Hume, David (1711–76)

2. Hume’s account of knowledge

Knowledge and probability. In Hume’s early works, the term ‘knowledge’ is restricted to what is certain, and the term ‘probability’ (in a wide sense) is used in the Treatise for all factual beliefs which might get revised. (Later Hume is more willing to talk with the vulgar, and use ‘knowledge’ less strictly, ‘probability’ more narrowly.) Knowledge in the strictest sense is confined to current sense impressions, along with ‘intuitions’ about the relations between currently sense-perceived qualities, and certain ‘relations of ideas’, namely those that are ‘demonstrably certain’. ‘Demonstration’ shows that some claims must be affirmed if we are to avoid self-contradiction. Factual claims are never of this ‘demonstrated’ sort. They are either, if restricted to current sense certainties, known immediately, or, if they make a claim about what goes beyond the evidence of current sensation, then they will be less than certain. Some past tense factual claims will be based on memory (and we can of course misremember), but most factual claims depend upon some inference which takes us from the evidence of current sense impressions and memories to a belief about what we have not, or not yet, ourselves observed. Hume’s main concern in Book I of the Treatiseand in the Enquiry Concerning Human Understanding is to examine these latter inferences, and distinguish the relatively well-supported from the less well-supported of them. Of particular interest to him are those that involve reliance on other people’s testimony, and his famous discussion of the acceptance of reports of miracles (Enquiry Concerning Human Understanding: X) puts all the main principles of his epistemology to work. (This discussion was originally intended for inclusion in the Treatise.)

Impressions and ideas. Hume’s most general epistemological principle is that all ‘ideas’, the contents of our thought, derive from more lively ‘impressions’, the contents of our sense experience and emotional experience. This is put forward at first as an empirical thesis, but it is also used as a normative principle. Impressions are either ‘of sense’ or, like passions, ‘of reflection’. Ideas for which impression-credentials cannot be produced are dubbed ‘false’, ‘imagined’, or ‘pretended’ ideas. Hume takes this line with the idea of a vacuum, and of duration without change. We may think we have such ideas, he argues, but we cannot really have them. It is as if ideas, simply by being ideas, make some implicit claim to having impressions as their source, and, if such claims cannot be made good, they are dismissed as false pretenders. Ideas which go well beyond their impression data are termed ‘fictions’. These include the idea of the continuous existence through time of physical things we have only intermittently observed, and of the identity of our own minds, with their everchanging perceptions. In accepting such ideas we ‘disguise the variation’ and ‘remove the interruption’ (Treatise: 254) in the data on which such ideas are based. This constructive contribution of the human mind, in its interpretation of the data of its experience, was highlighted also by Immanuel Kant, who promoted Hume’s ‘fictions’ into a priori ‘categories’.

Abstract ideas. Hume follows Berkeley in taking every idea to be fully particular or ‘individual’, and to refer to something of which we have had or might have an impression. Abstractness and generality get into ideas by the use to which they are frequently put in our thinking. A particular idea becomes general, on a given occasion for a given thinker, if, for the purposes in hand, any other idea from a certain set could be substituted for it – if it is serving as a representative of that group of resembling ideas. The words of the thinker’s language label these sets (are ‘annexed’ to them), and fix the sort of ‘resemblance’ determining their membership. So if our purpose is to consider a claim about all triangles, we will begin with an ‘image’ before our minds of, say, an equilateral triangle, but we may and should substitute other ideas of other types of triangles to test the claim. As Hume puts it, it is as if the thinker had ‘a whole intellectual world of ideas’ available to them, all of them ideas of particulars, but ‘collected together’ in groups and ‘plac’d under a general term with a view to that resemblance which they bear to each other’. We show a capacity to ‘pick out such as were most proper for our purpose’ (Treatise: 24). Hume terms this instinct for relevance ‘a kind of magical faculty in the soul’. At times he speaks as if the crucial resemblances are recognized before the resembling ideas are ‘plac’d under a general term’; at other times it seems as if the prior availability of such terms is crucial to the operation of abstraction, as he analyses it. To treat language as a great aid to abstraction and generalization, rather than an indispensable precondition, would sit better with Hume’s views about the role of generalization in causal inference, since he attributes causal inference to animals who lack language. He says that ‘men…surpass animals in reasoning’, and that ‘books and conversations enlarge the sphere of one man’s experience and thought’ (Enquiries: 107).

Causal inference to new beliefs. Hume defines belief as ‘a strong and lively idea deriv’d from a present impression related to it’ (Treatise: 105). The derivation is by causal inference; the relation is causation. The inference is a move of the mind from some impression of sense (or memory) to a conclusion about what is not (or not yet) observed. The conclusion inherits liveliness, ‘vivacity’, or ‘the belief feeling’ from its impression-premise. An inference such as ‘this is bread, so it will nourish’ is never a ‘demonstration’, since there would be no logical self-contradiction in affirming the premise but denying the conclusion. Memory may assure us that in the past bread has nourished us, and current sense impressions may assure us that this seems to be bread, but what assures us that this bread will act on us as past bread has acted? We can try to convert the causal inference ‘this is bread; past bread has nourished us; so this will nourish us’ into a ‘demonstration’ by supplying an extra premise claiming that ‘instances of which we have no experience must resemble those of which we have had experience’ (Treatise: 89); but to claim that we know this, that it is certain, would be false pretence. It itself cannot be ‘demonstrated’. Moreover, if we try to make it a matter of ‘probability’, itself a conclusion of an inference whose premise is that in the pastinstances which we had not yet experienced turned out to resemble those we had already experienced, we would have an inference of the same form as the original one, from evidence about the observed to a conclusion about the unobserved. To try to support the ‘presumption’ that ‘the course of nature always continues uniformly the same’ by pointing out that it has so continued is to give an answer that ‘gives still occasion to a question of the same kind, even in infinitum’ (Treatise: 91). We cannot turn our original inference into a sound demonstration. Either we add a premise about what nature ‘always’ does, which is merely dogmatically asserted, and so get a valid demonstration one of whose premises is a mere ‘presumption’, or we add less general premises for which we can provide empirical verification, but which leave the inference non-demonstrative in form.

Hume in Enquiry Concerning Human Understanding presents this argument as if it provides ‘sceptical doubts concerning the operations of the understanding’. In the Treatise it is simply a phase of a longer positive argument to the conclusion that ‘’tis only so far as it [causation] is a natural relation, and produces an union among our ideas, that we are able to reason upon it, or draw any inference from it’ (Treatise: 94). This is the substance of what the Enquiry Concerning Human Understanding calls the ‘sceptical solution’ to the ‘sceptical doubts’. For Hume the scientist of human nature it is simply a fact about us, and a fortunate one, that like other animals we do make causal inferences; unlike them, however, we are able to realize exactly what we are doing, and to formulate rules to help us do it better, so that our predictions will be less often falsified.

The Treatise account of our belief formation through causal inference is not only explanatory but also normative, since it ends in rules for improving our inferences. Hume shows how we can see causal inferences as cases of mental association, and explains just what is special about association by ‘causation’. Originally in the Treatise (Bk I, Pt I, Sect. IV), he had simply listed causation as one of the three ‘natural relations’ or associative principles. But its effects are found to be special. Only causation, among the forms of association, can form a new belief – not merely, like the other principles of association, revive an old one, or make it more vivid. Hume undertakes to analyse the causal relation as we perceive it, so as to show how it can guide inference to new factual beliefs. He takes it to require contiguity, rather than distance, in space (one of the other principles of association), contiguity in time in the form of immediate priority in time, and some other more mysterious element that we perceive as ‘necessary connexion’.

It is not until near the end of his long account that Hume can demystify that important ingredient in our idea of cause, although ‘hints’ had been given eight sections earlier, when Hume had ‘discover’d’ a ‘new relation betwixt cause and effect, …Constant Conjunction’ (Treatise: 87; original emphasis). Although ‘from the mere repetition of any past impression [or successions of impressions?], even to infinity, there never will arise any new original idea, such as that of necessary connexion’, he had written, ‘it wou’d be folly to despair too soon’, and in the end it turns out, on his story, that the experience of constancy of temporal conjunction does produce something from which the idea of necessity can be derived, namely inference, or constrained belief, what Hume calls ‘the determination of the mind’. His prediction that ‘Perhaps t’will appear in the end, that the necessary connexion depends upon the inference, instead of the inference’s depending on the necessary connexion’ (Treatise: 88) turns out to be correct. Necessity ‘lies only in the act of the understanding’ (Treatise: 166). Causal necessity has been reduced to the inferability of one member of a constantly conjoined pair of events from the other. So Hume then offers us two alternative definitions of causation, one as constant conjunction, the other as inferability.

In Section V of the Enquiry Concerning Human Understanding Hume says that the ‘custom’ that determines our causal inferences has ‘equal weight and authority’ with the ‘reason’ which is at work in demonstrative inferences. Yet he calls his naturalistic solution ‘sceptical’, and in Section XII of the Enquiry Concerning Human Understanding refers to the sceptic’s ‘triumph’ in such an analysis of causal inference. In the Treatise the ‘rules’ offered for improving our causal inferences are characterized as having the status only of ‘unphilosophical probability’, or prejudice, albeit a reflexive case of prejudice. Sceptics may be pleased, Hume writes, when they observe ‘a new and signal contradiction in our reason…the following of general rules is a very unphilosophical species of probability; and yet ‘tis only by following them that we can correct this, and all other unphilosophical probabilities’ (Treatise: 150). Scholars continue to disagree about the sense of ‘sceptic’ in which Hume sees himself as one of these sceptics, and about what most pleased him in his account of causal reasoning.

Testimony, credulity, and miracle reports. Hume’s descriptive account covers bad as well as good causal inferences and probability estimations, and it is as part of his attempt to separate the better from the worse inferences that he looks at our reliance on human testimony, the inference from the fact that someone says they witnessed something to the conclusion that they did witness it, or, more generally from the fact that they say that they believe something, and have good evidence to do so, to the conclusion that they do believe it, and do have the evidence they claim for it. A large part of what we call ‘education’, Hume says, is simply believing what we are told, especially if it is repeated over and over again. (This for him confirms how sensitive we are to repetition.) Such uncritical credulity is out of conformity to Hume’s epistemic rules. ‘The wise man…proportions his belief to the evidence’ (Enquiries: 110). The evidence on human lying, and on human error, should give us pause before believing all we hear. When what we hear goes counter to the most widely established ‘constant conjunctions’, when it is something that it suits someone’s purpose that we believe, when it caters to our own suspect ‘inclination to the marvellous’, and especially when all these factors are present, the wise person is on their guard. Miracles, claimed violations of laws of nature on which religions are founded, do combine all these factors. ‘What greater temptation than to appear a missionary, a prophet, an ambassador from heaven?’ (Enquiries: 125). Self-delusion and ‘pious frauds’ are to be expected in such cases, from both reporter and hearer of the report. To anyone who looks calmly at the human record of miracle reports and at the rival religions founded on such reports, and who then weighs the probability that a reporter of such a miracle is lying or deluded against the ‘proof’ (constancy of conjunction), which is the support for the ‘law of nature’ that supposedly has been violated, the conclusion will be clear. ‘We may conclude that the Christian Religion not only was at first attended by miracles, but even at this day cannot be believed by any reasonable person without one.’ Hume here does his typical reflexive turn – as he applies causal reasoning to itself, so he here applies his account of belief in miracles to that very belief itself. (This idea is celebrated in the title of a modern challenge to theism, John Mackie’s The Miracle of Theism.)

Bodies and minds. Book I, Part IV of the Treatise explores ‘sceptical and other systems of philosophy’. The influence of Berkeley is apparent in Hume’s discussion of our belief in lasting physical things, but Hume, unlike Berkeley, cannot appeal to God’s notions of physical things to explain their individuation and identity. Belief in lasting bodies pervades our interpretation of current sensations, and so infects the terminology in which most of our causal inferences are formulated. So if these beliefs are, as Hume claims, mere ‘fictions’, based on ‘false suppositions’ (Treatise: 217), then our causal inferences will also be contaminated. Belief in lasting material objects (‘bodies’), Hume claims, helps us to shore up our belief in constant conjunctions, since we just assume that although we may not have observed some particular cause or effect, it nevertheless happened. Lasting things serve as homes for unperceived events, and we need to believe in them to sustain our faith in ‘nature’s’ regular course of action. But some uses of causal analysis, in particular those used to examine the causal relation between physical stimuli and our sense organs, and the supposed relation of ‘secondary’ to ‘primary’ properties, can seem to destroy our belief that in sense perception we get reliable information about lasting bodies. There is an unstable relation between our trust in causal inference and our presumption that there are lasting physical things that are involved in causal relations with one another, and with our minds. The first part of the conclusion to Book I of the Treatise explores this instability.

Hume’s account in the Treatise (Bk I, Pt IV, Sect. VI) of what we believe about the identity of our own minds over time, an account supplemented in Book II of the Treatise, when he turns to personal identity ‘as it regards our passions’ (Treatise: 253), is less sceptical than the account of belief in lasting mind-independent bodies. However, in the ‘Appendix’ he reports that his hopes that a contradiction-free account could be given of our self-conceptions and the sort of identity we have reason to ascribe to ourselves (Treatise: 633) had been dashed when he realized that, in the section on personal identity, he had involved himself in a ‘labyrinth’. Commentators do not agree on just what the perceived difficulty was. He had rejected the idea of a simple self, possessor of its perceptions, since there is no impression from which such an idea could be derived. The ‘true idea’ of any human mind is of a ‘system of different perceptions or different existences, which are link’d together by the relation of cause and effect, and mutually produce, destroy, influence, and modify each other’ (Treatise: 261). The identity of a mind is ‘only a fictitious one, and of a like kind to that which we ascribe to vegetables and animal bodies’ (Treatise: 259). The most proper analogy, however, is not to bodies but to a ‘republic or commonwealth…as the same individual republic may not only change its members, but also its laws and constitution; in like manner the same person may vary his character and disposition, as well as his impressions and ideas, without losing his identity’ (Treatise: 261). This account of the ‘fiction’ we tell ourselves about our (and our nation’s) identity, and of the factual supports which give it believability, had seemed a satisfactory substitute for the rejected rationalist view that we know ourselves to be simple thinking substances, with an unchanging essence, persisting uninterruptedly through time. But the ‘Appendix’finds the account given of ‘the principle of connexion’ that binds the ‘different existences’ together to be ‘very defective’. Hume’s ‘scepticism’ (Treatise: 633) about his original account of the achievable coherence and empirical well-groundedness of our ideas of ourselves continues to spawn a wealth of interpretative literature.

On Hume’s own account, if it takes very ‘refin’d and elaborate’ or ‘metaphysical’ reasonings (Treatise: 268) to see the difficulties the philosophical sceptic finds, then even the mind that has seen them usually ‘quickly forgets’ them. His own purported attitude to our questionable beliefs about the world and ourselves is, by the end of the conclusion of Book I of the Treatise, that of the ‘true sceptic’, who, while not forgetting his philosophical doubts, is nevertheless as diffident about them as about his earlier convictions. In both cases he defers to his readers to point out to him any errors they find in his reasoning. Plenty of readers have taken up Hume’s invitation, especially when it comes to his Treatise writings on personal identity. In the Enquiry Concerning Human Understanding he dropped this ‘difficult’ topic, and very much curtailed his treatment of belief in lasting bodies and a ‘material world’ (Sect. XII, Pt I). ‘A blind and very powerful instinct of nature’ is said to make us have and preserve a belief in lasting external objects, and trust our own senses to give us reliable information about such objects. ‘The profounder and more philosophical sceptics’ will be able to shake our blind faith, producing ‘momentary amazement and confusion’, but if we ask the sceptic ‘what he proposes by all these curious researches?’ ‘if any durable good or benefit to society could ever be expected to result’ (Enquiries: 159) from dwelling on his arguments, then, Hume says, the sceptic is the one who will be embarrassed.

Social epistemology. Hume’s discussions of human knowledge in his essays and in the History of England keep the question ‘Do you expect durable good and benefit for society to come from your epistemological researches, or are you merely amusing yourself?’ very much in the forefront of their author’s mind. His question becomes ‘When and how do various branches of knowledge grow?’ not ‘Is knowledge possible?’ Where his earlier epistemology was ‘abstruse’, analytical, and general, his later epistemology is much less general, and not at all abstruse. His interest shifts to social epistemology. Do the sciences thrive better in a republic than in a monarchy? What sort of representation does a guinea make, and would it make a different representation if it were made of copper, not gold? What relation is there between commercial trade between nations and exchange of ideas? What has the invention of the printing press done to the life of the mind, and what other inventions make or might make comparable changes? Epistemology merges into social philosophy and ethics.