Mill, John Stuart: Logic and metaphysics

J.S.Mill Logic and metaphysics John Skorupski ENLIGHTENMENT AND ROMANTICISM IN MILL’S PHILOSOPHY Mill’s importance as one of the major figures of nineteenth-century politics and culture, and the current interest in him as a moral and political philosopher, are both so great that they make it hard to see him in another aspect—as a leading contributor to the British tradition of epistemology and metaphysics. Yet it was the System of Logic (1843) that first established his reputation; and his views in this field remain as interesting and relevant as his better known views in ethics and politics. Throughout his intellectual life Mill sought to weave together the insights of enlightenment and Romanticism. He applied Romantic idealism’s moral understanding to utilitarianism’s concepts of character, imagination and purpose, freedom and reason, human good. To German Romanticism he owes one of his master themes—that of the culture of human nature as a whole, both in its diverse spontaneity and in its rational autonomy. But the metaphysical and psychological foundations of his thought lie securely in the naturalistic empiricism of the British school—and, moreover, in its radical and associationist, rather than its conservative and innatist, wing. Thus the deepest questions about Mill’s philosophical success turn on the possibility of such a synthesis, and on how far he achieved it. Reasons for doubt centre on two great issues, which at bottom are linked. Must not naturalism subvert reason, as Kant thought? And must not a natural science of man—a scientific psychology—subvert the understanding from within, the moral psychology of autonomy and expressive spontaneity, which Mill shares with the ‘Germano-Coleridgeans’ (as he called them)? Both questions lead us to the System of Logic, the first to its analysis of deductive and inductive reasoning, the second to its treatment of freedom and determinism. They remain important issues in contemporary philosophy. Beyond these two questions about the coherence of Mill’s thought, there is a third—whether his associationist psychology can be reconciled with the idea of determinate human potentiality he requires for his account of ‘man as a progressive being’. But this question does not have the direct contemporary significance that the other two have and will not be discussed here. NATURALISM AND SCEPTICISM Naturalism is the view that the mind is an entirely natural entity—a part of the natural order. But if the mind is only a part of nature, it seems that no real knowledge of the natural world can be a priori. Either all real knowledge is a posteriori, grounded in experience, or there is no real knowledge. That this consequence genuinely follows from naturalism Mill and Kant would have agreed. The difference between them was that Mill, unlike Kant, thought that knowledge could be grounded on such a basis. Thus the purpose of the System of Logic was, firstly, to spell out the full extent of epistemological empiricism—that is, of the view that all knowledge is a posteriori—and, secondly, to show how knowledge is possible on that basis. Its greatest achievement is the radicalism and the rigour with which it presses through to the first of these objectives. Before we examine the System it will be useful to draw some further broad contrasts between Mill, Kant, David Hume and Thomas Reid. Unlike Hume, Kant or even Reid, Mill shows no interest at all in scepticism. He agrees with Kant that naturalism must remove a priori grounding from the common-sense principles which Reid presents. He does not agree with Reid that those principles are ‘innate’, but he also sees—though with less clarity, as we shall see—that even if they were, that would not give them an epistemological, as against a merely psychological, independence of experience. Yet he does not unleash a sceptical attack on reason as Hume does, nor does he see any need to defend reason against such an attack. In fact he sees no crisis of reason, and it is this that sets him apart most fundamentally from the Critical legacy of Kant. Here he is loyal to his Benthamite inheritance. The Benthamites wished to distance themselves from Hume’s scepticism just as firmly as the Scottish common-sensists did— unlike the latter, though, they felt no call to respond systematically to it. Mill agreed with them: in this respect his attitude was English rather than Scottish. More broadly, however, both Mill and Reid belong in that British naturalistic camp which believes that no serious philosophical lesson can be drawn from scepticism. For neither of them thinks—as Hume did—that sceptical arguments are sound and significant, that they show something of negative importance about reason. The difference between them is rather that Reid believes that scepticism stems solely from an erroneous theory of mind, and can be entirely defused by denying the existence of the objects postulated by that theory— ideas—while Mill in contrast never engages with scepticism at all. If one thinks that scepticism is both unanswerable and unserious this may be the true philosophic wisdom. Whether it is wisdom or evasion is a question which keeps on returning in philosophy. But whatever view one takes on it one will misunderstand the System of Logic if one does not grasp that it is a work of what a contemporary philosopher in the same tradition, W.V.O.Quine, has called ‘naturalized epistemology’. One of the ways in which it is, is precisely this: that it neither raises nor seeks to answer sceptical questions. THE ANALYSIS OF LANGUAGE: LOGIC AND MATHEMATICS CONTAIN REAL INFERENCES Fundamental to the System is a distinction Mill draws between ‘verbal’ and ‘real’ propositions, and correspondingly, between ‘merely apparent’ and ‘real’ inferences. He applies it with greater strictness, and addresses his thesis, that merely apparent inferences have no genuine cognitive content, with greater resolution, than anyone had done before. We can best reconstruct it by starting with the notion of a merely apparent inference. An inference is merely apparent when no real inferential move has been made. For this to be so, the conclusion must literally have been asserted in the premisses. In such a case, there can be no epistemological problem about justifying the apparent inference—there is nothing to justify. A verbal proposition can now be defined as a conditional proposition corresponding to a merely apparent inference. (This is not the whole of Mill’s distinction, for he also counts elementary inferences and propositions concerning identity as verbal; but we can ignore that for the moment.) The distinction corresponds, as Mill himself notes, to that which Kant makes between ‘analytic’ and ‘synthetic’. (Kant formulates it for affirmative predicative propositions, and so does Mill, but the version given in the previous paragraph can be used to avoid this limitation.) But there is also a broader notion of analyticity, which has been influential in this century and which may also be thought to be implicit in some, though not all, of Kant’s formulations. It defines an analytic truth as one from whose negation a contradiction can be deduced, with the help, where necessary, of definitional transformations, and using principles of logic alone. In the broad sense of ‘analytic’, it becomes a trivial truth that logical principles are analytic. But what is then no longer trivial is the crucial thesis that analytic propositions have no genuine cognitive content, and hence pose no epistemological problem. If we keep to Mill’s understanding of the distinction between ‘verbal’ and ‘real’, which corresponds to the narrower Kantian notion of analyticity, pure mathematics, and logic itself, contain ‘real’ propositions and inferences with genuine cognitive content. The clear recognition of this fact is the chief philosophical achievement of the System of Logic. For if Mill is also right in holding that naturalism entails that no real proposition is a priori, then he has shown the implications of naturalism to be radical indeed. Not only mathematics but logic itself will be empirical. To demonstrate that logic and mathematics contain real propositions Mill has to embark on an extensive semantic analysis of sentences and terms (he calls them ‘names’), of syllogistic logic, and of the so-called ‘Laws of Thought’. His analysis has many imperfections and he never unifies it in a fully general account. But he does supply the foundations of such an account, and in doing so takes the empiricist epistemology of logic and mathematics to a wholly new level. The starting point is a distinction between the denotation and connotation of names. Names, which may be general or singular, denote objects and connote attributes of objects. (Attributes may themselves be denoted by ‘abstract’ names, though the term is misleading, for Mill conceives attributes nominalistically). A general name connotes attributes and denotes each object which has those attributes. Most singular names also connote attributes, but their grammatical construction indicates that they denote just one object if they denote at all. There is however an important class of singular names—‘proper names’ in the ordinary sense, such as ‘Dartmouth’—which denote an object without connoting any property. Identity propositions which contain only such non-connotative names, as ‘Tully is Cicero’, are in Mill’s view verbal. They lack content in the sense that, according to Mill, the only information conveyed is about the names themselves: ‘Tully’ denotes the same object as ‘Cicero’ does. Mill’s point is that there is no fact in the world to which ‘Cicero is Tully’ corresponds; a thought similar to that which inspired Wittgenstein’s treatment of identity in the Tractatus. The obvious difficulty about assimilating them to verbal propositions on this basis is that knowledge that Cicero is Tully is not a priori. We cannot know the proposition to be true just by reflecting on the meaning of the names— whereas Mill’s overall inten-tion is that the class of verbal propositions should be identical with the class of propositions which are innocuously a, priori because empty of content. He does not seem to notice this difficulty. The meaning of a declarative sentence—‘the import of a proposition’—is determined by the connotation, not the denotation, of its constituent names; the sole exception being connotationless proper names, where meaning is determined by denotation. (Again there is something puzzling here, for it needs to be explained how this thesis about the meaning of proper names is to be reconciled with the aposteriority of ‘Cicero is Tully’.) Mill proceeds to show how the various syntactic forms identified by syllogistic theory yield conditions of truth for sentences of those forms, when the connotation of their constituent names is given. Armed with this analysis he argues that logic contains real inferences and propositions. (He assumes that to assert a conjunction, A and B, is simply to assert A and to assert B. He defines A or B as ‘If not A, then B, and if not B, then A’. ‘If A then B’ means, he thinks, ‘The proposition B is a legitimate inference from the proposition A’.) His strategy is an admirably forceful pincer movement. One pincer is an indirect argument. If logic did not contain real inferences, all deductive reasoning would be a petitio principii, a begging of the question—it could produce no new knowledge. Yet clearly it does produce new knowledge. So logic must contain real inferences. The other pincer is a direct semantic analysis of the supposed ‘axioms’ of syllogistic reasoning and of the laws of thought. It shows them to be real and not merely verbal. The execution of this strategy is flawed, because Mill mixes it up with an interesting but distinct objective. He wants to show that ‘all inference is from particulars to particulars’. The point is to demystify the role general propositions play in thought. He argues that in principle they add nothing to the force of an argument; particular conclusions could always be derived inductively direct from particular premisses. Their value is psychological. They play the role of ‘memoranda’, summary records of the inductive potential of all that we have observed, and they facilitate ‘trains of reasoning’ (as e.g. in ‘This is A, all As are Bs, no Bs are Cs, so this is not C’). Psychologically they greatly increase our memory and reasoning power, but epistemologically they are dispensable. As Mill presents it, this thesis is tied in with his rejection of ‘intuitive’ knowledge of general truths and with his inductivism (which we shall come to shortly). For it assumes the illegitimacy of hypothesizing general propositions, as against generalizing to them from observation of singular conjunctions. Beneath this, however, there lies a deeper and obscurer sense in which a radical empiricist must hold that all inference is from particulars to particulars. For consider the inference from ‘Everything is F’ to ‘a is F’. Is it a real or merely apparent inference? It is impossible to hold it real if one also wishes to argue that real inferences are a, posteriori. But the only way of treating it as verbal which is open to Mill is to treat the premiss as a conjunction: ‘a is F and b is F and…’ If that approach is precluded then all that remains is to deny that ‘Everything is F’ is prepositional—it must, rather, express an inferential commitment. Both approaches are very close to the surface in Mill’s discussion of the syllogism, and he comes closest to the latter when he emphasizes that a general proposition is ‘a memorandum of the nature of the conclusions which we are prepared to prove’. But these issues about generality (and about conditional propositions) do not emerge clearly in his analysis; like his treatment of proper names and of identity, they were destined to make a decisive appearance on the agenda of philosophical logic only later, in the twentieth century. MILL’S EMPIRICIST VIEW OF LOGIC, GEOMETRY AND ARITHMETIC Though Mill’s treatment of generality and the syllogism is somewhat confused and opaque, he is quite clear cut in holding the laws of contradiction and excluded middle to be real—and therefore a posteriori—propositions. He takes it that ‘not P’ is equivalent in meaning to ‘it is false that P’; if we further assume the equivalence in meaning of P and ‘It is true that P’, the principle of contradiction becomes, as he puts it, ‘the same proposition cannot at the same time be false and true’. ‘I cannot look upon this’ he says ‘as a merely verbal proposition. I consider it to be, like other axioms, one of our first and most familiar generalizations from experience.’ He makes analogous remarks about excluded middle, which turns—on these definitions—into the principle of bivalence, ‘Either it is true that P or it is false that P’. A truly radical empiricism! After this it is not surprising to find the same broad strategy applied to mathematics. If it was merely verbal, mathematical reasoning would be a petitio principii. Moreover a detailed semantic analysis shows that it does contain real propositions. Mill provides brief but insightful empiricist sketches of geometry and arithmetic. On geometry he is particularly good. The theorems of geometry are deduced from premisses which are real propositions inductively established. (Deduction is itself of course a process of real inference.) These premisses, where they are not straightforwardly true of physical space, are true in the limit. Geometrical objects—points, lines, planes—are ideal or ‘fictional’ limits of ideally constructible material entities. Thus the real empirical assertion underlying an axiom such as ‘Two straight lines cannot enclose a space’ is something like ‘The more closely two lines approach absolute breadthlessness and straightness, the smaller the space they enclose’. Mill applies his distinction between denotation and connotation to show that arithmetical identities such as ‘Two plus one equals three’ are real propositions. Number terms denote ‘aggregates’ and connote certain attributes of aggregates. (He does not say that they denote those attributes of the aggregates, though perhaps he should have done.) ‘Aggregates’ are natural not abstract entities—‘collections’ or ‘agglomerations’ individuated by a principle of aggregation. This theory escapes some of the famous but rather unfair criticisms Frege later made of it, but its viability none the less remains extremely doubtful. The trouble is that the respects in which aggregates have to differ from sets if they are to be credibly natural, and not abstract, entities are precisely those in which they seem to fail to produce a fully adequate ontology for arithmetic. (One can for example number numbers, but can there be aggregates of aggregates, or of attributes of aggregates, if aggregates are natural entities?) However this may be, Mill’s philosophical programme is clear. Arithmetic, like logic and geometry, is a natural science, concerning a particular department of the laws of nature—those concerning the compositional properties of aggregates. The upshot is that the fundamental principles of arithmetic and geometry, as well as of logic itself, are real. Given epistemological empiricism, it follows that deductive reasoning, ‘ratiocination’, is empirical. Mill has provided the first thoroughly naturalistic analysis of meaning and of deductive reasoning itself. He distinguishes his own view from three others—‘conceptualism’, ‘nominalism’ and ‘realism’. ‘Conceptualism’ is his name for the view which takes the objects studied by logic to be psychological states or acts. It holds that names stand for ‘ideas’ which make up judgements and that ‘a proposition is the expression of a relation between two ideas’. It confuses logic and psychology by assimilating propositions to judgements and attributes of objects to ideas. Against this doctrine Mill insists that All language recognises a difference between a doctrine or opinion, and the fact of entertaining the opinion; between assent, and what is assented to… Logic, according to the conception here formed of it, has no concern with the nature of the act of judging or believing; the consideration of that act, as a phenomenon of the mind, belongs to another science. (7:87) He traces conceptualism to the seventeenth century: it was introduced by Descartes, fostered by Leibniz and Locke, and has obscured the true status of logic—which is simply ‘the Science of Science’—ever since. The nominalists—Mill cites Hobbes—hold that logic and mathematics are entirely verbal. Mill takes this position much more seriously than conceptualism and seeks to refute it in detail. His main point is that nominalists are able to maintain their view only because they fail to distinguish between the denotation and the connotation of names, ‘seeking for their meaning exclusively in what they denote’ (7:91). Nominalists and conceptualists both hold that logic and mathematic can be known nonempirically, while yet retaining the view that no real proposition about the mindindependent world can be so known—but both are confused. But what if one abandons the thesis that no real proposition about the mind-independent world can be known a The nineteenth century 86 priori? The realists do that—they hold that logical and mathematical knowledge is knowledge of universals existing in an abstract Platonic domain; the terms that make up sentences being signs that stand for such universals. This is the view Mill takes least seriously—but versions of it were destined to stage a major revival in philosophy, and semantic analysis would be their main source. In fact in the contemporary use of the term, Mill is himself a nominalist—he rejects abstract entities. That is why he treats aggregates as concrete objects, and attributes as natural properties rather than universals. But, just as severe difficulties lie in the way of treating the ontology of arithmetic in terms of aggregates rather than classes, so there are severe difficulties in the way of treating the ontology of general semantics without appealing to universals and classes, as well as to natural properties and objects. We can have no clear view of how Mill would have responded to these difficulties had they been made evident to him. But we can I think be fairly sure that he would have sought to maintain his nominalism. However, the central target of Mill’s attack is the doctrine that there are real a priori propositions. What, he asks, in practice goes on, when we hold a real proposition to be true a priori? We find its negation inconceivable, or that it is derived, by principles whose unsoundness we find inconceivable, from premisses whose negation we find inconceivable. Mill is not offering a definition of what is meant by such terms as ‘a priori’, or ‘self-evident’; his point is that facts about what we find inconceivable are all that lends colour to the use of these terms. They are facts about the limits, felt by us from the inside, on what we can imagine perceiving. Mill thought he could explain these facts about unthinkability, or imaginative unrepresentability, in associationist terms, and spent many pages claiming to do so. They are not very convincing pages, but that does not affect his essential point, which is this: the step from our inability to represent to ourselves the negation of a proposition, to acceptance of its truth, calls for justification. Moreover, the justification itself must be a, priori if it is to show that the proposition is known a, priori. (Thus Mill is prepared for example to concede the reliability of geometrical intuition: but he stresses that its reliability is an empirical fact, itself known inductively.) At this point, Kant could agree. To vindicate the possibility of synthetic a priori knowledge calls, he claims, for nothing less than transcendental idealism. But without synthetic a priori knowledge, knowledge as such becomes impossible. The very possibility of knowledge requires that there be a priori elements in our knowledge. THE METHODS OF INDUCTION AND THEIR STATUS The System of Logic in contrast sets out to vindicate in general terms the possibility of a scheme of scientific knowledge which appeals at no point whatever to an a priori principle. One point in the opposition between Critical and naturalistic epistemology, as we have noticed, is the latter’s refusal to take seriously pure sceptical arguments; but naturalistic epistemology also has two other ingredients—an appeal to a natural, or in Mill’s word ‘spontaneous’, agreement in propensities to reason, and what may be called an ‘internal’ vindication of these fundamental reasoning propensities. All three ingredients are present in the System of Logic. For Mill, the basic form of reasoning—epistemologically, historically and psychologically—is enumerative induction, simple generalization from experience. This is the diposition to infer to the conclusion that all As are B from observation of a number of As which are all B. (Or to the conclusion that a given percentage of all As are B from observation of that percentage of Bs among a number of As.) We spontaneously agree in reasoning that way, and in holding that way of reasoning to be sound. The proposition ‘Enumerative induction is rational’ is not a verbal proposition. But nor is it grounded in an a priori intuition. All that Mill will say for it is that people in general, and the reader in particular, in fact agree on reflection in accepting it. It is on that basis alone that he rests its claim. Mill’s problem of induction, the problem he wants to solve, is not Hume’s. In sidestepping the purely sceptical question about induction, Mill uses the analogy of a telescope which Thomas Reid had also used in a similar context—though in Reid the telescope is Reason as against Common Sense, while in Mill it is Scientific as against spontaneous induction: Assuredly, if induction by simple enumeration were an invalid process, no process grounded on it would be valid; just as no reliance could be placed on telescopes, if we could not trust our eyes. But though a valid process, it is a fallible one, and fallible in very different degrees: if therefore we can substitute for the more fallible forms of the process, an operation grounded on the same process in a less fallible form, we shall have effected a very material improvement. And this is what scientific induction does. (7:567–8) Mill’s aim is to provide the telescope. The problem he starts from is not a sceptical but an internal one—why is it that some inductions are more trustworthy than others? Why is a single instance, in some cases, sufficient for a complete induction, while in others, myriads of concurring instances, without a single exception known or presumed, go such a very little way towards establishing a universal proposition? Whoever can answer this question…has solved the problem of induction. (7:314) Mill’s answer takes the form of a natural history of the ‘inductive process’. The point is to show how that process is internally vindicated by its actual success in establishing regularities, and how it eventually gives rise to more searching methods of investigation. Mankind begins with ‘spontaneous’ and ‘unscientific’ inductions about particular unconnected natural phenomena or aspects of experience. Generalizations accumulate, interweave and are found to stand the test of time: they are not disconfirmed by further experience. As they accumulate and interweave, they justify the second-order inductive conclusion that all phenomena are subject to uniformity, and, more specifically, that all have discoverable sufficient conditions. In this less vague form, the principle of general uniformity becomes, given Mill’s analysis of causation, the Law of Universal Causation. This conclusion in turn provides (Mill believes) the grounding assumption for a new style of reasoning about nature—eliminative induction. In this type of reasoning, the assumption that a type of phenomenon has uniform causes, together with a (revisable) assumption about what its possible causes are, initiates a comparative inquiry in which the actual cause is identified by elimination. Mill formulates the logic of this eliminative reasoning in his well-known ‘Methods of Empirical Inquiry’. His exposition is rather garbled but he was right to be proud of it, for it did show how effective eliminative reasoning can be. His picture of the interplay between enumerative and eliminative reasoning, and of the way it entrenches, from within, our rational confidence in the inductive process, is elegant and penetrating. The improved scientific induction which results from this new style of reasoning spills back on to the principle of Universal Causation on which it rests, and raises its certainty to a new level. That in turn raises our confidence in the totality of particular enumerative inductions from which the principle is derived. In short, the amount of confidence with which one can rely on the ‘inductive process’ as a whole depends on the point which has been reached in its natural history—though the confidence to be attached to particular inductions always remains variable. The fundamental norm of scientific reasoning, enumerative induction, is not a merely verbal principle. But what can it mean to deny that it is a priori? Mill says that we learn ‘the laws of our rational faculty, like those of every other natural agency’, only by ‘seeing the agent at work’. He is quite right: we can find out what our most basic reasoning dispositions are, only by critical reflection on our practice. This reflective scrutiny of practice is, in a certain sense, an a posteriori process. It examines dispositions which we have before we examine them. Having examined our dispositions, we reach a reflective equilibrium in which we endorse some—and perhaps reject others. We endorse them as sound norms of reasoning. But at this point the sceptic will ask by what right we do so—and Mill rules his question out of order. So denying that the fundamental principle of induction is a priori comes down, it seems, to just that ruling. Might it not just as well have been said that the principle is a priori? But that would suggest that there was some further story, Platonic or transcendental, to be had, which explained and legitimated our reasoning practice, and this is what Mill denies. So too does the sceptic: the fact that the sceptic and the naturalist agree on that hardly shows why it is all right to rule the sceptic’s question out of order. It seemed obvious to Mill’s epistemological critics, whether they were realists or post-Kantian idealists, that this was evasive: naturalism could seem to differ from scepticism only by being uncritical. Where Hume launches a sceptical assault on reason, Mill opens up all our beliefs to an empirical audit. Hume takes deduction for granted and raises sceptical questions about induction. Mill takes the legitimacy of natural reasoning propensities for granted, but he questions the aprioricity of deduction. Hume and Mill are both naturalistic radicals—but in quite different ways. Mill leaves no real principle of deduction, no common-sense belief entrenched—with one telling exception. The exception is our disposition to rely on the deliverances of memory, which he acknowledges, in the manner of Thomas Reid, to be ‘ultimate’. But, with this exception, the only ultimate principle that survives in Mill’s science of science is enumerative induction. The whole of science, he thinks, can be built by this single instrument. HYPOTHESIS This is Mill’s inductivism—the view that enumerative induction is the only ultimate method of inference which puts us in possession of new truths. Is he right in thinking it to be so? The question produced an important, if confused, controversy between him and William Whewell (1794–1866). Their disagreement concerned the role of hypotheses. Whewell argued that the Hypothetical Method was fundamental in scientific inquiry: the method in which one argues to the truth of an hypothesis from the fact that it would explain observed phenomena. Mill had read Whewell’s History of the Inductive Sciences (1837), and he could hardly fail to be aware of the pervasiveness of hypotheses in the actual process of inquiry, or of their indispensableness in supplying working assumptions—their ‘heuristic’ value, Whewell called it. The same point—the indispensableness of hypotheses in providing lines of inquiry—had been emphasized by the Frenchman Auguste Comte, with whose Cours de philosophie positive (which began to appear in 1830) Mill was also familiar. But what Mill could not accept was that the mere fact that an hypothesis accounted for the data in itself provided a reason for thinking it true. He denied that the Hypothetical Method constituted, in its own right, a method of arriving at new truths from experience. Yet Whewell’s appeal was to the actual practice of scientific reasoning, as observed in the history of science. An appeal of that kind was precisely what Mill, on his own naturalistic principles, could not ignore. The disposition to hypothesize is spontaneous, so why should it not be recognized as a fundamental method of reasoning to truth, as enumerative induction is? Mill’s refusal to recognize it is not arbitrary. The essential point underlying it is a powerful one: it is the possibility that a body of data may be explained equally well by more than one hypothesis. What justifies us in concluding, from the fact that a particular story would, if true, explain the data, that it is a true story? Other stories may equally explain the data. Mill places great emphasis on the increasingly deductive and math-ematical organization of science—that is quite compatible with his inductivism, and indeed central to it. But he takes the ‘Deductive Method’ of science to involve three steps: ‘induction’, ‘ratiocination’, and ‘verification’. A paradigm, in his view, is Newton’s explanation of Kepler’s laws of planetary motion. Induction establishes causal laws of motion and attraction, ratiocination deduces lower level regularities from them in conjunction with observed conditions, and verification tests these deduced propositions against observation. (Though this was not, and did not need to be, the historical order of inquiry.) But the Hypothetical Method suppresses the first of the three steps, the induction to ascertain the law; and contents itself with the other two operations, ratiocination and verification; the law which is reasoned from being assumed, instead of proved. (7:492) Mill agrees that it is legitimate to do this when the hypothesis in question has effectively been shown, by eliminative reasoning, to be the only one consistent with the facts. He allows various other cases of apparently purely hypothetical reasoning which are, in his view, genuinely inductive. When all such cases have been taken into account, we are left with pure cases of the Hypothetical Method, in which the causes postulated are not directly observable, and not simply because they are assumed to operate—in accordance with known laws, inductively established—in regions of time or space too distant to observe. What are we to say of such hypotheses? For example of the ‘emission’ theory, or the ‘undulatory’ theory of light? They cannot be accepted as inductively established truths, not even as probable ones: an hypothesis of this kind is not to be received as probably true because it accounts for all the known phenomena; since this is a condition sometimes fulfilled tolerably well by two conflicting hypotheses; while there are probably many others which are equally possible, but which from want of anything analogous in our experience, our minds are unfitted to conceive. (7:500) Such an hypothesis can suggest fruitful analogies, Mill thinks, but cannot be regarded as yielding a new truth itself. The data do not determine a unique hypothesis: it is this possibility, of underdetermination, which stops him from accepting hypothetical reasoning as an independent method of achieving truth, even though it is a mode of reasoning as spontaneous as enumerative induction. In seeing the difficulty Mill is certainly on sound ground. What he does not see, however, is how much must be torn from the fabric of our belief if inductivism is applied strictly. Thus, for example, while his case for empiricism about logic and mathematics is very strong, it is his methodology of science which then forces him to hold that we know basic logical and mathematical principles only by an enumerative induction. And that is desperately implausible. So it is an important question whether the difficulty can be resolved—and whether it can be resolved within a naturalistic framework, which does not yield to idealism. If naturalism can endorse the hypothetical method, it can develop a more plausible empiricism about logic and mathematics than Mill’s. But the ramifications of his inductivism are even wider, as becomes apparent if we turn to his general metaphysics. THE DOCTRINE OF THE RELATIVITY OF KNOWLEDGE Mill sets this out in his Examination of Sir William Hamilton’s Philosophy (1865). Sir William Hamilton (1791–1856) was a Scotsman who sought to moderate the views of Reid and Kant. He was a philosopher of subtlety and erudition (or even pedantry), the last eminent representative of the school of Scottish common sense, and a ferocious controversialist. Mill deemed him a pillar of the right-thinking intellectual establishment, ripe for demolition. But by the time the Examination appeared Hamilton’s death had made it impossible for him to reply—a fact which predictably caused Mill some regret. For the present-day reader, however, what is more regrettable is that Mill’s discussion of general metaphysical issues should be cast in so polemical a form. It means that important issues, particularly on the nature of logic and thought, remain shrouded in obscurity. Mill does however give himself space to develop his view of our knowledge of the external world. He begins by expounding a doctrine which he rightly takes to be generally accepted (in his time) on all sides. It affirms that all the attributes which we ascribe to objects, consist in their having the power of exciting one or another variety of sensation in our minds; that an object is to us nothing else than that which affects our senses in a certain manner; that even an imaginary object is but a conception, such as we are able to form, of something which would affect our senses in some new way; so that our knowledge of objects; and even our fancies about objects, consist of nothing but the sensations which they excite, or which we imagine them exciting, in ourselves. (9:6) This is ‘the doctrine of the Relativity of Knowledge to the knowing mind’. But there are two forms in which it may be held. According to one of the forms, the sensations which, in common parlance, we are said to receive from objects, are not only all that we can possibly know of the objects, but are all that we have any ground for believing to exist. What we term an object is but a complex conception made up by the laws of association, out of the ideas of various sensations which we are accustomed to receive simultaneously. There is nothing real in the process but these sensations. (9:6) According to the other, there is a real universe of ‘Things in Themselves,’ and… whenever there is an impression on our senses, there is a ‘Thing in itself,’ which is behind the phaenomenon, and is the cause of it. But as to what this Thing is ‘in itself,’ we, having no organs except our senses for communicating with it, can only know what our senses tell us; and as they tell us nothing but the impression which the thing makes upon us, we do not know what it is in itself at all. We suppose (at least these philosophers suppose) that it must be something ‘in itself, but all that we know it to be is merely relative to us, consisting in the power of affecting us in certain ways.(9:7) The first form (omitting from it the appeal to laws of association) corresponds to what is meant by ‘phenomenalism’ as the term is often used by philosophers today—though it was not so used in Mill’s time. Reid’s point, that sensations are not representative mental images but states of mind, does not contradict the doctrine of the Relativity of Knowledge, any more than his thesis that we perceive physical objects does. For on his account sensations, states of sensory consciousness, do mediate between the objects that excite them and the beliefs about those objects which are prompted by them—they are themselves distinct from both the objects and the beliefs. I cannot perceive without sensing. But I can sense without perceiving. For example I may have a visual sensation which prompts me to believe that I am seeing a red triangle on a green field. It is then apparently true to say, in an obvious and legitimate sense, that what I am immediately aware or conscious of is my visual sensation. That remains true even if I am perceiving no red triangle because no red triangle exists. Or, if one objects to talk of consciousness of a state of consciousness, one may simply say that my immediate visual consciousness is of a red triangle on a green field—in a sense in which that can be true though there is no such triangle. This is already enough to make epistemology, in Mill’s phrase, the ‘Interpretation of Consciousness’. The very fact of consciousness seems to impose the doctrine of the Relativity of Knowledge. To escape it, something more counter-intuitive would be required than the sensible points Reid makes about perception and sensation. It is notoriously difficult to pin down what that might be. Perhaps what is needed is nothing less than a denial that sensation is a category ontologically distinct from that of judgement and dispositions to judge: there is no irreducible category of Pure Experience. But Mill, at any rate, questions the irreducible status of sensation no more than Reid did. And he thinks it must follow that—whether or not we actually make an inductive inference from sensations to objects beyond sensation—such an inference is, epistemologically speaking, required. Is this too hasty? Is it dogmatism on Reid’s part simply to point out that we do form particular beliefs prompted by particular sensations, beliefs which we just do regard as rational? Cannot these specific cognitive dispositions be defended naturalistically, if the general disposition to make enumerative inductions can? But there is a difference. If we are immediately conscious only of states of affairs of one kind (our own sensory states) and on that basis form beliefs about states of affairs of a quite distinct kind (states of external physical objects) then some warrant is required. Reid needs to show why such a warrant does not have to rely on inductive inference—even though it licenses a belief in a state of affairs on the basis of immediate consciousness of a quite distinct state of affairs. He must call on warrants which are neither deductive nor inductive. And this requires support from ideas in the theory of meaning which had not yet been formed. We shall return to the point. MATTER AND MIND Mill sets about the notion of an ‘external’ object in great style: What is it we mean, or what is it which leads us to say, that the objects we perceive are external to us, and not a part of our own thoughts? We mean, that there is concerned in our perceptions something which exists when we are not thinking of it; which existed before we had ever thought of it, and would exist if we were annihilated; and further, that there exist things which we never saw, touched, or otherwise perceived, and things which have never been perceived by man. This idea of something which is distinguished from our fleeting impressions by what, in Kantian language, is called Perdurability; something which is fixed and the same, while our impressions vary; something which exists whether we are aware of it or not, and which is always square (or of some other given figure) whether it appears to us square or round—constitutes altogether our idea of external substance. Whoever can assign an origin to this complex conception, has accounted for what we mean by the belief in matter. (9:178–9) To assign this origin Mill postulates that after having had actual sensations, we are capable of forming the conception of Possible sensations; sensations which we are not feeling at the present moment, but which we might feel, and should feel if certain conditions were present, the nature of which conditions we have, in many cases, learnt by experience. (9:177) These various possibilities are the important thing to me in the world. My present sensations are generally of little importance, and are moreover fugitive: the possibilities, on the contrary, are permanent, which is the character that mainly distinguishes our idea of Substance or Matter from our notion of sensation. These possibilities, which are conditional certainties, need a special name to distinguish them from mere vague possibilities, which experience gives no warrant for reckoning upon. Now, as soon as a distinguishing name is given, though it be only to the same thing regarded in a different aspect, one of the most familiar experiences of our mental nature teaches us, that the different name comes to be considered as the name of a different thing. (9:179–80) We may speak of sensation conditionals of the form, ‘If such and such sensations were to occur, then such and such other sensations would occur with a given degree of probability’. (It need not always be certainty.) They express Mill’s famous ‘Permanent Possibilities of Sensation’. ‘Permanent’ is slightly misleading, for there is of course a change in the ‘Permanent’ possibilities of sensation whenever there is change in the world. Mill also uses other terms—‘certified’, ‘guaranteed’. We regularly find that whole clusters of sensation conditionals are true together, whenever some other sensory condition obtains. Thus whenever we experience that condition, we are justified in forming all the conditional expectations expressed in that cluster of conditionals. Moreover, as well as finding simultaneous correlations between certified possibilities of sensation, that is, between the truth of any sensation conditional in a set and the truth of any other in the set, we also find ‘an Order of succession’. Whenever a given cluster of certified possibilities of sensation obtains, then a certain other cluster follows—a certain other set of sensation conditionals becomes true. ‘Hence our ideas of causation, power, activity…become connected, not with sensations, but with groups of possibilities of sensation’ (9:181). But even if our reflective concept of matter—as the external cause of sensations—can be explained on psychological principles, it remains open for someone to accept the proposed origin for the concept, while also holding that good grounds can be given for thinking it to have instances. He or she will say that a legitimate inference can be made from the existence of the Permanent Possibilities and their correlations to the existence of an external cause of our sensations. It is at just this point that Mill’s inductivism comes in. Such an inference would be a case of hypothetical reasoning, to an explanation of experience which transcended all possible data of experience; and that is just what Mill rejects: ‘I assume only the tendency, but not the legitimacy of the tendency, to extend all the laws of our own experience to a sphere beyond our experience’ (9:187). So the conclusion that matter is the permanent possibility of sensation follows from the combination of the doctrine of the Relativity of Knowledge and inductivism. If matter is the permanent possibility of sensation, what is mind? Mill considers that ‘our knowledge of mind, like that of matter, is entirely relative’. Can the mind then also be resolved into ‘a series of feelings, with a background of possibilities of feeling’? He finds in this view a serious difficulty: to remember or expect a state of consciousness is not simply to believe that it has existed or will exist; it is to believe that I myself have experienced or will experience that state of consciousness. If, therefore, we speak of the Mind as a series of feelings, we are obliged to complete the statement by calling it a series of feelings which is aware of itself as past and future; and we are reduced to the alternative of believing that the Mind, or Ego, is something different from any series of feelings, or possibilities of them, or of accepting the paradox, that something which ex hypothesi is but a series of feelings, can be aware of itself as a series. (9:194) Mill is unwilling to accept ‘the common theory of Mind, as a so-called substance’: nevertheless, the self-consciousness involved in memory and expectation drives him to ‘ascribe a reality to the Ego—to my own Mind—different from that real existence as a Permanent Possibility, which is the only reality I acknowledge in Matter’ (9:208). If we discount, this conscientious uncertainty about what to say of the self, the tendency of Mill’s analysis is towards the view that all that exist are experiences in a temporal order. Yet he claims, like others before and after him, that this metaphysics is consistent with common-sense realism about the world. Phenomenalism, he thinks, leaves common sense and science untouched. In particular, minds and experiences are still properly to be seen as a part of the natural order. But are the experiences referred to, in the phenomenalist’s analysis, the very same as those referred to in common sense and scientific talk (call this ‘naturalistic’ talk)? If they are not, then we have yet to be told what they are. Then suppose they are the same. In naturalistic talk, we make reference to subjects and their experiences—and also to physical objects and their properties. Psychology, including Mill’s psychology, seeks to establish causal correlations among experiences and their physiological antecedents and consequents. But if phenomenalism is right, only the experiences are real. Mill thinks we are led to that by the very standards of reasoning recognized in a naturalistic ‘science of science’, or ‘system of logic’. If he is right, then the naturalistic vision of the world, which sees minds as part of a larger causal order, is self-undermining. For if we are led to the conclusion, that only states of consciousness are real, by an application of naturalism’s own standards, then that conclusion has to be understood on the same level as the naturalistic affirmation that states of consciousness are themselves part of a larger causal order external to them—and therefore as inconsistent with it. Causal relations cannot exist between fictional entities which are mere markers for possibilities of sensation. So either naturalism undermines itself or there is something wrong with Mill’s inductivist analysis of our natural norms of reasoning, or with his endorsement of the doctrine of the Relativity of Knowledge, or both. It is not our business to diagnose the situation further here. But it should not be assumed that Mill’s most fundamental tenet— his naturalistic view of the mind—can be safeguarding solely by rejecting inductivism and endorsing the hypothetical method. The result would be a philosophy which postulates the external world as an inference to an hypothetical explanation of pure experience. Something still fails to ring true in that. More is needed: backing for the view mentioned earlier, a view which may be thought of as in the spirit of Thomas Reid, though he did not give it the necessary backing—the view, namely, that there are norms which are neither inductive nor deductive, but which defeasibly warrant experience-based assertions about the physical world. MORAL FREEDOM The necessary backing, showing how such defeasible warrants can obtain, could be provided only by a philosophy which treats concepts, and the meanings of expressions in a language, as constituted by rules of use. This conception of concept and meaning is not present in Mill, though it could be reached by a sound progression from his naturalistic analysis of logic, and his rejection of conceptualism, nominalism and realism. Its growth can probably best be dated to the next generation of philosophers after Mill, and to the agenda of problems which they developed (partly at least in response to and reaction against Mill) at the end of the century; thus, to pragmatism, empirio-criticism, as also in some respects to neo-Kantianism and British idealism. That same conception opens up the possibility of new responses to Mill’s problem about the method of inference to the best hypothetical explanation—which was that in cases of genuine hypothesis we cannot be certain that there is a single best explanation. Though the roots of the required conception of concept and meaning date to that period, how best to formulate it is still an open issue. Moreover the most difficult (though not unconnected) question for the naturalist, that of giving an account of reasons, still remains. As we have seen, it stands out as an obstacle for Mill when he needs to account for the authority of fundamental norms of reasoning. And it also stands out when he tries to show how, on the naturalistic view, it is possible for human beings to be morally free. This was a central issue for Mill. He deals with it as it appears in the classic question of freedom and determinism. His commitment to determinism was complete. But the conclusion drawn by others from that doctrine, that we have (in Mill’s phrase) no ‘power of self-formation’, and hence are not responsible, properly speaking, for our character or our actions, would have destroyed the very centre of his moral convictions. Power to determine one’s own purposes and hold to them, responsibility for one’s actions, are at the heart of Mill’s ideal of life. ‘Moral freedom’, the ability to bring one’s desires under the control of a steady rational purpose, is a condition of self-realization, of having a character in the full sense at all. So he must show how causally conditioned natural objects can also be rationally autonomous agents. The sketch of a compatibilist solution which he provides in the System of Logic is brief but penetrating. He thought it the best chapter in the book. It is certainly a worthy contribution to the great empiricist tradition, which dismisses the problem as a perennially tempting confusion, to be dissolved by careful analysis. To describe determinism as the doctrine of ‘Philosophical Necessity’ is, Mill thinks, misleading. Not just because of the general empiricist point that causation is not compulsion but for a more subtle reason; because ‘in common use’ only causes which are irresistible, whose operation is ‘supposed too powerful to be counteracted at all’ are called necessary: There are physical sequences which we call necessary, as death for want of food or air; there are others which, though as much cases of causation as the former, are not said to be necessary, as death from poison, which an antidote, or the use of the stomach-pump, will sometimes avert…human actions are in this last predicament: they are never (except in some cases of mania) ruled by any one motive with such absolute sway, that there is no room for the influence of another. (8:839) This is a general distinction, but Mill is right to think it important for the analysis of free action. It can be applied to motives; an action caused by an irresistible motive is plainly not free. Without the distinction between resistible and irresistible causes, determinism turns into fatalism. We lose the sense of our moral freedom, which rests on the conviction that the motives on which we in fact acted were resistible. We fall into the idea that we have no power over our character; no ability to resist motives which we dislike or to choose to act on those which we admire. Now incompatibilists will concede that changes in our character may result from behaviour which is itself caused by the wish to change our character. But they will not concede that this is a true case of ‘self-formation’, because they think the wish to change our character is heteronomous: it comes from without. And they think that follows simply from the fact that the wish is determined, ultimately if not proximately, by external circumstances. So Mill has to show that while the wish must indeed be determined, that does not entail heteronomy. It can still be my wish. He cannot answer simply by invoking the distinction between resistible and irresistible motives. For a motive might perfectly well not be irresistible, in that it could be blocked by other motives—yet still be heteronomous. Something has to be added if we are to move from the idea of my motives being resistible, in the sense that they could be trumped by conflicting motives, to the stronger idea that I have the power to resist motives. That idea is the idea of rationality: the ability to recognize and respond to reasons. I act freely if I could have resisted the motive on which I in fact acted had there been good reason to do so. A motive which impairs my moral freedom is one that cannot be defeated by a cogent reason for not acting on it. The differ-ence between a heteronomous agent, driven by conflicting motives which are capable of checking each other, and an autonomous agent who himself or herself resists the motive, lies in the fact that the latter responds to, and acts on, reasons. Acting from good reason is still acting from a motive which is causally determining. What matters is how the motive determines: it must be so related to the facts, as they are believed to stand by the agent, as to constitute a good reason and it must also be the case that the agent acts on it as a reason. The same holds for the will to alter our character. It must indeed always be caused, and hence caused ultimately by circumstances we cannot help. But it still satisfies the conditions of moral freedom, if it results from our grasping that there is reason to change ourselves, and not, say, from indoctrination or obsession. Moral freedom for Mill is the ability to act on good reasons, as autonomy is for Kant; though Mill does not highlight the point as Kant rightly does. And of course it is not, for Mill, transcendental. It is something I may have to a greater or less degree. I am more or less free overall, according to the degree to which I can bring my motives under scrutiny and act on the result of that scrutiny, I can make myself more free, by shaping my motives or at least by cultivating the strength of will to overcome them A person feels morally free who feels that his habits or his temptations are not his masters, but he theirs: who even in yielding to them knows that he could resist; that were he desirous of altogether throwing them off, there would not be required for that purpose a stronger desire than he knows himself to be capable of feeling…. we must feel that our wish, if not strong enough to alter our character, is strong enough to conquer our character when the two are brought into conflict in any particular case of conduct. And hence it is said with truth, that none but a person of confirmed virtue is completely free. (8:841) The person of confirmed virtue is the person who can conquer desires, when there is reason to do so, by a virtuous habit of willing. One must be careful not to assume that Mill, because he is an empiricist, is in the Humean tradition which holds reason to be a slave of the passions. We have already emphasized the difference between Hume’s scepticism about both theoretical and practical reason, and the naturalistic epistemological stance taken in different ways by Reid and Mill. In the practical as in the theoretical case Mill is best understood not as denying the existence of categorical reasons, in the sceptical style of Hume, but as quietly naturalizing them. He takes it that talk of principles of theoretical and practical reason, and of our recognition of such principles, is justified. But the fact remains that he does not explain how it is justified, if nothing is a priori. He does not dramatize the issue, as Kant’s Critical philosophy does. And he does not confront the Critical questions: what is it for a reason to exist, what is it to grasp a reason, how can reason be efficacious? But, on his own showing these questions must be answered, in a fashion compatible with naturalism, if we are to make sense of ordinary categories—reasoning, inferring, deliberating, deciding—categories which involve thinking of agents and reasoners as free followers of rationally given norms. For while inference does seem to be a causal process it yet also appears to be something more than, or incommensurable with, a causal process. It seems to involve the acausal consciousness of a rule of reason. Precisely the same can be said for the rationalizing relation between motive or deliberation, and free action or choice. NATURALIZED EPISTEMOLOGY This Kantian argument, that naturalism cannot account for the rationality of experience, thought and action, was taken up by Mill’s idealist successors in Britain, led by T.H.Green. They elevated Hume, who was held to have perceived it, over Mill, who was felt to have hidden his face from it, or been unable to grasp it. And certainly Mill offers no thoroughgoing examination of what, on his own philosophy, the status of fundamental norms of reason is—how they can have objective authority. He merely takes it that they do. The same applies to Reid. Like Reid, Mill assumes that any cognitive disposition which is ‘ultimate’, original or spontaneous thereby underpins an objective norm. We have seen this in his treatment of beliefs based on memory, as well as beliefs based on induction, and the same applies to his well-known derivation of what is desirable from what is desired ‘in theory and in practice’. The similarity between Reid and Mill, on this fundamental point of naturalistic epistemological method—the appeal to spontaneous dispositions which survive critical reflection—is easy to miss. It is obscured by the undeniably important disagreement between them in what they actually place on their respective lists of fundamental principles, as also by the intrusion of the controversy between innatist and associationist psychology. But terms like ‘ultimate’, ‘spontaneous’, ‘natural’, etc. need not mean ‘innate’—one must not confuse a phenomenon which is important for epistemological method—that of finding a principle naturally obvious— with a particular psychological explanation of its origin. The divergence between Hume’s sceptical naturalism, together with its Kantian and idealist sequels, and the naturalistic epistemology taken for granted by Reid, Mill and others became a great divide in nineteenth-century philosophy. The most influential philosophers in the heyday of the analytic movement in the twentieth century stand also in the Critical rather than the naturalistic epistemological tradition. More recently, however, that has changed, largely through the influence of Quine. ‘Naturalized epistemology’ is once more influential. The questions concerning it remain the same. Today, as in Mill’s time, one can ask whether there is any route open to the naturalist, between Humean scepticism and Kantian idealism. If there is, it requires a sharper distinction than Mill, or indeed Quine, makes between norms and facts. Mill argued soundly from the naturalistic premiss, that no factual statement can be a priori. But the same is not true of normative statements. For all that grounds the objectivity of norms is reflective equilibrium and convergence—as indeed Mill’s epistemological method implies. Having recognized this crucial point, one may innocently concede that statements about fundamental norms are—in a way Mill himself could have accepted, that is, a way which requires no transcendental or platonic mystery—a priori. Norms constitute the concepts which order our thought. Rationality, grasp of concepts, consists in sensitivity to them; it cannot therefore belong in the realm of the factual, any more than the norms themselves do. If this conception of concept and meaning can be made out, we can endorse Mill’s naturalistic view of man. BIBLIOGRAPHY Citations of passages from Mill are by volume and page number of the Collected Works of John Stuart Mill, ed. J.M.Robson (London and Toronto: University of Toronto Press and Routledge & Kegan Paul, 1963–). The following volumes have been cited: 4.1 VII, VIII, A System of Logic, Ratiocinative and Inductive: Being a Connected View of the Principles of Evidence and the Methods of Scientific Investigation, textual editor J.M.Robson; introduction by R.F.McRae, 1973. 4.2 IX, An Examination of Sir William Hamilton’s Philosophy and of the Principal Philosophical Questions discussed in his Writings, textual editor: J.M. Robson, introduction by A.Ryan, 1979. See also: 4.3 Ryan, A. J.S.Mill, London: Routledge and Kegan Paul, 1974. 4.4 Scarre, G. Logic and Reality in the Philosophy of John Stuart Mill, Dordrecht: Kluwer, 1989. 4.5 Skorupski, J. John Stuart Mill, London: Routledge, 1989.