No part of A Treatise of Human Nature has excited as much critical attention, or inspired as much disagreement, as Part III of Book I, in which Hume presents his accounts of causation and nondemonstrative or probable reasoning. It is the most significant and influential general discussion of metaphysical and epistemological issues in Hume’s philosophy, as well as the most extensive. It is also, in the view of many critics, among the least well understood sections of the Treatise. The influence of Hume’s discussion on subsequent developments in analytic philosophy rests on two main arguments: his argument concerning probable or causal inference, to which is often attributed the first posing of ‘Hume’s problem of induction’; and his so-called ‘regularity’ account of causation.1 These arguments are familiar to every student of philosophy. They are the acknowledged starting points of much intrinsically interesting and important philosophical discussion. Unfortunately, a good deal of this discussion, while of undoubted philosophical merit, bears little relation to the sorts of questions in which Hume is interested. This is perhaps unsurprising since the majority of critics have chosen either to ignore Hume’s sustained general discussion of the causal relation, or to treat his arguments as independent contributions to philosophical discussions which Hume did not address, and which, in some cases, he clearly wished to avoid. Hume’s account is a good deal richer than conventional interpretation would suggest. The whole should be read as a contribution to his new science of man – his attempt ‘to introduce the experimental method of reasoning into moral subjects’. In the course of his investigation he treats of a number of interconnected notions, including probable inference, causality, knowledge and necessity, presenting a number of mutually-supportive arguments, before bringing his story to a climax with his putative ‘definitions’ of causation. His arguments concerning induction or probable reasoning2 and causation emerge as part of his sustained treatment of what is, for him, the single most important natural operation of the mind: the transition made ‘from the idea or impression of one object to the idea or belief of another’ (T.92). Hume’s intention, in so far as it can be gathered from his own words, is to ‘explain fully’ the relation of cause and effect, upon which such inferences depend, and to consider the idea of causation in order to ‘see from what origin it is deriv’d’ (T.74).
Commentators have, in the main, preferred to consider Hume’s so-called ‘problem of induction’ and his self-styled ‘definitions’ of causation as independent and free-standing pieces of argumentation.3 The standard appreciation of these arguments is typified in Robert J. Fogelin’s description of them as Hume’s ‘sceptical attack on induction and his attempted regularity definition of causation’.4 I find this interpretation, at best, superficial and, in some respects, quite misleading. It has been challenged in a number of quarters.5 A growing number of philosophers recognise that it must be replaced. But, as yet, no really satisfactory alternative picture has emerged. The new view of Hume as a realist or ‘sceptical realist’ about causation and the external world has seemed an attractive prospect to many, but, to my mind, lacks the sort of textual support required of a satisfactory general interpretation.6 The absence of any consensus among scholars as to what general philosophical vision unifies Hume’s thinking on these matters has meant that the sceptical view remains the prevalent one among philosophers. As I say, I believe that this view should be replaced. But it must be replaced by an interpretation which is sensitive to all aspects of Hume’s thinking, including his scepticism. I propose what I believe is a more balanced appreciation of Hume’s arguments. By setting them firmly within the context of his general account of the causal relation, I hope to avoid the worst of the errors of interpretation noted above. In Chapter Four, I describe the interpretive impasse which has emerged in recent years among scholars concerned with Hume’s theory of causation, and suggest a way of moving beyond it. In this chapter, I focus in detail on the early sections of Part III, giving special attention to Hume’s argument concerning reasoning from experience, his so-called ‘problem of induction’.
It must, in fairness, be conceded that the conventional sceptical readings of Hume on induction and causation are not without textual support. Nevertheless, they face a number of difficulties. I will briefly note a few of them. First, the two arguments emerge as part of an extended discussion which not only presupposes, but openly endorses the practice of causal reasoning. Hume goes on to offer eight normative rules of causal reasoning, stating the conditions under which inductive inferences are warranted. These rules emerge as a result of the reflexive application of causal reasoning. Second, neither argument contains any explicit reference to scepticism. The crucial Section VI of Part III, in which Hume presents his ‘sceptical attack on induction’, in fact contains not a single direct reference to scepticism of any sort (T.88–91). The radically sceptical thesis that inductive argument is unwarranted is never presented. Hume does, indeed, discuss ‘sceptical systems’, but he does so in Part IV, and then without reference to any specific sceptical doubts concerning causal inference. The discussion of Part IV, like that of Part III, makes consistent use of obviously causal terms. These points must, at the very least, raise a prima facie difficulty for the view that Hume is a radical sceptic about causal reasoning.
There is also good textual evidence against the view that Hume adopts a regularity theory of causation. As we saw in Chapter One, Hume distinguishes between the mere redescription of regularities among events and their explanation in terms of ‘laws and forces’ (E.14). He believes there to be ‘hidden and ‘conceal’d’ forces in nature which the natural scientist might hope, ‘in some degree’, to discover. Hume’s science of man is targeted on explaining the operations of the understanding in terms of the ‘simplest and fewest’ general causes. These facts are alone enough to place the conventional view under considerable suspicion. But other problems face these readings. There is the obvious interpretive difficulty of treating as independent two arguments which are not only parts of the same extended account, but are evidently intended to be mutually-supportive and interdependent. The argument concerning probable reasoning prepares the ground for the analysis of causation which follows it. Hume’s explanation of the idea of necessary connection treats it as a product of causal inference. Necessary connection is what distinguishes causal from non-causal or accidental sequences of events. To ignore the obvious relatedness of these arguments to each other, and to the general drift of the discussion in Part III, is already to risk the ‘displacement’ of Hume’s arguments into other, philosophically unfriendly, contexts.7
In this chapter, I will argue against the widely held view that Hume is a radical sceptic about the reasonableness of induction. This view receives its most detailed recent defence in the influential writings of D.C. Stove.8 According to Stove, Hume argues for the sceptical thesis that no inductive argument increases the probability of its conclusion on the basis of the tacitly held assumption that only deductively valid arguments make their conclusions more probable. Since this assumption is false, Stove argues, the most Hume can be said to have shown is that inductive arguments do not render their conclusions as probable as deductive ones. This extremely negative evaluation of Hume’s argument has not gone unchallenged. A number of commentators now take the opposing view that the target of Hume’s negative argument concerning induction is the deductivist model of probable reasoning.9 On this view, Hume argues for the replacement of the deductivist model which Stove ascribes to him with one that allows for the full reasonableness of inductive arguments. Both interpretations share an account of the sort of reasoning in which Hume is interested in Section VI. Stove believes that Hume argues for the worthlessness of inductive argument on the basis of the assumption that the only arguments which are good or reasonable are deductively valid ones. Proponents of the non-sceptical interpretation agree that Hume’s concern is with the deductivist model of probable inference, but contend that he argues only for the relatively modest conclusion that inductive arguments, though eminently reasonable, are not deductively valid. There are good textual and historical grounds for rejecting both of these interpretations. I propose their replacement with something rather different. In my view, Hume’s concern is not at all with the normative question of the reasonableness or otherwise of inductive inference. He certainly does not argue directly for the sceptical thesis that inductive argument is unwarranted. I have already given reasons for thinking such a move unlikely. Furthermore, I do not believe that this is a conclusion which Hume is entitled to draw. It is to his credit that he does not do so. Hume is more concerned with answering an important empirical question which arises naturally within his science of man. Any interpretation of Hume’s philosophical writings must begin by acknowledging the fundamental concerns which he means his work to address. Chief among these is his concern with the origin of central human ideas and beliefs. His interest, in this case, is in the question of what determines us to form beliefs concerning unobserved objects and events (T.92). His answer need not commit him to any definite view as to the reasonableness of the inferences which take us ‘beyond the senses’ in forming such beliefs (T.74). It is still open to Hume to argue for the warrantedness of inductive argument, and this, I find, is precisely what he does.
Part III of Book I of the Treatise concerns the nature of our reasonings about matters of fact. Hume begins by distinguishing two groups of what he terms ‘philosophical relations’: those ‘such as depend entirely on the ideas’, and those ‘such as may be chang’d without any change in the ideas’ (T.69). In Part I Section V, Hume identified seven philosophical relations which, he claimed, constituted all the ways in which the mind can compare its objects. He now indicates that they can be divided up into two classes. Relations of the first sort, that is to say, relations of resemblance, contrariety, degrees in quality, and proportions in quantity or number, fall within the province either of intuition, or, in the case of quantitative or numerical relations, demonstration. In intuition, we are immediately aware of the relation between ideas; in demonstration, our awareness of a relation is mediated by a chain of reasoning. The demonstrative sciences of algebra and arithmetic are, according to Hume, the only ones in which we can ‘carry on a chain of reasoning to any degree of intricacy, and yet preserve a perfect exactness and certainty’ (T.71). All four of these relations depend solely on the comparison of ideas, and, for Hume, only those relations which depend entirely on our ideas can be ‘the objects of knowledge and certainty’ (T.70). Knowledge, according to Hume, concerns only ‘that evidence, which arises from the comparison of ideas’ (T.124). Our judgments concerning matters of fact, on the other hand, involve the tracing of the relations of identity, contiguity in time and place, and cause and effect (T.73), relations which may change without any change in ‘the ideas’, and which yield conclusions which are merely probable and do not amount to knowledge.
It is in relations of the latter sort that Hume is chiefly interested. He claims that of these three relations, the only one ‘that can be trac’d beyond our senses’ to inform us of absent or unobserved ‘existences and objects’ is causation (T.74). The relations of identity, and contiguity in time and place, are, properly speaking, a matter of ‘perception rather than reasoning’, involving ‘a mere passive admission of the impression thro’ the organs of sensation’ (T.73). In none of the judgments we make concerning them do we go beyond what is immediately present to the senses. All probable reasoning is based on the relation of cause and effect. Hume reiterates the point in the Enquiry, observing that all transitions from observed to unobserved matters of fact are ‘reasonings’ founded on the relation of cause and effect (E.26). It soon emerges that it is probable reasoning, or ‘reasoning from cause and effect’, that interests Hume. His broad purpose in the corresponding passages of the Enquiry is, he writes, ‘to enquire what is the nature of that evidence which assures us of any real existence and matter of fact, beyond the present testimony of our senses, or the records of our memory’ (ibid.). This would also be a fair characterisation of Hume’s main concern in Part III of Book I of the Treatise. In the Enquiry, Hume finds the question of the nature of causal reasoning to arise very naturally from the distinction he draws between ‘relations of ideas’ and ‘matters of fact’ (E.25–6). In the Treatise, he comes to the question in a rather more roundabout manner. Having dealt with ‘the objects of knowledge’ in the opening section, Hume devotes the remaining sections to considering, from a number of perspectives, the nature of the one relation capable of producing a connection ‘as to give us assurance from the existence of one object, that ‘twas follow’d or preceded by any other existence or action’ (T.73–4).
Section II of Part III is entitled ‘Of probability; and of the idea of cause and effect’. Hume later explains that he is using ‘probability’ in an unusually broad sense in order to include all nondemonstrative arguments in his treatment. These include not only those arguments from causes and effects which are still ‘attended with uncertainty’ (T.124), but those that amount to what Hume terms ‘proof. Probability is, therefore, inclusive of all our reasonings concerning matters of fact. Hume draws the distinction in this way in order to oppose probable belief, of all sorts, to knowledge, which is the exclusive province of intuition and the demonstrative sciences. To mark ‘the several degrees of evidence’ which attend judgments of the first sort, Hume distinguishes probability from proof. When he writes of probability in its narrower sense he has in mind judgments the evidence for which is still uncertain. Hume goes on to treat of this sort of probability in Sections XI and XII of Part III. But in its broadest sense, ‘probability’ also includes judgments which are ‘entirely free from doubt and uncertainty’, and should, strictly speaking, be regarded as proofs. It is with judgments of this sort that Hume is most concerned. Probable inferences based on an observed constant conjunction of events and a present impression of one of the conjuncts are attended by the greatest degree of certainty, and produce the strongest, or most vivacious, kind of causal belief. The liveliness or vivacity of a belief is lost through the observation of ‘mixed’ or ‘contrary experiments’, resulting in what Hume terms ‘imperfect belief (T.135). Hume appears to have seen probability in terms of a weakening or dropping-off of proof. He recognises that many of our probable beliefs are like this. Both probabilities and proofs are instances of causal or probable reasoning, broadly understood, and, as such, depend upon the relation of causation. Causation turns out to be the only relation upon which we can reason which does not depend entirely on the ideas.
Identity and contiguity do not count as reasoning at all since, in the case of these relations, both the objects are present to the mind along with the relation. Only the causal relation takes the mind beyond what is immediately present to the senses, ‘to discover the real existence or the relations of objects’ (T.73). In Section II, Hume announces his intention to consider the nature and origin of the idea of this relation. It is, Hume says, ‘impossible to reason justly, without understanding perfectly the idea concerning which we reason; and ‘tis impossible perfectly to understand any idea, without tracing it up to its origin, and examining that primary impression, from which it arises’ (T.74). Since there appears to be no quality which is common to all causally-related objects, the idea of causation must be derived from some relation among objects. Hume finds that, whatever the objects which we denominate cause and effect, they are always found to be contiguous in both time and place. Even distant objects, which appear productive of each other, are commonly presumed to be related by a chain of causes ‘contiguous among themselves’ (T.75). He also notes that there is always a ‘Priority of time in the cause before the effect’ (T.76). Hume, however, makes clear that contiguity and succession, though ‘essential’ to causality, do not provide a ‘compleat idea of causation’.10 An object may be contiguous and prior to another without being considered its cause. There is, he writes, ‘a Necessary Connexion to be taken into consideration; and that relation is of much greater importance, than any of the other two above-mention’d’ (T.77).
Hume recognises that there is a problem of distinguishing between accidental and non-accidental or causal sequences of events. A sequence of events could be related by both contiguity and succession without being considered causal. Events in a sequence might be related in both of these ways, but ‘merely coincidentally’. One way of putting this is to say that the features of contiguity and succession, though necessary to our idea of the causal relation, are not sufficient. There must, in other words, be some third feature in virtue of which we distinguish causal from accidental regularities. This feature, according to Hume, is ‘necessary connection’. We would not consider a given sequence to be causal unless we believed there to be a connection between cause and effect, such that, when the cause occurs, the effect must follow. It is the certainty of such a connection that convinces us that a given sequence is causal, and not just an accidental succession of events. We believe there to be a necessary connection between cause and effect. However, when Hume attempts to discover the origin of the idea of this relation, by examining the ‘known qualities’ of objects, he finds no relations but those of contiguity and succession. We have no impression of the necessary connection between cause and effect. Let us take as an instance of a causally related pair of events Hume’s favourite example of one billiard ball striking another. We observe the movement of the first ball, its contact with the second, and the movement of the second ball across the table. But we do not experience the necessary connection between these distinct events. All we experience is the contiguity and succession of events. We have no impression of one event happening because or as a result of another. Our idea of necessity cannot, therefore, be derived from sense experience. Since Hume has been frustrated in his direct search for an impression of necessity, he is compelled, he says, to ‘beat about all the neighbouring fields, without any certain view or design’, in the hope that an examination of some other questions ‘will perhaps afford a hint, that may serve to clear up the present difficulty’ (T.78).
He sets out to consider two ‘neighbouring’ questions of significance: ‘For what reason we pronounce it necessary, that every thing whose existence has a beginning, shou’d also have a cause?’ and ‘Why we conclude, that such particular causes must necessarily have such particular effects; and what is the nature of that inference we draw from the one to the other, and of the belief we repose in it?’ (ibid.). Hume pursues these questions throughout the next fourteen sections of the Treatise. They mark an important change of tack, away from the search for an original impression of necessary connection in sense experience, and towards an examination of the nature of the inferences we make from cause and effect and, crucially, of what determines us to make them. Hume can find no single impression of sensation capable of providing the sort of grounding or ‘just foundation’ which an impression of necessity might be thought to provide for belief about the unobserved (T.90). The earlier part of his analysis established that sense experience was unable to reveal all that is essential to causality. In the absence of an appropriate impression of sensation, the attempt to explain the idea of necessity is, for the time being, set aside. Hume turns his mind to the question of how it is that the causal relation takes us ‘beyond our senses’ to conclusions about what is not currently observed. This marks the start of the remarkable causal investigation of probable argument, during which Hume shows how, contrary to initial expectation, the idea of necessity arises from the inferences we make, and not the other way around. It is with these issues in mind that Hume turns to the first of his two questions.
In Section III, Hume argues that we can never demonstrate the necessity of a cause without ‘shewing at the same time the impossibility there is, that any thing can ever begin to exist without some productive principle’. This latter proposition, he says, is ‘utterly incapable of a demonstrative proof (T.79). Since all distinct ideas are separable from each other, and the ideas of cause and effect are ‘evidently distinct’, the separation ‘of the idea of a cause from that of a beginning of existence, is plainly possible for the imagination (T.79–80). The possibility cannot be refuted ‘by any reasoning from mere ideas’ (T.80). It cannot therefore be demonstrated that it is ‘absolutely’ impossible for something to begin to exist without a cause. Hume draws the general conclusion that it is not ‘from knowledge or any scientific reasoning’ that ‘we derive the opinion of the necessity of a cause’ (T.82). In talking of ‘scientific reasoning’, Hume has in mind demonstration which, together with intuition, provides ‘the foundation of science’ (T.73). The opinion can thus arise only from ‘observation and experience’. Questions could certainly be raised about the adequacy of Hume’s perfunctory treatment of this argument.11 But Hume is more concerned to underline the latest shift in gear which he now feels entitled to make. It is, as I have said, the inference from cause to effect and the nature of its evidence that interests him. The foundation of all our reasonings concerning the causal relation is experience. We must therefore ask ‘how experience gives rise to such a principle?’ or, since, Hume tells us, the same answer will serve in both cases, ‘Why we conclude, that such particular causes must necessarily have such particular effects, and why we form an inference from one to another?’.
Hume begins Section IV by reminding his readers that although the mind, in reasoning from causes or effects, is carried beyond what is immediately present to the senses, ‘it must never lose sight of them entirely, nor reason merely upon its own ideas, without some mixture of impressions, or at least of ideas of the memory, which are equivalent to impressions’ (T.82). His concern now is with the important task of explaining how, by what, and under what conditions, the mind is determined to pass from the idea or impression of one object to the idea or belief of another (T.92). He turns first to the conditions under which we make such inferences. All reasonings from experience involve not only an idea or belief of something not immediately present to the senses, but an ‘immediate perception of our memory or senses’. When I have an impression of fire, I immediately infer the presence of heat. Although I do not have a present impression of heat, I have had previous impressions of heat and of its conjunction with fire. When I have the impression of fire, my mind forms the idea of its usual conjunct. The impression provides the basis for the inference. Any ‘chain of argument or connexion of causes and effects’ must be founded either on the authority of the memory or senses or on the testimony or impressions of others, themselves being ‘founded on those characters or letters, which are seen or remember’d’ (T.83). Without this foundation there would be neither belief nor evidence. A belief concerning something which is not present to the senses must, therefore, be related to a present impression or idea of memory, which, Hume says, is ‘equivalent to impressions’. However far our inferences take us from what is currently observed, they must start with some object or event which is either seen or remembered (T.82).
Hume is interested in the foundation of the transition the mind makes from observed to unobserved matters of fact. In Section V he considers what he calls ‘the first act of the judgment’ which, he thinks, provides the basis for causal reasoning (T.86). All reasonings concerning causes and effects involve both a present impression of the senses or idea of memory and ‘an idea of that existence, which produces the object of the impression, or is produc’d by it’ (T.84). The belief or assent which, Hume says, always attends the memory and senses ‘is nothing but the vivacity of those perceptions they present; and this alone distinguishes them from the imagination’ (T.86). Hume is preparing the ground for the account of causal inferences that follows in Section VI, and for the conclusion that inductive inference is determined not by reason, ‘but by certain principles, which associate together the ideas of these objects, and unite them in the imagination’ (T.92). The vivacity, or ‘force and liveliness’, of our ‘original’ perceptions constitutes the first act of judgment and ‘lays the foundation of that reasoning, which we build upon it, when we trace the relation of cause and effect’ (T.86). In reasoning concerning causes and effects vivacity is transmitted from the original impression to the idea or belief which is the conclusion of the inference. A belief is nothing more than a lively or vivacious idea, associated with a present impression (T.96). Hume highlights another important factor in augmenting the vivacity of perceptions and, so, in inducing belief. An idea of the imagination may acquire, by repetition alone, ‘such a force and vivacity, as to pass for an idea of the memory’ and have ‘the same influence on the mind as nature, and infixing the idea with equal force and vigour’ (T.86). The basis of any causal inference is an impression or idea of memory. We cannot, however, infer the existence of an unobserved object or event on the basis of the impression alone (T.86–7). We must also have had some experience of the repeated past conjunction of two objects, on which basis, Hume writes, ‘we always draw an inference from one object to another’ (T.88).
The stage is now set for Hume to turn to the important question of the nature of the transition the mind makes from an original impression or idea of memory to the idea of a connected cause or effect. This is the topic of the famous Section VI, ‘Of the inference from the impression to the idea’. Hume concluded Section V by claiming that it is the ‘force and liveliness’ of the original perception that accounts for its foundational role in all our reasonings from cause and effect. This is an important conclusion within Hume’s theory of belief and within the causal story he is telling. He next asks what it is that causes the transition the mind makes from impression to idea. His interest, as I said before, is in inferences ‘which are deriv’d from the relation of cause and effect, and which are entirely free from doubt and uncertainty’ (T.124). His first move is to deny that we make such inferences on the basis of an inspection of the qualities of ‘particular objects, and from such a penetration into their essences as may discover the dependency of the one upon the other’ (T.86). Hume uses the same argumentative strategy he used in arguing against the possibility of demonstrating that every event must have a cause. There is no object, Hume writes, ‘which implies the existence of any other if we consider these objects in themselves, and never look beyond the ideas which we form of them’ (T.86–7). Such an inference would amount to knowledge, the province of demonstration, and would imply ‘the absolute contradiction and impossibility of conceiving any thing different’ (T.87). However, since all ideas are clearly distinct and, therefore, separable, it is ‘evident there can be no impossibility of that kind’. The inference from present impression to unobserved object or event cannot be demonstrative or prior to the observed conjunction of the two events. It is perfectly possible for us to conceive, without self-contradiction, of an event having some cause or effect other than the one it has. When we pass from an impression to the idea of any object, ‘we might possibly have separated the idea from the impression, and have substituted any other idea in its room’ (ibid.). Hume’s point is not merely that one is conceivable without the other, but that nothing in the one implies or gives us grounds to believe in the existence of the other if we consider ‘the objects in themselves’.12 This is made clearer in the Abstract, where Hume writes that it ‘is not anything that reason sees in the cause, which makes us infer the effect’. Inferences from experience are not, in other words, based on the perception of the ‘hidden’ or ‘secret’ powers of objects. An inference of this sort would amount to a demonstration, but, since the mind ‘can always conceive any effect to follow from any cause’, it is clear that no such demonstration is possible (T.650). It is only ‘by EXPERIENCE’ that ‘we can infer the existence of one object from that of another’ (T.87).
Hume is interested in the nature of this experience. What are the material conditions under which we form inferences from the perceived to the unperceived? Hume argues that it is only after we have experienced the constant conjunction of objects that we infer one from the other (T.88). We remember that we have experienced frequent instances of the existence of one species of object, and also remember ‘that the individuals of another species of objects have always attended them, and have existed in a regular order of contiguity and succession with regard to them’ (T.87). When we remember the species of object which we call flame, for example, we naturally call to mind the ‘species of sensation’ which we call heat. We also call to mind their constant conjunction in all past instances: ‘Without any further ceremony, we call the one cause and the other effect, and infer the existence of the one from that of the other’ (ibid.). A third relation is now added to the relations of contiguity and succession which Hume discovered in Section II: ‘This relation is their CONSTANT CONJUNCTION. Contiguity and succession are not sufficient to make us pronounce any two objects to be cause and effect, unless we perceive, that these two relations are preserv’d in several instances’ (ibid.). Hume believes that he has discovered the source of the idea of necessity ‘which makes so essential a part’ of our idea of the causal relation. His earlier strategy of abandoning the search for an impression of necessity in experience is to be fully vindicated. We may now, he says, ‘see the advantage of quitting the direct survey of this relation’ since there are hopes that ‘by this means we may at last arrive at our propos’d end’ (ibid.).
Hume is quick to temper his optimism with a note of caution. The ‘new-discover’d’ relation of constant conjunction appears, at first glance, ‘to advance us but very little in our way’, for it seems evident that the ‘mere repetition’ of instances can ‘only multiply, but not enlarge the objects of our mind’ and so can never produce ‘any new original idea, such as that of a necessary connexion’ (T.88). How can the mere repetition of instances produce in us any new idea other than that of repetition itself? Hume has no immediate answer. He guilefully suggests that it might, in the end, appear ‘that the necessary connexion depends on the inference, instead of the inference’s depending on the necessary connexion’ (ibid.). His real interest, at this juncture, lies in explaining how an experience of the constant conjunction of two sorts of thing causes us, on observing an instance of the one sort, to form an idea or belief concerning the existence of an instance of the other. He has already established that the inference must be founded on experience. He next asks: ‘Whether experience produces the idea by means of the understanding or by the imagination; whether we are determin’d by reason to make the transition, or by a certain association and relation of perceptions’ (T.88–9). If reason determined us, Hume says, it would proceed upon the principle ‘that instances, of which we have had no experience, must resemble those, of which we have had experience, and that the course of nature continues always uniformly the same’ (T.89). Hume asks whether this proposition is founded on arguments derived from either knowledge or probability. It is easily seen that there can be no demonstrative or knowledge-yielding arguments to prove the principle since it is possible at least to ‘conceive a change in the course of nature; which sufficiently proves, that such a change is not absolutely impossible. To form a clear idea of any thing is an undeniable argument for its possibility, and is alone a refutation of any pretended demonstration against it’ (ibid.). Could probable argument be the source of the principle? Hume reminds us that probability, ‘as it discovers not the relations of ideas, consider’d as such, but only those of objects, must in some respects be founded on the impressions of our memory and senses, and in some respects on our ideas’ (ibid.). Were there no such mixture the inference would be merely conjectural. The relation of cause and effect is the only one which can take us beyond the impressions of the senses and memory. The idea of cause and effect can be derived only from experience, ‘which informs us, that such particular objects, in all past instances, have been constantly conjoin’d with each other: And as an object similar to one of these is suppos’d to be immediately present in its impression, we thence presume on the existence of one similar to its usual attendant’ (T.90). Probability is, therefore, on this account, ‘founded on the presumption of a resemblance betwixt those objects, of which we have had experience, and those, of which we have had none; and therefore ‘tis impossible the presumption can arise from probability’ (ibid.). The same principle cannot ‘be both the cause and effect of another’. It cannot, in other words, be established on the basis of an argument whose acceptability depends upon its presumption.
Hume shows that the uniformity principle can be founded neither on demonstrative nor probable arguments. Not only does reason fail us in the discovery of ultimate connections between causes and effects, but ‘even after experience has inform’d us of their constant conjunction, ‘tis impossible for us to satisfy ourselves by our reason, why we shou’d extend that experience beyond those particular instances, which have fallen under our observation. We suppose, but are never able to prove, that there must be a resemblance betwixt those objects, of which we have had experience, and those which lie beyond the reach of our discovery’ (T.91–2). Since probability is itself founded on this presupposition, any attempt to prove it by probable argument ‘must be evidently going in a circle, and taking that for granted, which is the very point in question’ (E.35–6). Hume turns to the other faculty of mind which he recognises as a source of ideas and belief, the imagination. He has already ‘taken notice of certain relations, which make us pass from one object to another, even tho’ there be no reason to determine us to that transition’ (T.92). We may, he says, establish it as a general rule that whenever the mind uniformly makes a transition without any reason, it is influenced by these relations. This is the case in the present instance. Reason cannot show us the connection between one object and another, ‘tho’ aided by experience, and the observation of their constant conjunction in all past instances’. We do, nevertheless, form beliefs about unobserved events in just these circumstances. Hume draws the conclusion that whenever the mind ‘passes from the idea or impression of one object to the idea or belief of another, it is not determin’d by reason, but by certain principles, which associate together the ideas of these objects, and unite them in the imagination’ (ibid.).
Hume presents versions of this argument in the Abstract and in Section IV of his Enquiry concerning Human Understanding. The fuller and, in some ways, clearer version of the Enquiry nevertheless preserves much the same structure and draws the same apparently sceptical conclusion. A good deal, however, is left out, and it is only in the Treatise that Hume presents the argument within the context of his general discussion of the causal relation. I have attempted to show how much richer the version of the Treatise is for the nourishment it receives from the other parts of Hume’s treatment of probability. In the Abstract, Hume gives the following account of his theme:
The celebrated Monsieur Leibnitz has observed it to be a defect in the common systems of logic, that they are very copious when they explain the operations of the understanding in the forming of demonstrations, but are too concise when they treat of probabilities, and those other measures of evidence on which life and action intirely depend, and which are our guides even in most of our philosophical speculations. In this censure, he comprehends the essay on human understanding, le rechereche de la verite, and l’art de penser. The author of the treatise of human nature seems to have been sensible of this defect in these philosophers, and has endeavoured, as much as he can, to supply it. (T.646–7)
Hume rightly thought his treatment of probability and his argumert concerning induction or probable reasoning to be among the outstanding and most original achievements of his philosophy. The conclusion of that argument was that we are determined, not by reason, but by certain principles of the imagination, to pass from an impression of one object to the idea or belief of another. Custom, and not reason, is, according to Hume, ‘the guide of life’ (T.652). The importance of this conclusion to Hume’s philosophy is unquestioned, but, beyond this, there is little critical agreement as to how his argument is to be understood. In this section and the next, I will critically examine a number of the most influential readings. I begin with the conventional interpretation that Hume presents a radically sceptical evaluation of the reasonableness of inductive argument.
On this view, Hume is supposed to have argued for the unwarrantedness or epistemic worthlessness of inductive argument. In doing so, we are told, he became the first philosopher to pose what has become known as the problem of induction. Stroud defends one such view. He describes Hume’s rejection of ‘reason’ or ‘the understanding’ as the source of causal inference as ‘his most famous sceptical result’. Hume rejects reason on the grounds that no inductive inference is ‘reasonable or rationally justifiable’, and, in doing so, ‘condemns as unjustifiable a whole mode of inference or pattern of reasoning’.13 Even after the experience of the constant conjunction of two sorts of events, and the observation of one of the conjuncts, it is no more reasonable to believe that the second conjunct will occur than not. The premises of inductive arguments do not, in other words, render their conclusions more reasonable. The most influential, and, to my mind, best-worked out argument for this view is presented by D.C. Stove. Stove argues that Hume ‘presupposes an exclusively deductive ideal of reason’ which leads him to reject as unreasonable or unwarranted any argument that is not deductively valid. On the basis of the tacitly held assumption that only deductively valid arguments increase the probability of their conclusions, Hume is supposed to have argued for the highly sceptical thesis that no inductive argument makes its conclusion more probable. It is to this account that I turn first.
Stove reads Hume as accepting the false thesis that only deductively valid arguments are reasonable or make their conclusions more probable. He illicitly assumes the truth of this thesis in arguing for the negative conclusion that premises concerning observed objects are ‘irrelevant’ to the probability of conclusions concerning unobserved objects.14 Stove’s argument is complex and detailed, but its main elements are easily summarised. His treatment focuses on the nature of the presupposition of uniformity upon which, Hume argues, all probability is founded. Sometimes, Stove writes, ‘when we say of an argument from p to q, that it presupposes r, our meaning is as follows: that, as it stands, the argument from p to q is not valid, and that, in order to turn it into a valid argument, it would be necessary to add to its premisses the proposition r’. When Hume suggests that all probable arguments ‘presuppose’, or are ‘founded on the supposition’, that unobserved instances will resemble observed ones, it is in this sense, Stove believes, that the term ‘presuppose’ is used. Inductive arguments presuppose the uniformity principle in the sense that this principle would be required to turn a invalid inductive inference into a valid deductive one. Stove invites us to consider a predictive inductive inference, typical of the sort Hume is interested in, from the premises ‘This is a flame, and all of the many flames observed in the past have been hot’ to the conclusion This is hot’.15 The argument, Stove writes, ‘is invalid as it stands. Nor could it be turned into a valid argument without the addition of some further premiss which will have the effect of saying that (at least in respect of heat) flames yet unobserved resemble observed flames’. The addition of the uniformity principle is needed if this argument is to be ‘turned into a valid one’.16 What Hume wants to show, according to Stove, is that the premises of inductive arguments do not make their conclusions more probable. But all he has shown, at least on the basis of those premises which he makes explicit, is that inductive arguments never produce conclusions which are deductively valid: what Stove calls ‘inductive fallibilism’. Hume draws his unwarranted sceptical conclusion on the basis of a suppressed premise that only deductively valid arguments lend support to their conclusions. Since this premise is false, all Hume has shown is that inductive arguments do not render their conclusions as probable as deductive ones.
What Hume means to say, according to Stove, is that inductive arguments are invalid as they stand, and that ‘it would be necessary, in order to turn them into valid arguments, to add to their premisses the Resemblance Thesis’ or uniformity principle.17 Hume is a ‘deductivist’ in that he tacitly assumes that only deductively valid arguments are reasonable or warranted. This is the only mode of inference in which Hume is interested. Stove draws attention to the critical section of Hume’s argument. Hume argues that if reason determined us to make the transition in question it would proceed upon the principle or supposition that nature is uniform (T.89). Stove’s argument is plausible because it appears to make sense of Hume’s use of this principle. His idea is that Hume thinks of the uniformity principle as a ‘middle term’ without which the inference would be deductively invalid and, so, on the basis of the tacit premise, unwarranted. Hume is supposed to have established the unreasonableness of inductive argument by arguing that the uniformity principle can arise from neither probable nor demonstrative reasoning. This makes sense if we assume, with Stove, that Hume is presupposing a deductivist model of probable argument. Without this additional premise, inductive arguments remain invalid, and, thus, unreasonable. The principle is needed to turn a deductively invalid argument into a valid one.
Stove’s interpretation has been influential. There are, nevertheless, good textual grounds for resisting his conclusion, a number of them, I think, quite decisive. Stove goes on to argue that Hume’s crucial deductivist premise is false, and that his conclusion is, therefore, unwarranted. I do not take issue with this part of Stove’s argument. I am more interested in the allegation that Hume is committed to deductivism and, on those grounds, to the highly sceptical conclusion that all inductive arguments are unreasonable. If it can be shown that Hume did not hold the critical premise that only deductively valid arguments make their conclusions reasonable then Stove’s case will have been effectively shortcircuited. What is the textual and historical evidence for thinking that Hume was committed to deductivism? Stove rests his claim on the sense he assigns to ‘presuppose’ in his translation of the crucial passages of Section VI (T.89–90). It has seemed obvious to many critics, Stove among them, that all inductive inferences are invalid as they stand and that something like the uniformity principle would be needed to make them valid. It would, therefore, seem likely that when Hume writes that if reason determined us it would proceed upon the basis of the uniformity principle, he means that the principle would be required to make probable arguments valid. Hume would, therefore, appear committed to the view that only deductively valid arguments are reasonable. This, however, is to take far too restrictive a view of the possible senses in which the principle could be presupposed in probable argument.18 Stove is on relatively solid ground when it comes to recasting the principle as a ‘medium’ or ‘middle term’ in an inference. In the first Enquiry, Hume writes that probable argument requires ‘a medium, which may enable the mind to draw such an inference, if indeed it be drawn by reasoning and argument’ (E.34). Where I believe Stove goes wrong is in assuming, without argument, that in talking about a medium or middle term Hume must have in mind an ‘exclusively deductive’ form of inference.
Hume acknowledges Locke as the source of the distinction between demonstrative and probable argument drawn first in the Treatise and reiterated in the Enquiry (E.56). There is, therefore, good reason to suppose that an understanding of Locke’s views on probable argument will shed some light on Hume’s own. What we find in Locke is not only a toleration of middle terms in probable argument, but an explicit statement of the role they play. Locke writes that probability ‘is nothing but the appearance of such an Agreement, or Disagreement, by the intervention of Proofs, whose connexion is not constant or immutable, or at least is not perceived to be so, but is, or appears for the most part to be so, and is enough to induce the Mind to judge the proposition to be true, or false, rather than the contrary’.19 It differs from knowledge in terms not of the ‘intervention’, or otherwise, of ‘proofs’, but of the nature and strength of the connexion between the proofs.20 A demonstration shows the agreement or disagreement of ideas by the intervention of proofs which have ‘a constant, immutable, and visible connexion with one another’.21 It is the strength of the relation between the proofs, or intermediate ideas, that gives a demonstrative argument its force. Proofs, according to Locke, are merely those ‘intervening Ideas, which serve to shew the Agreement of any two others’.22 He also writes of these proofs as ‘mediums’, remarking that in a demonstration there must be a remembrance of ‘the Intuition of the Agreement of the Medium, or intermediate Idea, with that we compared it with before’.23 In demonstration, the relations between the ideas which intervene in showing the agreement of any two others are intuitive. Locke writes that there are also ideas ‘whose Agreement, or Disagreement, can no otherwise be judged of, but by the intervention of others, which have not a certain Agreement with the Extremes, but a usual or likely one: And in these it is, that the judgment is properly exercised, which is the acquiescing of the Mind, that any Ideas do agree, by comparing them with such probable Mediums’.24
Far from taking mediums or middle terms to be restricted to use in demonstrative arguments, Locke accords them an explicit role to play in probable argument. In any probable chain of reasoning the intervening ideas or mediums are related not intuitively, but only ‘for the most part’.25 Hume’s explicit endorsement of the distinction Locke draws between probable and demonstrative argument, gives us some reason for resisting Stove’s assumption.26 But it is not decisive. We have shown only that Stove is wrong simply to assume that in talking of ‘mediums’ or ‘middle terms’ Hume must have had in mind an ‘exclusively deductive’ mode of inference. If we are to show decisively that the crucial deductivist premise is one that Hume did not and could not have held we must look in more detail at the text.
Hume is supposed to have recognised only one form of inference. There are strong textual grounds for rejecting this view. It is evident from the early sections of Part III that Hume was concerned to distinguish demonstrative and nondemonstrative of probable modes of reasoning. Stove could hardly deny that Hume identifies both probable and demonstrative forms of inference. His response is to suggest that Hume’s concern in making the distinction is not with the ‘degree of conclusiveness’ of the arguments, but with ‘the epistemological character of the premises’.27 The kind of interest which Hume displays in his lengthy discussions of probability in Sections XI to XIII of Part III is dismissed by Stove as ‘an empirical, psychological interest, rather than a logico-philosophical and evaluative one’.28 The distinction he draws is ‘a material and descriptive one’ concerned with distinguishing those arguments the premises of which are necessarily true, and can be known independently of experience, from those the premises of which are contingent. He does not recognise a second form of inference the conclusions of which differ in their degree of conclusiveness. The model of inference he has in mind is that of a deductively valid argument with contingent premises. Hume’s considered position remains, for Stove, that there are no inductive arguments which lend something less than full support to their conclusions.
Much of the evidence for holding that Hume distinguishes two modes of inference seems equivocal in the light of Stove’s rejoinder. There, remains, nevertheless, evidence enough to show that Hume did not always talk of ‘probability’ or ‘probable argument’ in the sense suggested by Stove. Hume’s argument concerning ‘scepticism with regard to reason’ is based on the supposition that a set of premises can lend a conclusion a degree of support or probability varying between 0 and 1 (T. 180–3). Hume shows how the assurance attending the conclusion of an extended chain of reasoning diminishes in proportion to the number of successive ‘new examinations’ that attend it, until, at last, there remains ‘nothing of the original probability’ (T. 182). Let us suppose that we have formed an original judgment to which we assign the probability 1. Now suppose that we consider the probability of error involved in our calculation: ‘we are oblig’d by our reason to add a new doubt deriv’d from the possibility of error in the estimation of the truth and fidelity of our faculties’ (ibid.). But this new estimation is itself founded on probability and so weakens further the evidence, the next estimation of which is weakened by another doubt of the same kind, and so on in infinitum. No matter what the strength of our original conviction ‘it must infallibly perish by passing thro’ so many new examinations, of which each diminishes somewhat of the force and vigour’ (T. 182–3). Hume elsewhere treats probability in terms of a diminution or dropping-off of proof (T.135). Stove unsatisfactorily dismisses the sections on probability as ‘an inessential part of Hume’s philosophy of induction’.29 Nevertheless, they suggest quite strongly that Hume considered the premises of some arguments to lend a merely probable degree of support to their conclusions. The sections should be read as attempts to explain how probable arguments can supply differing degrees of warrant for their conclusions.
A number of other passages support the view that Hume did not subscribe to the extreme form of inductive scepticism Stove attributes to him. Hume writes in the Enquiry that a wise man ‘proportions his belief to the evidence … He considers which side is supported by the greater number of experiments: to that side he inclines, with doubt and hesitation; and when at last he fixes his judgment, the evidence exceeds not what we properly call probability’ (E. 110–1). It is ‘directly contrary to the rules of just reasoning’ to give assent to the weaker evidence, in preference to the stronger (E.109). These passages, drawn from Hume’s famous Section X on miracles, obviously presuppose that the force of probable reasonings can vary. The ‘contrariety of evidence’ may ‘diminish or destroy the force of any argument, derived from human testimony’ (E.l 12–3). It is perfectly clear, from these and other passages, that Hume believes that some arguments lend a probable degree of support to their conclusions. In contrast with demonstrative arguments, which are either ‘irresistible’, or have ‘no manner of force’, probable arguments can and do vary in force, as ‘difficulties can take place, and one argument counter-bal lance another, and diminish its authority’ (T.31). It would require a significant distortion of the text to read Hume as saying anything else.
The weight of textual evidence against Stove’s position now looks impressive. The difficulties faced by his reading emerge clearly from an examination of the structure of Hume’s argument.30 Hume provides a summary of the argument in the Abstract. His first move is to argue that since it is not anything ‘that reason sees in the cause, which makes us infer the effect’, knowledge of causal relations can be gained only through experience. An a priori inference would amount to a demonstration, but, since ‘the mind can always conceive any effect to follow from any cause’ and ‘whatever we conceive is possible’, it is evident that no demonstration from cause to effect is possible. Whenever a demonstration takes place, ‘the contrary is impossible, and implies a contradiction’ (T.650–1). But in the case of inductive inference, it is never absolutely impossible for an argument to have true premises but a false conclusion. The striking thing about this argument is that Hume might easily have used it, at this early stage, to conclude, on the basis of the deductivist assumption, that all inductive arguments are unreasonable: the sceptical thesis Stove presents as the over-all conclusion of Hume’s argument. Stove argues that Hume presupposes ‘an exclusively deductive ideal of reason’. If this were so, then Hume could straightforwardly derive the sceptical conclusion that we are not determined by reason to make inferences from observed to unobserved, on the basis of the conceivability of an inductive inference having true premises but a false conclusion. He does not do so. Hume, instead, argues that the determination of inductive inference by reason requires that the uniformity principle be supported by argument, and considers the different sorts of argument from which this supposition could be derived. There are, of course, two possible sources of support for the principle that nature is uniform, including, Hume argues, probable argument. This strategy would be remarkable in a philosopher who subscribed to the view that only deductively valid arguments are good ones. Yet Hume not only insists on drawing the distinction between probable and demonstrative arguments, but goes on to argue that we cannot prove ‘by any probable arguments, that the future must be conformable to the past’ (T.651).
So far, I have argued that Stove’s position lacks both textual and historical support. It cannot, however, be dismissed on these grounds alone. Stove makes the stronger claim that Hume’s argument concerning induction is intelligible only on the basis of his subscribing to the false thesis that deductively valid arguments alone increase the probability of their conclusions. Hume’s claim that all probable arguments must proceed on the basis of the uniformity principle appears to makes sense if we think of the principle as an additional premise, a ‘middle term’, necessary to turn a deductively invalid argument into a valid one. There are, once again, good grounds for resisting Stove’s conclusion. In the first place, it is clearly not the case that the uniformity principle is necessary to turn an invalid inductive argument into a valid one. As Stroud points out, there are indefinitely many ways of adding premises to make ‘a previously invalid argument deductively valid’, of which the simplest would be to add the conclusion to the premises.31 Stroud complains that, on this understanding of ‘presuppose’, Hume’s requirement is a purely logical one which could be satisfied trivially without reference to the basis upon which one could know or have reason to believe the conclusion. It is not clear that Stroud is right to separate the question of logical validity from the ‘epistemic’ notion of certainty in this way.32 The conclusion of a deductively valid argument can be regarded as certain relative to the truth of its premises. Nevertheless, once we acknowledge Hume to be making what is, at least in part, an epistemic, rather than a purely logical point, the question of what assumptions about inference are being made by Hume seems more open.33 Hume’s statement of the principle in the Treatise seems strong enough to support the view that Hume is interested in what could turn a good inductive argument into a good deductive one. Hume writes that if reason determined us to make the inference, ‘it would proceed upon that principle, that instances, of which we have had no experience, must resemble those, of which we have had experience’ (T.89). Taken literally, however, such a principle would transform not only good inductive inferences into deductively valid ones, but bad inductive inferences too. Clearly, some other interpretation is desirable. But can we account for the epistemic character of the supposition without characterising its role in purely logical terms? Why, in other words, should Hume think that probable arguments presuppose the uniformity principle unless he is operating with a deductivist model of inference? I have already given grounds for rejecting Stove’s interpretation. It remains to be seen whether or not a workable alternative picture is available.
A promising approach has been sketched by Stroud. He answers the question of what Hume means by saying that inferences from the observed to the unobserved are ‘founded on the supposition’ of the uniformity principle as follows: ‘To say that an inference is “founded” on a particular supposition is to say at least that no one will be justified in inferring the conclusion from the premisses unless he is also justified in believing the supposition on which the inference is “founded”.’34 This initial formulation is broad enough to accommodate a number of different views about the character of the inference, and the nature of the justification involved. Why might Hume think that inductive inferences are founded on the uniformity principle in the above sense? One possibility, we have seen, is that Hume thinks that inferences from the observed to the unobserved are invalid without it. If Hume believes that the uniformity principle is needed in order to turn an inductive argument into a deductively valid one, then he must also believe that the conclusion of an inference is only reasonable or justified if it is logically implied by its premises. If this is all Hume means then, as Stroud points out, he can only be said to have shown that ‘if no one is ever justified in believing a proposition unless he is justified in believing something that logically implies it, then no one is ever justified in believing anything about the unobserved’, and this falls significantly short of the sceptical conclusion Stroud and others attribute to him.35 It is difficult to see how Hume could have thought this argument to have the sceptical import widely claimed for it. As we have seen, there are other reasons for rejecting Stove’s interpretation which do not involve ascribing to Hume the radically sceptical conclusion that inductive arguments are unreasonable or rationally unjustifiable. It is, therefore, highly desirable that we replace it with some more balanced and sympathetic account.
Stroud argues that Hume need not have in mind an exclusively deductive model of inference. Hume exploits another aspect of the ‘traditional conception of reason’ which allows him to form a truly sceptical conclusion without making the unwarranted assumption that all reasons must be ‘deductively sufficient’.36 Let us suppose, with Stroud, that someone who has experienced a constant conjunction of events and is presently observing one of the events forms the belief that its usual conjunct will now occur. This person might believe for a bad reason, ‘completely unconnected with his past and present experience’, or have made a lucky guess: ‘So something else must be true of him as well. It would seem that, if he is to be reasonable in believing that a B will occur, he must somehow take his past and present experience with respect to As and Bs as a good reason to believe that a B will occur.’37 But even if he does believe that what he has experienced gives him good grounds for believing that the second event will occur, and even if this last belief is true, it still does not follow that the man’s belief is reasonable or rationally justified.38 For the belief that event B will occur to be reasonable, the agent must also reasonably believe that what he has experienced provides grounds for inferring event B.39 If he cannot show that he has good reasons for this belief, then he cannot, on this understanding of reasonableness, be said to have made a rationally justified inference from observed to unobserved at all. This ‘self-conscious’ and ‘potentially regressive’ aspect of the notion of reason may well, Stroud thinks, be ‘what Hume is focussing on in the traditional conception’.40
By focusing on this aspect of reasonableness, Hume could support his claim that reasonable belief about the unobserved requires something more than past observation of constant conjunctions of events and a present impression of one of those events, without committing himself to the deductivism attributed him by Stove. A ‘fully rational agent’, on this view, is ‘not one who proceeds rationally only at the last step’.41 The conception of reason Stroud ascribes to Hume involves the self-consciousness that one has proceeded reasonably at every step of the argument. Without this awareness the idea or belief which is the conclusion of a causal inference arises not from reason but as a result of the operation of some other faculty of mind. Hume can draw his sceptical conclusion concerning belief in unobserved matters of fact without assuming that the only good reasons are deductively sufficient ones. His conception of reason requires that everything to which we appeal in argument itself be shown to be reasonable. For reasonable belief in B, it is not enough to have experienced a constant conjunction of As and Bs, and to have a present impression of an A – one must also reasonably believe that one’s past and present experiences make it reasonable to infer the existence of B. This, however, is something in which one could never reasonably believe. Any support for the principle that constant conjunctions will continue uniformly in future would have to be derived from experience. But any inference from experience is reasonable only on the basis of one having reason to believe that past experiences provide reason to infer conclusions about future and unobserved events. Any attempt to derive the uniformity principle by probable argument would, therefore, take for granted the very point at issue. Since there is no reason to believe that observed instances provide grounds for belief about the unobserved, there can be no reasonable inferences from the observed to the unobserved.
One obvious interpretive difficulty, which Stroud acknowledges, concerns the construction he places on Hume’s claim that inference from observed to unobserved matters of fact is ‘founded on the supposition’ of the truth of the uniformity principle. Stroud characterises the presupposition as the belief that what is and has been observed is good reason for belief in the unobserved.42 This, however, is not directly equivalent to what Hume says is required for reasonable belief in these circumstances. What is needed, according to Hume, is a principle stating that ‘those instances of which we have had no experience, must resemble those, of which we have had experience’ (T.89). The strength of this claim might be thought to give some credence to the deductivist interpretation. Hume, however, gives a less strident formulation in the Enquiry. Here he characterises the principle as the belief ‘that the future will be conformable to the past’ (E.35), or the past be a ‘rule for the future’ (E.38). This, Stroud argues, ‘comes close to the claim that one must reasonably believe that what is and has been observed can be relied on as a guide to the future’.43 Stroud’s claim could, indeed, be read as a loose paraphrase of what Hume has in mind when he writes that instances of which we have experience must resemble those of which we have no experience: ‘To say that the murderer must have only four toes on the left foot is to indicate that what you already know is good or conclusive reason to believe that about the murderer, and not just that he does have only four toes on the left foot.’44 No one who has observed a constant conjunction of A and B, and is presently observing A, will reasonably believe that B will occur unless he also reasonably believes that what he has experienced is ‘a good or conclusive’ reason for believing B. Stroud’s interpretation is preferable to Stove’s not only in avoiding some of the structural anomalies we found in Stove’s deductivist translation of Hume’s argument, but also in giving a plausible account of the different ways in which Hume puts the claim that inference from the observed to the unobserved is ‘founded’ on the uniformity principle.
Stroud is not, of course, the only philosopher to reject the key assumptions Stove makes with respect to the target of Hume’s argument. A number of critics have rejected both the claim that Hume subscribed to the view that only deductively valid reason are good ones, and the thesis that Hume took a radically sceptical view of induction.45 These authors take the opposing view that the target of Hume’s argument concerning induction is the deductivist model of probable argument. While they agree with Stove as to the mode of inference in which Hume is interested, they argue that Hume wants to replace the deductivist model of inference with one which allows for the reasonableness of inductive argument. Far from being a deductivist, or subscribing to a thesis remotely like the one Stove describes, Hume is concerned with the refutation of deductivism. Beauchamp and Rosenberg argue that Hume should be read as arguing for what both they and Stove term ‘inductive fallibilism’, and that this position ‘is not, as some have alleged, trivial’.46 In their view, “’Hume’s scepticism concerning rationalism” is a measured and proper antidote to the excess of that philosophical view’. Hume’s ‘larger purpose’ is to attack the ‘rationalistic’ conception of reason. According to Scott Arnold, the whole point of Hume’s discussion of whether the uniformity principle can be established by argument ‘is to see whether invalid predictive-inductive inferences can be “cured” of their invalidity by being transformed into acceptable deductively valid counterpart inferences’.47 Hume’s ‘sceptical fire’ was directed ‘at those who would claim more for causal (inductive) reasoning than it could deliver’.48 ‘Inductive fallibilism’, according to Scott Arnold and Beauchamp and Rosenberg, alone describes the sceptical position attributable to Hume. The problem of induction ‘is simply not to be found in Hume’s philosophy’.49 It is to this position that I now turn.
I noted at the outset that the conventional view of Hume as a radical sceptic about induction faced a number of serious interpretive difficulties. The justification of many of the claims of the Treatise depends on some inductive inferences being warranted. The ‘experimental method of reasoning’ which Hume proposes introducing into the moral sciences is inductive. He could not, therefore, call into question the reasonableness of inductive reasoning without also impugning the method of the Treatise. Hume not only makes inductive inferences, but formulates a set of rules by which to regulate the judgment in causal reasoning. These rules state the conditions under which inductive inferences are justified. He contrasts the ‘philosophical’ nature of these reasonings with those bom of ‘superstition’ and ‘prejudice’. Not only does Hume allow that there are arguments which give less than full support to their conclusions, but he also endorses those beliefs which are proportioned to the evidence. In the Enquiry, he attacks the reliance on testimony in religious belief, arguing that the evidence resulting from testimony ‘admits of a diminution, greater or less, in proportion as the fact is more or less usual’ (E.113). On occasion, Hume makes his endorsement of the inductive method explicit. A man who concludes another person to be near him ‘when he hears an articulate voice in the dark, reasons justly and naturally; tho’ that conclusion be derived from nothing but custom’ (T.225). The inference from effect to cause ‘is not only a true species of reasoning, but the strongest of all the others’ (T.97n). This evidence is hard to reconcile with the claim that Hume was a radical sceptic about induction.
Indeed, Hume does not present the conclusion of Section VI as sceptical and shows little sign of recognising the destructive implications widely attributed to it. His intentions in Part III do not seem to be sceptical at all. The upshot of his general discussion of the causal relation is not a general pessimism about the prospects for inductive reasoning, but a set of normative rules by which to judge of causes and effects. When Hume does invoke the spectre of scepticism, it is not to endorse the beliefs of ‘that fantastic sect’ (T.183), but to counsel a ‘true’ or ‘mitigated’ form of scepticism, giving rise not to the radical rejection of our natural ‘rational’ practices, but to a due ‘diffidence’, in both our doubts and our convictions (T.273). He acknowledges that his philosophy contains a strong sceptical dimension, but there is no suggestion, either in Section VI or elsewhere, that this scepticism is such as to issue in a radically sceptical evaluation of probable argument. I will say more about the general nature of Hume’s scepticism in later chapters. My focus in this section will be non-sceptical or anti-rationalist readings of Hume’s argument concerning induction. The above points must, at the very least, raise a prima facie difficulty for the conventional sceptical interpretation of Hume’s conclusion that inductive inferences are not determined by reason.
All anti-rationalist readings trade, in one way or another, on the distinction Hume is supposed to have drawn between a rationalist conception of reasoning, marked by logical necessity and comparable to the deductive model of inference Stove ascribes to him, and reasoning understood more broadly to include all sorts of probable argument. According to Beauchamp and Rosenberg, Hume’s argument concerning induction uses ‘reason’ in the former ‘stipulatively restricted sense’.50 Hume shows ‘first that demonstrative reasoning does not yield factual results and, second, that induction is not marked by the logical necessity attending demonstrative reasoning’.51 Hume’s argument is aptly characterised as ‘a frontal assault on rationalist assumptions that at least some inductive arguments are demonstrative’.52 He shows that inductive reasoning can provide neither ‘self-evident certainty’, nor the logical necessity uniquely characteristic of demonstrative reasoning.53 The rationalist model of inference must, therefore, be replaced by a model which allows for the reasonableness of inductive inference. The full force of Hume’s scepticism is directed at the rationalist model of reasoning which, he argues, is unable to account for the reasonableness of induction.
There is some textual support for the view that Hume, on occasion, insisted upon a distinction between different conceptions of reasoning. When he writes of reasoning as ‘a kind of cause, of which truth is the natural effect’ he means the faculty of mind responsible for the discernment, by intuition or demonstration, of relations of ideas (T.180). It is equally clear that Hume uses ‘reasoning’ in a much broader way to include any sort of transition or inference to a vivacious idea or belief. It is in this sense that Hume writes of the ‘reasoning faculty of brutes’ and animals, as well as of human creatures. The actions of an animal in avoiding fire or shunning strangers ‘proceed from a reasoning, that is not in itself different, nor founded on different principles, from that which appears in human nature’ (T. 176–7). This usage accords with that of those philosophers ‘who have divided human reason into knowledge and probability’ (T.124). Despite his inclusive use of ‘reasoning’ in these and other places in Part III, where Hume ‘directly discusses the nature of induction’, he is supposed to have restricted reason to ‘a priori reason’ or ‘the discernment of ideas and their relations (i.e., to deductive reasoning and intuitive derivation of nonsynthetic a priori propositions)’.54 The target of Hume’s negative argument concerning induction is not, it is suggested, the institution of induction itself, but the narrow rationalistic conception of probable argument.
Even if we allow that Hume used ‘reason’ in more than one distinct sense, what is the evidence for believing that Hume used ‘reason’ in the restrictive rationalistic sense in the crucial passages of Section VI? I will consider three main objections to this suggestion, each of which has been used to discredit the anti-rationalist view, and which, together, constitute decisive evidence against it.55 One difficulty for the view is that when Hume is not discussing induction directly, he either uses ‘reason’ in the broad sense to include both demonstrative and probable argument, or makes it explicit that he is doing otherwise.56 For example, in introducing his discussion of scepticism with regard to reason, Hume makes clear that it is with reason as it regards the ‘demonstrative sciences’ that he is concerned (T.180). The same procedure is in evidence when Hume announces his intention to use ‘probability’ in a narrower-than-usual sense in Section XI of Part III. Hume warns his readers that he will be using the term in a special sense, noting that ‘in the precedent part of this discourse’ he has followed the ‘method of expression’ of those philosophers who, with Locke, ‘have divided human reason into knowledge and probability’ (T.124). He gives no such warning in the crucial Section VI.
Hume makes a point of telling us when he is deviating from the usage of philosophers, such as Locke, who divide reason into knowledge and probability. It would be reasonable, therefore, to expect that when Hume does not give word to the contrary, he is following the Lockean usage which he recognises as standard. I noted earlier that there was good reason to suppose that an understanding of Locke’s views on demonstration and probable argument would shed light on Hume’s own. Hume acknowledges Locke as the source of the distinction he draws between probability and demonstration. The faculty of reason, according to Locke, operates with both knowledge and probability. It is responsible for probable and demonstrative inference, and consists, according to Locke, ‘in nothing but the Perception of the connexion there is between the Ideas, in each step of the deduction, whereby the Mind comes to see, either the certain Agreement or Disagreement of any two Ideas, as in Demonstration, in which it arrives at Knowledge; or in their probable connexion, on which it gives or with-holds its Assent, as in Opinion’.57 In both these cases, ‘the Faculty which finds out the Means, and rightly applies them to discover Certainty in the one, and Probability in the other, is that which we call Reason’.58
Reason, according to Locke, is the faculty of mind responsible for the perception of the connections between intermediate ideas ‘in each step’ of a probable or demonstrative argument. The mind, at times, ‘cannot so bring its Ideas together, as by their immediate Comparison, and as it were Juxtaposition, or application one to another, to perceive their Agreement or Disagreement’.59 In such cases, ‘it is fain, by the Intervention of other Ideas (one or more, as it happens) to discover the Agreement or Disagreement, which it searches; and this is that which we call Reasoning’.60 Demonstrative reasoning involves a chain of argument ‘in every step’ of which ‘there is an intuitive knowledge of the Agreement or Disagreement, it seeks, with the next intermediate Idea, which it uses as a Proof.61 Probable reasoning is the tracing of a chain of ideas, or proofs, the connections between which hold only ‘for the most part’. Reasoning, then, is a chain of ideas the first idea of which is related ‘in every step’, either intuitively or ‘for the most part’, to the last. If Hume were using ‘reason’ in the restrictive sense stipulated by Beauchamp and Rosenberg then this would mark a significant departure from ordinary usage, as Hume understood it. Hume, however, gives us no clue that this is what he has in mind in this context. In Section XI, he tells us that he has followed the common ‘method of expression’ in ‘the precedent part’ of the Treatise, and this, of course, includes his famous argument concerning induction.
A second difficulty for the anti-rationalist interpretation concerns the way in which Hume structures his argument concerning induction. This, as we saw earlier, suggests very strongly that Hume recognised two sorts of inference, both of which he considered to be possible sources of the uniformity principle. Hume proposes to consider ‘all the arguments’ upon which the uniformity principle could be founded. There are, of course, only two possibilities: demonstration and probable argument. Since neither form of argument can establish that there is a ‘conformity betwixt the future and the past’, and since all inductive inference presupposes this principle, Hume concludes that the mind is not ‘determin’d by reason’ in making these inferences. As I noted, had Hume been concerned purely with reason as a faculty of demonstrative argument, he could have concluded, on the basis of the conceivability of an inductive inference having true premises and a false conclusion, that this sort of inference is not one that reason determines us to make. In demonstrative argument, the contrary is impossible and implies a contradiction. Hume would have no need to argue that the uniformity principle can be supported by neither demonstrative nor inductive argument. Yet this is his strategy in the second part of his argument. Even if Hume had been successful in showing that the uniformity principle could be given a probable form of support, this would be of no relevance to the claim that ‘at least some inductive arguments are demonstrative’ – the possibility he is supposed to be arguing against. Probable arguments produce a lesser ‘degree of evidence’ than demonstrative ones (T.124) and so can provide neither the self-evident certainty nor the logical necessity that Beauchamp and Rosenberg say ‘uniquely characterize’ demonstrative reasoning. The anti-rationalist view is unable to account for either Hume’s insistence on considering both demonstrative and probable argument as the source of the principle, or the presence of an argument to the conclusion that the uniformity principle cannot, in Hume’s words, ‘arise from probability’ (T.90). This is strong evidence for the view that Hume’s target is not the rationalist conception of probable argument.
A third and, I think, decisive objection concerns the terms in which Hume presents the conclusion of his argument. Hume argues that ‘even after the observation of the frequent or constant conjunction of objects, we have no reason to draw any inference concerning any object beyond those of which we have had experience’ (T.139). In the Abstract, he concludes that ‘[w]e can give no reason for extending to the future our experience in the past, but are entirely determined by custom, when we conceive an effect to follow from its usual cause’ (T.654). Hume’s conclusion is that we can have no reason whatsoever for believing that events of which we have had experience will resemble those of which we have had none. This is a much stronger conclusion than the one attributed to him by Beauchamp and Rosenberg. In their view, Hume concludes merely that ‘there can be no demonstrative arguments to prove, that those instances, of which we have had no experience, resemble those, of which we have had experience’.62 This modest conclusion, which Hume in fact draws part-way through the second part of his argument, and then only in reiteration of his ‘foregoing method of reasoning’, could hardly be said to warrant Hume’s own assessment of the originality of his conclusion. Having restated the conclusions that there is nothing in any object which affords us a reason to draw any conclusion beyond it, and that even after the observation of a constant conjunction of events, we have no reason to draw any inference beyond experience, Hume writes: ‘I say, let men be once fully convinc’d of these two principles, and this will throw them so loose from all common systems, that they will make no difficulty of receiving any, which may appear the most extraordinary’(T. 139). The strength of this claim would scarcely be credible in a philosopher who had done no more than argue for a form of inductive fallibilism.63
Hume is arguing for a conclusion radical enough to throw men ‘loose from all common systems’. This conclusion is not directed at a narrow rationalistic conception of probable argument. Hume argues against the claim that we can have demonstrative knowledge of causal relations at the first stage of his argument. What bears the brunt of Hume’s attack is not the rationalistic view, but the prevailing Lockean view of reason as a faculty of intellectual ‘light’ or perception, a guide for conduct and a source of ‘certain definite principles of action’.64 In Locke’s view, probability, as well as knowledge, is a product of the faculty of reason. Reason, in other words, is what determines us to make the transition from a past conjunction of events and a present impression of one of those events to the idea of its regular conjunct. According to Locke, ‘what our own and other Men’s constant Observation has found always to be after the same manner’ provides us ‘with reason’ to conclude them to be ‘the Effects of steady and regular Causes’. Finding the proposition ‘That Fire warmed a Man’ to be ‘agreeable to our constant experience’, we are ‘put past doubt, that a relation affirming any such thing to have been, or any predication that it will happen again in the same manner, is very true’.65 This is the view that Hume rejects. If reason determined us, after the constant observation of the conjunction of fire and warmth, to infer one from the other, it would do so on the basis of the supposition of the uniformity principle. But the proposition that nature is uniform cannot be derived from reason and, so, the transition from the impression of fire to the idea of warmth is not one that the faculty of reason could determine us to make. When the mind passes from the impression of one object to the idea or belief of another, ‘it is not determin’d by reason, but by certain principles, which associate together the ideas of these objects, and unite them in the imagination’ (T.92). It is custom, rather than reason, which supplies the missing step in the chain of argument. In the Enquiry, Hume writes that ‘even after we have experience of the operations of cause and effect, our conclusions from that experience are not founded on reasoning, or any process of the understanding’ (E.32). A few pages later he adds that ‘it is not reasoning which engages us to suppose the past resembling the future, and to expect similar effects from causes which are, to appearance, similar’ (E.39). It is not, in other words, ‘by any process of argument or ratiocination’ that we come to believe that instances of which we have had experience will resemble those of which we have had none. All inferences from experience involve a step ‘taken by the mind which is not supported by any argument or process of the understanding’ (E.41). Reasoning, for Locke, involves the tracing of a chain of ideas the first idea of which is related ‘in every step’ to the last. All probable argument, however, depends upon the supposition of uniformity, and, as Hume shows, there is no argument, either demonstrative or probable, from which that supposition could be derived. We have, therefore, ‘no reason to draw any inference concerning any object beyond those of which we have had experience’ (T.139). Hume concludes that it is not ‘reason, which is the guide of life, but custom’ (T.652).
Any probable argument involves a step which proceeds not from argument or ratiocination, but from custom and instinct. We are ‘determined by Custom alone to suppose the future conformable to the past’ (T.652). It cannot, therefore, be reason that determines us to make the transition from present impression to idea or belief. Hume, like Locke, considers reason to be the faculty of mind responsible for the formation of arguments, demonstrative and probable. Locke believed that in probable argument reason ‘perceives the probable connexion of all the Ideas or Proofs one to another, in every step of a Discourse, to which it will think Assent due’.66 Hume shows that in any reasoning from experience there is an intermediate step taken by the mind which ‘is not supported by any argument’. This step is supplied not by reason but by a strong natural instinct. Reason cannot, therefore, perceive the connection of all the ideas, in every step of a probable argument. This is a conclusion of the first importance in Hume’s philosophy. It marks his rejection of the Lockean view of reason as the dominant guide to human conduct and belief. Reason is neither dominant, nor autonomous, as Locke supposed. Custom alone determines the mind to suppose the future conformable to the past and, however easy this step may appear, ‘reason would never, to all eternity, be able to make it’ (T.652). All our reasonings concerning causes and effects ‘are deriv’d from nothing but custom’ (T.183).
Hume arrives at the position known as ‘Hume’s scepticism about induction’. It is clear, however, that Hume did not consider the upshot of his argument to be scepticism. The reason for this is that Hume has no intention of calling into question the epistemic warrant of induction, at least, not in Section VI of Part III. In fact, Hume’s plan for the section is clearly set out and carefully prosecuted. His concern is with the straightforward empirical question of what causes or ‘determines’ us to make inductive inferences, and to form beliefs about the unobserved (T.92). He finds that it is not the faculty of reason, but custom or the imagination that is ultimately responsible for our adoption of induction. All inferences from experience involve a mediate step which no argument can support. Since Hume is working with a conception of reason which requires the self-consciousness that one has proceeded reasonably at every step of a chain of argument, he can conclude that it is not reason that causes us to make inductive inferences. Hume’s answer need not commit him to any position as to the reasonableness of inductive inference. Indeed, it is quite clear that Hume does not assimilate the question of the reasonableness of induction to the question of whether or not induction is a product of the faculty of reason. He does not suggest that inductive arguments are reasonable or justified only if determined by reason, rather than by custom or instinct. The important evaluative question of the justification of induction goes unasked, at least until Part IV, of Book I.
Hume’s conclusion concerns the causes of inductive inference rather than the question of its justification. Should we then conclude, with Beauchamp and Rosenberg, that the problem of induction is not to be found in Hume’s philosophy? It is clear that the problem Hume is addressing in Section VI is not the problem of the justification of induction. It does not follow from the fact of our inductive inferences being determined by ‘custom or a certain instinct of our nature’ (E.159) that we are unjustified in making them. Nevertheless, many subsequent attempts to show that our reliance on induction can be justified without presupposing that reliance are clear responses to Hume’s argument. His argument quite obviously militates against the possibility of producing a supporting argument for induction which does not itself rely on the practice. Hume, however, avoids suggesting that the reasonableness of induction depends on its justification by argument. Indeed, he is at pains to stress that although probable arguments are grounded in the imagination, rather than in reason, they are not to be regarded in the same way as the prejudices and superstitions which are the result of the weak and irregular operations of the imagination (T.225). He is aware of the normative question of the justification of induction, though he raises it only in Part IV of Book I. His argument concerns the apparent opposition between the conclusions formed from cause and effect and ‘those that persuade us of the continu’d and independent existence of body’ (T.231). When we reason causally, we are led to conclude, on the basis of arguments showing the mind-dependency of both primary and secondary qualities, that ‘neither colour, sound, taste, nor smell have a continu’d and independent existence’, and, thus, that ‘there remains nothing in the universe, which has such existence’ (ibid.). There is, therefore, a ‘direct and total opposition’ between reason and the senses. Although both operations are ‘equally natural and necessary in the human mind … in some circumstances they are directly contrary’ (T.266).
It is clear that Hume believes that the opposition between reason and the senses calls into question the epistemic value of inductive inference. He complains that the ‘intense view of these manifold contradictions and imperfections in human reason has so wrought upon me, and heated my brain, that I am ready to reject all belief and reasoning, and can look upon no opinion even as more probable or likely than another’ (T.268–9). Scepticism arises as a result of the direct opposition he claims to have discovered among the principles of mind which he describes as ‘natural and necessary’ (T.266). Induction is justified because it is a part of a framework of ‘permanent, irresistable, and universal’ human beliefs and practices (T.225). The principle of ‘the customary transition from causes to effects, and from effects to causes’ is among those ‘general and more establish’d properties of the imagination’ (T.267) without which ‘human nature must immediately perish and go to ruin’ (T.225). These principles are ‘the foundation of all our thoughts and actions’. Since they are universal, those philosophies which uphold principles opposed to ‘the other principles of custom and reasoning’ may ‘easily be subverted by a due contrast and opposition’ (ibid.). Such philosophies involve their proponents in selfcontradiction and, for that reason, are ‘observ’d only to take place in weak minds’ (ibid.). We are justified in adopting the principles of inductive reasoning because they are unavoidable, necessary and ‘useful in the conduct of life’. Hume’s scepticism concerning induction emerges only when this tidy framework of irresistible natural beliefs is itself threatened by the ‘direct and total opposition’ of two ‘equally natural and necessary’ operations of the human mind (T.266).
It is not possible ‘for us to reason justly and regularly from causes and effects, and at the same time believe the continu’d existence of matter’ (ibid.). Although the belief in the continued existence of matter is not among the ‘general and more establish’d’ principles of the imagination, being merely a ‘fiction’ or ‘illusion’ (T.217), it is nonetheless a necessary and indispensable part of out intellectual lives. Causal reasoning itself presupposes that objects in the physical world are distinct and lasting, and must, therefore, undermine itself in undermining the belief upon which it depends. One possible response to this opposition, which Hume considers, is to prefer neither, but successively assent to both, but if we do this then, Hume asks, ‘with what confidence can we afterwards usurp that glorious title [of philosopher], when we knowingly embrace a manifest contradiction?’ (T.266). If we are ignorant of this ‘deficiency in our ideas’ in ‘common life’ it is only because we are subject to ‘an illusion of the imagination’ (T.267). But to what extent ought we to yield to the illusions of the imagination? If we assent to ‘every trivial suggestion of the fancy’, then we will fall into such errors and absurdities that ‘we must at last become asham’d of our credulity’. If, on the other hand, we adhere only to ‘the general and more establish’d properties of the imagination’, the consequences are just as fatal, for, as Hume showed in Section I of Part IV, ‘the understanding, when it acts alone, and according to its most general principles, entirely subverts itself, and leaves not the lowest degree of evidence in any proposition, either in philosophy or common life’ (T.267–8). Should we then ‘establish it for a general maxim, that no refin’d or elaborate reasoning is ever to be receiv’d’? Hume thinks not, for to do so would be to ‘cut off entirely all science and philosophy’ (T.268). It would also involve us in self-contradiction, for ‘this maxim must be built on the preceding reasoning, which will be allow’d to be sufficiently refin’d and metaphysical’ (ibid.). To reject the notion of normativity in reasoning would, Hume contends, be self-contradictory, since the sceptical philosopher must himself rely on rules of reasoning in reaching his conclusions. Hume is still, therefore, prepared to give philosophy ‘the preference to superstition of every kind of denomination’ (T.271). But the philosopher who emerges from these sceptical enquiries is a philosopher of a different sort to the one who first set out on them. In recognising that he can save himself from ‘total scepticism’ only by means of a ‘singular and seemingly trivial property of the fancy’ (T.268), the true philosopher becomes ‘diffident of his philosophical doubts, as well as of his philosophical convictions’ (T.273). He no longer indulges reason in an unrestrained way, but tempers his scepticism with ‘a due deference to the public’ (T.274). This is not to say that the search for endorsable rules is abandoned. Our philosopher will look to endorse those rules which, like Hume’s rules of causal reasoning, survive the test of evaluation from their own point of view. Hume’s eight rules constitute a programme for enquiry to which the philosopher can self-consistently assent.67 These rules are natural habits of mind which endure their own reflexive examination in order to become rules. Hume recognises that any defence of inductive reasoning must, ultimately, involve inductive reasoning. Although we are initially justified in adopting induction because it is natural and irresistible, and because we know of nothing comparable, the justification is strengthened when the rules of inductive reasoning are found to survive critical survey by their own standards. Nevertheless, with this newly-acquired self-consciousness comes the awareness that these principles are themselves dependent upon certain of the more trivial suggestions of the fancy. The true philosopher is fully conscious of the limits of his own reason, and its dependence upon the fictions of the imagination. Hume arrives at a sort of meta-scepticism which involves a due diffidence concerning one’s own sceptical conclusions. This new way of doing philosophy is recommended because it can present only ‘mild and moderate sentiments’ (T.272). By weighing up our doubts as carefully as our convictions, we might still hope to ‘establish a system or set of opinions, which if not true (for that, perhaps, is too much to be hop’d for) might at least be satisfactory to the human mind, and might stand the test of the most critical examination’ (ibid.).
1In Chapter Four, I offer a detailed summary and critique of recent debate concerning Hume’s position as to the existence of causal connections ‘in the objects’. I suggest that the proposed replacement of what is sometimes termed the ‘standard’ sceptical view of Hume as a regularity theorist on causation with a realist or ‘sceptical realist’ view is unsatisfactory. The most detailed such view is perhaps that presented by Galen Strawson, The Secret Connexion: Causation, Realism, and David Hume, Oxford, Clarendon Press, 1989. His work builds significantly on that of John P. Wright, in his The Sceptical Realism of David Hume, Manchester, Manchester University Press, 1983.
2It should be noted that Hume does not describe inferences of this sort as ‘inductive’. He uses the term twice in the Treatise (T.27, T.628), and then, seemingly, to denote inference in general, rather than causal or probable inference in particular. Nevertheless, since Hume wrote, it has become usual to reserve the term for inferences of this sort, and, for the most part, I have been prepared to follow this pattern of usage. I will, however, note one peculiarity of Hume’s construal of arguments of this type. Hume does not construe inductive arguments primarily as inferences from particular samples to general or universal conclusions. His concern appears to be with predictive-inductive inferences, which is to say, with inferences to singular conclusions, ie., from ‘All As have been Bs’ to ‘This A is a B’. It is, however, obvious that Hume intends his conclusions also to apply to inductive inferences with general conclusions.
3See, for example, J.A. Robinson, ‘Hume’s Two Definitions of “Cause”’, in Philosophical Quarterly, XII, 1962; F.L. Will, ‘Will the Future be Like the Past?’, in Critical Assessments, London, Routledge and Kegan Paul, 1995, ed. Stanley Tweyman, vol. II, pp.3–17; and W.D. Oliver, ‘A Re-examination of the Problem of Induction’, ibid., pp. 18–28. Barry Stroud, op.cit., treats Hume’s account of inference from cause and effect, in Chapter Four of his book, as the negative phase of his story, and his account of causation, in Chapter Five, as the positive phase. P.J.R. Millican, ‘Hume’s Argument Concerning Induction: Structure and Interpretation’, Critical Assessments, vol. II, pp.91–146; and Robert J. Fogelin, Hume’s Scepticism in the Treatise of Human Nature, London, Routledge and Kegan Paul, 1985, are among those commentators sensitive to the structure of Hume’s argument as a whole.
4Robert J. Fogelin, op.cit., p.38. Proponents of the ‘standard’ or conventional sceptical view of Hume on causation and induction include John Passmore, op.cit.; Karl Popper, Logic of Scientific Discovery, New York: Basic Books, 1959, esp. p.369; Wesley Salmon, The Foundations of Scientific Inference, Pittsburgh, University of Pittsburgh Press, 1967; and D.C. Stove, Probability and Hume’s Inductive Scepticism, Oxford, Oxford University Press, 1973.
5The case against has been presented by, among others, Tom L. Beauchamp and Alexander Rosenberg, Hume and the Problem of Causation, Oxford, Oxford University Press,1981; N. Scott Arnold, ‘Hume’s Scepticism about Inductive Inference’, Journal of the History of Philosophy, vol. 21, no. 1, 1983, pp.31–55; and Janet Broughton, ‘Hume’s Skepticism about Causal Inferences’, Pacific Philosophical Quarterly, 1983, vol. 64, no. 1, pp.2–18.
6See, for example, Edward Craig, The Mind of God and the Works of Man, Oxford, Clarendon Press,1987, pp.69–130; Galen Strawson, op.cit.; and John P. Wright, op.cit. I discuss these accounts in Chapter Four.
7Robert J. Fogelin, op.cit., p.38.
8See Stove, op.cit.
9Leading proponents of this view include Beauchamp and Rosenberg, op.cit., and N. Scott Arnold, op.cit.
10According to Stroud, op.cit., pp.43–4, Hume must do more if he is to show that contiguity and succession are necessary conditions of two things being causally related to each other. Hume claims that ‘nothing can operate in a time and place, which is ever so little remov’d from those of its existence’ (T.75). Although distant objects may ‘sometimes seem productive of each other’ we still presume that there is a chain of causes ‘contiguous among themselves’. The problem for Hume is that if he is to show how we get the idea that all causally related objects are contiguous in time and place, then he must show that we always get an impression of contiguity when we identify a causal relation among objects. The question of what Hume presumes to be the case is not really relevant. A similar problem emerges in Hume’s treatment of succession. We do not always get an impression of the temporal priority of a cause over an effect. When we identify a causal relation between two moving billiard balls, for example, we ‘do not actually see the contact of two billiard balls to be slightly earlier than the beginning of the motion of the second ball’ (Stroud, op.cit., p.44).
11See, for example, Stroud, op.cit., p.47.
12Stove, op.cit., p.32, makes Hume’s stress on the conceivability of alternatives take all the weight of the argument. This overlooks Hume’s belief that an a priori inference from cause to effect would require the discernment of some connection between cause and effect such that one would imply the other.
13Stroud, op.cit., pp.52–3. According to Stroud, this argument plays an important part in what he characterises as the ‘negative phase’ of Hume’s philosophy. It was, he says, put forward for the ‘definite, positive purpose’ of advancing his naturalistic study of the principles of human nature. The need for such a study is seen ‘more clearly after the traditional theory of reason or belief hasbeenexploded’.
14Stove, op.cit., p.59.
15Ibid., p.43.
16Ibid. See also Stove, ‘Hume, Probability and Induction’ in Hume, ed. V.C. Chappell, New York, Doubleday, 1966, p.203.
17Stove, op.cit., 1973, p.44.
18A number of critics take this view. See, for example, P.J.R. Millican, op.cit., pp. 103–109; and Stroud, op.cit., p.256.
19John Locke, op.cit., IV.XV.l.
20I am indebted here to P.J.R. Millican, op.cit., p. 103–8.
21Locke, op.cit., IV.XV.l.
22Ibid., IV.II.3.
23Ibid., IV.XVII.15.
24Ibid., IV.XVII.16.
25Ibid., IV.XV.l.
26Millican, op.cit., p. 106, argues that Hume, like Locke, ‘is quite untainted by the now apparently common but always gratuitous assumption that only a demonstrative argument can contain a “middle term”’. Direct textual support for this view is thin on the ground, but Millican points out a passage from the Dialogues which lends some support to his view. Demea refers to the ‘mediums’ which the ‘accurate’ Newtonian philosopher Cleanthes uses in presenting his argument from design for the existence of God: ‘…still less can I approve of the mediums by which you endeavour to establish it. What! No demonstration of the being of a God! No abstract arguments! No proofs a priori!…. Can we reach no farther in this subject than experience and probability?’. See Hume, op.cit., 1990, p.54.
27Stove, op.cit., 1966, p. 198.
28Stove, op.cit., 1973, p. 120.
29Ibid.
30Millican, op.cit., p. 123, presents an outline of this argument which shows up the weaknesses in Stove’s structure-diagram of Hume’s argument.
31Stroud, op.cit., pp.255–6.
32See N. Scott Arnold, op.cit., p.35.
33Stroud, op.cit., p.256.
34Ibid., p.56.
35Ibid., pp.56–7.
36Ibid., p.60.
37Ibid., pp.60–1.
38Ibid., p.61.
39Ibid.
40Ibid., p.62.
41Ibid.
42Ibid, p.63. See also N. Scott Arnold, op.cit., p.36.
43Stroud, op.cit., p.63.
44Ibid.
45See Beauchamp and Rosenberg, op.cit.; N. Scott Arnold, op.cit.; Janet Broughton, op.cit.; and Beauchamp and Mappes, ‘Is Hume Really a Sceptic about Induction?’, American Philosophical Quarterly, vol. 12, no. 2, pp. 119–129.
46Beauchamp and Rosenberg, op.cit., p.74.
47Scott Arnold, op. cit., p.41.
48Ibid., p.55.
49Beauchamp and Mappes, op.cit., p. 119.
50Beauchamp and Rosenberg, op.cit., p.44. See also Annette Baier, A Progress of Sentiments, Cambridge, Mass., Harvard University Press, 1994, p.60, who writes that Hume’s interest in the crucial Section VI is ‘restricted to the faculty of intellectual intuition and demonstration, that which can discern “intelligible” connections’. Baier quotes in this context the OED definition of ‘intelligible’ as: ‘Capable of being apprehended only by the understanding (not the senses)’.
51Beauchamp and Rosenberg, op.cit., p.43.
52Ibid., p.41.
53Ibid., p.37.
54Ibid., p.43.
55See, for example, P.J.R. Millican, op.cit.’, and Don Garrett, Cognition and Commitment in Hume’s Philosophy, New York, Oxford University Press, 1997, pp.84–1.
56See Garrett, op.cit., pp.84–5.
57Locke, op.cit., IV.XVII.2.
58Ibid.
59Ibid., IV.II.2.
60Ibid.
61Ibid., IV.II.7.
62Beauchamp and Rosenberg, op.cit., p.44.
63Millican, op.cit., p. 136, observes that this claim would amount to no more than a reaffirmation of the ‘Lockean orthodoxy’. He cites Locke’s remark that ‘most of the propositions we think, reason, discourse, nay act upon, are such, as we cannot have undoubted Knowledge of their Truth’, op.cit., IV.XV.2.
64John Locke, Essays on the Law of Nature, Oxford: Clarendon Press, 1954, ed. W. von Leyden, p. 111.
65Locke, op.cit., 1979, IV.XVI.6.
66Ibid., IV.XVII.2.
67Annette Baier, op.cit., pp.93–96, shows how Hume is careful to observe each of the eight rules in the sections leading up to their endorsement.