it is true that one may focus one’s attention on particular sensations, as a means of gaining knowledge of one’s bodily position this activity is simply a charade. I can often be said to know that my legs are crossed, just in the sense that I do not try to get up without uncrossing them, and thus falling over. This knowledge is immediate. Alexander Technique requires concentrating attention on one’s posture, but the process is an elusive one which does not seem to involve an analogue of concentrating one’s gaze, or straining to catch what someone is whispering. An alternative approach for proponents of the perceptual model is to concede that no action is involved in the case of proprioception, but deny that this fact has any significance, arguing for instance that one should not expect directionality. Compare exploratory touch. If, as proponents of a perceptual model might maintain, there is a sense-organ for touch, then directionality would result from the sense-organ moving to remain in physical contact with the object – for instance I move my hand, and ensure that the cutaneous nerve-receptors continue to remain in contact with the object being tracked. In the case of proprioceptive touch, the response continues, the organs are already in the place which the information concerns, and so of course they do not need to be moved. However, „the place which the information concerns“ 30

Merleau-Ponty 1962: 93.

283 is within the body-subject, and as will be argued below, this makes a vital difference. The most persuasive line of objection finds exact parallels in the behaviour of perception and proprioception. If I can often be said to know that my legs are crossed, in the sense that I do not try to get up without uncrossing them, then in a similar way I might know through hearing it that there is a car behind me when I cycle along the road, in the sense that I do not try to turn right without indicating. In neither case, it may be argued, is attention required. On the other hand, the objection continues, forming the judgement „There is a car behind me” does require attention, and similarly for the judgement „My legs are crossed”. There is a related objection to the distinction between exploratory and proprioceptive touch. Proprioceptive touch occurs when I feel the rain on my face, but surely, it may be argued, exploratory touch „occurs“ at the same time, since I also find out about something in the world, viz. that it is raining. Why distinguish exploratory and passive touch, when for the other senses, active and passive do not mark distinct senses? Raindrops felt on my head are like a car coming into my field of vision; touching my head is like peering out of the window to catch sight of the car that has just disappeared up the road. I would respond that the difference is that the judgment „My legs are crossed” is always – barring exceptional circumstances such as radical proprio-blindness - redundant. In contrast, even when seeing and feeling do not actually involve action, there is always a possible action of looking or touching. One could gain further knowledge of the surface of the sphere by means of active exploratory touch, or turn to look at the car as it disappears up the road. There is no parallel here with proprioceptive touch or with proprioception in general. I avoid objects in my path automatically, just as I know that my legs are crossed; but it would be wrong to say that I „just know“ that there is a large tree in front of me, or I „just know“ where the piano is when I can remember where it is. I know that it is in front of me because I can see it or because I can remember where it is. Knowledge of the immediate environment is not immediate knowledge. There are further reasons for denying that proprioception is a sense. If it were, feedback about error would be required. However, there is no sensespecific feedback or correction in the case of proprioception - feedback comes entirely from the other senses. But there are more fundamental ways in which proprioception and the other senses are interdependent. For it is surely the case that sensory orientation assumes proprioception – in particular visual proprioception or visual kinaesthesis. One cannot orient one’s sense-

284 organs effectively without knowing, for instance, whether one is moving or stationary oneself – information derived from proprioception. This is an insight associated with the psychologist J.J. Gibson, who was responsible for familiarising us with the concept of visual proprioception. His work places a particular interpretation on the claim that proprioception is not one of the senses, but while endorsing aspects of his account, I will distinguish my own treatment from it. 3. The Gibsonian account and a distinct sense of „feel“ There is important common ground between the position of psychologist J.J. Gibson and the phenomenological tradition. Gestalt psychologists influenced Gibson as well as Merleau-Ponty, and a philosophical treatment of the body and self-consciousness should attempt to draw on each. Gibson formulated his view of proprioception in reaction to that of Sherrington and his contemporaries, who assumed that „each sense had to have 31 its specialised receptors that could excite corresponding sensory nerves” – a version of the perceptual model. In The Senses Considered as Perceptual Systems, Gibson sought to clear up a „deep theoretical muddle”: „The verb to sense can mean either to have a sensation or to detect something and the two meanings are radically different“, he argues, and he intends the latter meaning. To sense, whether ourselves or the objects around us, is to 32 detect things. In a discussion of the vestibular system, Gibson points out that there are „no introspectively clear impressions from this organ”, implying that neither perception nor proprioception need be founded on 33 sensation. Gibson treats proprioception as „a component of the function34 ing of all the perceptual systems”, and claims that proprioception and perception are interdependent. They are distinguished in terms of their function, and not by the receptors, sensory nerves, or sensations it involves; a perceptual system is not to be identified with a sensory modality, a specific channel of input which is transmitted by specialised sensory nerves to the brain, where it is processed. According to Gibson, perceptual systems are the means by which we actively pick up meaningful information about the environment: „All the perceptual systems are propriosensitive as well as exterosensitive, for they all provide information in their 35 various ways about the observers’ activities” . These two functions are 31

Gibson 1966: 33. Gibson 1979: 115; Gibson 1972. 33 Gibson 1966: 71. 34 Reed 1988: 227. 35 Gibson 1979: 115 32

285 interdependent: „an environment implies something that is surrounded, and therefore awareness of the environment implies an awareness of the body existing in the environment. Equally, an awareness of the body entails 36 some feeling of its relation to the surroundings” . Gibson regards vision as kinaesthetic in that it registers body movement as much as the muscle-joint-skin system and the inner ear; like these, 37 „vision obtains information about both the environment and the self“. He regards visual proprioception, the least automatic and „highest” variety of proprioception, as central to action-guidance, especially in any new task; the movement sensitivity of the visual system dominates that of the muscular and articular systems at least in manipulation and locomotion: „we see where we are going and the layout of the environment through which we are going at the same time…vision is kinaesthetic in that it registers movements of the body just as much as does the muscle joint system and 38 the inner ear system”. Visual kinaesthesis explains the perception of passive as well as active movement - how a passenger in a car, engaged in no overt activity, can perceive that it is they who are (passively) moving, and not the trees, buildings and road rushing past. Other senses too have a propriospecific component: „information about the self is multiple, and…all kinds are picked up concurrently…An individual not only sees himself, he hears his footsteps and his voice, he touches the floor and his tools, and when he touches his own skin he feels both his hand and his skin at the same time. He feels his head turning, his muscles flexing and his joints bending. He has his own aches, the pressures of his own clothing, 39 the look of his own eyeglasses - in fact, he lives within his own skin” . Gibson believes that all these varieties of proprioception are to be distinguished from perception, which they nevertheless accompany, on the basis of their function of self-sensitivity or egoreceptivity. A defence of a Gibsonian account of proprioception has recently been presented by Bermúdez in The Paradox of Self-Consciousness. Bermúdez maintains that perceptual experience does not only provide information about the external world, but is inextricably combined with self-specifying 36

Gibson 1968. Gibson 1979: 183. 38 Gibson 1966: 36, 1979: 185. Gibson's interest in visual proprioception grew out of his wartime experiments in training pilots, where it was found that while the plane tilts, the pilot's horizon or „straight ahead” does not slope but remains at eye level; whilst non-visual proprioception is often unreliable, the pilot’s visual world remains stable – see Reed 1988: 78. 39 Gibson 1979: 115. 37

286 information without which the former would be of little use; he believes that Gibson’s ecological approach to perception shows that perceptual experience 40 is a source of first-person nonconceptual contents. I am sympathetic to Bermúdez’ view that proprioception counts as a „genuine form of selfconsciousness“ and indeed with his claim that there is a close connection 41 between IEM and the essence of first-person judgments. His Gibsonian picture leads to some persuasive arguments in favour of the view that knowledge of the position of one’s own body is a continuous model based on continuous feedback from touch and sight and other sensations. However, I would argue against Bermúdez that Gibson’s claim that proprioception is the product of other senses may equally be interpreted as a rejection of the perceptual model. Furthermore, there is at least a tension between Gibson’s picture of proprioception as continuous with other perceptual modalities, and Bermúdez’ claim that proprioception is a „form of self-consciousness“. My suspicion is that the Gibsonian picture is too deflationary of selfconsciousness, since it fails adequately to acknowledge the IEM status of proprioception, and does not allow a distinctive ground which generates the phenomenon. Nonetheless, my principal concern is to reject the perceptual model of proprioception, and it is possible that a development of Gibson’s account will fulfil this requirement while acknowledging self-consciousness. I will conclude by offering some final objections to the perceptual model. Although Gibson distinguishes two senses of „sense“, it is also necessary to distinguish two senses of „feel“. The denial that proprioceptive knowledge is immediate - I do not have to do anything in order to have it, I „just know“, for instance, that my legs are crossed - arises, I will argue, from a misunderstanding of the concept of feeling characteristic of the image theory and perceptual model. Proponents of the perceptual model claim that in shifting my attention from the object being touched to the sensations which I enjoy while touching it, I am simply moving my attention from objects that lie outside one of my bodily boundaries, for instance the surface of a hand, to what is going on at or beneath that boundary. Hence according to Michael Martin our bodily experiences „have as part of their phenomenological 42 content that the region felt falls within one’s body“. Although I believe that 40

Bermúdez 1998: 114. Bermúdez 1998: 144; and elsewhere. Bermúdez's discussion of proprioceptive content is fertile and suggests many questions for further discussion, especially concerning the dual criteria for location and the concept of a hinge 1998: 154-61. 42 Martin 1995: 270, 273. He denies that bodily awareness is a matter of „nonperceptual states immediately caused by action on the body“. One target here may be the implausible Cartesian introspectionist view that bodily sensations are „nonperceptual sensory 41

287 content is essentially propositional and would therefore question the very notion of „phenomenological content“, I wish to focus on Martin’s assumption that the same sense of „feeling“ is involved in both exploratory and proprioceptive touch. I would argue in contrast that „feel“ has a quite different sense in the proprioceptive case. Exploratory touch aside, I do not „feel“ my body in the same way that I feel other „regions“. One can become aware of specific bodily feelings, but not in the way that an image or perceptual account requires. Feelings of pressure, proprioceptive touch, muscle tension or skin stretching do not involve feeling in the perceptual sense exhibited in exploratory touch. Normally - that is, when the subject is not suffering from proprio-blindness - the question „How did you know that your legs were crossed?“ is redundant. However, some attempts to answer it are misleading as well as otiose. „I felt that they were crossed“ mistakenly implies an action of touching. (Compare „How did you know that there’s a bump on your knee?“ „Because I felt it“.) „My legs felt (as if they were) crossed“ is also wrong, since it implies a feeling distinct from the knowledge that my legs are crossed. The only sensible answer to what in the normal case is the purely theoretical question „How do you know that your legs are crossed?“ is „I just know“. To return to the Wittgenstein passage quoted earlier, which defends this claim whilst apparently leaving space for a non-perceptual sense of „feel“: My lower arm is now lying horizontally and I should like to say I feel that; but not as if I had a feeling that always goes with this position...- rather as if the „bodily feeling“ of the arm were arranged or distributed horizontally, as e.g., a film of damp or of fine dust on the surface of my arm is distributed like that in space. So it isn’t really as if I felt the position of my arm, but rather as if I felt my arm, and the feeling had such and such a position. But that only means: I simply know how it is lying - without knowing it because....As I also know where I feel pain - but don’t know it because.....43

Wittgenstein here assimilates knowledge of bodily position with knowledge of the location of pain in a quite different way to Michael Martin’s proposal discussed earlier. But that assimilation constitutes a denial of the perceptual model also; for the claim that it is „as if the ‘bodily feeling’ of the arm experiences“; another might be the direct knowledge position. A perceptual view is pervasive in Bermúdez et al eds. 1995, and also in Bermúdez 1998, for instance p. 135. 43 Wittgenstein 1980 Vol I, para 786; the idea is also expressed in paras 784-5. See also Wittgenstein's discussion in his 1958 Part II, pp. 185-6; and Merleau-Ponty 1962: 93 on the localisation of pain.

288 were...distributed horizontally“ suggests that I am not perceiving an object at all. Wittgenstein’s implication might be that „I felt my arm“ in the proprioceptive sense can only mean „My arm wasn’t asleep/ frozen/ anaesthetised, and I just know how it is lying“. It suggests that the subject of proprioception is not, properly speaking, simultaneously an object of perception, and that proponents of the perceptual model of proprioception are mistaken in claiming this. It may be felt that nothing decisive depends on whether proprioception is regarded as a mode of perception. Certainly there is a case for claiming that „perception“, like „self-consciousness“, is a philosopher’s term of art, and that intuition has a restricted role in deciding whether a faculty counts as perceptual. However, the perceptual model is not innocuous. Its claim that I am simply moving my attention from objects that lie outside one of my bodily boundaries to what is going on at or beneath that boundary, makes it seem contingent that I have proprioceptive knowledge only of my own body. Although proponents of the perceptual model often acknowledge that proprioception is unique among modes of perception in providing knowledge only of one object and its parts, they cannot give an adequate explanation of why this should be. Bermúdez does make a serious attempt at such an explanation, arguing that the perceptual model meets identification and multiple-object constraints; but satisfaction of the latter constraint, whilst allowing him to reject the possibility of q-proprioception, results in the implausibly broad extension to the concept of self-consciousness noted 44 earlier. The perceptual model denies the conceptual interdependence of proprioception and bodily identity and thus embodies the residue of a traditionally Cartesian picture of bodily awareness. In contrast, the direct 45 knowledge account is consistent with the concept of a body-subject. The assumption that proprioception yields knowledge is the strongest reason for supporting a perceptual model, so it is important to qualify this assumption. It rests on the fairly remote possibility of error of such judgments as “My arms are folded”; though immediate, this judgment is not infallible. But the possibility of error should not be overstated. The subject’s knowledge of their posture and position is mostly so reliable as to be almost certain. Except in pathological cases, errors involve inattention, and are limited to rather complex situations - intertwining one’s hands for instance 44

Bermúdez 1998: 137-51, especially p. 144. Another example is Martin 1995: 267, who argues, I think inconsistently, that an account of „what ties the content of bodily awareness to a particular object...does not seem to be one that we can provide purely a priori” (p. 283). 45 These issues are pursued in Hamilton (forthcoming).

289 or to the errors of detail that result from the imprecise nature of propriocep46 tive knowledge discussed earlier. In contrast, it is hard to see how I could be mistaken that my legs are crossed, except where phantom-limb phenomena are involved. (Such phenomena often involve pain in the missing limb, but patients also (mis-) report its position too, for instance claiming that it 47 feels as if it is buckled under them. ) To adapt a well-known quotation, therefore, it seems that it cannot be said of me except perhaps as a joke, that I know that my legs are crossed. The joke, as with knowing that I am in pain, would be a feeble one – scarcely a joke at all. (One might also say „He knows that his legs are crossed“ when the subject has just woken up from an anaesthetic and is becoming unconfused.) When one compares proprioception and memory, then arguably memory is also a faculty which yields direct, immediate and reliable knowledge, and is not a mode of perception. But the kind of knowledge which proprioceptive judgments express is of a curiously muted variety, the knowledge-claims oddly unconvincing. However, to deny that proprioception mostly does yield knowledge would be to make the unconvincing assimilation of judgments of bodily position to avowals of sensation. Proprioception has an ambiguous status, therefore. It involves knowledge of an object - one’s body - yet that knowledge is groundless, at least in what are usually referred to as „internalist“ terms. In that sense, one could say, proprioception constitutes basic knowledge of one’s body.

46

The present writer once had the bizarre experience of resting his chin on his hands and thinking that the skin on his thumb was rough, when the roughness turned out to be the beard stubble he was rubbing. In „The Body-Body Problem“ and other articles in his 1999, Danto several times suggests that immediate bodily knowledge might be indubitable - though like Anscombe he does not use the term „proprioception“. 47 A more extreme example is A.R Luria's patient Zasetsky, „The Man with a Shattered World“, who received massive brain damage during World War II: „Sometimes when I'm sitting down I suddenly feel as though my head is the size of a table...When I close my eyes, I'm not even sure where my right leg is; for some reason I used to think (even sensed) it was somewhere above my shoulder“ (Luria 1972: 42-3).

290 REFERENCES Anscombe, G.E.M. 1963. Intention. Oxford: Blackwell. ——— .1981. „The First Person” in her Collected Philosophical Papers Vol II: Metaphysics and the Philosophy of Mind. Oxford: Blackwell. Baker, L. Rudder. 2000. Persons and Bodies: A Constitution View. Cambridge: Cambridge University Press. Bell, D. 1990. Husserl. London: Routledge. Bermúdez, J. et al (eds.).1995. The Body and the Self. Cambridge, MA: Bradford/MIT Press. ——— .1998. The Paradox of Self-Consciousness. Cambridge, MA: Bradford/MIT Press. Budd, M. 1989. Wittgenstein’s Philosophy of Psychology. London: Routledge. Candlish, S. 1996. „Wittgenstein and the Doctrine of Kinaesthesis”, Australasian Journal of Philosophy, 74: 581-97. Carnap, R. 1967. The Logical Structure Of The World, trans. R. George. Berkeley and Los Angeles: University of California Press. Cassam, Q. 1997. Self and World. Oxford: Clarendon. Cole, J. 1991. Pride and a Daily Marathon. London: Duckworth. ——— . and Paillard, J. 1995. „Living Without Touch and Peripheral Information about Body Position and Movement: Studies with Deafferented Patients” in Bermúdez et al eds. 1995. Danto, A. 1999. The Body/Body Problem. Berkeley: University of California Press. Evans, G. 1982. The Varieties of Reference. Oxford: Clarendon Press. Gibson, J.J. 1966. The Senses Considered as Perceptual Systems. Boston: HoughtonMifflin. ——— .1968 [Unpublished]. On the Difference between Perception and Propriocep tion. http://www.huwi.org/Gibson/difference.php (accessed 28/06/04)

291 ——— . 1972 [Unpublished]. Note on The Differences between a Sensory Modality and a Perceptual System. http://www.huwi.org/gibson/modality.php (accessed 28/06/04) ——— .1979. The Ecological Approach to Visual Perception, Boston, MA: Houghton Mifflin. Hamilton, A. 1991. „Anscombian and Cartesian Scepticism”, Philosophical Quarterly, 41:39-54. ——— .1995. „A New Look at Personal Identity”, Philosophical Quarterly, 45:332-49. ——— .2000. „The Authority of Avowals and the Concept of Belief”, European Journal of Philosophy, 7:20-39. ——— .2003. „‘Scottish Commonsense’ about Memory: A Defence of Thomas Reid’s Direct Knowledge Account”, Australasian Journal of Philosophy, 81: 229-45. ——— .(forthcoming) Memory and the Body: A Study of Self-Consciousness Husserl, E. 1989. Ideas Pertaining to a Pure Phenomenology and to a Phenomenological Philosophy, Second Book, trans. R. Rojcewicz and A. Schuwer. Dordrecht: Kluwer. Luria, A. 1972. The Man With A Shattered World, trans. L. Solotaroff. Harmondsworth, Middlesex: Penguin. McDonald, P. 2000. Descartes and Husserl: The Philosophical Project of Radical Beginnings. Albany, New York: SUNY Press. Martin, M. 1995. „Bodily Awareness: A Sense of Ownership”, in Bermúdez et al (eds.) 1995. Merleau-Ponty, M. 1962. Phenomenology of Perception, trans. C. Smith. London: Routledge and Kegan Paul. O’Shaughnessy, B. 1995. „Proprioception and the Body Image” in Bermúdez et al. (eds.) 1995. Reed, E. 1988. James J. Gibson and the Psychology of Perception. Yale: Yale University Press. Sacks, O. 1985. The Man Who Mistook His Wife For A Hat. London: Duckworth. Smith, B. and Smith, D. 1995. The Cambridge Companion to Husserl. Cambridge:

292 Cambridge University Press. Smith, B. 1995. „Common sense” in Smith, B. and Smith, D. 1995. Smith, D. 1995. „Mind and Body” in Smith, B. and Smith, D. 1995. Wittgenstein, L. 1958. Philosophical Investigations. Oxford: Blackwell. ——— .1969. The Blue and Brown Books. Oxford: Blackwell. ——— .1980. Remarks on the Philosophy of Psychology, two volumes. Oxford: Blackwell.

About the Authors Igor Douven is a lecturer of philosophy at the Erasmus University, Rotterdam David Eng is assistent professor of philosophy at California State University, Bakersfield Bob Hale is professor of philosophy at the University of Glasgow Steven D. Hales is associate professor at Bloomsburg University Andy Hamilton is a lecturer of philosophy at the Unversity of Durham Christian B. Miller is assistant professor of philosophy at Wake Forest University Bence Nanay is adjunct professor of philosophy at the University of California, Berkeley Duncan Pritchard is a lecturer in philosophy at the University of Stirling Sabine Roeser is an assistant professor at the Technische Universiteit, Delft Ron Rood is adjunct professor of philosophy at the Vrije Universiteit, Amsterdam Christian Weidemann is associate professor of philosophy at the University of Münster René van Woudenberg is professor of philosophy at the Vrije Universiteit, Amsterdam

Philosophische Forschung Philosophical Research ______________________________________________________________________

Georg Meggle (ed.)

Edited by Johannes Brandl • Andreas Kemmerling Wolfgang Künne • Mark Textor

Georg Meggle (Ed.)

Social Facts & Collective Intentionality

Ethics of Terrorism & Counter Terrorism

3-937202-08-0, Hardcover, 478 pp., EUR 138,00

3-937202-68-4, Hardcover, 410 pp., EUR 98,00

Social Facts & Collective Intentionality: the combination of these two terms refers to a new field of basic research. Working mainly in the mood and by means of Analytical Philosophy, at the very heart of this new approach are conceptual explications of all the various versions of Social Facts & Collective Intentionality and the ramifications thereof. This approach tackles the topics of traditional social philosophy using new conceptual methods, including techniques of formal logics, computer simulations and artificial intelligence. Yet research on Social Facts & Collective Intentionality also includes ontological, epistemological, normative and - last but not least - methodological questions. This volume represents the state of the art in this new field.

We are supposed to wage war against Terrorism – but exactly what we are fighting against in this war, there is nearly no consensus about. And, much worse, nearly nobody cares about this conceptual disaster – the main thing being, whether or not you are taking sides with the good guys. This volume is an analytical attempt to end this disaster. What is Terrorism? Are terrorist acts to be defined exclusively on the basis of the characteristics of the respective actions? Or should we restrict such actions to acts performed by non-state organisations? And, most important, is terrorism already by its very nature to be morally condemned?

Mark Siebel • Mark Textor (Hrsg.)

René van Woudenberg, Sabine Roeser, Ron Rood (Eds.)

Semantik und Ontologie

Beiträge zur philosophischen Forschung ISBN 3-937202-43-9, Hardcover, 445 pp., EUR 93,00

Der zweite Band der Reihe Philosophische Forschung spannt zwei Kerngebiete der Analytischen Philosophie zusammen: die Semantik und die Ontologie. Was sind die Grundbausteine unserer Ontologie? Wie beziehen wir uns sprachlich bzw. geistig auf sie? Diese und weitere Fragen werden von international renommierten Philosophen aus historischer und systematischer Perspektive diskutiert. Die Beiträge sind in Deutsch und English verfasst. Sie stammen von Christian Beyer, Johannes Brandl, Dagfinn Føllesdal, Dorothea Frede, Rolf George, Gerd Graßhoff, Peter Hacker, Andreas Kemmerling, Edgar Morscher, Kevin Mulligan, Rolf Puster, Richard Schantz, Benjamin Schnieder, Oliver Scholz, Severin Schröder, Peter Simons, Thomas Spitzley, Markus Stepanians, Ralf Stoecker und Daniel von Wachter.

ontos verlag Frankfurt • Paris • Ebikon • Lancaster • New Brunswick www.ontosverlag.com

Basic Belief and Basic Knowledge Papers in Epistemology ISBN 3-937202-70-6, Hardcover, 293 pp., EUR 89,00

Over the last two decades foundationalism has been severely criticized. In response to this various alternatives to it have been advanced, notably coherentism. At the same time new versions of foundationalism were crafted, that were claimed to be immune to the earlier criticisms. This volume contains 12 papers in which various aspects of this dialectic are covered. A number of papers continue the trend to defend foundationalism, and foundationalism’s commitment to basic beliefs and basic knowledge, against various attacks. Others aim to show that one important objection against coherentism, viz. that the notion of ‘coherence’ is too vague to be useful, can be countered.

ontos verlag