Abstract: I begin with the consideration of an argument which Moritz Schlick once gave in support of a verificationist theory of meaning. My interest in this argument has two main sources. Firstly, one finds in the literature of Logical Positivism no great wealth of argument in favour of verificationism, for in the heyday of the Vienna Circle it seems to have been felt sufficiently persuasive just to postulate the Verification Principle and display its anti-metaphysical consequences; in view of this, and the power of the Principle once accepted, any argument offering to demonstrate it from philosophically neutral premisses is a precious object. Secondly, we shall see reason to believe that a form of verificationism, differing in some respects from the original but resembling it in the affirmation of the complete monopoly in the theory of meaning of what is in principle knowabley is currently important in philosophical thought, and that it is sometimes supported by arguments bearing at least a superficial resemblance to that adduced by Schlick. Even if the resemblance should turn out to be no more than superficial, thinking about Schlick's argument should help us to get these more recent efforts into sharper focus. The essential paragraph is brief enough to be quoted in full:

Abstract: entities to serve as the meanings of linguistic expressions and then can say nothing of the nature of linguistic understanding except that it involves our 'grasping' those entities. But, as This content downloaded from 207.46.13.164 on Mon, 27 Jun 2016 05:15:40 UTC All use subject to http://about.jstor.org/terms

Abstract: Moliere's joke about the sleeping potion's 'dormative virtue' is standard fare in the philosophy of mind. One cannot, so it is claimed, explain someone's seeing, or remembering, or intending something by postulating a little man in the head who looks at an image, retrieves 'a memory trace, or forms a resolve. Taking this lesson to heart, writers like Attneave [I960], Fodor [I968], and Dennett [1978] have argued that explanations of a mental capacity can only avoid the emptiness of Moliere's dormative virtue by decomposing the capacity into a set of components which are more rudimentary. Postulating little men in the head is permissible, as long as the little men do not have the same full-blown abilities as the people they inhabit. Homunculi are all right, if they are stupid. Rorty ([1979], p. 236) has observed that the pattern which allegedly vitiates smart homunculus explanations can be found in other sciences, but without such disastrous results. An explanation of why the bodies of the solar system revolve as they do may with impugnity invoke the idea that the bodies are composed of atoms whose components exhibit the same pattern of motion. Similarly, the digestive capacities of an organism may be explained perfectly well by the fact that it is the host of a parasite which itself has those very abilities. If smart homunculi confer the kiss of death in the context of psychological explanation, then perhaps we have at last uncovered what is unique about the mind. Whatever might be said of planets and parasites, and whatever promise there may be in Brentano's idea of intentionality, it would appear that stupidity is the mark of the mental.2 I suspect, however, that no such distinctive feature is at hand. What goes for physics and biology also applies to psychology. What, then, is wrong with smart homunculi? Why must homunculi be stupider than we are, if 'Chinese box explanations', as we might call them, are permitted in other domains? Sometimes proferred explanations are empty because they rather obviously could not be true. The supposed problem with the evolutionary

Abstract: Principle (A) is a cousin of the one used by Edmund Gettier to construct his well-known counterexamples. It has received its share of abuse, but no one seems prepared to abandon it.I Indeed, its poorer (I would say bankrupt) relatives (e.g., S knows everything implied by what he knowswhether or not he knows it is implied) have sometimes been taken as axiomatic in systems of epistemic logic. The second principle, though less precise, and more difficult to state in any crisp way, is taken for granted by all of us who believe that knowledge is transmissible. It is hard to see how something like this principle could fail to be true if communication is possible and education occurs. For the way we know much of what we know is by hearing it from others who know. The following example convinces me that both principles are false.2 This is not to say that they do not hold in special circumstances, only that to regain their status as principles their scope must be restricted. George loves Bordeaux wines and he is especially fond of those from the Medoc region of Bordeaux. He has quite a remarkable palate and unerringly identifies a genuine Medoc as a Bordeaux, and specifically as a Medoc, when he tastes one. Strangely enough, though, given his general knowledgeability about wines, George is confused about Chianti. He has

Abstract: can readily be converted into a statement of someone's reason for acting: 'He took the medicine because he wanted to be healthy and thought that if he took it he would be healthy.' Aristotle used the notion of a 'practical syllogism' to make this connection. In De Motu chapter 7 he used the practical syllogism to explain reasons for acting and in Nicomachean Ethics, Books 6 and 7 he used it in discussing practical reasoning. (I am aware that in saying this I am using the term 'practical syllogism' in a wider sense than is favoured by some writers, for instance Hardie and Cooper. I shall return to this issue at the conclusion.) Sometimes there is practical reasoning which does not lead to action, as in weakness of will (EN Bk. 7). In this case the practical reasoning does not supply the reasons for the action performed. Conversely, there can be a reason for acting without any reasoning (MA ch. 7, 70Ia3Iff.). In order to understand Aristotle's account we need to look at what is common to both practical reasoning and reasons for acting and at what is peculiar to each.

Abstract: In his Dialogues concerning Natural Religion Hume has Cleanthes argue that there is nothing amiss with a series of causes being infinite since in such a series 'each part is caused by that which preceded it, and causes that which succeeds it'. 'Where then', he asks, 'is the difficulty?' To the suggestion that the difficulty lies not with any part of the series but with the series as a whole, Cleanthes replies that 'did I show the particular causes of each individual in a collection of twenty particles of matter, I should think it very unreasonable should you afterwards ask me what was the cause of the whole twenty'.I The same kind of argument is offered by Paul Edwards2 who, in considering the example of a pile of books stacked one on top of the other, has pointed out that even if the pile were infinite there would be no book that was not supported by another. But if no book were unsupported, there would be no danger of the pile collapsing, and hence no need of any 'unsupported supporter' to underpin the pile as a whole. Contrary to what proponents of the contingency argument might say, therefore, it is only a finite pile that would need support in order to avert a collapse; an infinite pile would require no support at all. The points made by Hume and Edwards, and later endorsed by Antony Flew,3 have also been underlined by Ronald Hepburn:

Abstract: One anti-abortion line of argument proceeds like this: (i) We should do to others as we are glad they have done to us; (2) I am glad that no one killed me when I was a fetus, sO (3) I should not kill those who are fetuses.' The argument involves the metaphysical assumption that I once existed as a fetus. Is that so? The answer to the question 'When did I begin to exist?' somehow seems to depend upon the answer to the question 'What am I?' But if this lastquestion is parsed as 'What kind of thing am I ?', we must proceed with caution. One possible answer to the question 'What kind of thing am I?' is 'chess player'. But 'chess player' is decidedly not a term that applies to me throughout my entire history or career; the time at which I became a chess player is considerably later than the time at which I began to exist. It may be said that any proper answer to the question 'What kind of thing am I ?' must be posed in terms of a 'sortal' expression that applies to me at every moment at which I exist.2 So-called 'phasesortals', terms that apply to things throughout only proper parts of their histories or careers, do not tell us what kind of thing a thing really is. To determine what kind of thing I am we need to be given a 'substance' sortal or an 'ultimate' sortal-a term that applies to me throughout my entire history or career. Quite obviously 'chess player' is out. But what is in? One popular candidate is 'person'. Assuming that those of us who are people are people throughQut our entire histories or careers, and assuming further that no fetus is a person, none of us (people) ever existed as fetuses. It is often assumed that a necessary condition for being a person is that a being have certain psychological capacities or abilities.3 (The

Abstract: In this paper I try to establish a conditional: if consciousness has a function then that function is not performed by any brain mechanism. As the reader will often find me making the assumption that consciousness has a function, I emphasize now that this is simply for conditional proof. In the first section I derive a principle of functional explanation which is the mainstay of my argument. This principle is a consequence of a familiar account of evolutionary function which I shall have to present. But, for the sake of brevity, I will simply state what is needed here, give an illustrative example and make comparative reference to the literature. In the second section I derive my main conclusion, in the final section I do so again but in the context of the Functionalist philosophy of mind.

Abstract: In Part I of the Parmenides Plato sets his talent to work against his own middle-period doctrine of transcendent Forms. The attack appears to be in earnest, and seemingly it is effective. Especially impressive is the proof that the theory of Forms yields an infinite regress: this is the Third Man argument, and Plato deploys it here in two different versions as if grimly to enforce recognition of an inescapable paradox. Certainly he himself never proclaims a solution. Consequently it is puzzling to find him later (as it seems) maintaining without any hint of apology a metaphysical position with precisely the objectionable feature that fell prey to the Third Man argument. This was the conception of a Form as the transcendent manifestation of a characteristic common to some set of sensible particulars,1 whose mutual resemblance derives from their resemblance to the Form. From this the Third Man brings to light, by a step which once taken must then be repeated, a further transcendent manifestation to account for the resemblance of the Form itself to the sensibles. But in the Timaeus, still by many believed to be later than the Parmenides, Plato speaks with calm assurance of physical things, and indeed of the entire physical universe, as likenesses made to resemble timeless divPne exemplars. Perhaps we should see this as a remarkable instance of the power of faith to remain unswayed by rational criticism: remarkable because in this case the faith itself is supposedly required by reason, and because it is also the believer himself who has gone on record delivering the rational onslaught. The candour of that self-exposing elenchus is not more amazing than the serenity with which its victim emerges as if untouched-a strange development indeed for Socrates' greatest-ever disciple.

Abstract: In Remarks on the Foundations of Mathematics, and elsewhere, Wittgenstein made some general statements about mathematical proof which can only be described as paradoxical. What he says is directed to proofs for which there is no general method of solution such as there is for statements of the form a x b = c where a simple calculation is sufficient. His concern is with propositions for which one cannot in advance of proof describe any procedure which will demonstrate them. Examples are: 'For all natural numbers n, 22' + I is prime', 'There is no greatest prime', 'Every equation of nth degree has exactly n roots', 'Any even number is the sum of two primes'. In an earlier study,' 'Proof and the Theorem Proved', I singled out a group of propositions for which his account of proof seemed correct, and not paradoxical. Examples given were the propositions a a =I', 'represents any number whatever', whose proofs provide justifi0

Abstract: i. When William addresses an English sentence to John, there are capacities and tendencies which John has which are not shared by monoglot Pierre, viz. the capacity to use the utterance as the basis for a systematic inference to some correlated psychological attitude presumably possessed by speaker William, and a tendency to modify his own psychological set in a way appropriate to the recognition thus achieved of William's attitude to the world. And not only does John as hearer have these attitudes and tendencies, but (in the primary and central case) both he and utterer William recognize that he does, and that William depends crucially upon them in hoping that his utterance will achieve its communicative goal. Systematically associated with an utterance of a sentence of a language, then, is a cluster of propositional attitudes-attitudes possessed by the utterer and inferred to by the hearer on the basis of the utterance, and attitudes adopted by the hearer in response to the utterance. Theorists vary in their accounts of just which attitudes are involved, and more particularly of just which should be singled out as crucial, inasmuch as they are the ones whose known association with the utterance is essentially exploited by speaker and hearer in achieving their conversational aim (thus, on such questions as to whether the only critically associated attitudes are ones possessed by the speaker and inferred to by the hearer, or whether attitudes adopted by the hearer too need to be reckoned as crucial). Further disagreement centres on the issue of the correct description of those community-wide psychological facts which presumably underlie the possibility of general reliance upon speaker/hearer association of utterances with crucial attitudes. But different accounts have a good deal in common. Everyone agrees that, whatever the crucial attitudes associated with a given utter-

Abstract: Suppose that we accept the following two claims: (a) Our common-sense psychology, which gives content to the everyday notions of 'belief', 'desire', 'pain' and so forth, embodies quite unacceptable dualist preconceptions, (b) the theoretic terms of a substantially misguided theory fail to refer. Then, we shall be inclined to conclude, there are strictly speaking no such states as the 'beliefs', 'desires' and such-like of common parlance. These things are to be eliminated from our ontology in favour of purely material states of the brain. Now, I take it that neither (a) nor (b) is self-evidently true: but equally, neither is patently absurd. And so, pending a detailed investigation of the ramifying issues underlying these two claims, we would expect the eliminative materialism to which they give rise to remain a genuine option in the philosophy of mind. It would be surprising indeed if we were to be absolved from the complex task of disentangling the issues underlying (a) and (b) by the production of an easy knock-down argument against the eliminative materialist. However, Nicholas Everitt has recently tried to produce just such an argument:1 perhaps not surprisingly, as I hope to show in this reply, his argument fails. Everitt characterizes the position he opposes as holding that 'in a literal and straightforward way there are no mental items, just as there are no unicorns'. But this is potentially misleading. Unicorns are entirely mythological, which is just to say that there is nothing in the world which answers in any way to talk of such beasts. Contrast, say, states of demoniacal possession: these again do not exist-but talk of such states, we may suppose, does misidentify a genuine class of states, namely forms of hallucinatory psychosis. On the eliminative materialist's view, talk of mental states is more akin to talk of possession than talk of unicornsinvolving the mishandling of genuine distinctions rather than pure mythology. Thus mental states are to be eliminated in favour of the brain states whose presence or absence is the reality which underlies whatever there is of value in the distinctions of our everyday psychology, just as states of possession are to be eliminated in favour of the psychoses whose presence or absence is the reality which underlies the miscategorizations of the demon theory. It would indeed be absurd to treat mental states on a par with unicorns-things to be unceremoniously eliminated, leaving no residuary legatees. In so far as our common-sense psychology is (within its limitations) a reasonably successful theory, those who favour its elimination owe us at least the ceremony of an explanation of its successes, an account of what real distinctions our folk theory has fumblingly grasped.

Abstract: Philosophers who devise fourth conditions in definitions of knowledge that p should investigate how their definitions might apply to cases where, if the subject did know thatp, that knowledge would be memory knowledge. Carl Ginet has done this.' More than this, by carefully developing a correlative four-part definition of memory knowledge that p, he has provided what may be the most nearly adequate traditional account of memory knowledge to have come along.2 I shall argue that, as a definition, the latter account fails-owing to the problematic way in which it incorporates the fourth condition in his general definition of knowledge.

Abstract: In his Interests and Rights: The Case Against Animals, 1 R. G. Frey advances a number of arguments which, he thinks, show that animals 'cannot have desires' (e.g., p. ioi). Some of these arguments turn on Frey's analyses of desires in terms of beliefs and beliefs in terms of thinking that certain sentences are true (pp. 53 ff.). Other arguments against the possibility of animals having desires are marshalled independently of these analyses. My present interest is limited to these latter arguments. Frey acknowledges that a critic of his view that animals cannot have desires will protest that, at the very most, the arguments that turn on his analysis of desire and belief show that animals cannot have some kinds of desires, not that they cannot have any. 'Specifically', Frey writes (p. ioi), 'this critic is likely to have in mind that, however many kinds of desires there are, there is a class of desires-let us call them "simple desires"which do not involve the intervention of belief . . .'. Thus, for example, Fido may, on the critic's view, simply desire a bone, and the attribution of this desire to Fido need not imply that Fido has any beliefs about anything. Now, it is, I think, doubtful that anybody, including Frey's patient critic, believes this of Fido, since it is unclear what it could mean to say that Fido desires a particular bone but that he does not believe anything whatsoever about the object he selects by way of fulfilling his desire. That is, it is unclear how one could make sense of the attribution of desires to some individual, A, while denying that A believes that what A selects to fulfil A's desire is what A desires (e.g., is a bone). But let us suppose, for the sake of argument, that this is possible. We are to imagine Frey's critic saying that Fido simply desires the bone. Frey, since he believes that animals cannot have desires, must believe that Fido cannot simply desire this. Why? Those who attribute simple desires to animals, Frey believes, are faced with a dilemma. Assuming that Fido is said to desire a bone, the question arises: 'Is (Fido) aware that it has this simple desire? It either is not so aware or it is' (p. I04). In either case, Frey believes, the attribution of simple desires to Fido in particular, and to animals in general, is impaled by a horn of the dilemma he poses. Let us first consider Frey's argument against those who would attribute simple desires to Fido but who would deny that Fido is aware that he has this desire. Here is Frey's principal argument against this view.

Abstract: Discussions of the Prisoner's Dilemma and of public goods commonly claim that, although there are circumstances in which the independent pursuit of rationally self-interested strategies will fail to achieve an optimal result for all interested persons, the intervention of a third party can succeed in securing that result. Assume that prisoners Red and Blue confront the conventional pay-off matrix for the alternative strategies of confession and non-confession. It is held that the attainment of the optimal result for all interested persons requires that White enforce a policy of nonconfession on both prisoners. What I wish to show is that this claim, that White's enforcement of such a policy can be regarded as a solution to the problem, is alternatively inconsistent with one of the following three statements upon the truth of which the Dilemma and its putative resolution depend: (i) all persons pursue rationally self-interested strategies; (ii) the enforcement of it is needed to bring about joint non-confession; or (iii) joint non-confession is optimal for all interested persons. In the circumstances of the Dilemma, it is clearly in the interest of each prisoner that a policy of non-confession be discriminatorily enforced on the other but not on himself. And it is thus in the interest of each of them to promise to make it worth White's while to enforce such a discriminatory policy. Is there any reason why White will not accede to the larger or more credible of the bribes promised him, and will refuse to enforce a discriminatory policy? He might decline to do so if he is not interested in the bribe. But his so declining is inconsistent with (i). Alternatively, he might decline to do so if the prisoners' joint non-confession is optimal for White as well. But if this is so and if White acts in a rationally self-interested manner, then he will not enforce the scheduled penalties and will make this intention known to the prisoners. Hence joint non-confession will occur spontaneously-contrary to (ii). Finally, he might decline to do so if enforcing a discriminatory policy will result in his being penalised by other persons, e.g. those who set the scheduled penalties. But if this is so, then those other persons must be deemed to have an interest in securing both prisoners' confessions-which is contrary to (iii). This last argument might be queried along the following lines. It could be objected that, even if joint non-confession were in the interest of those other persons as well, there would nevertheless still be a need for its enforcement to bring it about. The reason for this is that it would remain in each prisoner's interest unilaterally to confess, since his confession and his colleague's non-confession promise him a lighter penalty than does their joint non-confession. What is wrong with this objection is its implicationthat those who set the schedule of penalties are irrational. For if the prisoners' joint non-confession is indeed optimal for them, they act irrationally in scheduling a heavier penalty for unilateral non-confession than for unilateral confession.

Abstract: In the Logisch-philosophische Abhandlung Wittgenstein says so little about physics that commentators have had the greatest trouble to comment on it. (See, for example, Black (I964).) It even happened that a misprint in the first edition of this elliptic treatise went unnoticed for sixty years. The explanation seems to be that it did not occur to anyone that Wittgenstein in his philosophy of physics did little more than summarize parts of a few articles Mach wrote in his Popular-wissenschaftliche Vorlesungen. (Ryle (195 I) even said of Wittgenstein: 'He was influenced by Frege and Russell, not by Mach.') In order to show that Wittgenstein was directly influenced by Mach, I will confine myself to the following subjects on which Wittgenstein commented: (i) chronometry, geometry and problems of space and time (6.36I I, 6.36I I I), and (2) physical principles 6.32-6.34). Mach wrote on the first of these subjects in the last essay that appeared in the fourth edition of the aforementioned book, 'Eine Betrachtung fiber Zeit und Raum'. The second he treated in his articles 'Ueber das Princip der Erhaltung der Energie' and 'Werden Vorstellungen, Gedanken vererbt?' (Mach (I91o).) I am not concerned here with the influence of Part VI of Russell's Principles of mathematics on Wittgenstein's Abhandlung, nor with the influence of Hertz's Prinzipien der Mechanik on both works, I only want to point out that the similarity of Wittgenstein's remarks with some of Mach's popular-scientific agruments is so striking that one cannot but conclude that Wittgenstein had read at least some of the articles of Mach's Popular-wissenschaftliche Vorlesungen. First, there are Wittgenstein's comments on Kant's problem of the right and left hand which cannot be made to cover one another: (i) this problem occurs already in the plane, and even in one-dimensional space; (2) the right and left hand are in fact completely congruent; (3) a right-hand giove could be put on a left hand if it could be turned in four-dimensional space (cf. 6.36III). Mach made similar remarks, but he at least did not omit his source: according to Mach, M6bius remarked about I827 'that a linear figure abc, that can be seen as the symmetrical counterpart of a'b'c'mirrored in SS on the same straight line, can never be made to cover the latter on this line; for this purpose one has to take the figure out of the straight line and turn it; in order to do that at least two dimensions are needed, so a plane' (ibid., pp. 502-503). In addition Mach mentioned M6bius' generalisation, not only for triangles in a plane, but also for congruent three-dimensional bodies, that cannot be made to cover each other in any way in space: 'But one could

Abstract: Problems there may be about the interpretation of Peirce's distinction between type and token but nothing Professor Dipert says in his reply1 persuades me that my original distinction2 was ill-found. I argued that in general one section of a token can be replaced by the corresponding section from another of the same type; the fact that this cannot generally be done in the case of performances of music whereas it can be done in the case of a copy of a poem suggests that the type of which performances are the tokens is not the work but the interpretation. Dipert argues there are nonaesthetic cases where part of a token cannot be replaced by part of another of the same type. He instances removing the top part of a line of print from one fount and replacing it with the corresponding part from a different fount in such a way that the result is not recognizable as that, or any other sentence. Secondly he argues that the replacement of one part of a musical performance by the corresponding part of another is quite possible, contrary to what he believes to be my position. The first argument is easily disposed of. It is an ignoratio elenchi. Dipert has not shown that the result of his transplant is not a token of the same type but merely that it is not (immediately?) recognizable as such. The second argument is more interesting because it reveals either a lack of concern for or a lack of understanding of the distinctions that musicians commonly make, distinctions which I think can be given a rationale. Dipert maintains that we can 'almost certainly' interchange passages or at least a single chord between performances by Bernstein and Walter without aesthetic damage. This I would not deny providing that the performances are performances of the same interpretation. Both his examples and the general tenor of his argument suggest that Dipert intended to write 'interpretation' instead of 'performance' and I shall take this more plausible form of his criticism as the basis for discussion. If, as is virtually certain, Bernstein's interpretation is neither a copy of nor coincidentally identical with Walter's, then Dipert's claim is evidently intuitively wrong. I think that there are some interesting arguments which are relevant. First of all the very 'spacing' of a chord should be integral to an interpretation. The way that either of these conductors stresses or plays down the inner parts should cohere with an overall vision of the work. It is, of course, not to the point that any individual listener might fail to notice

Abstract: Is reference made to the Euston Arch in either (i) or (2)? If the answer for (2) is affirmative, then it follows immediately that it is possible to refer to what does not exist. (Of course it does not follow that, in general, any true negative singular existential statement of the form of (2) involves reference to the non-existent.) If the answer for (2) is negative, then the entailment of (2) by (i) is obscure. It needs more than just simplification to sustain it. Unless, that is, we provide a negative answer for (i) as well; and such an answer, though it has been given, is surely implausible. For if it is possible to refer to a past object and talk of its colour, its height, and its position, there is surely no semantic hurdle in going on to speak of its destruction and hence its no longer existing. Only the discovery of concealed traps or compelling theory should make us think otherwise. Stephen Read has recently argued that '. . . "exists" restricts its subject to an attributive use' and that '... the answer to the puzzle about the apparent reference to what was denied to exist, or for that matter, affirmed to, is that these phrases do not occur referentially, and do not refer, but occur attributively.'1 As a general answer, this seems to me to be quite wrong. The verb 'refer' suffers two ambiguities. First, it suffers agentinstrument ambiguity. This it shares with many verbs (e.g. 'cut', 'write', 'bite') which may be predicated of people in characterising their actions, or of the instruments they employ. Thus we may speak of people or knives cutting, people or pens writing, and people or teeth biting. So, we may speak of people (agents) or the words they use (instruments) as referring. What the agent means to be doing and what the instrument is actually doing often fail to coincide. What the knife may be cutting, the pen writing, or the teeth biting, need not be what the cutter, writer, or biter thinks. So it is again, with referring. The words may in fact refer to something other than the agent thinks he is referring to, may refer to something when he thinks they refer to nothing, and may refer to nothing

Abstract: In his article on 'The Normative Concept of Coercion' (Mind, 89, October I980), Cheyney Ryan makes a number of claims about the meaning of the word 'coercion' which seem to me very arguable or, in some cases, obscure. The first of these is that 'coercion' is the 'correlative' of 'liberty' (p. 48 I) or that, as Ryan puts it later, 'liberty' should be construed to mean 'absence of coercion' (p. 496). Should not 'liberty' also be construed, however, to mean absence of compulsion, constraint, restraint (perhaps), interference, hindrance (the correlative notion which appears in Ryan's quotation from Hobbes on page 48I), not to mention oppression, enslavement and so forth? Are not these, too, correlatives of liberty, each in its own unique way? An account of the concept of coercion would be unacceptably vague, it seems to me, if it were an account of all that of which the absence is liberty, since liberty comprises the absence of many other things besides coercion. To take just one example: children of school age are not at liberty not to attend school, but we call their attendance 'compulsory', not 'coercive' (or, if we do, then this is for some other reason than that they are not at liberty not to attend). Ryan himself notes on page 494, in his criticism of Dworkin, that '. . . words like "constraint", "compulsion", and "coercion" ' are not 'really interchangeable' and that each plays 'a unique role . . . in our language'. Surely, then, it should be important to him to distinguish ways in which not only coercion but also constraint, compulsion and so forth are absent from liberty. Otherwise, he would himself be using the word 'coercion' as a blanket term in just the way that he appears to be criticizing in Dworkin. Perhaps Ryan would say that, by contrast with coercion, compulsion and so forth are correlatives not of liberty but of freedom-a distinction to which he points on pages 496-497, where he says: '. . . a right to freedom, not liberty, (is) a right to be spared from preventive actions, not coercive ones'. But this method of distinguishing the concepts of freedom and liberty seems unworkable, for at least two reasons. First, prevention itself can sometimes be coercive, as Ryan notes on page 49I when he refers to '. . . instances of prevention, where P coerces Q into not doing A'. (Indeed, much of Ryan's article is concerned precisely with the question of how we should distinguish coercive from non-coercive prevention.) Thus, if prevention can sometimes be coercive, then freedom cannot be distinguished from liberty as being characterized by an absence of prevention rather than of coercion, since an absence of prevention would sometimes be an absence of coercion too. Second, even when prevention is not coercive, it does not necessarily imply a loss of freedom-or so, at least, Ryan himself appears to argue in note I on page 482, where he claims that '. . . a spectacular catch by a centrefielder may prevent a batter from winning the