Symposium on Being Known

BOOK SYMPOSIUM Being Known

By   Clarendon Press, 1999. x + 358 pp. £40.00 cloth, £15.99 paper

SUMMARY   New York University Being Known discusses some of the issues arising out of the Integration Challenge, the challenge of reconciling, in any given domain, a credible metaphysics for that domain with an acceptable epistemology for that domain. This task of reconciliation often requires revision in our existing philosophical theory of a particular domain. When it does so, we have a range of options. We may revise our metaphysics; or our epistemology; or both; or we may revise our conception of the relation between the two. More radical options include: the introduction of surrogate, allegedly problem-avoiding, truthconditions for sentences about the problematic domain; the rejection of truthconditions at all for the domain; the adoption of scepticism for thought about this domain; and, at the end of this road, rejection of the idea that the domain involves an intelligible subject-matter at all. There is a class of cases in which a good theory of intentional content for thought about a given domain will close the apparent gap that leads to the Integration Challenge in that domain. This is the class of cases in which concepts of the domain are epistemically-individuated, that is, individuated by the conditions under which certain contents containing them are knowledge. In these cases, there will be sufficient conditions for knowing a given content that are supplied by an appropriate theory of that intentional content. So, in these cases, the substantive task devolves into the task of developing a theory of intentional content for concepts of the domain in question. Two very different broad approaches exist, each of which is appropriate for different kinds of domain. One model, that of constitutively causally sensitive conceptions, suits those cases in which the capacity for causal explanation plays an essential role either in the truth of statements about that domain, or in our having concepts in that domain. Here we would not expect truth in the domain to be an a priori matter. The other model, that of implicitly known principles, applies in domains where causal explanation does not play such a 81 © Blackwell Publishers Ltd. 2001 2001, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA

PHIB42.2C01

81

3/15/01, 7:53 PM

role, and where truth is fundamentally an a priori matter. In these cases, implicitly known principles have contents which determine the truth-values of statements in the domain, and which also enter the thinker’s evaluation of sentences about the domain. In this second model, we do not have a causal epistemology. Being Known argues that the first model is appropriate for our understanding of the past, while the second model is appropriate for our understanding of metaphysical necessity. There are two elements in our understanding of past-tense contents. One is our implicit knowledge of the property-identity link, which is the natural generalisation of the principle: A thought (utterance) ‘Yesterday it rained’ is true if and only if yesterday had the same property as today is required to have for a present-tense thought (utterance) ‘It is now raining’ to be true when evaluated with respect to today. Taken together with the fact that many present-tense predications are predications of categorical properties that resist reduction to counterfactuals or to dispositions, the property-identity link implies that past-tense statements cannot generally be analysed in counterfactual terms. Since counterfactuals are employed by verificationist theories of the meaning of past-tense statements, this approach is fundamentally opposed to verificationism. The other component in the account of a thinker’s understanding of the past tense is a causal sensitivity of the thinker’s temporal impressions to instances of temporal relations themselves. This externally-characterised sensitivity is part of what makes the impression a temporal impression. Each of these two elements requires the other. The property-identity link itself involves the past tense, which could not be grasped unless the thinker were capable of temporal impressions. On the other hand, memory and temporal impressions give knowledge only because there is a uniform property any arbitrary time has to be for (say) ‘It is raining’ to be true with respect to that time. In the modal case, Being Known attempts to steer a middle course between David Lewis’s modal realism on the one hand, and various mind-dependent treatments of modality on the other. This middle course adopts a principlebased approach. It attempts to characterise a set of principles, the Principles of Possibility, which taken collectively give necessary and sufficient conditions for a description to represent a genuine possibility. The Principles of Possibility fall into three classes. First there is the Modal Extension Principle, which in outline requires that any genuine possibility respect the same rules for fixing the semantic values of concepts as fix those values in the actual world. Second, there are principles requiring that any genuine possibility respect what individuates the objects, properties and relations which it concerns. Finally there is a Principle of Plenitude, which has the consequence that something is possible provided it is not excluded by the other Principles of Possibility. These Principles of Possibility serve not only the metaphysics, but also the theory of understanding and of epistemology. To understand metaphysical 82 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

82

3/15/01, 7:53 PM

necessity is to have tacit knowledge of what is stated in the Principles of Possibility, and to draw on that knowledge in evaluating modal claims. A method of coming to know a modal truth can be ratified as yielding knowledge if there is a derivation of its soundness from the Principles of Possibility that collectively determine modal truth. The Principles of Possibility themselves are a priori. Though there are, as Kripke pointed out, necessary a posteriori truths, Being Known argues that every such case is one in which the truth can be explained as a consequence of truths each one of which is either modal and a priori, or is a posteriori and also non-modal. If this is correct, truth in the modal domain is fundamentally a priori. The source of modal truth must itself also be a priori, as the Principles of Possibility are. This summarises the first two-thirds of the book, and I have devoted the bulk of the summary to that part, since the commentators in this Symposium have concentrated on that first two-thirds. The final third of the book discusses the Integration Challenge in three other areas: that of self-knowledge and intentional content, where there is a need to reconcile the external individuation of much intentional content with the ability of thinkers to know the content of many of their mental states without checking on their external relations; that of the self itself, where some thinkers have been tempted by their correct recognition of the distinctive epistemological features of the first person to postulate an exotic, even transcendent, metaphysics of the self; and that of freedom, where it has proved difficult even to state a coherent metaphysics of the notion. From the discussion of all of these areas, Being Known draws the conclusion that we can learn more about the metaphysics and epistemology of any given domain by considering them not in isolation, but in the light of the relations they must bear to one another.

83 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

83

3/15/01, 7:53 PM

COMMENTS ON BEING KNOWN   The University of Virginia Christopher Peacocke takes up the challenge of presenting, for a given domain of knowledge, an account “that makes it intelligible how we have the knowledge in that domain that we do have” (p. 2). This is an interesting project, particularly because it is being undertaken by such a distinguished philosopher. My attempt to contribute to it will be limited to raising criticisms, repressing the fear that they may be symptomatic of a lack of understanding, in the hope that providing Peacocke an occasion for clarification will make them worth bringing up. In the case of knowledge about the past, Peacocke says that “a thinker’s understanding of the sentence ‘Yesterday it rained’ . . . involves his having the information that this past-tense thought is true just in case yesterday had the same property that today has to have for it to be true that it is raining today” (p. 59). While many thinkers capable of knowing perfectly well that it rained yesterday would be incapable of grasping that formula, Peacocke makes that irrelevant by adding that the required knowledge would be “implicit”. Still, to justify such an attribution of knowledge, someone ought to be able to understand and make explicit what is being known. And in the case of the past, this is to confront famous philosophical puzzles about ‘tense’. Peacocke says, “If the present time and some past time can have the same property, that of being a time at which rain occurs␣ .␣ .␣ .” (p. 62) as if they surely can. Is there then a property, being a time at which rain occurs, which both today and yesterday have? Or is it that yesterday had that property while today has it? Is there such a time as yesterday, or is it just that there was such a time, which no longer exists? One line of answer to these questions is that all times are equally real, and grammatical tense is primarily a device for locating the time at which the statement is being made relative to the time of the event described. Misunderstanding this device gives rise to the illusion that there are such properties as being present or being past. Ideally, one could say ‘tenselessly’ that time t1 is a time at which rain occurs, independently of indicating where the time of this saying is with respect to t1. This line is sometimes accused of ‘spatializing time’, though the charge is disputed. Peacocke says that his position is consistent “with a treatment of the future as open” and thus “cannot be described as ‘spatializing time’ ” (p. 116). This is not clear to me. Peacocke would apply the same rule to the future tense as to the past, that a thinker’s understanding of ‘Tomorrow it will rain’ involves his having the information that that future tense thought is true just in case tomorrow has the same property that today has to have for it to be true that it is raining today. So how does the future differ from the past? It might be 84 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

84

3/15/01, 7:53 PM

more difficult to know that the required property is possessed by tomorrow than it is for yesterday. But that would not be enough to warrant the conclusion that the future is not equally real. Peacocke says that his position is not consistent with the view expressed by Prior “that only the present is real”. He effectively rebuts Prior’s argument for this claim, charging that, “By this reasoning, one could equally argue that rain elsewhere is unreal” (p. 114). He does not consider a more historical argument according to which one knows the difference between remembering, present perception and anticipation. Of course ‘real’ has a wide range of uses. When a student leaves my office, our conversation remains a valued memory and in that sense is very real. But it no longer exists, and in that sense is not real. When a student has just left, and another is nearing the door about to enter, I am aware of a present (brief ) lull and of remembering a past conversation which once existed but no longer does and of anticipating a future conversation which will exist but does not yet. To mark this by calling the present lull real and the other two items unreal is not without problems. For one, there is the question whether this is saying of some thing that it is non-existent or rather, saying that there is no existent thing of such and such kind. But it is a plausible alternative to the ‘tenseless’ view. It is possible to perceive here and elsewhere at once, the speed of light not being a factor in such common observations. I can see directly that both here and across the street are equally real and on that basis ascend to the view that many distinct places are equally real. (We need not concern ourselves with all places, leaving the stars to the physicists.) The falsity of the view that all times are equally real can be seen by reflecting on familiar features of experience. There is nothing in spatial perception to correspond to the distinction between memory, present perception and anticipation. Whether or not the above position is true, it is implausible to suggest that ordinary folk who know that it rained yesterday ‘implicitly’ know whether it is. The competing philosophers of time all need to produce some epistemology, yes, to be complete philosophers. But it would be a rare extremist among this group who would hold that none of his opponents are capable of knowing that it rained yesterday. On Peacocke’s view, it seems that his philosophical opponents about the nature of time and tense must implicitly know that they are mistaken. There is an argument against scepticism to the effect that the sceptic’s claims that something is not knowable themselves somehow imply that that thing is knowable after all. This might be an effective strategy against a philosopher who holds that it is impossible to know anything about the past. That may indeed conflict with displays of competence in normal everyday discourse. But philosophical doctrines about such matters as the ontological status of the putative subjects or predicates of claims about the past are not likely to reach into the realm of common knowledge in that way. If the reply is that we should not take a philosopher’s pronouncements about time and existence as showing he does not know ordinary things even when the ‘implicit’ knowledge this requires is at odds with his pronouncements, then it would seem that the ‘integration challenge’ is not much of a 85 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

85

3/15/01, 7:53 PM

challenge. To say what ‘makes it intelligible that we have the knowledge we do have’ given some metaphysical conception of what there is to be known would seem to require little beyond saying that people, by virtue of functioning in the alleged state of things, thus implicitly know enough about it to so function. It can easily happen that people have real difficulty in getting to know whether it rained yesterday. Scientists studying a forest tract may get into disagreement over whether it rained there yesterday and one of them may present a winning argument for his side. Epistemologists have been interested in describing the sort of considerations which might be relevant in such a disagreement. Peacocke’s project must be very different from those. Turning to ‘our’ knowledge of modal truths, knowledge that some proposition is necessarily true or possibly true, it seems that there is no realm of common knowledge to provide a test for a theory about knowability of truths in the domain. Philosophers’ views as to the ‘reality’ of past present or future can be held with little connection with everyday questions about what happened yesterday. But those homely topics do at least involve the notions in question. One who finds no division into necessary and contingent truths will have to contend only with the more esoteric common sense of some philosophers. There is a common-sense consensus that some knowledge requires less stirring about the world than some other. Both arithmetic and geology can be learned from an armchair, but the latter required some people going to work with shovels and hammers. This vague gradation of degrees of ‘dependence on experience’, or ‘a priority’ of knowledge does not guarantee any corresponding agreement about necessity. Philosophers who claim to identify propositions which are ‘necessary’ but not ‘a priori’ or ‘a priori’ but not ‘necessary’ are writing in honey. Peacocke explains knowledge of modal truths in terms of knowing how to give acceptable descriptions of ‘possible worlds’. This terminology has the defect of suggesting such questions as whether the proposition that P is true in world W is itself necessarily true. The thought that it is true in some worlds but not in others suggests a vicious regress undercutting ever being able to say simply that any proposition X is true in a world Y. It would be true in Y in Y (perhaps) or some other worlds but not in others. To avoid this, we might say that ‘P is true in W’ is either necessarily true or necessarily false. That avoids regress, but at the cost of a disturbing application to such an example as ‘It is true in this world that Norfolk is the largest city in Virginia’. To call that necessarily true conflicts with the fact that it would ordinarily be taken as just a way of saying that Norfolk is the largest city in Virginia, which is a typical standard case of something non-necessary. The introduction of the ‘worlds’ terminology invites such conflict. This is not to say that the users of this terminology cannot make enough stipulations to avoid any inconsistencies. It is just that their sheer arbitrariness makes it doubtful that there could be any common sense knowledge claims turning on the answers. Peacocke holds that properly describing possible worlds involves principles of possibility and necessity such as that “A thought or proposition is possible iff it is true according to some admissible assignment” (p. 150) (for necessity, 86 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

86

3/15/01, 7:53 PM

“true in all admissible assignments”). He responds to a charge of circularity, saying “it may be objected that the only reason that we say the same rule is applied in determining the extension of bachelor as is applied in determining the extension of unmarried man is that it is necessary that all and only bachelors are unmarried men” (p. 159). This charge presupposes that there is a rule which is applied in determining both these extensions. Thus an unmarried human specimen of indeterminate sex either is or is not in this extension, no matter how much disagreement there might be as to the sex of that person. Perhaps this latter suggestion could be disavowed, leaving only the assumption that the extension is determined, to the extent that it is determined at all, by the same rule for the two distinct things ‘bachelor’ and ‘unmarried man’, whatever they may be. The extension of a genuine property1 is determined independently of whether any humans grasp it. There are exceptions to this rule, as in the case of the property of being a property which is grasped by people, but the extension of a property is not something determined by people following a rule. One can say that a rule is independent of people in that they did not choose it and cannot change it except in ways not under their direct control, as in the drifting of social norms. This is not enough to properly characterise genuine necessary relations between genuine properties. To analyse necessity in terms of concepts rather than properties and in terms of the fixity of rules leads to assumptions which make conventionalism about necessity more formidable than it should be. Among the leading realists about properties nowadays, the scientific realists, such ‘properties’ as being a bachelor would be the first to be rejected as mere shadows of words. From the standpoint of the assumption that necessity is to be based on the identity of the concepts in currency among speakers, the charge of circularity mentioned above is indeed not well founded and would seem to attribute to a necessarily obtaining state of affairs quite an implausible causal role. In his reply to the charge, Peacocke takes the line that the identity of the concepts bachelor and unmarried man need not invoke their having the same application in all possible circumstances: “identity of concepts may be explained in constitutive terms, in terms of what it is to possess one concept rather than another”. The availability of such explanation has been questioned by Quineans doubtful that we know what we would say in every circumstance. But suppose that we could know this about some person and some ‘concepts’ of his. Peacocke wisely stresses that The use of the identity condition for concepts is a use of modality by us as theorists. It is not a use by the thinker whose mastery of concepts is being characterized. The distinction is important. (p. 160) Indeed it is. It might be useful in the case of the past, to say that the ‘implicit’ knowledge allegedly possessed by the ordinary person aware that it 1. As opposed, for example, to the alleged property of being a class which is not a member of itself.

87 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

87

3/15/01, 7:53 PM

rained yesterday consists in his concepts being arranged in ways exhibited by him in his performances but perhaps known only by the theorists explaining those performances. Consider an example used by Grice and Strawson in their reply to Quine.2 (1) ‘My neighbour’s three year old child understands Russell’s Theory of Descriptions’ is held to be clearly not analytically false, while (2) ‘My neighbour’s three year old child is an adult’ is clearly analytically false. Grice and Strawson dismiss the suggestion that (2) might be defended in the case of an incredibly mature child. But the Quinean position here is quite strong. There is no good scientific basis for holding that people would be unable to accept (2). The judgement that the state of people’s conceptual scheme would preclude this could turn out to be mistaken. It could be replied, by Peacocke or anyone else, that such a mistake would not prove that the very idea of concept identity is useless. It might be possible (in some non-controversial meaning of this term) to distinguish between the programming of the man who refuses to consider that a three year old could be an adult but backs down on encountering a medical anomaly and another who complains that the usage has changed or become slack, or appeals to the legal definition of ‘adult’. Perhaps there could be a successful distinction between one who refuses to give up because he is inept about verbal issues, or stubborn or excessively legalistic, and one whose refusal is wired deeper in his neural circuitry. There then might be a successful scientific distinction between a man’s insistence that no bachelor is married and his apparently equally firm insistence that no bachelor is over twenty feet tall. The man himself might be utterly uncomprehending of the suggestion that it is of course possible for there to be a man of that height. He might insist that it is impossible. But ‘theorists’ could ascertain that the man’s programming would allow him to recognise a certain presentation (as in a science fiction film) as being of a twenty-five foot bachelor while none would allow him to recognise a married one. Perhaps certain inputs trigger unmarried and others trigger man and all the program for bachelor does is line these inputs up with the program for conjunction. Under twenty feet is not included in the programming in that man. Our man just does not know these facts about himself the way we theorists can. One would not have to be a Quinean to be sceptical whether such knowledge of speakers is possible. (And it would be easier to be sceptical if not a Quinean since one could be free of discomfort about denying the possibility of something.) Even if it were possible, it would seem to be a private feature of a speaker. Those who believed that whales were not mammals probably did not know that they suckle their young. But some who insist that they are for all that fish just have a different meaning for ‘fish’ from the scientific one. Is it that their concept of fish includes whales or that they are mistaken about the concept? They may not be mistaken about their concept, if that is a matter of having certain responses built into their programming. But is the extension of a concept determined by an individual’s concept or by some ‘social concept’?

2. ‘In Defense of a Dogma’, The Philosophical Review, 65 (1956).

88 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

88

3/15/01, 7:53 PM

Either way, it is not determined with logical exactitude, in the way the extension of a property is determined by the property. Everything is either F or not and not both, by the necessary truths of logic. But this notion of extension has been cast in doubt by such paradoxes as Russell’s about the ‘property’ of being a non-self-member. Many have concluded that it cannot be true that every property uniquely determines an extension and are ready to leave such determination as there is up to people, either individually or socially. Peacocke says that “the actual extension of an atomic observational shape concept diamond-shaped consists of all objects which have the same shape as those which, as they actually are, would produce a certain kind of experience in a properly perceiving subject” (p. 135) If we consider a familiar sorites-type series beginning with a universally accepted diamond shape and progressing by steps such that the unaided eye of observers cannot discern any difference between adjacent entries, we can by such steps make our way to any shape we choose, unable to ‘determine’ where we stopped encountering diamonds. To say that an assignment s is “admissible” only if it “assigns to diamond-shaped an extension which includes precisely those objects which, according to s, are of the same shape as those which, as they actually are, would produce a certain kind of experience in a properly perceiving subject”, suggests an unwarranted optimism about the prospects for precision about such a question as whether a given figure is diamond-shaped. Of course an assignment s for the extension of a concept will never fail to include precisely those objects which s counts as included. This triviality is surely not what Peacocke intends, in spite of his qualification “according to s”. But if it is a matter of s assigning precisely those objects which produce an experience which elicits ‘diamond-shaped’ from a ‘properly perceiving subject’ then we will need at least the qualification to a given occasion of observation. For in a sorites series, any human subject, however proper, will not be consistent in his application of ‘diamond-shaped’ on repeated runs through the series. Peacocke’s intention appears to be that, in describing a ‘possible world’ we must be constrained by the same rules for determining the extension of a concept as apply in the real world. When we say of a given percussion instrument that it is triangular but it is possible that it should be diamond-shaped, we must apply the same standard for diamond-shaped as we do in real life. But that is not a standard we can satisfy even in real life. Properly perceiving subjects will not determine the same extensions from one occasion to the next, let alone one world to the next. Of course, if we require that the assignments are always made from this world, then the variation in extensions will not be any greater than they are in real life. However, Peacocke seems sympathetic to a proposal by Nathan Salmon purporting to show that something can be necessary without being necessarily necessary. The argument begins with the assumption that it is a requirement for admissible specification of an alternate possible world that “if a particular table in fact originally came from a certain particular quantity of matter m, then according to any genuinely admissible assignment according to which 89 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

89

3/15/01, 7:54 PM

that table exists, the table originally came from a quantity of matter overlapping to some specified degree with that of m”. There could then be a quantity of matter m* such that it is impossible that the table came from m*. But . . . the following may be true: had the table had somewhat different origins it could then have come from m*. Something is possibly possible for the table which is not actually possible for it. (p. 196) The idea seems to be that for some actual table T which was made from matter M, a world description in which T is depicted as coming from altogether different matter is not admissible, but some slight differences might nonetheless be allowed. Setting aside the question where such a rule comes from, it would seem that our capacity to recognise suitable remnants of M would involve the usual sorites troubles. Still, we might be able to recognise M* as unacceptably different from M in any case. But now it seems that, from the standpoint of a world description featuring an acceptable difference from M, say M´, we might be able to find M* an acceptable difference from M´. Thus it would be possible for it to be possible for T to come from M* but not possible for T to come from M*. For the purposes of technical modal logic, it can be true in a world W that P is necessary even if there is a world in which P is false. The line is that P is true in all worlds ‘accessible’ to W. This appeal to an accessibility relation is a clever technical device, but there remains the question of giving it some interpretation applicable to what might be known to be necessary or possible. Here I have to confess to being at a loss to understand how stories about T originating from M, M´ or M* could have any such application. We can presumably tell a story in which T came from M and get credited with stating the actual facts. We can with equal facility tell stories in which T came instead from M´ and stories in which T came from M*. The M´ stories establish a possible difference in origin. The M* stories do not, but do establish a possible possible difference. We can of course say that if, say, a three molecule difference is ‘permissible’ then starting from M, M-3 will do, but M-6 will not, but starting from M-3, M-6 will be acceptable. There might even be some psychological truths to be found along this line. You might say, ‘We can see, from this world, that an M-6 (as source of T) world is out, but if we were in an M-3 world the M-6 world would look better.’ And if we were in a drunken stupor, 2 + 2 = 5 might look good. It has been suggested that the accessibility relation in technical modal logic could be explained in such psychological terms, with stipulations to rule out the drunken cases. No doubt it could be so interpreted. But it could be interpreted otherwise too, so that the accessibility relation would be transitive. It would seem that the account of knowledge in this area would then reduce to giving an account of how we learn to keep track of which word game is in play. Peacocke says that “It is a matter for further philosophical research whether . . . colour incompatibilities should be included among the constitutive properties of colours. . . . These issues␣ .␣ .␣ . merit separate books in their 90 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

90

3/15/01, 7:54 PM

own right” (p. 146). A book by Peacocke on colours would be highly valuable and might create a way of thinking about colours which made ‘Nothing is entirely red and entirely green at the same time’ seem to be clearly resolved by considerations of concept identity. But it would be excessively optimistic to assume that such research would be discovering the existence of determinate boundaries of concepts. As for real properties, the scientific realists are not much more enthusiastic about colours than they are about bachelorhood.

91 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

91

3/15/01, 7:54 PM

YES, BUT WHAT IS THE MOTHER OF NECESSITY?   The University of Sydney It’s a truism of philosophy that Realists must not postulate more than we could reasonably hope to know, while Anti-Realists must not leave us with so little that all knowledge is impossible. But balance is not easily come by—and even less in philosophy than in life. So philosophy continues to struggle over the hard cases, with neither the Realist nor the Anti-Realist able to score an easy victory. One of the most difficult cases, and the most contended, is that of alethic modality: the concepts of necessity and possibility. A quick sketch will serve to bring out the problem. Let us take for granted, for the moment, the standard Kripkean semantics. Then there are a set of possible worlds and an accessibility relation on those worlds that is reflexive, symmetric and transitive. This gives us our usual philosophical notions of possibility and necessity (S5). Suppose that we wish to be Realists about the ontology that this semantics appears to presuppose. Then we will think that an infinity of such possible worlds genuinely exists. But we should then also be Realists about the accessibility relation, taking it to be a model of our epistemic capacities with respect to such worlds. But that seems absurd: we do not have knowledge of what goes on in other possible worlds. So Realism seems doomed to failure. But what if we started by being modest about our epistemic capacities, limiting any accessibility relation to this world; what then would we get in the way of necessity? The answer is the disappointing one, that we get collapse into the propositional calculus. No modalities at all. The moral is that if we start out with our desired modal logic and try to read from it our ontic and epistemic commitments then it rapidly exceeds what is possible—while if we start with reasonable assumptions about what we can know and try to build our modal logic on that basis then we end up, seemingly, with no modal logic at all. And yet we have good reason to believe that we need modal concepts to make sense of both logical and mathematical truths and logical and mathematical inference. We are at an impasse. Clearly we need a new idea. I Christopher Peacocke’s Being Known offers a new idea for the solution of this problem. He argues that we have implicit knowledge of certain Principles of Possibility. These Principles are represented as constraints on the referential assignments that can be made to the concepts that we employ, so that an admissible assignment (under the Principles) represents a genuinely possible state of affairs. Thus an assignment is simply a function (in the mathematical 92 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

92

3/15/01, 7:54 PM

sense) from concepts to the referents that are appropriate to their semantic class: singular terms are mapped to objects; predicates are mapped to properties and relations; quantifiers are mapped to a domain of objects, etc. These are then built up into assignments to whole propositions and a specification is the set of all propositions that are true under an assignment. (Peacocke speaks of ‘Fregean Thoughts’ where I have used ‘proposition’ but we can be flexible with our terminology here.) What, then, makes an assignment admissible? According to Peacocke an assignment is admissible if it respects the rules that we use when we assign extensions to all concepts in the actual world. Thus if all horses are mammals in the actual world then this subset relation has to be preserved when we have a different referential assignment to horses and mammals. The network of reference assignments that are the result of the application of our semantic rules is then meant to determine, in a broad, structural way, the limits of what is to count as an admissible interpretation. In brief: it fixes the things that must be so, leaving free the things that could have been otherwise. Two worries surface almost immediately. Firstly, Peacocke does not mention the semantic rules that apply in our world to the very many terms that have a modal component—terms other than alethic possibility and necessity, such as believable, dangerous, attractive, provable, admissible, etc. All such terms must have their meaning fixed not by (what we could think of as) the first-order semantic rules, but only by the incomplete determination of the semantics at the first-order level. This is an odd, and not entirely natural, semantic division; it also creates concerns about a regress of determinations (admissible?) and, of course, circularity. It also means that Peacocke cannot be right when he imagines the assignment function as complete. But let us leave that problem aside for a moment. There is a second problem. Even if we restrict our attention to concepts with no modal component it is not obvious that we can make the distinction between concepts, or relations between concepts, that are fixed by our semantic rules and those that are not so determined and that therefore are allowed to vary. It is hard, for example, to see how mathematics is going to come out necessary in this way. Of course, we can always gerrymander the rules and the consequent admissible assignments, but that is hardly in the spirit of Peacocke’s proposal. There are meant to be natural assignments, readily grasped, corresponding to our ready grasp of modal truths. And yet there is something undeniably appealing about the idea of basing our understanding of modalities on permutations of the assignments found in the actual world. These advantages can be seen in a far simpler, and I think much clearer, example: the tautologies of propositional logic. Properly viewed, a truth-table evaluates a formula consisting of propositional variables, with each row of the table representing equivalence classes of substituted statements, equivalent in truth value. Yet there is an easy slide to a different view in which the table is seen as evaluating some actual proposition, with the different lines of the table representing the truth values that the proposition might have in other possible worlds. And it is only in this second, not entirely legitimate picture, that tautologies can be seen as having any real 93 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

93

3/15/01, 7:54 PM

connection with necessary truths.1 Certainly there is no way to read off necessity from a table that is evaluating a formula with propositional variables: what does truth under all substitutions have to do with other possible worlds? The alternative picture of truth-tables gives us a combinatorial breakdown of possible states of affairs and their co-variance. Their rows represent worlds in which snow is not white, grass is not green, the good prosper and evil-doers are punished. It is this combinatorial representation of other states of affairs that gives us a grip on modal truths. It also shows us why the argument that I gave at the beginning of this paper was too swift: it showed that we cannot build modalities on top of a PC base without encountering epistemic difficulties, but that does not mean that we cannot find modal truths within PC. We can, and we can use this to justify the traditional view of logical truths as a species of necessary truth. But having done so we can now turn the original problem on its head by asking what modal logic the tautologies of PC satisfy. This question was asked and answered by E.J. Lemmon as long ago as the late fifties: the truth table tautologies satisfy a system that he called S0.5.2 This example of PC tautologies provides, I think, a simple model of the kind of strategy that Peacocke wants to follow. Hold fixed the meanings of certain terms (concepts, in Peacocke’s way of speaking) which correspond to our logical vocabulary; allow the reference of other terms (those which occupy the variable position in the corresponding formula) to take other referents under the assignment function; the necessary truths are those that are true under every admissible assignment. Of course, Peacocke does a lot of the additional work necessary to make this strategy work—for, left to itself, it will fall far short of delivering everything that we want to count as a necessary truth. In particular he expends a good deal of effort to show that Kripkean a posteriori necessities of constitution and identity are secured by his analysis. But having one clear case where the strategy looks to be successful provides some reason to think that it can be extended and refined to cover the difficult cases as well.3

1. The reason the view is not entirely legitimate is that there may be relations between the component statements of a composite proposition that make some lines of the table strictly impossible—however this does not affect our understanding of tautologies as necessary truths. It merely means that some propositions that are necessary, though non-tautologous, will be wrongly counted contingent. See my Shadows of Necessity: Deductivism, Modality, and the Limiting of Reason (forthcoming) for more on this, and an example. 2. E.J. Lemmon ‘Is there only one correct system of modal logic?’, Aristotelian Society Supplementary Volume XXXIII (1956), pp. 23–40. Despite its low numbering this is still a T-based system. 3. There is a connection—which I will leave unexplored here—between John Etchemendy’s discussion of Tarski’s analysis of logical consequence, and in particular Vann McGee’s response to it in terms of set-theoretic surrogates for possible worlds, and Peacocke’s admissible assignments. Suffice it to say that the need for abstract objects to ensure the extensional adequacy of the standard (Tarskian) model-theoretic analysis of logical truth does not bode well for any attempt to make the non-abstract elements of the actual world stretch to all possibilities. It is overwhelmingly likely that Peacocke will have to take a very revisionary attitude to modal truth.

94 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

94

3/15/01, 7:54 PM

II But though there are reasons to be optimistic there is also cause for disquiet. Peacocke speaks confidently of Principles of Possibility without it ever becoming clear what—ontologically—these things might be. It is clear enough that they are not meant to be simply subjective rules that we choose to adopt: Peacocke rejects, rightly, I think, all, so-called, non-cognitivist views. The Principles of Possibility are meant to objectively exist; they stand to us as objects of our knowledge. Yet what are they? Are they to be seen as a species of law of nature—like the Fermi Exclusion Principle? Or are they more abstract than that, and more akin to an algebra, or mathematical induction, or the Reflection Principle in set theory? But however we classify them it seems natural to ask the question: must these Principles be as they are, or could they have been otherwise? In other words, are they contingent or necessary? And to that question no answer seems possible; for if they are contingent then other principles are possible and we have lost the ground for any genuine necessity; or the Principles could not have been otherwise and are therefore themselves necessary, and we must look for the ground of necessity in whatever it is that has determined that they be so. The only response that I can imagine to this Euthyphro-style argument is just to insist that the Principles themselves are neither contingent nor necessary—and that only serves to heighten the mystery that surrounds them (are we now saying that it is impossible that they be either contingent or necessary? And what might that mean?)4 But there is some reason to think that Peacocke would try to finesse this argument by seeking to apply these Principles to themselves. It is by this means that he hopes to get the iterated modalities that are needed for S4 or S5, so perhaps the same trick can be used to give a modal status to the principles. To sharpen the point, let us bundle all the modal principles together and call them P*. Then we can say that snow is white or snow is not white is true under all admissible assignments (or by dint of the auxiliary modal principles). Thus we have P* (snow is white or snow is not white). The question is, is it true that P* (P* (snow is white or snow is not white) )? And, unfortunately, the answer is that, no, it isn’t obviously true. It isn’t obviously true under all admissible assignments—we don’t even have a reason to believe that admissible is semantically fixed; nor does it seem to be a statement of essence or origin. At the heart of this problem is the stuff on which P* has to work, which are concepts and the functions mapping concepts to their referents. If these are meant to 4. It is worth saying that non-cognitivism—which, in the form that it has been given by Edward Craig and Simon Blackburn (and, earlier, Quine), claims that our belief that a proposition is necessary is a matter of its ‘epistemic entrenchment’, our unwillingness to give it up—is open to a similar, Euthyphro-style, objection. Why, we might ask, can the proposition not be given up? It cannot be because of any characteristic of the proposition itself. Nor can it make a distinction between other propositions that we might be unwilling to give up for other reasons and the ‘necessary’ ones. Non-cognitivists invert the natural order of explanation and replace metaphysical mysteries with epistemological ones. But the former are always preferable, for the latter rob us of the rationality that we need to solve any problem.

95 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

95

3/15/01, 7:54 PM

be non-abstract elements of the actual world (and they are, for otherwise the knowledge gap will open up for these abstracta, defeating Peacocke’s intentions) then there is no reason to think that they are necessary. We are perfectly free, as far as the Principles themselves are concerned, to map the concept of concept to something else. It is not one of the things that the admissible assignments hold constant. The moral of this is that Peacocke may be able to capture T-based concepts of necessity and possibility, but he probably cannot answer questions about the modal status of his Principles of Modality, nor can he reach all the way to S4 or S5. This is not a failing peculiar to his analysis, for it infects also the Positivist’s attempt to reduce necessities to analyticities. To see this let us note that, even if we were to agree that it is analytic that snow is white or snow is not white, it certainly does not seem to be true that it is analytic that it is analytic that snow is white or snow is not white.5 But I would further suggest that the only reason that it seems to be analytic that snow is white or snow is not white is because the statement seems to be necessary, and this puts a constraint on meaning. So I would also say that the reason the Principles that Peacocke gives seem to fix what is necessary and possible is because they themselves are constrained to respect antecedent modal facts. And indeed if that were not so it would be completely unobvious why the things that the Principles hold fixed cannot (and that ‘cannot’ has a strong modal sense) vary. I would, however, like to distinguish this problem from another kind of problem which sits near it in logical space: the problem of circularity. Peacocke has a great deal to say about this problem and my own view is that what he says is quite cogent. The charge might run as follows: no explication of modal notions is entitled to use, or advert to, any modal notions whatsoever: not consistency, not conceivability, not truth under all possible substitutions. To use such notions is to give an account that is circular, and consequently illegitimate. Charges of this kind are frequently made in philosophy, and on many occasions they are simply misdirected. In fact the accusation is usually made insensitive to the purposes of explanation. It is true that if one is trying to explain a modal concept to a sceptic who is convinced that there in no such thing, or to the neophyte who hasn’t yet grasped any aspect of the concept, then using modal notions to explicate modal notions would indeed be illegitimate, or at least, unhelpful. But if it is one’s belief that there are modal facts in the world—if that is not something that one is trying to establish—then there is no reason why one

5. It is useful to compare the failure of this way of getting iterated modalities in S4 with the way they are achieved in the standard Kripkean semantics. In the latter we do not ask whether it is true in all possible worlds that it is true in all possible worlds that snow is white or snow is not white, we simply move from a reflexive accessibility relation to a reflexive and transitive relation. It is this ease that makes Kripkean semantics the astonishingly flexible tool that it is. It should not be abandoned lightly. It is also worth noting, here, against the non-cognitivist’s attempted reduction of possibility to a (highly idealised) form of conceivability—an idealisation that I can’t see that he is entitled to—that it also will reach only as far as S4. (See G.E. Hughes and M.J. Cresswell’s Introduction to Modal Logic, (Methuen, 1968) p. 78.

96 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

96

3/15/01, 7:54 PM

should not use modal concepts in the explication of modal facts. There is nothing illegitimate in that.6 If Peacocke is occasionally sensitive to the charge of circularity it is, I suspect, because he has ambitions for something more far-reaching: he wants to refute the sceptic by showing that we have, incontrovertibly, modal knowledge, not just modal beliefs. But that is probably an impossible task. How can one show that there is a cognitive necessity for modal notions to one who is unwilling to grant the use of so much of our ordinary discourse? My own view is that we should be more ruthless with sceptics: if someone doubts that there is any legitimacy to modal notions then let him try to give a satisfactory account of the world without them. If he can, then so much the better, if not then we can move on. In general: best explanation wins. III But however we answer the question about the modal status of the Principles they must at least be true, and objectively true. Yet this raises again the question of how we know that they are true. Strangely, Peacocke has very little to say about this problem, being content to claim that we have a tacit understanding of the Principles when we make modal judgements. In effect this seems to say little more than that the Principles are a priori and innate—and if this is the answer then we might wonder if we have really made any progress beyond Leibniz, who argued for just such a view in the New Essays on Human Understanding.7 And perhaps this helps to place Peacocke’s view: that it is, in essence, a revival of seventeenth-century Rationalism. But if Rationalism foundered it was not because it suggested that there was innate knowledge but rather because it offered no reason to believe that what is innate is true. As far as I can see Peacocke faces the same problem. In fact we can sharpen the problem into a dilemma. Suppose that a belief is a candidate for being considered innate. If there is nothing in experience that acts to make it probable that the belief is true then we cannot count the belief as knowledge—we must rest content with thinking it an innate belief. On the other hand if there is something in our experience that serves to justify the belief then the belief may count as knowledge but there is no longer any

6. Suppose that one were trying to explain the direction of time. It would be absurd for someone to insist that no time-asymmetric processes could be used in the explanation, because ‘to do so would be circular’. Since it is obvious that time is asymmetric, all asymmetric processes (thermodynamic disequilibrium, K-meson decay, non-Unitary Quantum collapse, etc.) are relevant in explaining that fact. 7. Peacocke acknowledges Leibniz as a precursor (p. 171), but the only difference between them that he notes is that Leibniz believed that all necessary truths were a priori, whereas Peacocke believes, with Kripke, that some may be a posteriori. But Leibniz hints at an answer to the question that Peacocke resolutely ignores, for he suggests that innate ideas are placed in our souls by God and that we come to know them by some form of inner attention. And since they had their provenance in God, Leibniz may have thought that that was a sufficient reason to believe that they were true.

97 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

97

3/15/01, 7:54 PM

reason to think it innate, for we could have acquired it through that same experience. Either way, there is no such thing as innate knowledge.8 The whisperings of a modal conscience, no less than a moral one, need not be trusted. We might also wonder whether saying that a set of Principles is innate is any real advance on saying that knowledge of other possible worlds is innate. If the first is an unexplainable fact about us then we do not obviously need to revise our metaphysics to accommodate it. We can just stick with the innateness of the various modal axioms. My own view—to turn positive for a moment—is that the problems Peacocke addresses are not solvable without making severe adjustments in our epistemology. Much of the pressure on Realism comes from the insistence that we be in direct causal contact with all of the items that we wish to believe in. So Realists about mathematical truth face the charge that we do not seem to have direct causal contact with numbers or sets or functions; Realists about modal truths do not have the requisite contact with other possible worlds, etc. Granted this lack of direct causal contact how could we be said to have knowledge of such truths, even supposing that they are true? It is this knowledge gap that the Anti-Realist leverages to such effect.9 I suspect that it is this same exigent epistemology that is driving Peacocke’s concerns. But there is little reason, I suggest, to demand so much: direct causal contact may be sufficient for knowledge (ceteris paribus) but it is not necessary. The method of science is to postulate entities, develop the theory of same, and if the addition of such entities gives us the best overall explanation, then we are entitled to our tentative belief. Thus quarks were believed to exist long before there was direct evidence for them, likewise black holes. Experimental evidence certainly strengthens belief but it need not be negligible without it. We may say the same for the entities that enter into the truth-conditions for mathematics and logic. Since our scientific theories require us to employ both we are entitled to believe that they are true, and we are entitled therefore to believe in the ontology that makes them true. Again: best explanation wins. As a second strand to this, we may note that our modal beliefs are not just a matter of making judgements about what is necessary and what is possible. We also make judgements about what is contingent. Thus were I to see a branch broken from an oak tree here, I would judge it to be a contingent matter of fact because I can see that that oak tree, there, does not have its branch broken. They have similar essences and so where they differ must be in what is accidental. It is this that provides us with a great deal of our modal 8. There is a tiny escape route in this argument that is of little practical value in the present context but that must be mentioned. It is possible that a belief could be innate and such that subsequent experience indicates that, by a sheer stroke of good luck, the belief is true. In this case we could have acquired the belief through experience, it’s just that we didn’t. 9. I should confess here that if there is one piece of anti-realism that I cannot take at all seriously it is anti-realism about the past. For if there is one thing that we are obviously in direct causal contact with it is the past: it is simply the interior of the backward light cone—it is everything that may affect us. Moreover, the statement that the past is what once existed must be a candidate for Most Trivial Truth ever considered.

98 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

98

3/15/01, 7:54 PM

beliefs—though it is often neglected in discussions these days.10 Thus we make judgements about what is contingent based on a posteriori necessities (making them a posteriori contingencies) and we mesh this as best we can with our evolving ideas about logical truth and logical inference. The result is a heterogeneous body of modal beliefs that we try to bring together in a single modal system, S5. In fact the concept of contingency does a great deal to constrain our choice of modal logic, for it turns out that only certain systems allow the definition of such a concept.11 It is useful, I think, to make a distinction between the semantics of modality (which is best given by Kripke’s possible worlds) and the sources of our knowledge of modality. The problem of modal realism has been exacerbated by the too ready assumption that, if what it means for a statement to be necessary is that it be true in all possible worlds, then to know that the statement is necessary must mean that we have knowledge of other possible worlds. Far from this assumption being obviously true, it seems to me nearly obviously false. Our knowledge of modalities arises from our theorising about the actual world—and is susceptible to change as our theories change. It is just part and parcel of our fallible knowledge of the world around us. But the fact that it arises out of our knowledge of the actual world does not mean that it cannot be a genuine knowledge of necessity and possibility. Thus I suggest that in 1953 Quine slightly misidentified the ‘two dogmas of empiricism’ with devastating results for subsequent theorising. It is not the analytic/synthetic distinction that is in need of giving up but rather the a priori/a posteriori distinction. Our knowledge of the world is fallible and revisable in the light of new discoveries and explanation is indeed holistic, as Quine argued, but this does not mean that there are no necessary truths.12 It simply means that our knowledge even of modal truth is a posteriori in character. Thus it does not follow, as Quine believes it follows, that because even our ideas of logical truth are revisable that the truths of logic are not necessary truths. The real dogmas of empiricism are the belief that necessary truths must be known a priori and its dual, that a posteriori knowledge is only ever knowledge of contingencies. It is these dogmas that have made the problem of mathematical knowledge seem so intractable. IV But this brief sketch of a solution is in the opposite direction to Peacocke’s position in Being Known. Seen from a distance his general view resembles a 10. So it is neglected also by Peacocke. Being Known contains no index entry for contingency. 11. Essentially it constrains them to be T-based systems. See M.J. Cresswell’s ‘Necessity and Contingency’ in Studia Logica, 47 (1988), pp. 145–9. 12. In fact I think that only half of Quine’s Holism is plausible. It is right to say that all (or perhaps, almost all) statements are revisable in the light of recalcitrant experience—this is simply to say that no statement is a priori certain; but I suggest that it is deeply implausible to assert that any statement may be able to be retained in the light of recalcitrant experience. In most instances rationality will require the revision of a particular statement. Since I adopt only half of Quine’s Holism, my position may be called ‘Halfism’.

99 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

99

3/15/01, 7:54 PM

kind of Ontological Argument. Reflecting on the meanings of concepts in certain propositions we find that they imply that the propositions must be known, and thus must be true. But whereas it is uncontroversial that knowledge implies truth it is very controversial—indeed quite strange—to suggest that meaning implies knowledge. (It is this claim that Peacocke calls the Linking Thesis.) The only way to understand this thesis, I suggest, is to consider it not as a relation between the meaning of concepts and knowledge but rather as a relation between two different items of knowledge, thus: A knows the meaning of statement ‘p’ entails that A knows that p is true. The Linking Thesis (LT) is the claim that this statement is true for some p. I suggest that LT is never true under the usual meaning of the antecedent, because all that is really required for A to know the meaning of ‘p’ is for A to understand what would be the case were p true. But even in the case of the most plausible candidate for a priori status it is always possible that someone understand the proposition and yet it not be so. The only way that LT will have true instances is if a large number of externalist assumptions are imported into the meaning of ‘A knows the meaning of “p’’ ’ so that if ‘p’ turned out to be false it would also turn out that A did not really know the meaning of ‘p’ after all, appearances notwithstanding. And under these conditions we would never know whether A, or anyone else, really knows the meaning of ‘p’. Thus the price that we will pay for making LT true will be the loss of our grip on whether anyone understands the meaning of certain propositions. Rather than making the world epistemically accessible we will simply have infected our understanding of understanding with the inaccessibility of reality. One doesn’t put out fires by setting oneself alight. At least, it isn’t recommended.

100 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

100

3/15/01, 7:54 PM

COMMENTS ON PEACOCKE   The University of Birmingham I Much of Christopher Peacocke’s Being Known is devoted to careful and illuminating discussion of time and modality, of the self, self-knowledge and freedom. In each case he wants to respond to what he calls the ‘integration challenge’: to show how an account of the possession conditions for a concept can be compatible with our confidence that we possess knowledge of contents which contain that concept. Peacocke proposes a distinctive strategy for responding to this challenge, one which makes use of a distinctive range of new resources. Although I shall discuss some of his applications of his strategy, I shall be mostly concerned with trying to get clear about the abstractly presented overall strategy. Philosophers have traditionally assumed that a concept provides a rule which enables us to establish whether something falls under it. In that case, we can respond to the integration challenge by showing how we can be justified in our beliefs and assertions about different subject matters. If knowledge is justified true belief, we are then on target for an explanation of how knowledge is available too. Peacocke offers an alternative approach. Work on the analysis of knowledge since the 1960s has identified kinds of epistemic dependence and security other than those traditionally employed: whether we have knowledge may depend, not on what reasons we can offer in support of a belief, but rather on whether the belief was causally dependent on its subject matter, whether it was produced by a reliable belief forming process, whether, in forming it, the believer tracked the truth, and so on. Peacocke’s proposal is that we explain the concept of possession by showing that it yields judgements that express knowledge, thus exploiting the kinds of epistemic dependence alluded to above. If a concept is “epistemically individuated”, then possession of it is enough to ensure that (ceteris paribus) when we apply it to something in a “rationally non-discretionary” judgement, then ceteris paribus that judgement constitutes knowledge. The link has an intuitive appeal: according to Peacocke, judgement ‘aims at knowledge’; and many concepts are most directly manifested in judgement. This enables him, in dealing with propositions about the past, to rely upon the contingent causal sensitivity to past events provided by memory to explain how the concept of the past is possible. In obvious ways, this enables him to resist some traditional arguments for anti-realism about the past and to show how our concept of the past allows for events that are epistemically inaccessible. I am most concerned here with understanding the rather abstract formulations Peacocke provides and I shall examine, first, his claims about knowledge in Chapter 2, and, second, the role given to ‘externalist’ factors in his epistemology and metaphysics. 101 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

101

3/15/01, 7:54 PM

II First, unlike Edward Craig and others who explain the concept by reference to its role in identifying reliable sources of testimony, Peacocke insists we think of knowledge as the “aim” of judgement: a “correct” judgement that p is normally a case of knowing that p. Second, and here the links with recent developments in epistemology are clear, two sorts of elements are appealed to in explaining knowledge. One is akin to the idea that knowledge depends upon possession of reliable faculties of perception and memory (etc.): there must be an appropriate informational link between the experiences that entitle us to make the judgement and their objects. My judgement that I see a zebra must be appropriately causally sensitive to the presence of the zebra in front of me. The other is, roughly, that my making that judgement must depend upon no flawed background views. If the judgement is formed through inference, all of its premises must themselves be known; and all the presuppositions of my forming that judgement must be both true and properly introduced. For a wide class of cases, if the subject finds that she has no rational alternative but to make the judgement, and these further conditions are satisfied, then the judgement is an instance of knowledge. Setting aside judgements resulting from inference, we shall consider cases where the subject is “in the mode of taking her senses at face value”: if something looks like a zebra, that is what she will think that it is; if it looks like barn, she will be disposed to think that it is one; if she has a memory of seeing a zebra just before seeing a lion yesterday afternoon, she will be disposed to judge that this is what happened. And we will be concerned with “rationally non-discretionary” judgements: rationality does not permit the agent to withhold judgement in the content while retaining her grasp of the concepts contained in that content. Very roughly, Peacocke holds that such judgements count as knowledge unless one of the two conditions mentioned in the previous paragraph is not met: if it is not knowledge, it should face some possible defeater; if there is a possible defeater that has not been undermined, then the judgement would not be rationally non-discretionary. These claims are intriguing, not least in their refusal to follow much contemporary epistemology in prescinding from questions about the nature of judgement and the identity of concepts when discussing questions about knowledge and justification. I have some comments and questions, more requests for clarification than criticisms of the view. Peacocke’s discussion is abstract and I am particularly uncertain about the range and variety of the presuppositions of judgement that he discusses. First a comment on Peacocke’s rejection of the view that explanation of the concept of knowledge should focus on its role in identifying reliable informants of sources of testimony. It certainly supports his view that most successful Gettier counterexamples to analyses of knowledge introduce considerations that would lead someone to question their entitlement to judge the content in question. There can be no doubt that the post-Gettier literature has identified varieties of epistemic dependence that are directly relevant to correct judgement. Although Peacocke has obviously profited from the post-Gettier 102 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

102

3/15/01, 7:54 PM

literature, it is striking that his book contains very little discussion of knowledge sentences or thoughts about knowledge. The parallel with truth is striking. If we think that the concept of truth is manifested in our practice of assertion or judgement, we may adopt a substantive theory while denying that it is needed to explain the behaviour of the truth predicate. If our aim is simply to explain those thoughts which contain truth as a constituent, a minimalist theory might suit our needs. Similarly an account of judgement may provide a substantive theory of Gettier intuitions while an account of the role in thoughts of the concept of knowledge could lead to a very thin theory that, indeed, focuses on the identification of informants. One example can illustrate this. If I am aware that someone has made errors of calculation that cancel each other out, I may truly claim she knows the answer to the calculation. In this case, what is required for knowledge appears to be the truth of the ascriber’s presuppositions rather than the putative knower’s presuppositions: and it is the ascriber who is authorising acceptance of her answer. I can be correct to ascribe knowledge where the knower would be wrong to claim knowledge. In that case Peacocke’s theses concern the correctness of the subject claiming knowledge, not the correctness or otherwise of some third party ascribing knowledge to her. I am unclear how much of a catch-all Peacock’s conception of presupposition is. Consider someone at the zoo who adopts the mode of taking her senses at face value. On this basis, she correctly identifies the animal in the compound as a zebra. Unfortunately graffiti artists who delight in painting stripes on mules plague the zoo. Suppose, in addition, that our subject lacks the concepts of mule, paint, and, perhaps even, of fake. Once she realises what is happening, she will acquire these concepts pretty quickly and thus withdraw her judgement. As just noted, we, who ascribe knowledge to her, may withdraw the ascription because our presupposition that there were no fake zebras or painted mules present has been defeated but, presumably, this presupposition is not one that she shares. It seems implausible, in this case, that she knows that she sees a zebra, but it is not wholly clear how Peacocke would avoid this conclusion. Perhaps adoption of the mode of taking one’s senses at face value presupposes (falsely) that it is good to do so. But I am unsure how to reconcile that with Peacocke’s adoption of a relaxed notion of ‘rationally non-discretionary’ judgement that is relative to the practice of taking one’s senses at face value. Perhaps our subject simply lacks our rich conception of a zebra and so would be correct to classify painted mules as zebras. I am unclear of the line Peacocke would take.

III As Peacocke insists, this approach to epistemology is ‘externalist’. We take our temporal impressions as entitling us to make judgements of such forms as ‘X occurred around half an hour before Y more than three hours ago’. When philosophers explain how this entitlement can obtain, they will point to the 103 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

103

3/15/01, 7:54 PM

presence of mechanisms, of whose presence the subject may be fully unaware, which ensure that present memory impressions are appropriately causally dependent upon previous experience or activity (pp. 49, 90, 95). Since the concept of a temporal impression is not common property, this explanation may not even be intelligible to the subject. Whether I possess knowledge depends upon conditions which I may be incapable of monitoring: whether the required information links hold; whether my presuppositions are true; and whether they were properly acquired. He suggests that it is a consequence of this that many concepts have ‘externalist’ elements too, and I am unclear what this involves. A paradigm kind of externalism about concepts is offered by Putnam’s claims concerning ‘Twin Earth’: in this case, the identity of my concept appears to be a function of its extension. Some externalist theories of singular thoughts could hold that the identity of an individual concept (for example of London) depends upon the existence and character of a particular city. Does Peacocke’s story yield a form of externalism as strong as these? In what sense is it externalist? I want to approach this indirectly. As he acknowledges, Peacocke’s externalist approach to epistemology has affinities with reliabilism and may be thought to face similar difficulties. Pure reliabilist theories of knowledge have always seemed problematic, largely because the mere (contingent) fact that some method of belief formation is reliable fails to explain why we are entitled to rely upon it in forming judgements (p. 241); we need a way of excluding reliable but non-rational ways of forming beliefs (p. 26). At the very least an element of subjective reliability is required alongside objective reliability (pp. 51–2). In the cases that concern us, a basis is needed for explaining why we are entitled to (for example) take perceptual experience at face value. Peacocke suggests allusively that we must “advert to its membership of a kind some members of which possess correct contents in certain basic circumstances in which the subject of experience is, or was once, situated” (p. 27). Without much more careful explanation of how the explanation proceeds, it is hard to see why the irrational reliable methods could not share this feature. Much depends upon what makes circumstances ‘basic’. The passages in which Peacocke discusses externalist elements in our concept of the past seem to be relevant here. He claims that “It is only if a temporal impression has an externally individuated content—in this case, a content individuated by its relations to the passage of time itself—that it can yield, in appropriate circumstances, non-inferential knowledge with a temporal content” (p. 94). This develops the tentatively expressed claim that “It is hard to see how one could give an account of what it is to have an impression that something happened a certain time ago . . . without mentioning that person’s ability, in suitable circumstances, to track and mark out temporal intervals of roughly that length” (p. 94). So our possession of the ability to identify temporal intervals is required, in slightly different ways, both to explain how memory impressions are possible at all, and how we can possess non-inferential knowledge of temporal facts. But this is not something of which the knower need be aware. 104 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

104

3/15/01, 7:54 PM

The question I have here seeks further clarification of just what an account of possession conditions requires here. Related to these claims are some counterfactuals: if our judgements were not causally sensitive to the passage of time, or if we were not good at tracking and marking temporal intervals, we would, most likely, not possess temporal concepts, not obtain immediate knowledge of the past, or not possess memory impressions. But a non-externalist could, presumably, accept those counterfactuals, in which case Peacocke’s constitutive claims must surely go further. Or does the focus on knowledge ensure that any such counterfactuals are constitutive of concept possession?

105 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

105

3/15/01, 7:54 PM

THE PAST, NECESSITY, EXTERNALISM AND ENTITLEMENT   The issues raised by the three commentators on Being Known range so widely, and some of them are so fundamental, that it would take another book to address them all. So in writing this response, I have had to be highly selective. I have tried to select those issues on which clarification, further defence, and substantive further development of the positions in Being Known are likely to be of greatest interest. What follows is grouped by topic, rather than commentator, and the topics I have chosen are: the past and the property-identity link; necessity; and the relations between externalism and entitlement relations. I thank James Cargile, Adrian Heathcote and Christopher Hookway for asking so many pertinent questions.

The Past and the Property-Identity Link The property-identity link is the natural generalisation of this principle: A thought (utterance) ‘Yesterday it rained’ is true if and only if yesterday had the same property as today is required to have for a present-tense thought (utterance) ‘It is now raining’ to be true when evaluated with respect to today. James Cargile holds that the property-identity link is committed to some form of ‘tenseless’ treatment of the metaphysics of time. Here we distinguish such a metaphysics of time itself, which we are considering, from the epistemology and philosophy of mind for temporal attitudes, which we are not now considering. In attributing implicit knowledge of the property-identity link to those who understand the past tense, am I not, Cargile wonders, attributing implicit knowledge whose content is incompatible with the explicit denials of those who reject any kind of tenseless metaphysics? And is there not a danger that the attribution of implicit knowledge will trivialise the Integration Challenge? Surely meeting the Challenge must, in any given domain, involve more than simply “saying that people, by virtue of functioning in the alleged state of things, thus implicitly know enough about it to so function” (Cargile, supra, p. 86). I reply that the core content of the property-identity link is consistent with a metaphysics of time that rejects tenselessness. I take Cargile’s point that the formulations in Being Known are slanted in the direction of a tenseless metaphysic. But we can correct the slant. Cargile mentions the view that while there was such a day as yesterday, there is no longer in any legitimate sense 106 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

106

3/15/01, 7:54 PM

such a day. If we hold that view, we can still consistently accept the propertyidentity link in this form: The past-tense claim ‘It rained yesterday’ is true if and only if it there was a day, yesterday, and it then had the property which today has to have for ‘It is now raining’ to be true when evaluated with respect to today. This carries no implication that the day that existed yesterday in any sense still exists, even tenselessly, today. We could give a parallel future-tense version too. The believer in a tensed metaphysics of time, in accepting this fully past-tensed formulation, can thereby gain all the anti-verificationist, anticonstructivist consequences of the property-identity link. Because this seems to be a consistent position, the treatment of time in Being Known need not involve the attribution, to believers in a tensed metaphysic, of implicit knowledge of a content that they explicitly deny. Not that this would necessarily be an objectionable state of affairs: of two linguists disputing over the correctness of a proposed grammatical rule, at least one must be explicitly denying the content of a grammatical rule of which, arguably, he has tacit knowledge. The same applies to someone who explicitly offers an incorrect definition of ‘chair’, even though he has tacit knowledge of the correct definition, a knowledge that influences his classification of examples. As these last cases remind us, the attribution of implicit knowledge is answerable to what it can explain. The attribution of implicit knowledge in no way trivialises the task of addressing the Integration Challenge. Whatever implicit knowledge of a specific content is attributed in a particular proposal that attempts to meet the Integration Challenge in a given domain, the proposal needs substantiation by the citing of data that are best explained by implicit knowledge of precisely that content. In the case of thought about the past, I tried to motivate the attribution of implicit knowledge of the propertyidentity link by appeal to such matters as: our knowledge of the fact that a past-tense truth explains counterfactuals about the results of investigation, rather than conversely; our appreciation that strictly speaking those counterfactuals are not sufficient for the truth of the relevant past-tense statements (because of the possibility of interference); and so forth. I also emphasised that since the content of the implicit knowledge attributed itself used the past tense, the thinker must have some other fix on the past than that given in the implicit knowledge. That was supplied by the externalist treatment of temporal thought given later in the chapter on the past in Being Known. Without that additional element, the attribution of implicit knowledge of the propertyidentity link would not be by itself an adequate response to the Integration Challenge in the case of the past. Necessity I divide this discussion of necessity into two parts: into some (intended) clarification of my position, followed by some discussion of the fundamental issues 107 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

107

3/15/01, 7:54 PM

that Adrian Heathcote rightly raises about the metaphysical and epistemic status of the Principles of Possibility. (a) The Nature of the Target Notion, and Clarifications There is a prior question about the aim of the whole enterprise of addressing the Integration Challenge in the modal case. Are we really dealing here with a notion that has a life in our ordinary non-philosophical thought about the world? Cargile doubts it. In the case of modality, unlike that of the past, he writes, “there is no realm of common knowledge to provide a test for a theory about knowability of truths in the domain” (supra, p. 86). A little later he adds, “One who finds no division into necessary and contingent truths will have to contend only with the more esoteric common sense of some philosophers” (ibid.). I disagree. The metaphysical necessities are those that are to be held constant in the scope of any counterfactual speculation. This is why, as David Lewis notes, ‘necessarily p’ can be defined in terms of counterfactuals, as ‘If not-p were the case, then A & not-A would be the case’, for any arbitrarily chosen A.1 Anyone who engages in counterfactual reasoning and judgement in a rational and plausible fashion is thereby committed to the existence of a body of necessary truths. Counterfactual thought is also part of the apparatus of everyday thought about the world, and is not the province only of esoteric philosophy. The very notion of objectivity in its most basic application, to the perceptible world, makes essential use of counterfactuals. It involves knowledge of such truths as ‘If I were elsewhere and even if I were not perceiving it, that tree over there would still exist’. What is held constant in counterfactual speculation is not restricted to the nomologically true, either. A scientist who is trying to assess whether or not L is a law may argue ‘If L were a law, then p would be true; but not-p; hence L is not a law’. This kind of reasoning goes beyond the commonplace: but it is evidently a part of the empirical sciences, and not an outgrowth only of philosophy. All these points apply even before we mention practical reasoning, in which counterfactuals are equally indispensable. To abandon everything which has a commitment to a distinction between the necessary and the contingent would be to abandon both theoretical and practical reason. I now turn to two clarifications. (i) Heathcote is surely right to say that some linguistically atomic expressions, including for instance ‘credible’, ‘flexible’ and ‘dangerous’, have a modal component in their meaning. In such cases, given that our task is to explicate the meaning of modality, we cannot simply take for granted the rule that fixes their extension in the actual world, without presupposing some of what we were setting out to explain. For these cases, the principle-based approach would have to offer a two-stage treatment. At the first stage, one deals with 1. D. Lewis, ‘Counterfactuals and Comparative Possibility’, rp. in his Philosophical Papers Vol. II (Oxford University Press, 1986), p. 11.

108 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

108

3/15/01, 7:54 PM

concepts and expressions that do not have a specific modal component. One employs the Modal Extension Principle and the other Principles of Possibility to determine the truth-value of modal statements involving concepts treated at this first stage. At the second stage, one can then go on to give the rules for determining the extension of ‘credible’, ‘flexible’ and the like by drawing upon the rules for determining the extension of modal statements already determined at the first stage. This seems to me to correspond to a natural intuitive division in our ordinary understanding of such expressions: to know what it is for something to be flexible, one must first know what it is for it to flex. (ii) Cargile is worried about the use of the notion ‘p is true with respect to world w’ when a world is regarded as a set of propositions, Thoughts or sentences. He says that to avoid a problem of regress, “we might say that ‘p is true in w’ is either necessarily true or necessarily false” (supra. p. 86). I certainly am committed to saying what he says we might say. The possible worlds, in my treatment, are the genuinely possible specifications. Specifications are simply sets of propositions or Fregean Thoughts: and any given proposition or Thought is either an element of a given set, in which case it is necessarily so; or else it is not, and is necessarily not (like any other object that is not an element of a given set). Cargile writes that this has a disturbing application to such an example as ‘It is true in this world that Norfolk is the largest city in Virginia’. To call that necessarily true conflicts with the fact that it would ordinarily be taken as just a way of saying that Norfolk is the largest city in Virginia, which is a typical standard case of something nonnecessary (ibid). There is no problem. The proposition ‘Norfolk is the largest city in Virginia’ is necessarily an element of any set of which it is a member. If X is the set of all true propositions (formulable from a given set of atomic constituents), then ‘Norfolk is the largest city in Virginia’ is an element of X. It is also necessarily an element of X. But it is not necessary that X is the set of true propositions. Some set other than X might have been the set of true propositions. Some of the sets other than X that might have been the set of true propositions contain the proposition ‘Norfolk is not the largest city in Virginia’. Hence it is not necessary that the proposition ‘Norfolk is the largest city in Virginia’ is an element of the set of true propositions. That proposition about the size of Norfolk retains its contingency, consistently with the membership relation between elements and sets being necessary in any case in which it holds. (b) The Modal Status of the Principles of Possibility On the deeper issue of the status of the Principles of Possibility under my approach, I take first the question of their modal status. I hold that the Principles of Possibility are necessary. I also hold that their necessity can be shown to follow from their truth, in the presence of certain other truths. This is the position for which I argued on pp. 151–3 of Being Known. Here, by way 109 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

109

3/15/01, 7:54 PM

of illustration, is an argument for the necessity of one of the Principles of Possibility, the Modal Extension Principle (I take it in its ‘Unified’ form). The Unified Principle (‘the UMEP’) states that: An assignment s is admissible only if: for any concept C, the semantic value of C according to s is the result of applying the same rule as is applied in the determination of the actual semantic value of C. According to the principle-based treatment, the UMEP is necessary iff it holds under all admissible assignments. Therefore, to determine whether the UMEP is necessary, we have to ask: is there an admissible assignment to the concept admissible that violates the displayed UMEP? I argue that there is not. For we can apply the UMEP to the concept admissible itself. To do so, we have to ask after the rule that determines the extension of admissible in the actual world. The rule that determines the extension of admissible in the actual world is given by the set of principles formulated for admissibility in the principlebased conception. These are the rules that make the concept of admissibility the concept it is. (Hence I dissent from Heathcote’s remark that “we don’t even have a reason to believe that admissible is semantically fixed” (supra, p. 95). These rules include the UMEP itself. I take it also that, given the way the semantic value of the concept assignment is fixed in the actual world, any admissible assignment to that concept will assign it an assignment (and not, say, a cat or dog); and pari passu for the other concepts in the UMEP. Hence any admissible assignment to admissible will be one that makes the UMEP true. So the UMEP is true under all admissible assignments; that is, it is necessary. Heathcote has a further objection to this style of argument, to the effect that if concepts and the functions mapping them to their referents “are meant to be non-abstract elements of the actual world . . . then there is no reason to think that they are necessary” (ibid). In fact it was the rule (not the mapping to referents) that was meant to be necessary, but no matter—let us still consider whether there is equally an objection from concepts as non-abstract elements to the necessity of the rule for determining the concept’s semantic value. ‘Concept’ remains very much a term of art, and I agree that there are important purposes for which we need to consider concepts as mental particulars. These particulars will, however, certainly need to be grouped into types, under the equivalence relation of being uses of the same concept (in the abstract sense of ‘concept’ I used in A Study of Concepts and continue to employ).2 We can ask: what is necessary for two mental particulars to be uses of the same concept? It seems to me that (prescinding from indexicality, for which we need a more complex formulation) identity of the rule for fixing their semantic values is a necessary condition of their being uses of the same concept in the type sense. Moreover, it seems to me that this rule contributes to—perhaps even exhausts—what makes the mental particular a use of a given type concept. If all this is so, treating concepts as particulars in no way 2. A Study of Concepts (MIT Press, 1992).

110 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

110

3/15/01, 7:54 PM

damages the above style of argument for the necessity of the Principles of Possibility. The argument still goes through at the level of the type. Actually, it goes through at the level of the mental particulars too, if they are essentially of the conceptual kind that they actually are, as is quite plausible. I am inclined to hold that the self-application solution to the question of the modal status of the Principles of Possibility is the only philosophically acceptable answer open to a principle-based conception. This is so for general structural reasons that apply to any theory. Suppose a theory T states that for a given proposition to have some philosophically interesting property F is for it to follow from a set of principles S. Suppose also that when a proposition p has F, the proposition that p has F is itself also F. Then the theory T can be correct and general only if the proposition containing that double mention of F itself follows from the set of principles S. To show that that requirement is met in the case of a theory of modality must involve showing that the elements of S are necessary. In the different case of a theory of the a priori, it must correspondingly involve showing that any postulated principles that are the proposed source of a priori truth are themselves a priori true. This is a structural requirement that must be met by any general theory of this abstract form, whatever the target notion under consideration. I hold that the structural requirement can be met, and a principle-based account of modality can thereby be fully general, without there existing any ground of necessity deeper than the Principles of Possibility themselves. The Principles of Possibility are themselves already at ground level. (c) The Epistemic Status of the Principles of Possibility On the epistemic status of the Principles of Possibility, Heathcote rightly describes my view as a form of Rationalism. I do hold that the Principles of Possibility are a priori. Heathcote attributes to me also the view that the Principles are innate. This attribution is not justified by anything in Being Known. I do hold that someone who understands an operator for metaphysical necessity has tacit knowledge of the propositions stated in the Principles of Possibility. Tacit knowledge can, however, be acquired—it need not be innate. As it happens, this misattribution does not matter to Heathcote’s argument, for, if sound, the reasoning he offers would apply against any conception of a priori knowledge, whether innate or acquired. He writes, “If there is nothing in experience that acts to make it probable that the belief is true then we cannot count the belief as knowledge” (supra, p. 97). This reasoning seems to me to overlook the possibility that rationalists have always defended, viz. that there are ways of coming to know certain propositions which give an entitlement whose status as such is independent of the content of the content of experience. Experience may be necessary for access to such entitlements, but the entitlement itself is experience-independent. Proofs in logic and mathematics plausibly have this status. Of course this classical rationalist conception must be earned by further argument, a task for other places; here I just want to make clear that I do stand in this rationalist area of logical space. 111 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

111

3/15/01, 7:54 PM

Even if it is allowed that the rationalist conception has some genuine application to some of our knowledge, we still have to answer Heathcote’s question of why and how the Principles of Possibility are included in our a priori knowledge. In outline, I would argue that we can come to know the Principles of Possibility as follows. First, each ordinary modal proposition that we know seems to me to rest ultimately on a priori modal propositions, for the reasons elaborated in Being Known. We can always split up the grounds for any a posteriori modal proposition we know into grounds each of which is either modal and a priori, or a posteriori and non-modal. (What is excluded is modal a posteriori knowledge the grounds for which cannot be split up in this fashion.) If this is correct, the epistemically fundamental modal propositions are known a priori. At the second step, we go on to ask ‘What is the best explanation of the meaning of necessity that would accord with the truth of these modal propositions that are known a priori?’. I contend that the best explanation is that necessity conforms to the Principles of Possibility. This is an a priori abduction, from a priori data about the truth of certain modal propositions, to a conclusion about the best a priori explanation of why they are true. Such a priori abduction is familiar from other cases in mathematics and the abstract sciences, in which we use a priori abduction to discover new axioms, by considering a range of a priori cases that a new axiom would explain. If it is done properly, a priori abduction from a priori cases can yield a priori entitlement and a priori knowledge. This two-step process is a way in which we, as theorists, can come to know the Principles of Possibility without essentially relying for entitlement on the content of experience. The ordinary understander of modality does not, of course, have to go through such a reflective process in order to understand modality. The ordinary understander’s tacit knowledge may be acquired in the same way in which any other tacit knowledge that influences judgements, such as tacit knowledge of the definition of ‘chair’, or of the recursion for addition, may be acquired. Immersion of the learner in sufficiently many examples can generate an underlying state whose content explains the thinker’s classification of new examples. When this underlying state has the content of the Principles of Possibility, and has been acquired in ways that rule out other ‘nearby’ hypotheses about what ‘necessarily’ means, it will amount to tacit knowledge of the Principles of Possibility. If I understand Heathcote’s position correctly, he holds that there can be fundamentally a posteriori knowledge of necessities. If there can be, there ought to exist (or it ought at least be possible for there to exist) an posteriori necessity whose grounds cannot be split up into propositions that are either modal and a priori, or a posteriori and non-modal. I do not know of any such examples. Yet despite this fundamental divergence between us, I should note that my variety of rationalism about necessity does agree with other parts of Heathcote’s position. Modal knowledge is, on my view and in his phrase, ‘part and parcel’ of our ordinary fallible knowledge of the world. It is part and parcel thereof, precisely because of the inextricable role of necessity in counterfactual reasoning that I emphasised above. Sensible rationalists have also long emphasised the fallibility of a priori methods of coming to know. 112 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

112

3/15/01, 7:54 PM

Externalism and Entitlement Hookway raises some important questions about externalism and entitlement, on which I am glad to have the opportunity to say more. Before doing so, however, I want to set aside one construal of what I was attempting. Hookway wonders whether the epistemological theses in Being Known should really be taken as theses that concern whether a subject is correct to claim knowledge of a given proposition, rather than “the correctness or otherwise of some third party ascribing knowledge to her”. Hookway illustrates his intended distinction by saying that “If I am aware that someone has made errors of calculation that cancel each other out, I may truly claim she knows the answer to the calculation” (supra, p. 103). I disagree, and think Hookway is misclassifying this example. An error in a calculation means that the subject is relying on an intermediate false belief, and this prevents the final conclusion of the calculation from being known (at least by means of this calculation) to the person who is calculating. This is an instance of Harman’s ‘no false lemmas’ requirement, which is independently confirmed in many other examples.3 My theses about knowledge in Being Known are meant to be contributions to theory of the general conditions under which something of the form ‘x knows that p’ are true, regardless of how the subject of the attribution of knowledge is picked out. Certainly it may be rational for one person, in his informational state and circumstances, to claim ‘a knows that p’ but not rational for another person, in a different informational state or circumstances, to claim ‘b knows that p’, even when a = b. But the epistemological theses of Being Known are about the truth of knowledge-ascriptions, not about their rationality (though they may have consequences for rationality). If a = b, it is not possible for ‘a knows that p’ and ‘b knows that p’ to differ in truth-value. Hookway is also worried that some descriptions of the circumstances that show that a thinker’s presuppositions in making a judgement are unfulfilled may be descriptions that employ concepts the thinker may not have in advance of finding himself in this situation. I agree: besides the example he gives, we could also mention cases in which holograms are in the subject’s environment, a kind of object that may come as a complete surprise to the perceiver. I think all this shows is that the presuppositions involved in relying on an informational state should be cast in more general terms. The rational thinker who relies on an informational state in forming a rationally nondiscretionary judgement should withdraw that judgement when shown that the same method he is using could easily have led to false beliefs in his actual circumstances. This is consistent with a fuller, more detailed characterisation of how it could have misled him needing to employ concepts new to the thinker.

3. G. Harman, Thought (Princeton University Press, 1973), Ch. 3, Section 6, pp. 46–50. Note that it is not necessary to take perceptual knowledge as inferential in order to accept the correctness of the ‘no false lemmas’ requirement for knowledge. Harman’s principle seems to me correct, even though in my judgement its range of application is narrower than he holds.

113 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

113

3/15/01, 7:54 PM

Hookway’s most important questions are: (1) In what sense is the treatment of temporal thought in Being Known externalist? (2) How is this externalism (whatever it may be) consistent with my rejection, in Being Known, of purely reliabilist treatments of knowledge? I will do my best in the very limited space remaining. In outline, my position is this: (a) There is an explicable sense in the theory of intentional content in which the treatment of temporal concepts in Being Known is externalist. (b) There is a general principle, which bridges epistemology and the philosophy of mind, and which connects states with contents that are externally individuated with the conditions under which a thinker is entitled to make a certain kind of judgement. (c) A proper elaboration of this principle and its sources distinguishes such entitlements from anything purely reliabilistic. I take these claims in turn. (a) What makes it the case that a thinker’s impressions are impressions of temporal intervals and orderings are the complex relations in which they stand to temporal intervals and orderings themselves in the basic case in which we have a subject embedded in the world in the normal way that we take for granted in propositional-attitude explanation and ascription. The temporal orderings and relations are at the level of reference, not sense. The complex relations in question include the explanation of significant examples of the impressions by the temporal facts of which they are impressions. They also include the capacity of temporal impressions to explain (in the context of other suitable attitudes) the temporal properties and relations of the actions, either bodily or mental, of the subject who has the impressions.4 This answer seems to me largely parallel to the externalist answer one would give to the question of what makes an impression have a spatial content. As always, it is important to distinguish constitutive from modal claims. In both the temporal and the spatial case, it seems to me that the externalism should be formulated in constitutive rather than modal terms. In the famous Putnam example, not only is it true that the identity of the concept water depends on what is in the extension of the concept. It is also possible that a thinker stand in exactly the same relations to a different liquid, twin-earth water. In the case of spatial impressions, there are obvious problems in supposing the thinker could be in exactly the same internally-individuated states and stand in the same relations to different spatial magnitudes. If the thinker is in the same internally-individuated states in responding to different spatial magnitudes, won’t his actions be inappropriate in some circumstances for the 4. On explanation of relational properties of actions, see my ‘Externalist Explanation’, Proceedings of the Aristotelian Society, 93 (1993), pp. 203–30.

114 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

114

3/15/01, 7:54 PM

realisation of his intentions? And won’t he discover this, so that continuing identity of internal states will not be preserved across cases? If examples can be devised to get around this complication, all well and good: but it is not necessary that there be such modal divergence for the constitutive thesis to hold. There can be constitutive dependence on external intervals and orderings, in both the temporal and spatial cases, even if it is not possible for thinkers to be in exactly the same continuing internal states, and yet the contents of their impressions differ. In fact the kinds of considerations to which we have just alluded explain why there may be no modal divergence, even while the claim of constitutive dependence still holds good.5 Of course some temporal impressions, even in the basic case of a subject embedded in the world in the normal way, misrepresent, and misrepresent reliably. Just as we have the Müller-Lyer illusion in the spatial case, we have (amongst others) the reliable illusion that the first occasion on which one looks at a digital clock that is changing second-by-second, the changing digit that is seen first seems to last longer than its successors. The variety of externalism I have been outlining about spatial and temporal impressions will hold that these illusory cases have their content in virtue in part of their relations to cases that are not illusory. It is a familiar point that best interpretation may require the attribution of some incorrect contents, against a background of generally correct contents in the basic case of normal embedding. (b) We can formulate a Second Linking Thesis, in addition to the original Linking Thesis in Being Known. This Second Thesis links the theory of content for informational states with the conditions for entitlement or non-inferential knowledge. Here is the version for non-inferential knowledge: A conscious informational state can result in knowledge that p, when a thinker takes the state’s content at face value, only if the state’s having that content is constitutively dependent on its complex relations to the kinds of objects, properties and events mentioned in the intentional content p. The version for entitlement results from replacing ‘can result in knowledge that p’ by ‘entitles a thinker to judge that p’. The two versions are closely related: presumably a state can result in knowledge that p, when the state is taken at face value, only if being in that state entitles the thinker to judge that p. Being Known formulates the special case of this Second Linking Thesis for (i) the case of impressions of temporal relations and knowledge about temporal relations (p. 94). It also mentions (ii) the fact that merely sensational (nonrepresentational) properties of perceptions, if such there be, do not supply states that give reasons for non-inferential judgements about perceived objects and their properties (p. 95). Here are some further plausible illustrations of the general Second Linking Thesis. (iii) Berkeley suggested that the perception 5. On the distinction between constitutive and modal dependence, see ‘Externalist Explanation’, pp. 226–7, and M. Davies, ‘Individualism and Perceptual Content’, Mind, 100 (1991), pp. 461–84, esp. p. 463.

115 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

115

3/15/01, 7:54 PM

of depth depends in some cases on the more bluish shades of things that are more distant from the perceiver. Such a content of experience is not individuated by reference to depth itself. It also does not supply by itself a rational, non-inferential basis for judgements about depth. (iv) A perceptual impression of some other person as amused makes rational a judgement that she is amused. If her face were perceived merely as having certain spatial features and relations, not described in terms of an expressed emotion, there would be no rational, non-inferential entitlement to the judgement, nor to the possible knowledge, that she is amused. (v) We hear and see other utterances and inscriptions as having certain senses. The experience of hearing someone’s utterance as saying that Empedocles leaped is different from that of hearing him say (in German) that Empedokles liebt, even though the experience of the sounds uttered, non-semantically characterised, may be identical. If one hears an utterance as having only non-semantic properties, any transition to a judgement about what it means could only be inferential. I put forward the Second Linking Thesis as an explanatory hypothesis about a link between external individuation on the one hand, and the domain of epistemic notions including entitlement and knowledge on the other. The Second Linking Thesis goes far beyond the truism that an informational state can give an entitlement to non-inferential judgement that p only if it represents it as being the case that p. The Second Linking Thesis is offered as part of an explanation of how there can so much as be informational states that entitle a thinker to non-inferential judgements. I take it as plausible that in the examples (i)–(v), the contents of the informational states are externally individuated, in the sense I have outlined. They may do so in interestingly different ways (notably in the case of the emotions), but I contend that in none of these examples could one explicate the possession of the full intentional content of the reason-giving informational state that provides noninferential knowledge without adverting to the state’s relations respectively to temporal relations themselves, spatial relations themselves, depth itself, the emotions themselves (in this case maybe the subject’s own), and semantic properties themselves. What then is the explanation of why these states lead to an entitlement to make the relevant judgements? When a thinker is in the mode of taking the content of a given kind of informational state at face value, he is not questioning that he is in the kind of basic case which supplies the circumstances with respect to which the intentional content of those states is individuated. (This presumption, which is not any kind of premise in the thinker’s reasoning, can of course be defeated in the presence of reasons for doubting it.) In those circumstances, these intentional contents do bear those complex relations to what they represent that individuate the intentional contents. In these circumstances, they are generally correct (though any particular content may be incorrect: the entitlement relation is not conclusive). Hence there is an entitlement to take the informational state at face value. (c) This argument does not extend to anything a posteriori reliably correlated with the content of the informational state, however high that degree of reliability may be. It may be wholly reliable that if one event is further 116 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

116

3/15/01, 7:54 PM

from the subject than a second event, then light from the first took longer to reach the subject from the first than it did from the second. But this relational property about light is not in itself the feature of the basic case on which the individuation of spatial impressions has a constitutive dependence. What gives non-inferential reason for what depends in part upon what individuates the intentional content of the reason. Externalism in the theory of intentional content, far from leading to the conclusion that reasons are merely reliable indicators, is, when properly developed, incompatible with such a primitive externalism in epistemology.

117 © Blackwell Publishers Ltd. 2001

PHIB42.2C01

117

3/15/01, 7:54 PM