I believe that David Hume understands 'knowledge' to be a term with two quite distinct meanings. I take Hume quite seriously when he describes himself as a mitigated sceptic; he identifies many factual beliefs derived from factual inferences as cases of empirical knowledge.[1] Hume's scepticism about our ability to provide justifying reasons for factual inferences is an important element of his analysis of empirical knowledge, positioning Hume with recent "externalist" accounts of knowledge and distancing him from typical "internalist" accounts.[2]

Furthermore, I want to subject Hume's position on empirical knowledge to one of his more celebrated claims: "When we entertain, therefore, any suspicion, that a philosophical term is employed without any meaning or idea (as is but too frequent), we need but enquire, from what impression is that supposed idea derived?"[3] While some commentators locate this remark at the cornerstone of Hume's theory of knowledge, it is really an empiricist criterion of meaning; it is a precondition of Hume's philosophical theory of knowledge, for one can neither believe nor know a meaningless string of jargon. To state the issue using Hume's own jargon, how does Hume's own idea of knowledge fare under his criterion of meaning?[4] To put it another way, if we look to Hume for an analysis of 'knowledge,' seeking empirical evidence from which the supposed idea is derived, what do we find?

We all know that Hume goes to considerable lengths to provide just such an analysis of belief, attempting to isolate the impression(s) responsible for regarding some ideas as beliefs rather than mere conjectures or thoughts. Does Hume do anything along those lines to distinguish knowledge from belief? After all, philosophical terminology is the "jargon" that Hume wants to put to the test of his criterion of meaning. I think that the first Enquiry offers the rudimentary outline of such an analysis of empirical knowledge, and that Hume does so in order to block the common charge that his appeals to psychological explanations are no substitute for a proper account of empirical knowledge.

II

Norman Kemp Smith provides the standard interpretation of Hume as a moderate sceptic. Yet even on Smith's account of Hume, knowledge is quite distinct from factual belief: "We have seen how limited and narrow is the field to which Hume confines 'knowledge' and 'science'. What lies outside this field is at best merely probable."[5] If Hume really restricts knowledge to relations of ideas, his radical scepticism denies all knowledge of matters of fact. A 'matter of fact' is an idea which is believed "concerning the existence of objects or of their qualities" and which is believed as a conclusion from "reasonings from causes or effects."[6] Hume's epistemology would thus propose that all knowledge derives from experience while denying that there is knowledge by means of factual inference.

Any suspicion that Hume really denies the possibility of knowing such matters of fact--but somehow forgets to directly say so--is dashed by his blunt statement that "animals learn many parts of their knowledge from observation."[7] Either this remark means that animals engage in moral reasoning that yields genuine empirical knowledge, or it is a bizarre conclusion to an argument that minimizes the differences between human and brute "reasoning" while arguing that animals know some matters of fact through factual inference.

Hume's first direct admission of empirical knowledge occurs in the Abstract of the Treatise, where he suggests that specific matters of fact can be known, not just causal principles. He reiterates that such principles are known, but adds something further: "No matter of fact can be proved but from its cause or effect. Nothing can be known to be the cause of another but by experience."[8] Hume implies that knowledge of causal relations can provide knowledge, not just opinion, about future events. And he says as much; when one billiard ball is moving "on its axis" and about to strike another, "I know it will not stop after the shock."[9] The Treatise carries a blunt assertion that "even in common life, we may attain the knowledge of a particular cause merely by one experiment, provided it be made with judgment."[10] Thus, Hume does allow knowledge of matters of fact.

Elsewhere, a central premise of Hume's discussion of liberty and necessity is that "I know with certainty that he is not to put his hand into the fire and hold it there till it be consumed."[11] The proposition which he claims to know with certainty is clearly a matter of fact, not a relation of ideas. And like our firm belief that the sun will rise tomorrow or that "all men must dye," it exceeds 'probability.' Hume's willingness to say that he knows such things is no aberration; it is central to his argument about liberty and necessity that we know, with relative certainty, what various persons will do under various circumstances. By means of experience

we mount up to the knowledge of men's inclination and motives from their actions, expressions, and even gestures, and again descend to the interpretation of their actions from our knowledge of their motives and inclinations.[12]

Hume claims that we have knowledge of various causal principles, all falling within the sphere of human reason that the Treatise identifies as falling between knowledge and probabilities. In section IX of the first Enquiry, he insists that animals know some of the same causal principles.

Comparing the Treatise and Enquiries on these issues, one interpretation is that Hume's position on empirical 'knowledge' shifted around the time he wrote the Abstract. Whereas Book I of the Treatise denies the possibility of knowing contingent truths through factual inference, the Enquiries repeatedly recognize knowledge of such matters of fact. Another reading would be that Hume did not change his mind and that the Enquiries are simply clearer about his position. Infelicitous wording creates the impression of an earlier doctrine that there is no empirical knowledge. Hume actually allows two distinct senses of the term but overly emphasizes the more technical and restrictive Lockean sense in key passages of the Treatise.

It is my suggestion that once we deny the temptation to restrict Humean knowledge to relations of ideas, we find that his analysis of knowledge is inexorably bound up with his explanation of belief formation through factual inference. Curiously, on neither interpretation does Hume point to justification as a necessary condition for knowledge. Hume does not mean to challenge our knowledge of contingencies by showing that they are unjustified. On the contrary, he argues that justification is irrelevant to much of our empirical knowledge. In other words, we do not get a firm handle on Hume's analysis of knowledge until we grasp his analysis of empirical knowledge.

III

Hume's many references to empirical 'knowledge' might be dismissed as mere loose talk, and not evidence of any epistemological view, were it not for the precedent of his use of a strict and a loose sense of 'reason' and 'inference' and, correspondingly, of 'truth.'

According to the Treatise, "Reason is the discovery of truth or falsehood."[13] He describes its operation: "all kinds of reasoning consist in nothing but a comparison, and a discovery of those relations, either constant or inconstant," among our ideas.[14] Relations of ideas which are not known intuitively are known by inference, in which case they are "demonstratively certain."[15] The 'constant' relations are analytic and a priori; the 'inconstant' are synthetic and a posteriori. Hume's complex distinction between knowledge, proof, and probabilities gives way, in the Enquiries, to the streamlined dichotomy of demonstrative and moral (the latter sometimes identified as 'experimental'). Yet reason is consistently described as culminating in a "discovery" of new relations and their truth or falsity. Although the process of such discovery is very different for the two types of propositions, Hume allows both to produce true belief through a process of 'inference.' All factual beliefs that go beyond simple observation and memory are the products of factual inference, that is, of reason.

Hume trips over his own terminology when he allows that there are "reasonings concerning matter of fact . . . founded on the relation of cause and effect," yet any "inference" from cause to effect is "founded on experience" and it is "not, therefore, reason, which is the guide of life, but custom."[16] It is clear that two senses of 'reasoning' are present here. A matter of fact may be reached as a product of reasoning (an 'inference') in the loose sense, but here 'reason' must be understood in a stricter, technical sense as meaning demonstrative inference. Under the broad sense he allows that "reason is nothing but a wonderful and unintelligible instinct" by which we come to believe matters of fact; as such, reason is shared by humans and animals. Whether beast or not, "the inference he draws from the present impression is built on experience."[17] What the instinctive sort of inference shares with demonstrative reasoning is belief formation.

'Truth' is also defined differently for each type of reasoning. Far from reserving truth for relations of ideas, known intuitively or by demonstrative reasoning (i.e., knowledge in the strict sense), Hume defines truth disjunctively, recognizing true matters of fact secured under the looser sense of reasoning and inference:

Truth is of two kinds, consisting either in the discovery of the proportions of ideas, considered as such, or in the conformity of our ideas of objects to their real existence.[18]

The principle of contradiction provides a criterion of truth for constant ("real") relations of ideas. If we ascertain that the judgment does not capture a constant relation, a different criterion of truth is needed. For these, Hume seems to advocate a correspondence theory of truth.

Reason is the discovery of truth or falsehood. Truth or falsehood consists in an agreement or disagreement either to the real relations of ideas, or to real existence and matter of fact.[19]

For all his supposed scepticism, Hume thinks that human reason, generating instinctive belief based on past experience, generally leads us to factual beliefs which are true. And, as we saw earlier, Hume says that we have 'knowledge' of many such matters of fact.

In short, by delineating two species of reason and two attendant species of truth, Hume sets the stage for two species of knowledge. Since a priori and a posteriori judgments are radically distinct in both genesis and criterion of truth, the failure of any judgment to attain evidence "from knowledge" leaves open the possibility of epistemic success under the weaker criterion of truth.

IV

Hume's first Enquiry discussion of the reason of animals is prefaced by the methodological point that any "operations of the understanding" invoked to account for our "experimental reasonings" must follow the same pattern as any account "in all other animals." In other words, if we hold that other animals reason but we do not demand justifactory conditions on the idea of empirical knowledge for those animals, then we cannot introduce justifactory conditions in an account of human inference culminating in empirical knowledge. Often construed as a further confirmation of Humean naturalism,[20] this section actually argues that our behaviorist criteria for regarding animals as knowledgeable about matters of fact apply equally well in the case of human knowledge. Indeed, Hume's references to 'knowledge' as the outcome of animal inference are the most significant difference between this section of the first Enquiry and the corresponding pages in the Treatise.

Hume's argument can be reconstructed as having four stages. First, on some occasions we observe that an animal, such as a horse or a dog, correctly "infers some fact beyond what immediately strikes his senses." The animal's ability to make the inference "is altogether founded on past experience." Based on a combination of present sense impressions and memories associating the present impression with associated events, factual inference provides the animal with knowledge of some additional fact. Second, the animal has this empirical knowledge through instinctive, non-reflective inference. For we do not observe anything to provide us with the idea that the animal arrives at this empirical knowledge "through any process of argument or reasoning." Third, we are invited to substitute human children for animals in the above. Fourth, we are invited to continue the analogy by substituting adult humans "in their ordinary actions and conclusions" for animals in the above. Thus, Hume argues that we have no impressions to support the idea that there are real differences in the process by which humans and animals move from a state of ignorance about various matters of fact to a state of belief, including empirical knowledge.

It is worth noting the pains taken by Hume in the closing paragraph of section IX to emphasize that the discussion concerns cases of knowledge. Section IX repeatedly identifies some of the inferential beliefs of "brute beasts" to be knowledge. So Hume's conclusion is not merely a repetition of this earlier points about belief formation, as many commentators suppose. Since there is no evidence that animals supplement their beliefs with some process of reasoning or argument in order to "treasure up a knowledge" of various causes and effects, there is no basis for the idea that the beliefs we arrive at through experimental reasoning are converted to genuine knowledge only by becoming "objects of our intellectual faculties." Philosophers engage in mere "jargon" when they import such a supposition into their analysis of the idea of empirical knowledge.

It is also worth noting that section IX is part of a larger pattern of references to knowledge, culminating in the second Enquiry's discussion of the proper relationship between reason and sentiment. As he says near the end of the second Enquiry, reason "conveys the knowledge of truth and falsehood."[21] Nothing about this later discussion of morals restricts knowledge to relations of ideas; since reason's scope explicitly includes "showing us the means of attaining happiness or avoiding misery," and since these means are never known a priori, Hume clearly means that even instinctive inference (reason in the looser sense) separates true from false matters of fact. The remaining question is how Hume distinguishes empirical knowledge from factual belief; what impression might ground the distinction?

V

We are apt to read old texts in light of today's prejudices, and a common prejudice among contemporary philosophers is the position that an analysis of knowledge must include some condition over and above mere true belief. Although Hume's powerful critique of rationalist epistemology generates the need for an alternative model of knowledge, Hume refuses to supply the sort of analysis that many of us expect to see. He takes great pains to block any attempt to interpret empirical knowledge as justified belief, or highly warranted belief, or anything else that requires a belief to be a factual inference from justifying reasons. In short, he examines the process of belief formation but steadfastly refuses to offer any "philosophical justification" of those beliefs.[22] Paradoxically, Hume holds that we "employ our reason" (employ reliable processes of reasoning) without having reasons.

At a crucial point in the argument, he suddenly shifts the investigation from the question of justification to the psychological question of how we come to believe these contingencies. Hume regards this shift as a significant philosophical breakthrough:

This is a very curious discovery, but leads us to others, that are still more curious. . . . Do I nothing but conceive the motion of the second ball? No surely. I also believe that it will move. What then is this belief? And how does it differ from the simple conception of any thing? Here is a new question unthought of by philosophers.[23]

Is this a question for 'philosophers' only in the quaint sense that includes moral philosophy, including psychology? Content to pursue a psychological explanation of belief formation, Hume refuses to provide logical grounds for factual inferences. So Hume cannot place any such requirement on empirical knowledge. From the perspective that regards such grounds as the only sort that can justify a belief, Hume seems to be a radical Pyrrhonian.[24] But of course Hume takes great pains to make it clear that he rejects any move to suspend or doubt factual beliefs simply because they are "unjustified."

Hume's position is supposedly flawed on a second count. He fails to recognize that 'knowledge' is an evaluative term in a way that 'belief' is not. In Flew's metaphor, Hume seems to have burdened himself with a "dead albatross," the view that "there can be no such thing as a good experiential reason."[25]

Taking these complaints against Hume in order, the next stage of my argument is to draw out the implications of recognizing that Hume simply did not proceed under any demand for justifying reasons. As he summarizes his own view in the Abstract, we "employ our reason only because we cannot help it. Philosophy would render us entirely Pyrrhonian, were not nature too strong for it."[26] From Hume's perspective, "because we cannot help" but employ this 'reason' or factual inference, to expect justifying reasons for the resulting factual beliefs is to surrender to scepticism.[27] In both the Treatise and first Enquiry, Hume offers a distinction between having an idea and having a belief, and then explores the mechanism of belief fixation in the face of contrary experiences. But when Adam's belief that one billiard ball will move another is elevated to the status of knowledge, that status seems to consist in nothing more than Adam's having a true idea which is more firmly and intensely conceived than the rival and false idea that it will not be moved. Yet Hume clearly says that we often know such things. How, then, does Hume distinguish between belief and empirical knowledge? Given his examples of children and animals, Hume's account cannot invoke the presence of justifying reasons.

Here, I follow the tradition that runs from Kemp Smith to Barry Stroud; Hume "naturalizes" epistemology, explicitly inviting us to abandon our search for the rational foundation of our factual beliefs, contenting ourselves with a an account of belief formation and fixation.[28] Furthermore, Hume's position is externalist, not internalist; introspection about our reasons plays no role in what makes experimental reasoning good or bad: "no reasoning or process of the thought and understanding is able either to produce or to prevent" factual belief.[29]

Internalists part company with externalists on the question of whether "psychological states that can be detected by introspection" are a necessary ingredient in a theory of knowledge; internalists hold that introspectible states are to be invoked either as grounds for the justification of beliefs, as a criterion of adequacy for such grounds, or as the basing relation that must hold between a belief and its adequate grounds.[30] Hume denies the introspectibility of cognitive processes by which we produce beliefs (grounds), of the propensity of our processes to produce true beliefs (adequacy), and of the support relation between a new belief and its supporting beliefs (basing relation).[31] Because Hume denies that we assign empirical knowledge to children and animals by reference to factors external to their epistemic perspective, we are mistaken to invoke anything more than neuro-physiological processes as causal foundations of belief when accounting for our own empirical knowledge. Hume's first Enquiry is thoroughly externalist, denying that introspectible factual inferences play any role in justifying a belief.[32]

Taking our lead from these recent externalists, what external relationship between an epistemic agent and the world is invoked by Hume as evidence of empirical knowledge? If Hume is to be consistent in distinguishing belief from knowledge, it must be some set of impressions, some body of "experience and observation," considered in the context of "careful and exact experiments."[33] But what "experiments" does Hume point to, independent of the perspective of the relevant cognitive agent, when elevating a reported belief to the status of knowledge? After all, he advises us to look to "experience and observation . . . as in every other kind of evidence."[34] More specifically, Hume points to two factors, namely the circumstances and the character of the person involved: "the ignorance and inexperience of the young are here plainly distinguishable from the cunning and sagacity of the old."[35] A number of such "circumstances" are enumerated in the peculiar footnote at the end of section IX of the first Enquiry.

If section IX of the first Enquiry signals Hume's shift from the topic of belief fixation to the topic of empirical knowledge, section X's treatment of testimony is the obvious place to look for further details. We are not disappointed, as when Hume explicitly contrasts the ignorance of inhabitants of India and Sumatra concerning "a situation quite unknown" with the situation of an inhabitant of Moscow.[36] Hume's discussion of our need to evaluate the reliability of testimony introduces an externalist perspective on the question of whether a specific agent is likely to have formed a true or false inferential belief. His recommendation is that any assessment of the reliability of factual inference depends on our evaluation of both the agent's character and the agent's exposure to "constant and uniform experience" concerning the issue at question.

The Indian Prince who refuses to believe that water solidifies in cold weather "reasons justly" but lacks the appropriate relation to the relevant "situation" or "state of nature."[37] There is no reason to fault either the propensity of his inferential processes to produce true beliefs (adequacy) or the support relation between his denial and his supporting beliefs (basing relation). The prince's internal perspective on his own cognitive processes has nothing to do with the question of whether his belief is knowledge. His factual inference is at fault only because he lacks experience with a "situation" (external ground) that has bearing on the truth of the resulting belief.[38] Part II of the essay on miracles goes on to enumerate several other ways in which the character of the agent (adequacy and basing relation) undermines our confidence in the agent's factual beliefs, including the agent's motivation to lie to us.

Hume's various remarks on circumstance and character go a long way toward relieving Hume of his "dead albatross," provided we do not expect either Book I of the Treatise or the first Enquiry to explicitly address the normative commitment involved in saying that a person 'knows' a matter of fact. Obviously, Hume cannot introduce any such dimension into his position on empirical knowledge until he presents the basic principles his moral philosophy, particularly the proposal that "the useful qualities of mind are virtuous because of their utility."[39] It is somewhat odd that Hume does not explicitly expound on the virtue of wisdom--after all, he regularly distinguishes between the wise, who "proportion" belief to the available evidence, and dunces and fops, who form rash prejudices--but we can guess what he must say to be consistent with his general position. Elsewhere in Hume's collected essays, Hume argues that "industry, knowledge, and humanity are linked together by an indissoluble chain" and are unusually advantageous in both public and private life.[40]

Hume insists that mental qualities are the true object of the sentiment of approbation. He also holds that intellectual habits, evidenced only by their effects, are part of what we esteem "when a man is called virtuous . . . [for] his social qualities."[41] Even within a context that restricts knowledge to relations of ideas, the Treatise argues that certain types of "unphilosophical probability" or factual inference are to be avoided as founded on "changeable, weak, and irregular" principles; "wise men" avoid inferential transitions founded on such principles. Hume's justification for preferring the more "permanent" principles? They have proven more "useful in the conduct of life."[42] Consistent with his general theory of normative judgment, Hume cites utility in order to incite our approbation. And surely the "catalogue of laudable qualities" presented in the second Enquiry contains several features (e.g., good sense and discernment) associated with the mental character of those who elsewhere are praised as reasoning well.[43] Insofar as intellectual habits and character are central to Hume's externalist analysis of knowledge, and intellectual virtue cannot be separated from what we evaluate whenever we evaluate someone, Hume is not subject to the charge that his system cannot introduce the normative dimension that distinguishes between empirical knowledge and foolish belief.

Finally, I should admit that I cannot locate any passage in which Hume explicitly states that the sentiment of approbation applies strongly to those cases of inferential belief that we deem 'knowledge' and weakly or not at all in those cases we deem 'belief.' Nonetheless, Hume has everything he needs to dispel the charge that his epistemology cannot recognize assessment of factual beliefs derived from factual inferences. Thus, Hume has an analysis of empirical knowledge consistent with his own criterion for the meaningful employment of philosophical terms.

Notes

1. By 'factual inference' I mean inferences classified by Hume as ones having conclusions about matters of fact and real existence; by 'factual belief' I mean the ideas believed as a natural result of such inferences; by 'empirical knowledge' I mean knowledge about matters of fact, including knowledge arrived at by factual inference. For a precedent, see John W. Carroll, "Humean Justified Belief," The Philosophical Quarterly 48 (1998): 373-78.

2. See Kihyeon Kim, "Internalism and Externalism in Epistemology," American Philosophical Quarterly 30 (1993), pp. 303-16.

3. David Hume,Enquiries Concerning Human Understand and concerning the Principles of Morals, ed. L.A. Selby-Bigge, Third edition, revised by P.H. Nidditch (Oxford: Clarendon Press, 1975, hereafter EHU and EPM), p. 22 (II). See also Hume's Treaties of Human Nature, ed. Selby-Bigge, Second edition, revised by Nidditch (Oxford: Clarendon Press, 1978, hereafter T), p. 3; at p. 7 he calls it "the first principle . . . in the science of human nature." Hume's punctuation is sometimes updated in quotations, and section numbers follow page references.

4. Hume's formulation of the criterion is fatally flawed, and thus my raising of the question of how he understood empirical knowledge is quite independent of any endorsement of the resulting position.

5. Norman Kemp Smith, The Philosophy of David Hume (New York: Macmillan, 1960), p. 365; see also pp. 349-63. Another version of this interpretation holds that for Hume, "Knowledge is exclusively conversant with Relations of Ideas, and probability only with matters of fact." (Farhang Zabeeh, Hume: Precursor of Modern Empiricism [The Hague: Martinus Nijhoff, 1960], p. 85.)

6. T, p. 94 (I, iii, 7). As such, knowledge that some sugar is sweet need not concern a matter of fact in Hume's technical sense. Having sampled sugar, I recall the experience rather than reason "from causes or effects."

7. EHU, p. 108 (IX).

8. T, p. 654 (Ab.).

9. T, p. 655 (Ab.).

10. T, p. 104 (I, iii, 8).

11. EHU, p. 91 (VIII, i); this paragraph added by Hume in 1777 (edition O). In the language of the Treatise, this would be knowledge by proof, as opposed to knowledge by intuition or demonstration.

12. EHU, pp. 84-85 (VIII, i).

13. T, p. 458 (III, i, 1). In this paragraph, Hume explicitly includes a posteriori judgments. See also p. 180.

14. T, p. 73 (I, iii, 2).

15. EHU, p. 25 (IV, i); see also T, p. 70 (I, iii, 1).

16. T, pp. 651-52 (Ab.).

17. T, pp. 176-79 (I, iii, 16); see EHU, pp. 104-8 (IX).

18. T, p. 448 (II, iii, 10).

19. T, p. 458 (III, i, 1); see also p. 416 (II, iii, 3).

20. Antony Flew, Hume's Philosophy of Belief (New York: Humanities Press, 1961), p. 166.

21. EPM, p. 294 (App. I).

22. John Biro, "Hume's New Science of the Mind," in David Fate Norton, ed., The Cambridge Companion to Hume (Cambridge University Press, 1993), p. 38.

23. T, p. 652 (Ab.).

24. C. D. Broad, "Hume's Theory of the Credibility of Miracles," in Human Understanding: Studies in the Philosophy of David Hume, ed. by Alexander Sesonske and Noel Fleming (Belmont, CA: Wadsworth, 1965), p. 97.

25. Flew, pp. 89-90.

26. T, p. 657 (Ab.).

27. Such a reading fuels recent attempts to interpret Hume as a sceptical realist; e.g., John P. Wright, The Sceptical Realism of David Hume (Minneapolis: University of Minnesota Press, 1983).

28. Barry Stroud, Hume (London: Routledge and Kegan Paul, 1977).

29. EHU, p. 47 (V, Part I); this claim expands on Hume's clear statement that "our conclusions from experience are not founded on reasoning or any process of the understanding" (EHU, p. 32 [IV, Part II]).

30. Kim, pp. 306-7.

31. Similar points are made by Biro, "Hume's New Science," p. 45.

32. Kim, p. 305; Kim is only concerned with the classification of contemporary analyses of epistemic justification and knowledge, and any application of his criteria to Hume is my own interpretation. Parallels to Hume's position can be seen in work of Fred Dretske and D. M. Armstrong: Fred Dretske, "Conclusive Reasons," Australasian Journal of Philosophy 49 (1971), pp. 1-22; D. M. Armstrong, Belief, Truth, and Knowledge (Cambridge: Cambridge University Press, 1973).