الرئيسية A Companion to Epistemology
A Companion to EpistemologyJonathan Dancy, Ernest Sosa
Epistemology - the theory of knowledge and of justified belief - has always been of central importance in philosophy. Progress in other areas of philosophical research has often depended crucially on epistemological presuppositions. This Companion, with well over 250 articles ranging from summary discussions to major essays on topics of current controversy, is the first complete reference work devoted to the subject. All the main theoretical positions in epistemology are discussed and analysed, tougher with the different categories of knowledge itself - scientific, historical, mathematical, a priori, moral and so on - and the special problems associated with these. There are also many entries covering individual concepts, arguments and problems, short definitions of technical terms, and biographical articles. With its unrivalled coverage of the field, supported by a comprehensive index and an extensive cross-referencing, the Companion is likely to remain the standard reference work in epistemology from undergraduate level upward for the foreseeable future.
Awful, unusable conversion to epub from some other format. Pieces of multiple pages displayed on each page. Basically unreadable. The second Blackwell Companion epub I've seen lately that's messed up like that.
27 February 2015 (15:36)
* * * title : A Companion to Epistemology Blackwell Companions to Philosophy author : Dancy, Jonathan. publisher : Blackwell Publishing Ltd. isbn10 | asin : 0631192581 print isbn13 : 9780631192589 ebook isbn13 : 9780631224822 language : English subject Knowledge, Theory of. publication date : 2000 lcc : BD161.C637 2000eb ddc : 121 subject : Knowledge, Theory of. Page iii A Companion to Epistemology Edited by Jonathan Dancy and Ernest Sosa Blackwell Companions to Philosophy Page iv Disclaimer: Some images in the original version of this book are not available for inclusion in the netLibrary eBook. Copyright © Blackwell Publishers Ltd, 1992, 1993 Editorial organization © Jonathan Dancy and Ernest Sosa, 1992, 1993 First published 1992 First published in paperback 1993 Reprinted 1994 (twice), 1996 (twice), 1997, 1998, 1999, 2000 Blackwell Publishers Ltd 108 Cowley Road Oxford OX4 1JF, UK Blackwell Publishers Inc. 350 Main Street Malden, Massachusetts 02148, USA All rights reserved. Except for the quotation of short passages for the purposes of criticism and review, no part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. Except in the United States of America, this book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated without the publisher's prior consent in any form of binding or cover other than that in which it is published and without a similar condition including this condition being imposed on the subsequent purchaser. British Library Cataloguing in Publication Data A CIP catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A Companion to epistemology / edited by Jonathan Dancy and Ernest Sosa p. cm. (Blackwell companions to philosophy) Includes bibliographical references and index. ISBN 0631172041 ISBN 0631192581 (pbk) 1. Knowledge, Theory of. I. Dancy, Jonathan. II. Sosa, Ernest. III. Series BD161.C637 1992 9232205 121dc20 CIP Typeset in 9.5 on 11pt Photina by Alden Multimedia Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall This book is printed on acid-free paper Page v Contents Contributors vii Introduction xiii A Companion to Epistemology 1 Index 528 Page vii Contributors Peter Achinstein Johns Hopkins University Felicia Ackerman Brown University Laird Addis University of Iowa Linda Alcoff Syracuse University J. V. Allen University of Pittsburgh Robert F. Almeder Georgia State University William P. Alston Syracuse University C. Anthony Anderson University of Minnesota Robert Audi University of NebraskaLincoln Michael Ayers Wadham College, Oxford David Bakhurst Queen's University at Kingston, Canada Thomas Baldwin Clare College, Cambridge David Bell University of Sheffield Akeel Bilgrami Columbia University Graham Bird University of Manchester David Bloor University of Edinburgh David Blumenfeld Georgia State University Laurence BonJour University of Washington Clive Borst Keele University L. S. Carrier University of Miami Albert Casullo University of NebraskaLincoln R. M. Chisholm Brown University Lorraine Code York University, Ontario L. Jonathan Cohen The Queen's College, Oxford Page viii Stewart Cohen Arizona State University John J. Compton Vanderbilt University Earl Conee University of Rochester John Cottingham University of Reading Robert Cummins University of Arizona Edwin Curley University of Illinois at Chicago Fred D'Agostino University of New England, Australia Vrinda Dalmiya Montana State University Jonathan Dancy Keele University Fred Dretske Stanford University Catherine Z. Elgin Lexington, Massachusetts Edward Erwin University of Miami Richard Feldman University of Rochester Richard Foley Rutgers University Dagfinn Føllesdal University of Oslo and Stanford University Graeme Forbes Tulane University Richard Fumerton University of Iowa Don Garrett University of Utah Margaret Gilbert University of Connecticut Carl Ginet Cornell University Hans-Johann Glock University of Reading Alan H. Goldman University of Miami Alvin I. Goldman University of Arizona Jorge J. E. Gracia State University of New York at Buffalo Richard E. Grandy Rice University A. C. Grayling St. Anne's College, Oxford John Greco Fordham University Patrick Grim State University of New York at Stony Brook Charles Guignon University of Vermont Susan Haack University of Miami Page ix P. M. S. Hacker St John's College, Oxford Andy Hamilton Keele University Peter H. Hare State University of New York at Buffalo Gilbert Harman Princeton University Adrian Heathcote Australian National University John Heil Davidson College, North Carolina Risto Hilpinen University of Miami Jaakko Hintikka Florida State University Christopher Hookway University of Birmingham Jim Hopkins King's College London Paul Horwich Massachusetts Institute of Technology Bruce Hunter University of Alberta Terence Irwin Cornell University Frank Jackson Australian National University Jerrold J. Katz City University of New York Jaegwon Kim Brown University Richard F. Kitchener Colorado State University Peter D. Klein Rutgers University Hilary Kornblith University of Vermont John Lachs Vanderbilt University Keith Lehrer University of Arizona Noah M. Lemos DePauw University Ernest LePore Rutgers University J. H. Lesher University of Maryland Paisley Livingston McGill University Steven Luper-Foy Trinity University, Texas Marilyn McCord Adams University of California, Los Angeles Gregory McCulloch University of Nottingham Scott MacDonald University of Iowa David McNaughton Keele University Page x David B. Martens Mount Royal College, Calgary Jack Meiland University of Michigan Phillip Mitsis Cornell University J. N. Mohanty Temple University, Philadelphia Jame Montmarquet Tennessee State University Paul K. Moser Loyola University at Chicago Alexander Nehamas Princeton University Anthony O'Hear University of Bradford M. Okrent Bates College, Lewiston, Maine George Pappas Ohio State University Christopher Peacocke Magdalen College, Oxford David Pears Christ Church, Oxford Michael Pendlebury University of the Witwatersrand Philip Pettit Australian National University Alvin Plantinga University of Notre Dame Leon Pompa University of Birmingham Richard Popkin Washington University, St Louis, Missouri John F. Post Vanderbilt University Nicholas Rescher University of Pittsburgh G. A. J. Rogers Keele University Jay Rosenberg University of North Carolina R. M. Sainsbury King's College London Wesley C. Salmon University of Pittsburgh Donald W. Sherburne Vanderbilt University Sydney Shoemaker Cornell Universty Robert K. Shope University of Massachusetts at Boston Harvey Siegel University of Miami John Skorupski University of St Andrews Brian Skyrms University of California at Irvine R. C. Sleigh University of Massachusetts at Amherst Page xi Steve Smith Wolfson College, Oxford P. F. Snowdon Exeter College, Oxford Elliott Sober University of Wisconsin, Madison Tom Sorell University of Essex Roy A. Sorensen New York University Ernest Sosa Brown University Edward Stein New York University Mark Steiner Hebrew University, Jerusalem Matthias Steup St Cloud State University Charlotte Stough University of California at Santa Barbara Steven K. Strange Emory University Barry Stroud University of California, Berkeley Robert S. Tragesser Columbia University John Troyer University of Connecticut Thomas Tymoczko Smith College, Northampton, Massachusetts James Van Cleve Brown University Jonathan Vogel Amherst College, Massachusetts Douglas N. Walton University of Winnipeg Kenneth R. Westphal University of New Hampshire Samuel C. Wheeler III University of Connecticut Nicholas P. White University of Michigan, Ann Arbor Michael Williams Northwestern University R. S. Woolhouse University of York Page xiii Introduction The present Companion, like the majority of the other volumes in the Blackwell Companions to Philosophy series, is organized as a standard reference book, with alphabetically arranged articles of varying length (anything from 250 to 3500 words) on leading theories, thinkers, ideas, distinctions and concepts in epistemology. It aims for a broad readership, while recognizing that the nature of contemporary epistemology inevitably imposes restrictions on this. In some other areas of philosophy it remains feasible to design a book which is largely accessible to the general reader; in epistemology, however, the main readership is likely to be students from undergraduate level upwards, as well as professional philosophers, and it is to them that the Companion is primarily addressed. A minority of topics resist treatment other than at an advanced level: they have not for that reason been excluded, lest coverage of the area become incomplete. But the vast majority are accessible to all levels of the intended readership. Not all entries will be comprehensible on their own: at least, not to the inexperienced reader. This is where the cross-referencing system comes in. I have used two interrelated methods of guiding readers from one entry to another. Within the text itself terms or names occur in small capitals; this will often occur where reference is made to DESCARTES, or to REALISM, for example. This means that there is an entry on this person or topic, and that it would be worthwhile having a look at it for present purposes. The mere fact that there is an entry on this person or topic, however, is not sufficient for me to flag it. Not all references to Descartes or to realism are significant. What is more, a person or topic may not be flagged in this way on its first occurrence in an entry; I may wait for the best moment, as it were. And sometimes one and the same person or topic is flagged more than once in the same entry, where there has been a long gap or I think it particularly appropriate for some other reason. Most of the flagging that is done within the body of the text is of this form; a word or phrase is highlighted in the sentence, as I highlighted DESCARTES above. In doing this, I have not insisted slavishly that the word highlighted be exactly the same one as the headword that the reader is effectively being referred to. For instance, I may direct the reader to an entry on realism by flagging the remark that Santayana was a REALIST. Sometimes, however, I was unable to work the cross-reference into the text in this way. On these occasions it is inserted at the end of the relevant sentence or paragraph. There are also cross-references to be found at the end of most entries. These fulfil two functions. First, they enable me to point out areas to which the present entry is related, but which have not occurred significantly in the text. Second, they enable me Page xiv to insist a bit that you should consider again looking at an entry that has already been flagged in the text. So if you see a person or topic flagged both within and at the end of an entry, you can take it that I think you really should have a look at it. The Blackwell series of Companions is conceived as related primarily to Anglo-American philosophy. The topics the editors chose to cover were selected with this in mind. But this does not mean that other traditions are completely ignored. There is an entry on Indian epistemology and, as well as a general entry on continental epistemology, there are many entries on individual thinkers in that tradition. We do not pretend, however, to give that tradition as detailed coverage as we give to the one which is our main focus. It might be thought that the jacket illustration is symptomatic of our general approach. Here we have the solitary thinker working in private. Isn't he a wonderful example of the CARTESIAN approach to epistemology which is so characteristic of the Anglo-American analytic tradition, and which is so vehemently rejected on the Continent? There is some truth in this, which we will come to in a moment. There are two points to be made against it. First, the attempt to escape from the clutches of the Cartesian paradigm is as common within the analytic tradition as it is outside. Second, our solitary thinker is not as solitary as all that. He is reading a book, which could be taken to show that he is not relying entirely on his own resources, as the Cartesian mind is supposed to do (see REID, TESTIMONY). Against this, one could point out that the picture exemplifies a conception of knowledge as something to be gained by rational enquiry and perception rather than in practical life and action. This 'logocentrism' may be a more insidious feature of the Cartesian approach, and certainly the emphasis on practice and action is distinctive of Continental epistemology (see for example HEIDEGGER), as is an emphasis on social considerations. One difficulty the editors faced in deciding which topics to cover derived from the interconnectedness of philosophical areas. Epistemology can be to some extent separated from adjacent areas, but only with a justified sense of artificiality. The nearest areas are metaphysics, philosophy of mind and philosophy of science. These gave us two problems, one theoretical and one practical. The practical one was that in considering whether to include an entry on a topic, we had to ask ourselves whether there would be an entry on it in one of the other Companions, and if so how our entry should be related to that one. At the limit, we have an entry on natural science, an area which will on its own occupy a large part of one Companion. But there are many other occasions where the shortness of our coverage here is caused by our sense that the major entry on this topic should not appear in a Companion to Epistemology. The theoretical one was that there are many occasions where views in epistemology are dependent on views in metaphysics or in the philosophy of mind, and we could not hope to cover everything equally well. Contributors were asked to concentrate on epistemology, and the entries have been written accordingly. When reading entries on individual thinkers, therefore, you should bear in mind that these entries do not pretend to be complete accounts of their subject's work in philosophy; they are concentrating on the epistemology as far as that is possible. The same applies to topics. The entry on natural science is concerned only with the epistemology of science, the Page xv entry on religious belief limits itself to epistemological considerations, and so on. The limitation to epistemology is normally implicit rather than explicit; otherwise every entry would have to be headed 'X's epistemology' or 'the epistemology of Y'. This Companion has two editors, divided by the Atlantic (and rejoined by electronic mail). Its general shape was conceived during a very pleasant weekend which I spent in Providence, RI, in Spring 1989. Thereafter, I relied on Ernest Sosa for a constant stream of suggestions about who in the US we might approach as potential contributors a stream that was evidence of his enviable knowledge of the profession. UK contributors were my responsibility. Beyond that, the detailed editing of contributions has been my province, though I am very grateful to my co-editor for help and advice on the occasional knotty points that arose. I am, of course, equally grateful to our contributors for being willing to undertake what in many cases was a fairly thankless and far from easy task and for the openness with which so many of them received my suggestions for changes to suit my own idea of how things should be. I have had many occasions to express my appreciation of the professionalism of the profession. Finally, I want to thank my wife Sarah, who helped me with various aspects of the editing process, and my son Hugh, who spent two weeks last autumn turning entries into computer-readable form. For a while this Companion was a family affair. JONATHAN DANCY KEELE, FEBRUARY 1992 Page 1 A A Priori/A Posteriori The a priori/a posteriori distinction has been applied to a wide range of objects, including concepts, propositions, truths and knowledge. Our primary concern will be with the epistemic distinction between a priori and a posteriori knowledge. The most common way of marking the distinction is by reference to KANT'S claim that a priori knowledge is absolutely independent of all experience (see A PRIORI KNOWLEDGE). It is generally agreed that S's knowledge that p is independent of experience just in case S's belief that p is justified independently of experience. Some authors (Butchvarov, 1970; Pollock, 1974), however, find this negative characterization of a priori knowledge unsatisfactory and have opted for providing a positive characterization in terms of the type of justification on which such knowledge is dependent. Finally, others (Putnam, 1983; Chisholm, 1989) have attempted to mark the distinction by introducing concepts such as necessity and rational unrevisability rather than in terms of the type of justification relevant to a priori knowledge. One who characterizes a priori knowledge in terms of justification which is independent of experience is faced with the task of articulating the relevant sense of experience. Proponents of the a priori often cite 'intuition' or 'intuitive apprehension' as the source of a priori justification. Furthermore, they maintain that these terms refer to a distinctive type of experience that is both common and familiar to most individuals. Hence, there is a broad sense of experience in which a priori justification is not independent of experience. An initially attractive strategy is to suggest that a priori justification must be independent of sense experience. But this account is too narrow since memory, for example, is not a form of sense experience but justification based on memory is presumably not a priori. There appear to remain only two options: provide a general characterization of the relevant sense of experience or enumerate those sources which are experiential. General characterizations of experience often maintain that experience provides information specific to the actual world while non-experiential sources provide information about all possible worlds. This approach, however, reduces the concept of non-experiential justification to the concept of being justified in believing a necessary truth. Accounts by enumeration face two problems: (1) there is some controversy about which sources to include on the list; and (2) there is no guarantee that the list is complete. It is generally agreed that perception and memory should be included. INTROSPECTION, however, is problematic. Beliefs about one's conscious states and about the manner in which one is appeared to are plausibly regarded as experientially justified. Yet some, such as Pap (1958), maintain that experiments in imagination are the source of a priori justification. Even if this contention is rejected and a priori justification is characterized as justification independent of the evidence of perception, memory and introspection, it remains possible that there are other sources of justification. If it should be the case that clairvoyance, for example, is a source of justified beliefs, such beliefs would be justified a priori on the enumerative account. The most common approach to offering a positive characterization of a priori justification is to maintain that in the case of basic a priori propositions, understanding the proposition is sufficient to justify one in believing that it is true. This approach faces two Page 2 pressing issues. What is it to understand a proposition in the manner which suffices for justification? How does such understanding justify one in believing a proposition? Proponents of the approach typically distinguish understanding the words used to express a proposition from apprehending the proposition itself and maintain that it is the latter which is relevant to a priori justification. But this move simply shifts the problem to that of specifying what it is to apprehend a proposition. Without a solution to this problem, it is difficult, if not impossible, to evaluate the account since one cannot be sure that the requisite sense of apprehension does not justify paradigmatic a posteriori propositions as well. Even less is said about the manner in which apprehending a proposition justifies one in believing that it is true. Proponents are often content with the bald assertion that one who understands a basic a priori proposition can thereby 'see' that it is true. But what requires explanation is how understanding a proposition enables one to see that it is true. Difficulties in characterizing a priori justification in terms either of independence from experience or of its source have led some to introduce the concept of necessity into their accounts, although this appeal takes various forms. Some have employed it as a necessary condition for a priori justification, others have employed it as a sufficient condition, while still others have employed it as both. In claiming that necessity is a criterion of the a priori, Kant held that necessity is a sufficient condition for a priori justification (see A PRIORI KNOWLEDGE). This claim, however, needs further clarification. There are three theses regarding the relationship between the a priori and the necessary which can be distinguished: (1) If p is a necessary proposition and S is justified in believing that p is necessary, then S's justification is a priori; (2) If p is a necessary proposition and S is justified in believing that p is necessarily true, then S's justification is a priori; and (3) If p is a necessary proposition and S is justified in believing that p, then S's justification is a priori (see NECESSITY, MODAL KNOWLEDGE). (2) and (3) have the shortcoming of settling by stipulation the issue of whether a posteriori knowledge of necessary propositions is possible. (1) does not have this shortcoming since the recent examples offered in support of this claim by Kripke (1980) and others have been cases where it is alleged that knowledge of the truth value of necessary propositions is knowable a posteriori. (1) has the shortcoming, however, of either ruling out the possibility of being justified in believing that a proposition is necessary on the basis of testimony or else sanctioning such justification as a priori. (2) and (3), of course, suffer from an analogous problem. These problems are symptomatic of a general shortcoming of the approach: it attempts to provide a sufficient condition for a priori justification solely in terms of the modal status of the proposition believed without making reference to the manner in which it is justified. This shortcoming, however, can be avoided by incorporating necessity as a necessary but not sufficient condition for a priori justification as, for example, in Chisholm (1989). Here there are two theses which must be distinguished: (1) If S is justified a priori in believing that p, then p is necessarily true; and (2) If S is justified a priori in believing that p, then p is a necessary proposition. (1) has the shortcoming of precluding the possibility of being a priori justified in believing a false proposition. (2), however, allows this possibility. A further problem with both (1) and (2) is that it is not clear whether they permit a priori justified beliefs about the modal status of a proposition. For they require that in order for S to be justified a priori in believing that p is a necessary proposition it must be necessary that p is a necessary proposition. But the status of iterated modal propositions is controversial. Finally, (1) and (2) both preclude by stipulation the position advanced by Kripke (1980) and Kitcher (1980) that there is a priori knowledge of contingent propositions. The concept of rational unrevisability has also been invoked to characterize a priori justification. The precise sense of rational unrevisability as well as its relationship to a priori justification have been presented in different ways. Putnam (1983) takes rational unrevisability to be both a necessary and Page 3 sufficient condition for a priori justification while Kitcher (1980) takes it to be only a necessary condition. There are also two different senses of rational unrevisability that have been associated with the a priori: (1) a proposition is weakly unrevisable just in case it is rationally unrevisable in light of any future experiential evidence; and (2) a proposition is strongly unrevisable just in case it is rationally unrevisable in light of any future evidence. Let us consider the plausibility of requiring either form of rational unrevisability as a necessary condition for a priori justification. The view that a proposition is justified a priori only if it is strongly unrevisable entails that if a non-experiential source of justified beliefs is fallible but self-correcting, it is not an a priori source of justification. Casullo (1988) has argued that it is implausible to maintain that a proposition which is justified non-experientially is not justified a priori merely because it is revisable in light of further non-experiential evidence. The view that a proposition is justified a priori only if it is weakly unrevisable is not open to this objection since it excludes only revision in light of experiential evidence. It does, however, face a different problem. To maintain that S's justified belief that p is justified a priori is to make a claim about the type of evidence that justifies S in believing that p. On the other hand, to maintain that S's justified belief that p is rationally revisable in light of experiential evidence is to make a claim about the type of evidence that can defeat S's justification for believing that p rather than a claim about the type of evidence that justifies S in believing that p. Hence, it has been argued by Edidin (1984) and Casullo (1988) that to hold that a belief is justified a priori only if it is weakly unrevisable is either to confuse supporting evidence with defeating evidence or to endorse some implausible thesis about the relationship between the two such as that if evidence of kind A can defeat the justification conferred on S's belief that p by evidence of kind B then S's justification for believing that p is based on evidence of kind A. See also A PRIORI KNOWLEDGE; ANALYTICITY; INTUITION AND DEDUCTION; KANT. Bibliography BonJour, L.: The Structure of Empirical Knowledge (Cambridge, MA: Harvard University Press, 1985). Butchvarov, P.: The Concept of Knowledge (Evanston: Northwestern University Press, 1970). Casullo, A.: 'Revisability, reliabilism, and a priori knowledge', Philosophy and Phenomenological Research 49 (1988), 187213. Chisholm, R. M.: Theory of Knowledge 3rd edn (Englewood Cliffs: Prentice-Hall, 1989). Edidin, A.: 'A priori knowledge for fallibilists', Philosophical Studies 46 (1984), 18997. Kitcher, P.: 'A priori knowledge', Philosophical Review 89 (1980), 323. Kripke, S.: Naming and Necessity (Cambridge, MA: Harvard University Press, 1980). Pap, A.: Semantics and Necessary Truth (New Haven: Yale University Press, 1958). Pollock, J.: Knowledge and Justification (Princeton: Princeton University Press, 1974). Putnam, H.: ' "Two dogmas" revisited', in his Philosophical Papers 3 vols. Vol. 3, Realism and Reason (Cambridge: Cambridge University Press, 1983), 8797. ALBERT CASULLO A Priori Knowledge The contemporary discussion of a priori knowledge has been largely shaped by KANT (1781). Central to his discussion are three distinctions. The first is an epistemic distinction, which divides knowledge into two broad categories: a priori and a posteriori. Kant's characterization of a priori knowledge as knowledge absolutely independent of all experience requires some clarification. For he allowed that a proposition known a priori could depend on experience in at least two ways: (1) experience is necessary to acquire the concepts involved in the proposition; and (2) experience is necessary to entertain the proposition. It is generally accepted, although Kant is not explicit on this point, that a proposition is known a priori only if it is justified independently of experiential evidence (see A PRIORI/A POSTERIORI). The second distinction is the metaphysical distinction between necessary and contingent prop- Page 4 ositions. A necessarily true (false) proposition is one which is true (false) and could not have been false (true). A contingently true (false) proposition is one which is true (false) but could have been false (true). An alternative way of marking the distinction characterizes a necessarily true (false) proposition as one which is true (false) in all possible worlds. A contingently true (false) proposition is one which is true (false) in only some possible worlds including the actual world (see NECESSARY/CONTINGENT). The final distinction is the semantical distinction between analytic and synthetic propositions. This is the most difficult to characterize since Kant offers several ostensibly different ways of marking the distinction. The most familiar states that a proposition of the form 'All A are B' is analytic just in case the predicate is contained in the subject; otherwise, it is synthetic (see ANALYTICITY). Utilizing these three distinctions, Kant went on to defend three theses which are at the core of the contemporary debate: (1) the existence of a priori knowledge; (2) a close relationship between the a priori and the necessary; and (3) the existence of synthetic a priori knowledge. In defending the existence of a priori knowledge, Kant did not attempt to analyse the concept of justification which is independent of experience. Instead, he (1781, p. 42) offered a criterion for distinguishing a priori knowledge from a posteriori knowledge: 'if we have a proposition which in being thought is thought as necessary, it is an a priori judgment'. Since Kant took it to be evident that there are necessary propositions which are known, the existence of a priori knowledge is quickly established. This defence of the existence of a priori knowledge, however, is inextricably tied to his account of the relationship between the a priori and the necessary. The operative principle appears to be that all knowledge of necessary propositions is a priori. Kant (1781, p. 11) also endorses the converse of this principle: all a priori knowledge is of necessary propositions. The conjunction of these two principles, however, does not entail that the categories of the necessary and the a priori are coextensive, since it does not entail that all necessary propositions are knowable. Kant's defence of the existence of synthetic a priori knowledge gives special prominence to mathematics. For it was the principles of arithmetic and geometry that provided his most enduring examples of necessary propositions which are arguably synthetic (see MATHEMATICAL KNOWLEDGE). Much of the recent work on a priori knowledge can be seen as either disputing or defending one of these three Kantian positions. Recent attacks on the existence of a priori knowledge fall into three general camps. Some, such as Putnam (1979) and Kitcher (1983), begin by providing an analysis of the concept of a priori knowledge and then argue that alleged examples of a priori knowledge fail to satisfy the conditions specified in the analysis. Attacks in the second camp generally proceed independently of any particular analysis of the concept of a priori knowledge but focus, instead, on the alleged source of such knowledge. Benacerraf (1973), for example, argues that the faculty of intuition which is alleged by some proponents of the a priori to be the source of mathematical knowledge cannot fulfil that role. A third form of attack is to consider prominent examples of propositions alleged to be knowable only a priori and to show that they can be justified by experiential evidence. MILL'S view that mathematical propositions can be inductively justified has received some support from Kitcher (1983) and Casullo (1988a). An alternative strategy is provided by QUINE (1963), who maintains that mathematical propositions can be justified only in so far as they are part of a larger theory which has a satisfactory match with experience. Recent work in modal logic has renewed interest in the topic of necessary truth. Accompanying this renewed interest has been a re-examination of Kant's views on the relationship between the necessary and the a priori. It has become a common theme in recent work to reiterate that the a priori/a posteriori distinction is an epistemic one, while the necessary/contingent distinction is a metaphysical one. Hence, it cannot be assumed without further argument that they Page 5 are coextensive. Furthermore, Saul Kripke (1980) has forcefully argued that there are necessary a posteriori propositions as well as contingent a priori propositions. Several recent analyses of a priori knowledge, such as Kitcher (1983), have the consequence that some contingent propositions are knowable a priori. The Kantian position that has received most attention is the claim that some a priori knowledge is of synthetic propositions. Initially, there were two different reactions. Some did not dispute the general claim but were concerned exclusively with some of Kant's particular examples of alleged synthetic a priori knowledge. FREGE, for example, challenged only the claim that the truths of arithmetic are synthetic (see MATHEMATICAL KNOWLEDGE). Others, such as AYER (1946), disputed the general claim and tried to establish that all a priori knowledge is of analytic propositions. A third, more radical, reaction came from Quine (1963), who challenged the cogency of the analytic/synthetic distinction. Given the close relationship between the a priori and the analytic forged by Kant's critics, some view Quine's attack as calling into question the cogency of the a priori/a posteriori distinction as well. The claim that there exists a priori knowledge is clearly the most fundamental of the three Kantian theses. Evaluating Kant's defence of the first thesis, however, requires addressing the second thesis regarding the relationship between the necessary and the a priori. The third thesis, although important, is less fundamental. For, on the one hand, if there is no a priori knowledge, the question of whether there is synthetic a priori knowledge does not arise. On the other hand, if the analytic/synthetic distinction is not a cogent one, the issue of the synthetic a priori again does not arise. It might be thought that the demise of the analytic/synthetic distinction calls into question the cogency of the a priori/a posteriori distinction as well. But it is difficult to see how this could be defended short of identifying the a priori with the analytic or uncritically assuming some necessary connection between the two concepts. Hence, our primary concern will be to review briefly the case for and against a priori knowledge. Kant's defence of the claim that mathematical propositions are knowable only a priori exemplifies a general pattern frequently utilized by proponents of the a priori. They begin by maintaining that there is a class of propositions whose members all have a particular feature. They then go on to argue that no proposition having this feature can be known on the basis of experience. Hence, if there is knowledge of the propositions in question, such knowledge must be a priori. In Kant's case, the class consists of mathematical propositions and the feature is necessity. Let us grant the claim that mathematical propositions are necessary and consider the key claim that experience cannot provide knowledge of necessary propositions. The phrase 'knowledge of necessary propositions' masks a crucial distinction between knowledge of the general modal status of a proposition as opposed to knowledge of its truth value (see NECESSITY, MODAL KNOWLEDGE). The basis of Kant's (1781, p. 43) contention that knowledge of necessity is a priori is the observation that 'Experience teaches us that a thing is so and so but not that it cannot be otherwise.' This observation, however, establishes at most that the general modal status of necessary propositions cannot be known on the basis of experience. It does not support the conclusion that the truth value of a necessary proposition cannot be known on the basis of experience. For it allows that experience can provide knowledge that a thing is so and so. Hence, Kant's observation fails to support his key claim that knowledge of mathematical propositions, such as that 7 + 5 = 12, is a priori. For this is a claim about knowledge of the truth value of such propositions rather than a claim about knowledge of their general modal status. A proponent of the a priori can retreat at this point and maintain that even if it has not been established that knowledge of the truth value of necessary propositions is a priori, nevertheless a case has been provided for maintaining that knowledge of the general modal status of a proposition is a priori. This contention, however, seems to rest solely on Page 6 the assumption that experience can provide information only about the actual world. Although this assumption derives some plausibility from the observation that one cannot 'peer' into other possible worlds, it conflicts with the fact that a good deal of our scientific knowledge goes beyond what is true of only the actual world. And yet we are not at all tempted to think that such knowledge is a priori. Consequently, if a posteriori knowledge of some non-actual worlds is possible, it remains to be shown why such knowledge of all non-actual worlds is not possible. Another feature of mathematical propositions, as well as others, which is often cited in support of the claim that knowledge of them is a priori is their alleged immunity from empirical disconfirmation. It is argued that if experiential evidence justifies us in believing mathematical propositions then it must be possible for experiential evidence to justify us in rejecting such propositions. But, the argument continues, we would not regard any experiences as justifying us in rejecting a mathematical proposition. Ayer (1946), for example, invites us to consider a situation where we count what we had taken to be five pairs of objects and find that they amount only to nine. He contends that in such a situation we would not reject the proposition that 2 × 5 = 10, but would explain away the discrepancy as merely apparent by invoking whatever empirical hypothesis fits best with the facts of the situation. It should be noted, however, that it is a standard feature of scientific practice to explain away isolated cases of apparent disconfirming instances to well-established generalizations by invoking some auxiliary hypotheses. Hence, more needs to be said at this point to substantiate the claim that mathematical propositions are immune from empirical disconfirmation. If a scientific principle which has received favourable support in the past were suddenly faced with a large number of apparent disconfirming instances and attempts to explain away those instances as merely apparent fail because the empirical hypotheses invoked to explain them away were not supported by independent tests, then it is evident that experience would have provided sufficient justification for rejecting the principle. Hence, in assessing Ayer's contention that experience cannot provide sufficient justification for rejecting a mathematical principle, one must consider a situation which incorporates the features present in the case of disconfirming the scientific principle: (1) a large number of apparent disconfirming instances to a mathematical principle; and (2) independent tests which fail to support the auxiliary hypotheses introduced to explain away the disconfirming instances as merely apparent. It has been argued (Casullo, 1988a) that in such circumstances it is unreasonable to dismiss the experiential disconfirming evidence as merely apparent since the bulk of one's evidence indicates that it is genuine. A third feature of mathematical propositions often cited in support of the claim that they are knowable only a priori is their alleged certainty. It is argued that if a mathematical proposition were justified on the basis of experiential evidence, its justification would be inductive in character. Since no inductive justification can confer certainty on its conclusion, it is concluded that mathematical propositions are knowable only a priori. The task facing proponents of the argument, however, is to specify the sense in which mathematical propositions are certain. It might be thought that the deductive character of mathematical PROOF supplies the needed answer. But there are several problems with this answer. The most obvious is that the conclusion of a mathematical proof is known with certainty only if the premisses from which the proof begins are known with certainty. But the deductive character of mathematics provides no account of the sense in which basic mathematical propositions are known with certainty. Furthermore, a priorists typically maintain that it is only basic mathematical propositions and their obvious consequences that are known with certainty. Hence, the question which must be addressed is the sense in which basic mathematical propositions are certain. It has often been maintained that epistemically basic propositions Page 7 are certain in the sense that a mistake regarding them is not possible. On this account, S's belief that p is certain just in case necessarily if S believes that p, then it is true that p. It is evident that this sense of certainty is trivially satisfied by any necessary truth which S believes. Consequently, it does not underwrite the claim that only propositions known a priori are certain. An alternative manner of specifying the requisite sense of certainty is to invoke the degree of support a proposition enjoys. A proposition which has the highest degree of support is one which is not open to future disconfirmation. More precisely, p is certain for S just in case there is no epistemically possible situation in which S would be less justified in believing that p. On this account of certainty, the argument faces the same difficulty as the earlier argument based on immunity to empirical disconfirmation. For if mathematical propositions are not immune to such disconfirmation, there are epistemically possible situations in which S would be less justified in believing them (see CERTAINTY). In closing let us consider two sceptical arguments against the a priori. Some writers, such as PUTNAM (1983), have taken QUINE'S (1963) claim that 'no statement is immune to revision' as denying the existence of a priori knowledge. Clearly, there are two separate issues involved in evaluating this contention: (1) the correctness of Quine's claim; and (2) the bearing of the claim, if correct, on the existence of a priori knowledge. Since we have already argued that there is reason to doubt that mathematical propositions are immune from experiential disconfirmation, let us grant (1) and consider (2). Clearly, if Quine's claim is to bear on the a priori, it is minimally necessary that the following thesis be true: (3) If S knows a priori that p then p is rationally unrevisable. The plausibility of (3) rests on the idea that a priori knowledge is independent of experience. It is alleged that if a proposition is susceptible to empirical disconfirmation then it is not independent of experience in the requisite sense. It has been argued (Casullo, 1988b) that there is reason to be dubious about this line of argument. For the statement that S knows that p, independently of experience appears to entail only: (1) S has justification for believing that p which is sufficient for knowledge; (2) this justification is independent of experience; and (3) the other conditions for knowledge are satisfied. But (1), (2) and (3) are compatible with (4) the possibility of experiential evidence which defeats the nonexperiential justification S has for believing that p (see A PRIORI/A POSTERIORI). A recurrent concern of those who resist endorsing a priori knowledge is that the existence of such knowledge appears mysterious. If there is a priori knowledge, then, presumably, it has its source in some human cognitive processes. But proponents of the a priori say little about these processes or the manner in which they produce a priori knowledge. At best, reference is made to processes such as 'intuition' or 'intuitive apprehension' along with the claim that they are familiar to anyone who has ascertained the validity of a step in logical proof. This response has two shortcomings. From the fact that there may be a distinctive phenomenological experience which occurs when one ascertains the validity of a step in a proof, it does not follow that these experiences accompany or are constitutive of the operation of a distinctive cognitive process. Furthermore, it is questionable whether invoking such processes explains how we are justified in believing mathematical or logical principles. For example, it is sometimes claimed that the intuitive apprehension of abstract entities is in some way analogous to the perception of physical objects. Benacerraf (1973) has drawn attention to one significant problem with such claims. Perception is a process which involves causal interaction between perceivers and the objects of perception. Abstract entities, however, are incapable of standing in causal relations. Given this disanalogy, some alternative explanation of how intuitive apprehension produces a priori knowledge is necessary. In summary, we have found that a number of traditional arguments in support of the existence of a priori knowledge as well as several sceptical arguments against it are Page 8 inconclusive. Proponents of a priori knowledge are left with the task of (1) providing an illuminating analysis of a priori knowledge which does not involve strong constraints which are easy targets of criticism; and (2) showing that there is a belief-forming process which satisfies the constraints provided in the analysis together with an account of how the process produces the knowledge in question. Opponents of the a priori, on the other hand, must provide a compelling argument which does not either (1) place implausibly strong constraints on a priori justification; or (2) presuppose an unduly restrictive account of human cognitive capacities. See also ANALYTICITY; EMPIRICISM; INTUITION AND DEDUCTION; KANT; LOGICAL POSITIVISM; MATHEMATICAL KNOWLEDGE; MILL. Bibliography Ayer, A. J.: Language, Truth and Logic 2nd edn (London: Gollancz, 1946). Benacerraf, P.: 'Mathematical truth', Journal of Philosophy 70 (1973), 66179. Casullo, A.: 'Necessity, certainty, and the a priori', Canadian Journal of Philosophy 18 (1988[a]), 4366. Casullo, A.: 'Revisability, reliabilism, and a priori knowledge', Philosophy and Phenomenological Research 49 (1988[b]), 87213. Kant, I.: Critique of Pure Reason (1781) trans. N. Kemp Smith (London: Macmillan, 1964). Kitcher, P.: The Nature of Mathematical Knowledge (Oxford: Oxford University Press, 1983). Kripke, S.: Naming and Necessity (Cambridge, MA: Harvard University Press, 1980). Putnam, H.: 'What is mathematical truth?', in his Philosophical Papers 3 vols. Vol. 1, Mathematics, Matter and Method 2nd edn (Cambridge: Cambridge University Press: 1979), 6078. Putnam, H.: ' "Two dogmas" revisited', in his Philosophical Papers 3 vols. Vol. 3, Realism and Reason (Cambridge: Cambridge University Press, 1983), 8797. Quine, W. V.: 'Two dogmas of empiricism', in his From A Logical Point of View 2nd edn (New York: Harper & Row, 1963), 2046. ALBERT CASULLO Abduction Inductive reasoning tests hypotheses against experience: typically, we derive predictions from hypotheses and establish whether they are satisfied. An account of induction leaves unanswered two prior questions: How do we arrive at the hypotheses in the first place? And on what basis do we decide which hypotheses are worth testing? These questions concern the logic of discovery or, in Charles S. Peirce's terminology, abduction. Many empiricist philosophers have denied that there is a logic (as opposed to a psychology) of discovery. Peirce, and followers such as N. R. Hanson, insisted that there is a logic of abduction. The logic of abduction thus investigates the norms employed in deciding whether a hypothesis is worth testing at a given stage of inquiry, and the norms influencing how we should retain the key insights of rejected theories in formulating their successors. See also INDUCTION; PEIRCE. Bibliography Hanson, N. R.: Patterns of Discovery (Cambridge: Cambridge University Press, 1958). Peirce, C. S.: Collected Papers vol. VII, ed. A. Burks (Cambridge, MA: Harvard University Press, 1958), pp. 89164. CHRISTOPHER HOOKWAY Absurdity An absurdity is any obviously, patently, or otherwise undeniably false proposition, such as 0 = 1 or, for some proposition p, the proposition p & not-p. Absurdities play the most important role in reductio ad absurdum arguments conducted in classical logic. One wants to demonstrate p. Assume that not-p. Show that not-p implies a false proposition A. Since any proposition that implies a false proposition is false, not-p is false, so that not-not-p is true, and not-not-p is logically equivalent to p. That is, from the Page 9 fact that not-p implies A, one infers that p. Such a demonstration of p would be controversial if the falsehood of A were in doubt. Hence it would be best if A were patently or uncontroversially false, i.e. absurd. Bibliography Kneale, W. and Kneale, M.: The Development of Logic (Oxford, Clarendon Press, 1962). ROBERT S. TRAGESSER Academy (Plato) PLATO (c.429347 BC) founded his school, named after a district of Athens, between 387 and 367 BC. Studies centred on philosophy, mathematics and science. The conjecture that Plato's Republic accurately describes its curriculum, however, seems wrong. It focused on Platonic thinking through the headships of Speusippus and Xenocrates (d. 314). It was later dominated by the scepticisms of Arcesilaus (d. 241) and Carneades (d. 129). In the first century BC it was dominated by a Platonism that was assimilated to the views of ARISTOTLE and STOICISM. Its subsequent history is unclear, and its activity apparently ceased with the closing of the pagan schools by Justinian in 529 AD. Bibliography Zeller, E.: Die Philosophie der Griechen in ihrer geschichtlichen Entwicklung, 6th edn. (Hildesheim, 1963), II.1.2 and III.1.12. NICHOLAS P. WHITE Act/Object Analysis According to the act/object analysis of experience, every experience with content involves an object of experience to which the subject is related by an act of awareness (the event of experiencing that object). This is meant to apply not only to perceptions, which have material objects (whatever is perceived), but also to experiences like hallucinations and dream experiences, which do not. Such experiences none the less appear to represent something, and their objects are supposed to be whatever it is that they represent. Act/object theorists may differ on the nature of objects of experience, which have been treated as properties, Meinongian objects (which may not exist or have any form of being), and, more commonly, private mental entities with sensory qualities. (The term 'sense-data' is now usually applied to the latter, but has also been used as a general term for objects of sense experiences, as in the work of G. E. MOORE.) Act/object theorists may also differ on the relationship between objects of experience and objects of perception. In terms of REPRESENTATIVE REALISM, objects of perception (of which we are 'indirectly aware') are always distinct from objects of experience (of which we are 'directly aware'). Meinongians, however, may simply treat objects of perception as existing objects of experience. See also ADVERBIAL THEORY; DIRECT REALISM; EXPERIENCE; REPRESENTATIVE REALISM; SENSE-DATA. Bibliography Ayer, A. J.: The Foundations of Empirical Knowledge (London: Macmillan, 1940). Jackson, F.: Perception: a Representative Theory (Cambridge: Cambridge University Press, 1977). Moore, G. E.: Philosophical Studies (London: Kegan Paul, 1922). Perkins, M.: Sensing the World (Indianapolis: Hackett, 1983). MICHAEL PENDLEBURY Adverbial Theory In its best-known form the adverbial theory of experience proposes that the grammatical object of a statement attributing an experience to someone be analysed as an adverb. For example, (I) Rod is experiencing a pink square is rewritten as Rod is experiencing (pink square)-ly. This is presented as an alternative to the Page 10 ACT/OBJECT ANALYSIS, according to which the truth of a statement like (l) requires the existence of an object of experience corresponding to its grammatical object. A commitment to the explicit adverbialization of statements of experience is not, however, essential to adverbialism. The core of the theory consists, rather, in the denial of objects of experience (as opposed to objects of perception) coupled with the view that the role of the grammatical object in a statement of experience is to characterize more fully the sort of experience which is being attributed to the subject. The claim, then, is that the grammatical object is functioning as a modifier, and, in particular, as a modifier of a verb. If this is so, it is perhaps appropriate to regard it as a special kind of adverb at the semantic level. See also EXPERIENCE; REPRESENTATIVE REALISM. Bibliography Chisholm, R. M.: Perceiving: A Philosophical Study (Ithaca, NY: Cornell University Press, 1957). Clark, R.: 'Sensing, perceiving, thinking', in Essays on the Philosophy of Roderick M. Chisholm ed. E. Sosa, Grazer Philosophische Studien (1981), 27395. Cornman, J.: Perception, Common Sense, and Science (New Haven: Yale University Press, 1975). Ducasse, C. J.: Nature, Mind, and Death (La Salle: Open Court, 1951). MICHAEL PENDLEBURY Agnosticism There are two forms of agnosticism: weak and strong. Consider theism: the proposition there is such a person as God an almighty and all knowing and wholly good creator of the world. A weak agnostic is someone who believes neither that there is such a person nor that there is not. In this respect an agnostic is to be contrasted with an atheist, who holds that there is no God, and a theist, who holds that there is. So the theist affirms theism; the atheist denies it; and the agnostic withholds it, having no view as to whether or not this proposition is true. A strong agnostic adds that it isn't possible to know or have a justified belief about the truth of theism, so that no one else should have a view on it either. ALVIN PLANTINGA Alston, William P. (1921) Alston has contributed to epistemology on many topics: the analysis of justification and knowledge, the foundationalismcoherentism and internalismexternalism controversies, epistemic principles, religious epistemology, perception and numerous others. He is known both for his own positions and for incisively developing distinctions now important in the literature. His early papers on FOUNDATIONALISM distinguished levels of justification and thereby showed that even if one is not directly (non-inferentially) justified in the second-order belief that one is justified in believing p, one may be directly justified in believing p. Since foundationalists as such need not require second-order justification regarding basic beliefs, this distinction undercuts much criticism previously considered decisive against foundationalism in all forms. In distinguishing many grades of privileged access, Alston also showed that neither foundationalists nor other epistemologists must regard INFALLIBILITY or some version of Cartesian certainty as the only alternatives to COHERENTISM in accounting for the varieties of justification. Regarding justification in general, Alston draws a contrast between deontological and 'strong position' notions. Roughly, the former treat justification as fulfilment of epistemic duty, the latter as a matter of being in a good position with respect to the truth of p, e.g. being able simply to see that p is true. He argues that much of the literature on justification fails to take account of this distinction, and he shows how the distinction can explain major disagreements. For instance, the former conception goes well with an internalist view, since one has introspective access to grounds of one's obligations, such as a memory of having promised to do some- Page 11 thing or a conviction that lying is wrong; the latter suggests an at least partly externalist view on justification, since one does not in general have such access to the reliability of one's position with respect to discerning the truth of p (see VIRTUE EPISTEMOLOGY; RELIGIOUS BELIEF; EXTERNALISM/INTERNALISM). Alston's own account of justification is a distinctive blend of internalism and externalism: if I justifiedly believe p, I must both have an appropriate access to my grounds and, by virtue of them, be in a good position vis-à-vis the truth of p. This condition normally also holds for knowledge, but Alston has rigorously argued that under special conditions knowledge is possible without justification and thereby on grounds to which one lacks access. For knowledge, as for justification, perception is a paradigmatic source. In both cases, moreover, first-order success is possible without second-order success; e.g. one can know that p without knowing that one does, or even that one's source, say perception, is reliable. But can we know or justifiedly believe perception is reliable? In discussing epistemic circularity, Alston argues that while a kind of circularity is implicit in plausible attempts to show the reliability of perception, it does not prevent one's justifiedly believing, or even knowing, that perception is reliable. To this extent, at least, scepticism is answerable. In recent work, Alston has pursued at least three major epistemological projects. He has defended the theory of appearing as an account of perception. He has developed a doxastic practice approach in metaepistemology, arguing, along lines suggested by the work of REID, that justification is rooted in a certain kind of social practice. And, using these resources and many others in and outside epistemology, he has built an account of the possibility of perception of, and thereby justified beliefs about, God. Writings "Varieties of privileged access', American Philosophical Quarterly 8 (1971), 22341. "Two types of foundationalism', Journal of Philosophy 73 (1976), 16585. 'Concepts of epistemic justification', The Monist 68 (1985), 5789. 'An internalist externalism', Synthese 74 (1988), 26583. Epistemic Justification (Ithaca, NY: Cornell University Press, 1989). (This contains all of the above.) Perceiving God (Ithaca, NY Cornell University Press, 1991). ROBERT AUDI Analyticity The true story of analyticity is surprising in many ways. Contrary to received opinion, it was the empiricist Locke rather than the rationalist Kant who had the better informal account of this type of a priori proposition. Frege and Carnap, represented as analyticity's best friends in this century, did as much to undermine it as its worst enemies. Quine and Putnam, represented as having refuted the analytic/synthetic distinction, not only did no such thing, but, in fact, contributed significantly to undoing the damage done by Frege and Carnap. Finally, the epistemological significance of the distinction is nothing like what it is commonly taken to be. LOCKE'S account of analytic propositions was, for its time, everything that a succinct account of analyticity should be (Locke, 1924, pp. 3068). He distinguishes two kinds of analytic propositions, identity propositions in which 'we affirm the said term of itself', e.g. 'Roses are roses', and predicative propositions in which 'a part of the complex idea is predicated of the name of the whole', e.g. 'Roses are flowers' (pp. 3067). Locke calls such sentences 'trifling' because a speaker who use them 'trifles with words'. A synthetic sentence, in contrast, such as a mathematical theorem, states 'a real truth and conveys with it instructive real knowledge' (pp. 3078). Correspondingly, Locke distinguishes two kinds of 'necessary consequences', analytic entailments where validity depends on the literal containment of the conclusion in the premiss and synthetic entailments where it does not. (Locke did not originate this concept-containment notion of Page 12 analyticity. It is discussed by Arnauld and Nicole, and it is safe to say it has been around for a very long time (cf. Arnauld, 1964, pp. 5965).) KANT'S account of analyticity, which received opinion tells us is the consummate formulation of this notion in modern philosophy, is actually a step backwards. What is valid in his account is not novel, and what is novel is not valid. Kant repeats Locke's account of concept-containment analyticity, but introduces certain alien features, the most important being his characterization of analytic propositions as propositions whose denials are logical contradictions (Kant, 1783, p. 14). This characterization suggests that analytic propositions based on Locke's partwhole relation or Kant's explicative copula are a species of logical truth. But the containment of the predicate concept in the subject concept in sentences like 'Bachelors are unmarried' is a different relation from the containment of the consequent in the antecedent in sentences like 'If John is a bachelor, then John is a bachelor or Mary read Kant's Critique'. The former is literal containment whereas the latter is, in general, not. Talk of the 'containment' of the consequent of a logical truth in the antecedent in cases like our example is only metaphorical, a way of saying 'logically derivable'. Kant's conflation of concept containment with logical containment caused him to overlook the issue of whether logical truths are synthetic a priori and the problem of how he can say mathematical truths are synthetic a priori when they cannot be denied without contradiction. Historically, the conflation set the stage for the disappearance of the Lockean notion. Frege, whom received opinion portrays as second only to Kant among the champions of analyticity, and Carnap, whom it portrays as just behind Frege, were jointly responsible for the disappearance of concept-containment analyticity. FREGE was clear about the difference between concept containment and logical containment, expressing it as like the difference between the containment of 'beams in a house' and the containment of a 'plant in the seed' (Frege, 1953, p. 101). But he found the former, as Kant formulated it, defective in three ways: it explains analyticity in psychological terms; it does not cover all cases of analytic propositions; and, perhaps most important for Frege's logicism, its notion of containment is 'unfruitful' as a definitional mechanism in logic and mathematics (Frege, 1953, pp. 1001). In an invidious comparison between the two notions of containment, Frege observes that with logical containment 'we are not simply taking out of the box again what we have just put into it' (ibid., p. 101). To overcome these shortcomings, Frege defines analytic propositions as consequences of laws of logic plus definitions, consequences that 'cannot be inspected in advance' (ibid., pp. 4, 1001). This definition makes logical containment the basic notion. Analyticity becomes a special case of logical truth, and, even in this special case, the definitions employ the power of definition in logic and mathematics rather than mere concept combination. CARNAP, attempting to overcome what he saw as a shortcoming in Frege's account of analyticity, took the remaining step necessary to do away explicitly with LockeanKantian analyticity. As Carnap saw things, it was a shortcoming of Frege's explication that it seems to suggest that definitional relations underlying analytic propositions can be extra-logical in some sense, say, in resting on linguistic synonymy. To Carnap, this represented a failure to achieve a uniform formal treatment of analytic propositions and left us with a dubious distinction between logical and extra-logical vocabulary. Hence, he eliminated the reference to definitions in Frege's explication of analyticity by introducing 'meaning postulates', e.g. statements such as '("x) (x is a bachelor ® x is unmarried)' (see Carnap, 1965, pp. 2229). Like standard logical postulates on which they were modelled, meaning postulates express nothing more than constraints on the admissible models with respect to which sentences and deductions are evaluated for truth and validity. Thus, despite their name, meaning postulates have no more to do with meaning than any other statement expressing a neces- Page 13 sary truth. In defining analytic propositions as consequences of (an expanded set of) logical laws, Carnap explicitly removed the one place in Frege's explication where there might be room for concept containment, and with it, the last trace of Locke's distinction between semantic and other 'necessary consequences'. QUINE, the staunchest critic of analyticity of our time, performed an invaluable service on its behalf albeit one that has gone almost completely unappreciated. Quine made two devastating criticisms of Carnap's meaning postulate approach which expose it as both irrelevant and vacuous. It is irrelevant because, in using particular words of a language, meaning postulates fail to explicate analyticity for sentences and languages generally, that is, they do not define it for variable 'S' and 'L' (Quine, 1953, pp. 334). It is vacuous because, although meaning postulates tell us what sentences are to count as analytic, they do not tell us what it is for them to be analytic (ibid., p. 33). Received opinion has it that Quine did much more than refute the analytic/synthetic distinction as Carnap tried to draw it. Received opinion has it that Quine demonstrated there is no distinction, however anyone might try to draw it. But this, too, is incorrect. To argue for this stronger conclusion, Quine had to show that there is no way to draw the distinction outside logic, in particular, in linguistics. In the absence of a particular theory in linguistics corresponding to Carnap's, Quine's argument had to take an entirely different form. Some inherent feature of linguistics had to be exploited in showing that no theory in this science can deliver the distinction. But the feature Quine chose was a principle of operationalist methodology characteristic of the school of Bloomfieldian linguistics. Quine succeeds in showing that meaning cannot be made objective sense of in linguistics if making sense of a linguistic concept requires, as that school claims, operationally defining it in terms of substitution procedures which employ only concepts unrelated to that linguistic concept. But Chomsky's revolution in linguistics replaced the Bloomfieldian taxonomic model of grammars with the hypothetico-deductive model of generative linguistics, and, as a consequence, such operational definition was removed as the standard for concepts in linguistics. The standard of theoretical definition which replaced it was far more liberal, allowing the members of a family of linguistic concepts to be defined with respect to one another within a set of axioms which state their systematic interconnections the entire system being judged by whether its consequences are confirmed by the linguistic facts. Quine's argument does not even address theories of meaning based on this hypothetico-deductive model (Katz, 1988b, pp. 22752; Katz, 1990, pp. 199202). PUTNAM, the other staunch critic of analyticity, performed a service on behalf of analyticity fully on a par with, and complementary to, Quine's. Whereas Quine refuted Carnap's formalization of Frege's conception of analyticity, Putnam refuted this very conception itself. Putnam put an end to the entire attempt, initiated by Frege and completed by Carnap, to construe analyticity as a logical concept (Putnam, 1962, pp. 64758; 1970, pp. 189201; 1975a, pp. 13193). However, as with Quine, received opinion has it that Putnam did much more. Putnam is credited with having devised science fiction cases, from the robot cat case to the twin earth case, that are counter examples to the traditional theory of meaning. Again, received opinion is incorrect. These cases are only counter examples to Frege's version of the traditional theory of meaning. Frege's version claims both (1) that sense determines reference, and (2) that there are instances of analyticity, say, typified by 'cats are animals', and of synonymy, say typified by 'water' in English and 'water' in twin earth English. Given (1) and (2), what we call 'cats' could not be non-animals and what we call 'water' could not differ from what the twin earthers call 'water'. But, as Putnam's cases show, what we call 'cats' could be Martian robots and what they call 'water' could be something other than H2O. Hence, the cases are counter examples to Frege's version of the theory. Page 14 Putnam himself takes these examples to refute the traditional theory of meaning per se because he thinks other versions must also subscribe to both (1) and (2). He was mistaken in the case of (1). Frege's theory entails (1) because it defines the sense of an expression as the mode of determination of its referent (Frege, 1952, pp. 5678). But sense does not have to be defined this way, or in any way that entails (1). It can be defined as (D). (D) Sense is that aspect of the grammatical structure of expressions and sentences responsible for their having sense properties and relations like meaningfulness, ambiguity, antonymy, synonymy, redundancy, analyticity, and analytic entailment. (Katz, 1972; 1990, pp. 21624) (Note that this use of sense properties and relations is no more circular than the use of logical properties and relations to define logical form, for example, as that aspect of grammatical structure of sentences on which their logical implications depend.) (D) makes senses internal to the grammar of a language and reference an external matter of language use typically involving extra-linguistic beliefs. Therefore, (D) cuts the strong connection between sense and reference expressed in (1), so that there is no inference from the modal fact that 'cat' refers to robots to the conclusion that 'Cats are animals' is not analytic. Likewise, there is no inference from 'water' referring to different substances on earth and twin earth to the conclusion that our word and theirs are not synonymous. Putnam's science fiction cases do not apply to a version of the traditional theory of meaning based on (D). The success of Putnam's and Quine's criticisms in application to Frege's and Carnap's theory of meaning together with their failure in application to a theory in linguistics based on (D) creates the option of overcoming the shortcomings of the LockeanKantian notion of analyticity without switching to a logical notion. This option was explored in the 1960s and 1970s in the course of developing a theory of meaning modelled on the hypothetico-deductive paradigm for grammars introduced in the Chomskyan revolution (Katz, 1972). This theory automatically avoids Frege's criticism of the psychological formulation of Kant's definition because, as an explication of a grammatical notion within linguistics, it is stated as a formal account of the structure of expressions and sentences. The theory also avoids Frege's criticism that concept-containment analyticity is not 'fruitful' enough to encompass truths of logic and mathematics. The criticism rests on the dubious assumption, part of Frege's logicism, that analyticity should encompass them. (Benacerraf, 1981, p. 25). But in linguistics where the only concern is the scientific truth about natural languages, there is no basis for insisting that concept-containment analyticity encompass truths of logic and mathematics. Moreover, since we are seeking the scientific truth about trifling propositions in natural language, we will eschew relations from logic and mathematics that are too fruitful for the description of such propositions. This is not to deny that we want a notion of necessary truth that goes beyond the trifling, but only to deny that that notion is the notion of analyticity in natural language. The remaining Fregean criticism points to a genuine incompleteness of the traditional account of analyticity. There are analytic relational sentences, for example, 'Jane walks with those with whom she herself strolls', 'Jack kills those he himself has murdered', etc., and analytic entailments with existential conclusions, for example, 'I think', therefore, 'I exist'. The containment in these sentences is just as literal as that in analytic subjectpredicate sentences like 'Bachelors are unmarried'. I will now show how a theory of meaning constructed as a hypothetico-deductive systematization of sense as defined in (D) overcomes the incompleteness of the traditional account in the case of such relational sentences. (For a treatment of the existential sentences, see Katz, 1988a.) Such a theory of meaning makes the principal concern of semantics the explanation of sense properties and relations like synonymy, Page 15 antonymy, redundancy, analyticity, ambiguity, etc. Furthermore, it makes grammatical structure, specifically, sense structure, the basis for explaining them. This leads directly to the discovery of a new level of grammatical structure, and this, in turn, makes possible a proper definition of analyticity. To see this, consider two simple examples. It is a semantic fact that 'male bachelor' is redundant and that 'spinster' is synonymous with 'woman who never married'. In the case of the redundancy, we have to explain the fact that the sense of the modifier 'male' is already contained in the sense of its head 'bachelor'. In the case of the synonymy, we have to explain the fact that the sense of 'spinster' is identical to the sense of 'woman who never married' (compositionally formed from the senses of 'woman', 'never' and 'married'). But in so far as such facts concern relations involving the components of the senses of 'bachelor' and 'spinster' and in so far as these words are syntactic simples, there must be a level of grammatical structure at which syntactic simples are semantically complex. This, in brief, is the route by which we arrive a level of decompositional semantic structure that is the locus of sense structures masked by syntactically simple words. Discovery of this new level of grammatical structure was followed by attempts to represent the structure of the senses found there. Without going into the details of sense representation, it is clear that, once we have the notion of decompositional representation, we can see how to generalize Locke's and Kant's informal, subjectpredicate account of analyticity to cover relational analytic sentences. Let a simple sentence S consist of a n-place predicate P with terms Tl, . . . , Tn occupying its argument places. Then: (A) S is analytic in case, first, S has a term Ti which consists of an m-placed predicate Q (m > n or m = n) with terms occupying its argument places, and second, P is contained in Q and, for each term Tj of T1, . . . , Ti1, Ti+l, . . . , Tn, Tj is contained in the term of Q which occupies the argument place in Q corresponding to the argument place occupied by Tj in P. (Katz, 1972) To see how (A) works, suppose that 'stroll' in 'Jane walks with those with whom she herself strolls' is decompositionally represented as having the same sense as 'walk idly and in a leisurely way'. The sentence is analytic by (A) because the predicate 'stroll' (the sense of 'stroll') contains the predicate 'walk' (the sense of 'walk') and the term 'Jane' (the sense of 'Jane' associated with the predicate 'walk') is contained in the term 'Jane' (the sense of 'she herself' associated with the predicate 'stroll'). The containment in the case of the other terms is automatic. The fact that (A) itself makes no reference to logical operators or logical laws indicates that analyticity for subjectpredicate sentences can be extended to simple relational sentences without treating analytic sentences as instances of logical truths. Further, the source of the incompleteness is no longer explained, as Frege explained it, as the absense of 'fruitful' logical apparatus, but is now explained as mistakenly treating what is only a special case of analyticity as if it were the general case. The inclusion of the predicate in the subject is the special case (where n = l) of the general case of the inclusion of an n-place predicate (and its terms) in one of its terms. Note that the defects Quine complained of in connection with Carnap's meaning postulate explication are absent in (A). (A) contains no words from a natural language. It explicitly uses variable 'S' and variable 'L' because it is a definition in linguistic theory. Moreover, (A) tells us what the property is in virtue of which a sentence is analytic, namely, redundant predication, that is, the predicational structure of an analytic sentence is already found in the content of its term structure. Received opinion has been anti-Lockean in holding that necessary consequences in logic and language belong to one and the same species. This seems wrong because the property of redundant predication provides a non-logical explanation of why true statements made in the literal use of analytic Page 16 sentences are necessarily true. Since the property ensures that the objects of the predication in the use of an analytic sentence are chosen on the basis of the features to be predicated of them, the truth conditions of the statement are automatically satisfied once its terms take on reference. The difference between such a linguistic source of necessity and the logical and mathematical sources vindicates Locke's distinction between two kinds of 'necessary consequences'. Received opinion concerning analyticity contains another mistake. This is the idea that analyticity is inimical to science. In part, the idea developed as a reaction to certain dubious uses of analyticity such as Frege's attempt to establish LOGICISM and Schlick's, Ayer's and other LOGICAL POSITIVISTS' attempts to deflate claims to metaphysical knowledge by showing that alleged synthetic a priori truths are merely empty analytic truths (Schlick, 1949, p. 285; Ayer, 1946, pp. 7187). In part, it developed as also a response to a number of cases where alleged analytic, and hence, necessary, truths, e.g. the law of excluded middle, had subsequently been taken as open to revision. Such cases convinced philosophers like Quine and Putnam that the analytic/synthetic distinction is an obstacle to scientific progress. The problem if there is one is not analyticity in the concept-containment sense, but the conflation of it with analyticity in the logical sense. This made it seem as if there is a single concept of analyticity which can serve as the grounds for a wide range of a priori truths. But, just as there are two analytic/synthetic distinctions, so there are two concepts of CONCEPT. The narrow Lockean/Kantian distinction is based on a narrow notion of concept on which concepts are senses of expressions in the language. The broad Fregean/Carnapian distinction is based on a broad notion of concept on which concepts are conceptions often scientific ones about the nature of the referent(s) of expressions (Katz, 1972, pp. 4502, and curiously, Putnam, 1981, p. 207). Conflation of these two notions of concept produced the illusion of a single concept with the content of philosophical, logical and mathematical conceptions but with the status of linguistic concepts. This encouraged philosophers to think that they were in possession of concepts with the content to express substantive philosophical claims, e.g. such as Frege's, Schlick's, Ayer's, etc., and with a status that trivializes the task of justifying them by requiring only linguistic grounds for the a priori propositions in question. Thus, there is no need to reject the analytic/synthetic distinction in toto to prevent analyticity from being put to dubious uses. All that is necessary is to keep the original, narrow distinction from being broadened. This insures that propositions expressing the content of broad concepts cannot receive the easier justification appropriate to narrow ones. Accordingly, in so far as the wholesale rejection of the analytic/synthetic distinction was based on a concern about dubious philosophy, particularly the possibility of blocking scientific progress, it threw out the baby with the bath water. Finally, there is an important epistemological implication of separating the broad and narrow notions of analyticity. Frege and Carnap took the broad notion of analyticity to provide foundations for necessity and a priority, and, hence, for some form of rationalism, and nearly all rationalistically inclined analytic philosophers followed them in this. Thus, when Quine dispatched the FregeCarnap position on analyticity, it was widely believed that necessity, a priority, and rationalism had also been despatched, and, as a consequence, that Quine had ushered in an 'empiricism without dogmas' and NATURALIZED EPISTEMOLOGY. But given there is still a notion of analyticity which enables us to pose the problem of how necessary, synthetic a priori knowledge is possible (moreover, one whose narrowness makes logical and mathematical knowledge part of the problem), Quine did not undercut the foundations of rationalism. Hence, a serious reappraisal of the new empiricism and naturalized epistemology is, to say the least, very much in order (Katz, 1990). See also A PRIORI KNOWLEDGE; INTUI- Page 17 TION AND DEDUCTION; KANT; LOCKE; MATHEMATICAL KNOWLEDGE; PHILOSOPHICAL KNOWLEDGE. Bibliography Arnauld, A.: The Art of Thinking (Indianapolis: Bobbs-Merrill, 1964). Ayer, A. J.: Language, Truth, and Logic (London: Gollancz, 1946). Benacerraf, P.: 'Frege: The last logicist', in Midwest Studies in Philosophy 6 (Minneapolis: University of Minnesota Press, 1981). Carnap, R.: 'Meaning postulates', in his Meaning and Necessity, 2nd edn (Chicago: University of Chicago Press, 1965). Frege, G.: 'On sense and reference', in Translations from the Philosophical Writings of Gottlob Frege, trans. P. T. Geach and M. Black (Oxford: Blackwell, 1952). Frege, G.: Foundations of Arithmetic (Oxford: Blackwell, 1953). Kant, I.: Critique of Pure Reason (1781) trans. N. Kemp Smith (London: Macmillan, 1964). Kant, I.: Prolegomena to any Future Metaphysic (1783) trans. L. W. Beck (Indianapolis: Bobbs-Merrill, 1950). Katz, J. J.: Semantic Theory (New York: Harper & Row, 1972). Katz, J. J.: Cogitations (New York: Oxford University Press, 1988[a]). Katz, J. J.: 'The refutation of indeterminacy', Journal of Philosophy 85 (1988[b]), 22752. Katz, J. J.: The Metaphysics of Meaning (Cambridge, MA: MIT Press, 1990). Locke, J.: An Essay Concerning Human Understanding (Oxford: Clarendon Press, 1924). Putnam, H.: 'It ain't necessarily so', Journal of Philosophy 59 (1962), 65871. Putnam, H.: 'The meaning of "meaning"', in Language, Mind, and Knowledge: Minnesota Studies in the Philosophy of Science (Minneapolis: University of Minnesota Press, 1975[a]). Putnam, H.: 'Is semantics possible?' (1970); reprinted in his Mind, Language, Reality (Cambridge: Cambridge University Press, 1975[b]), 13952. Putnam, H.: Reason, Truth, and History (Cambridge: Cambridge University Press, 1981). Quine, W. V.: 'Two dogmas of empiricism', in his From a Logical Point of View (Cambridge, MA: Harvard University Press, 1953). Schlick, M.: 'Is there a factual a priori?', in Readings in Philosophical Analysis eds H. Feigl and W. Sellars (New York: Appleton-Century-Crofts, 1949). JERROLD J. KATZ Anamnesis 'Recollection', or anamnesis has several roles in PLATO'S epistemology. In the Meno (806) it is invoked to explain the behaviour of an uneducated boy who answers a geometrical problem that he has never heard. At the same time it is used to solve a paradox about inquiry and learning. In the Phaedo it is said to explain our possession of concepts, construed as knowledge of Forms, which we supposedly could not have gained from experience. Recollection also appears in the Phaedrus, but is notably absent from important presentations of Plato's epistemological views in the Republic and other works. Bibliography Gosling, J. C. B.: Plato (London: Routledge, 1973), ch. 16. Gulley, N.: Plato's Theory of Knowledge (London: Methuen, 1962), ch. 1. White, N.P.: Plato on Knowledge and Reality (Indianapolis: Hackett, 1976), chs 23. NICHOLAS P. WHITE Antinomy An antinomy occurs when we are able to argue for, or demonstrate, both a proposition and its contradictory (see PRINCIPLE OF CONTRADICTION). but where we cannot now fault either demonstration. We would eventually hope to be able 'to solve the antinomy' by managing, through careful thinking and analysis, eventually to fault either or both demonstrations. Many paradoxes are an easy source of antinomies. For example, Zeno gave some famous let us say logical-cum-mathematical arguments which might be interpreted as demonstrating that motion is impossible. But our eyes as it were demonstrate motion (exhibit moving things) all the time. Where did Zeno go wrong? Where do our eyes go Page 18 wrong? If we cannot readily answer at least one of these questions, then we are in antinomy. In the Critique of Pure Reason KANT gave demonstrations of the same kind in the Zeno example they were obviously not the same kind of both, e.g. that the world has a beginning in time and space, and that the world has no beginning in time or space. He argues that both demonstrations are at fault because they proceed on the basis of 'pure reason' unconditioned by sense experience. Bibliography Beck, L. W.: 'Antinomy of pure reason', in Dictionary of the History of Ideas ed. P. P. Wiener (New York, Charles Scribner's Sons, 1973), vol. 1. Kant, I.: Critique of Pure Reason trans. N. Kemp Smith (London: Macmillan, 1964). Kneale, W. and Kneale, M.: The Development of Logic (Oxford: Clarendon Press, 1962). ROBERT S. TRAGESSER Apodeictic A proposition p is apodeictic when it is demonstrable in a sense entailing not only that p is true, but also that it is not possibly false. For Aristotle, an apodeictically true proposition was one inferred by formallogical syllogism from incontrovertibly true premisses (Posterior Analytics, 1.71b72c). Sometimes 'apodeictic' is used loosely to mean that a proposition is recognized to be beyond dispute; and sometimes it is taken to mean that it must be true (without any reference to demonstration). ROBERT S. TRAGESSER Aporia Any difficult problem that arises when we are trying to extend our knowledge of a matter, and that threatens seriously to impede our further progress, is called an aporia, especially when there seem to be equally strong arguments for and against any solution. An ANTINOMY is an especially nasty aporia. Bibliography Peters, F. E.: Greek Philosophical Terms (London: University of London, 1967). ROBERT S. TRAGESSER Apperception This is LEIBNIZ'S term for inner awareness or self-consciousness, in contrast with 'perception' or outer awareness. He held, in opposition to DESCARTES, that adult humans can have experiences of which they are unaware; experiences which may affect what they do, but which are not brought to self-consciousness. Indeed there are creatures, such as animals and babies, which completely lack the ability to reflect on their experiences, and to become aware of them as experiences of theirs. The unity of a subject's experience, which stems from his capacity to recognize all his experiences as his, was dubbed by KANT the transcendental unity of apperception. This apprehension of unity is transcendental, rather than empirical, because it is presupposed in experience and cannot be derived from it. Kant used the need for this unity as the basis of his attempted refutation of scepticism about the external world. He argued that my experiences could only be united in one self-consciousness if at least some of them were experiences of a law-governed world of objects in space. Outer experience is thus a necessary condition of inner awareness. See also INTROSPECTION. Bibliography Kant, I.: Critique of Pure Reason trans. N. Kemp-Smith (London: Macmillan, 1964). Leibniz, G. W.: New Essays on Human Understanding (1704) trans. P. Remnant and J. Bennett (Cambridge: Cambridge University Press, 1981), esp. pp. 537. DAVID MCNAUGHTON Aquinas, Thomas (1225-74) Theologian and philosopher, born near Naples. Aquinas's fundamental epistemic category is that of cognition (cognitio). He endorses the Page 19 Aristotelian view that the soul is potentially all things and holds that cognition is its actually becoming a given thing or, as he sometimes puts it, its being assimilated to that thing in a certain way (Summa theologiae Ia.12.4, 17.3, 76.2.AD4, 84.2.AD2). His account of this basic notion of cognition involves both a metaphysical account of the two relata in the relation of cognitive assimilation the human soul and the objects of human cognition and a psychological theory identifying the sorts of powers the soul must possess and the processes it must engage in if cognitive assimilation of this sort is to be possible. According to Aquinas's metaphysics, the reality to which human beings are assimilated in cognition is made up of basic elements (particular substances and accidents) that are joined together in various ways to form complexes (accidents inhering in particular substances). He holds that the soul must possess cognitive powers capable of rendering it isomorphic with an external reality of this sort. The intellective soul's activity of understanding (intellectus) allows human beings both to grasp, via sense perception, the natures of substances and accidents and to link them together into complexes (subject-predicate propositions). But cognition is not restricted to the sort of intake of information made possible by sense perception and understanding; human beings are also able, by virtue of a distinct activity of discursive thought (ratio, ratiocinatio), to acquire new cognition of things by drawing inferences from things already cognized. Aquinas's strictly epistemological views are to be found within this broad metaphysically-and psychologically-oriented account of cognition (Commentary on Aristotle's Posterior Analytics Prologue; Summa theologiae Ia.7586; Disputed Questions on Truth I, X). Aquinas's most detailed epistemological reflections occur in the context of his discussion of the propositional attitude scientia, which he conceives of as the paradigm of knowledge. To have scientia with respect to a given thing is to have complete and certain cognition of its truth; that is, to hold a given proposition on grounds that guarantee its truth in a certain way. Following Aristotle, Aquinas holds that grounds of this sort are provided only by demonstrative syllogisms, and so he maintains that the objects of scientia are propositions one holds on the basis of demonstrative syllogisms. To have scientia with respect to some proposition p, then, is to have a particular sort of inferential justification for p (Commentary on Aristotle's Posterior Analytics I.4). Now Aquinas holds that because the sort of justification essential to scientia is inferential, it is also derivative: scientia acquires its positive epistemic status from the premisses of the demonstrative syllogism and the nature of the syllogistic inference. Hence, he holds a principle of inferential justification according to which one is justified in holding the conclusion of some demonstration only if one is justified in holding the demonstration's premisses. The premisses that ground scientia are not only logically but also epistemically prior to the conclusion (Commentary on Aristotle's Posterior Analytics I.6). Aquinas argues that our justification for holding the premisses of demonstrative syllogisms cannot in every case be inferential: some propositions must have their positive epistemic status not by virtue of an inference (per demonstrationem) but non-inferentially, by virtue of themselves (per se) (Commentary on Aristotle's Posterior Analytics I.4, 7). Propositions that are known by virtue of themselves (per se nota) are Aquinas's epistemic first principles, the foundations of his foundationalist account of scientia. He offers two sorts of argument for his epistemological FOUNDATIONALISM. The first proceeds by attacking rival accounts of justification, concluding that inferential justification is possible only if there is non-inferential justification. If one holds that all justification is inferential and if a person is inferentially justified in holding some proposition only if he is justified in holding the premisses of the relevant inference, then one is committed to an infinite regress of justification (see INFINITE REGRESS ARGUMENT). If the regress is linear, then there can be no justification since one cannot possess an infinite number of distinct inferences. But if one tries Page 20 to avoid this sceptical result by maintaining that the regress of justification circles back on itself in such a way that an inferentially justified conclusion appears as part of the (proximate or remote) justification for itself, then one is committed to absurdities such as that one and the same proposition can be at once epistemically both prior and posterior to some other proposition and that some proposition can be epistemically prior to itself. Aquinas concludes that if inferential justification is to be possible, there must be noninferential justification (Commentary on Aristotle's Posterior Analytics I.78). His second argument for foundationalism rests on his positive characterization of the nature of non-inferential justification. He holds that certain propositions (immediate propositions) are knowable per se by virtue of the fact that their predicates belong to the definition (ratio) of their subjects (Commentary on Aristotle's Posterior Analytics I.5). We are non-inferentially justified in holding propositions of this sort because when we are aware that a proposition's predicate belongs to the definition of its subject, we are directly aware of the proposition's necessary truth and cannot be mistaken about it (Commentary on Aristotle's Posterior Analytics I.7, 19, 20,44; Summa theologiae Ia.17.3.AD2). To be aware that a given proposition is of this sort we must conceive its subject and predicate, which requires us to have attained explicit grasp of the real natures referred to by the subject and predicate terms (Commentary on Aristotle's Posterior Analytics I.2, 4, II.8). For example, when we have grasped the real nature human being (the real definition of which is rational animal), we cannot help but be aware of the necessary truth of the proposition 'A human being is an animal'. Aquinas holds, however, that attaining explicit grasp of the real natures of things can be difficult, and so he holds that not all propositions in which the predicate belongs to the definition of the subject are actually known to be of that sort. The subjects and predicates of certain purely formal, a priori propositions (e.g. those of logic and mathematics) are more easily cognized by human beings (because of their relative independence from matter), and so most people will recognize first principles of this sort as such and be non-inferentially justified in holding them. The subjects and predicates of a posteriori propositions (e.g. those of natural science), however, are accessible only with difficulty, and hence objective first principles of this sort may not be recognized as such. Aquinas holds that each of us has experience of being directly acquainted with the necessary truth of at least some first principles and that this provides us with sufficient reason for thinking that there is non-inferential justification (Commentary on Aristotle's Posterior Analytics I.4, 25, 41; Commentary on Boethius's De Trinitate 5). This account of non-inferential justification requires him to give an account of our cognitive relations to universal real natures (such as human being), the elements out of which complex (propositional) knowledge is built. He thinks of his account as resolving an ancient epistemological puzzle. The puzzle is how human beings, whose senses provide access to a world of irreducibly particular corporeal objects, can have cognition of universals. Aquinas rejects Platonist and Neoplatonist solutions that postulate the possibility of some sort of extrasensory contact with independent, immaterial universals. He not only rejects the existence of universals of this sort, but also holds that because human beings are by nature unified corporeal substances whose natural form of access to the world is through the bodily senses, all human cognition arises from sense perception. Aquinas's solution to the epistemological puzzle is a theory of intellective abstraction: cognition of universals, like all human cognition, originates from sense perception, and so from the external world of material particulars; but human beings possess a cognitive capacity (in particular what he calls an active intellect), which acts on sensory data to produce intelligible universals. We cognize the universal real natures that constitute the subjects and predicates of epistemic first principles when we possess actually intelligible species or forms abstracted by this mechanism from the material conditions that render them merely potentially intelligible Page 21 (Commentary on Aristotle's Posterior Analytics II.20; Summa theologiae Ia.79, 846; Disputed Questions on Truth X.6). According to Aquinas, then, we have paradigmatic knowledge when we hold a first principle by virtue of seeing that its predicate belongs to the definition of its subject (i.e. when we possess understanding intellectus of it) or hold a proposition on the basis of a demonstrative syllogism the premisses of which we hold in that way (i.e. have scientia with respect to it). He recognizes, however, that these conditions restrict paradigmatic knowledge to a very narrow range, and he allows that there is knowledge other than paradigmatic knowledge. First, following the ancient Greek distinction between demonstrative and dialectical reasoning, he allows that dialectical (probabile, persuasoria) reasoning can provide epistemic justification. Reasoning of this sort is distinguished by virtue of its producing conclusions that are not certain but merely probable. So-called probable arguments rely on premisses that are not necessary and certain but possess some positive epistemic status (propositions held by most people, on good authority, inductive grounds, etc.) and make use of broadly inductive argument forms (enumerative induction, analogy, probabilistic argument forms, etc.). Justificatory grounds of this sort give rise not to scientia but to opinion (opinio) or belief (fides), and Aquinas holds that one can be justified in holding propositions one holds in this way (Commentary on Aristotle's Posterior Analytics Prologue, I.44; Commentary on Boethius's De Trinitate 2.1.AD5; Summa theologiae IIaIIae.2.1, 2.9.AD3; Summa contra gentiles I.9).