David Hume’s influence is felt throughout Anglophone philosophy, most especially in epistemology. Typically, students and scholars of Hume’s epistemology focus on whether judgments and sentiments, among other considerations, can count as knowledge. For normative judgments, this focus leads to thorny questions about grounding, truthmaking, and non-cognitivism. This is rough terrain, and the approach foregrounds metaphysical matters that Hume took pains to avoid.
Accordingly, this piece will shift attention back to epistemology. The shift will come by looking at several modes of inference. Two modes are well known: demonstrative inference pertains to relations of ideas, and causal inference pertains to matters of fact. I explore an additional mode in perceptual inference – perception leads to existential judgment – as well as verdictive inference, which plays a crucial role in Hume’s moral sentimentalism and aesthetic theory. As this essay shows, switching the focus from judgment to inference leads to a more comprehensive view of Hume’s naturalism.
Ideas, Impressions, & Judgments
Any study of David Hume’s epistemology begins with distinctions. Hume writes:
Every one will readily allow, that there is a considerable difference between the perceptions of the mind, when a man feels the pain of excessive heat, or the pleasure of moderate warmth, and when he afterwards recalls to his memory this sensation, or anticipates it by his imagination (E 2.1).
Impressions are “our more lively perceptions, when we hear, or see, or feel, or love, or hate, or desire, or will” whereas ideas are “less forcible and lively” (E 2.3).
Impressions are further divided into external and internal sensations. External sensations are impressions that result from the five senses, while those that arise from an awareness of the operation of our minds are internal sensations. The latter include whatever we feel, desire, or will (E2.5), that is, objects of our passions, emotions, and sentiments (Owen 1999, 67; Schmitter 2021, §4). Hume maintains that “all our ideas or more feeble perceptions are copies of our impressions, or more lively ones” (E 2.5). This leads to the copy principle: “it is impossible for us to think of any thing, which we have not antecedently felt, either by our external or internal senses” (E 7.4).
The copy principle is easier to understand given a third distinction. Complex ideas, like “golden mountain,” are comprised of simple ideas. Complex ideas do not require correspondingly complex impressions (E 2.5). What is important is that for each simple idea there is a matching impression. Suppose a man’s first impression of gold came from looking at his mother’s wedding ring, and later he had the impression of a mountain while gazing at Half Dome. By putting the two simple ideas together he forms the complex idea “golden mountain.” According to the copy principle, all of our ideas are similarly composed of and restrained by our impressions.
Hume explains how we get from impressions to concepts (Lolordo 2021), and from concepts to judgments of knowledge. His views are largely based on Locke (1975), who used ‘idea’ to cover all items of the mind. Hume’s ideas are varied, including concepts, judgments, and the memories of previous impressions, but they are distinct from impressions. All impressions are “an original existence” (T 2.3.3) whereas ideas resulting from external sensations are representations. Right away, we can see a difference between judgments that are a result of perception as opposed to normative judgments that are a result of sentiment. Both are based on impressions, but only in the case of perception does the impression represent an external object.
In considering normative judgments, it is important to note that Hume distinguishes between judgments and sentiments (M App1.21). Strictly speaking, judgments in Hume’s narrow sense are limited to the understanding. Hume also uses ‘judgment’ in a broader sense, as in “suspense of judgment” (E 5.1; see also D 8.12, 12.5, 12.6, 12.32). Occasionally he even uses the term in reference to sentiments (T 3.3.1.27; ST 19, 23, 28; P 2.16). A judgment in this wider sense is a potential object of belief. Later, I explore judgments of “blame or approbation” (M App 1.3) as well as judgments of “beauty and deformity” (ST 16). Since I am using ‘judgment’ broadly, I do not bestow any properties on sentiments that Hume seems unwilling to bestow, such as truth and falsity.
Judgments of the Understanding & Perceptual Inference
Judgments of the understanding are of two types, each with its own form of reasoning (E 4.1-2). Relations of ideas are necessarily true, and their denial is impossible. They require demonstrative inference, which is truth-preserving, never taking us from true judgments to false judgments. Relations of ideas are also internally grounded since they are either intuitive or can be demonstrated from intuitions (if it feels too much like merely assuming the advantage of internalism about justification and warrant, Hume’s dissatisfaction with externalism is conveyed in the rejection of perceptual inference as a form of causal inference). In contrast, the denial of any matter of fact is possible, and they can only be established by causal inference. Causal inference covers inferences to causal claims or laws of nature as well as inferences from causal claims or laws of nature, as in the case of the prediction and explanation of events. Either way, causal inference falls short of demonstration (E 4). At best, matters of fact are externally grounded, depending on other matters of fact and ultimately objects (though Hume does not speak of “grounding” judgements explicitly).
Hume was also interested in judgments with two qualities. The first quality is that they have historically been offered as candidates for philosophical knowledge. Second, they fit easily into judgments of the understanding. These intractable judgments include laws of nature, for instance, and the famous problem of induction, which maintains there is no non-circular way of justifying laws of nature. Since laws of nature are complex, my point is made with the more modest example of perceptual inference: “I perceive Cindy, so she is home.” The inference begins with something mind-dependent yet ends with a judgment about the world. It is non-demonstrative, for it is possible to have perceptions without existence, as in dreams or hallucinations. Perhaps the inference is causal, but if I start filling in the details with the other needed premises, I inevitably include propositions that are less plausible than my original judgment that Cindy is home (Lewis 1996).
Hume regarded demonstration as the gold standard of knowledge. This is partly because the psychological story of belief meshes with the justificatory story. Demonstration is a learning process.
A demonstration, if just … can never be a difficulty. ‘Tis either irresistible, or has no manner of force. … Demonstrations may be difficult to be comprehended, because of the abstractedness of the subject; but can never have any such difficulties as will weaken their authority, when once they are comprehended. (T 1.2.2.7)
Contrast this with the link between smoke and fire. If I build a fire, I can predict there will be smoke; if I see smoke, I may infer there is a fire. These are also natural inferences I make. Yet, as in the case of perception, the inferences are not demonstrative, and the content of the judgment is not solely a function of the content of my evidence. The story that connects smoke and fire is complex, and its connection with my simple causal inferences is not apparent. The problems of perception and induction are examples of a more general philosophical malady: Many apparent judgments of knowledge are not judgments of the understanding.
Skeptical Naturalism & Natural Belief
How do we fit judgments into our system of knowledge given they do not conform to our models? Putting the question this way makes it seem as if we are smuggling unjustified beliefs into the system. It is important to note that natural beliefs are paradigm cases of knowledge. Consider perceptual inference. We cannot perform scientific experiments if we don’t take perceptual inference for granted. Science informs us about the inference, but science depends on it in the first place. In showing how natural belief and inference fit into our system of knowledge, Hume establishes the possibility of knowledge even if we concede that we have no standard reply to skeptical arguments.
Skeptical naturalism has two parts. The skeptical part admits that there is no standard reply to philosophical worries like the problems of perception and induction. By a standard reply, I mean a reply that points to a false premise or invalid inference. Since Hume has no standard reply to skeptical arguments, he provides a concessive response, conceding that when it comes to judgments of the understanding, the skeptic has an advantage. The naturalism-part demands that a reply to skeptical challenges is not needed. Natural beliefs are resistant to argument, both for and against. “Nature is always too strong for principle” (E 12.23).
For example, suppose we debate about the problem of induction and whether belief in the law of gravity is warranted. Skepticism about laws of nature may even emerge the winner in our debate, but at the end of the discussion we all go out the door, not the window (D 1.5). Candidates for natural belief include judgments about external objects, causality, and personal identity (Kemp Smith 1941). In a popular and influential body of work (1995, 2014), Paul Russell extends Hume’s naturalism to moral responsibility and judgments of blame. Though Russell remains critical of skeptical naturalism, I hope to resurrect the view by shifting the emphasis from judgment to inference. I begin with natural belief and expand to natural inference.
Hume’s natural beliefs are similar to Ludwig Wittgenstein’s hinge commitments (1969). Whether natural beliefs are items of knowledge or a kind of proto-knowledge is for our purposes “entirely verbal” (D 12.8n11; E 12). Either way they are essential to knowledge. Given the range of views of natural belief, I don’t provide an analysis, but list features that natural beliefs have in common (Campbell 2017).
Natural beliefs are matters of common-sense, and are near-universally held. Unless we are in a philosophy seminar, doubts about natural beliefs are dismissed as linguistic abnormalities, and not taken as genuine reports. As Wittgenstein notes, if I were to say “this mountain didn’t exist half an hour ago” it is such a “strange statement that it is not clear what I mean” (1969, §237). One could have a philosophical doubt about the existence of the external world or the law of gravity but not a practical or scientific doubt (§§19, 259). In the case of natural beliefs, even when there are expressions of doubt, they are seldom taken in earnest.
Arguments in favor of natural beliefs are not easily forthcoming, and appear either question begging or disconnected with our reasons for belief. Were someone to deny that he had a body, it is unclear “what it would mean to try to convince him that he had one” (§257), or even what he “would still allow to be counted as evidence” (§231). Wittgenstein asks: “Does my telephone call to New York strengthen my conviction that the earth exists?” (169, §210). It seems not. If not, my belief that the earth exists is unlikely the result of argument. There are reasons, to be sure, but they are not our reasons for belief, and the details of those reasons – if there are details – are evasive.
Natural beliefs are immune to skeptical argument. This is evidenced by the fact that skeptical arguments rarely have a lasting effect. We have philosophical doubts about natural beliefs but not practical doubts. In reply, one might worry that natural beliefs provide examples of easy knowledge, meaning almost any judgment could potentially be allowed into the system. The first feature of natural belief rules this option out. Natural beliefs are a matter of common sense. If there is reason for practical doubt, the judgment loses its natural belief status. Another potential problem is that if we regard natural belief as shared judgments, Hume seems committed to innate ideas. This concern diminishes if natural belief is a product of natural inference, as I propose.
Natural beliefs are adduced by a quasi-immediate inference. An immediate inference is a one-premise, valid argument. Immediate inferences are demonstrative. For example: All women are mortal, so Madonna is mortal. Provided ‘all’ has existential import, the conclusion must follow the premise. In most cases, perceptual inference is instantaneous and without notice, yet perceptual inference is not truth-preserving. It is quasi-immediate, meaning phenomenologically similar to immediate inference but non-demonstrative. Wittgenstein writes:
From its seeming to me – or to everyone – to be so, it doesn’t follow that it is so. What we can ask is whether it can make sense to doubt it. (1969, §2)
Perceptual inferences are treated as if they were demonstrative. Someone calls on the phone asking for Cindy, whom I perceive in my visual field. I immediately say “Yes, Cindy is home,” on the basis of perceptual inference. The question isn’t merely, What warrants such inferences? There are lots of accounts that take us from perception to judgments of existence. The problem is they are not our stories. I do not, at least not always, wait for the dictates of science before knowing that Cindy is home. I look, I see, and I know. It is similar with causal inferences, as well as with inferences to judgments of blame and of beauty.
Judgments of Blame
According to free will skepticism, all judgments of blame are unwarranted because no one has free will, and no one is morally responsible for his behavior. Hume regards this as another form of skepticism, similar to skepticism about the external world or the laws of nature. In each case, skeptical naturalism is the response. Skeptical naturalism about free will, as I envisage it, maintains that our propensity to blame (and praise) is a product of natural sentiment and inference, and for this reason such judgments are immune to skeptical argument. Consequently, arguments for free will skepticism are futile. Not unsound or invalid, but ineffective.
Hume draws two analogies to make his point. First, attempting to convince someone that no judgment of blame is warranted is like attempting to convince a man “lying under the racking pains of the gout” that there is no evil in the world using “specious and sublime” arguments (E 8.34). That “would surely more irritate, than appease” him (E 8.34). Adducing a judgment of blame requires forming a sentiment. Without the sentiment, the inference and judgment are incomplete. Since arguments have little effect on sentiments, they have little effect on judgments of blame.
One might retort that we are not at our philosophical best when suffering from gout, but Hume offers another, more convincing analogy between physical and moral ills. He writes:
A man who is robbed of a considerable sum; does he find his vexation for the loss any wise diminished by these sublime reflections? Why then should his moral resentment against the crime be supposed incompatible with them? (E 8.35)
Here the comparison is between the anguish of losing money in a robbery and the sentiment of blame toward the robber. If we don’t expect the “remote and uncertain speculations” (E 8.35) of skeptical arguments to undermine our feelings in the former case, why expect them to undermine our sentiments in the latter case? Judgments of blame are not judgments of the understanding, so it is no surprise that skeptical arguments against them will be better than those in their favor.
A crime is a blameworthy act not limited to the legal realm proper, according to Hume’s term of art (M App1). Our judgments of crime arise naturally under the right conditions like our perceptual inferences to external objects, our suffering of physical pain upon illness, and our feeling of anguish when we are the victim of theft. Hume asks: “Are such remote and uncertain speculations able to counterbalance the sentiments, which arise from the natural and immediate view of the objects?” I respond: No. Just as skeptical arguments are unlikely to prevent us from making judgments about external objects. Being natural, “these sentiments are not to be controuled or altered by any philosophical theory or speculation whatsoever” (E 8.35). In brief, philosophical doubt cannot remove natural beliefs, at least not for extended periods of time.
It is easy to overplay the role of sentiment in moral sentimentalism, suggesting that reason plays no role, which is not the case. Since judgments of blame are neither relations of ideas nor matters of fact, they are not entirely adduced by reasoning but in the right context, sentiments and judgments follow. Hume writes:
The mind of man is so formed by nature, that, upon the appearance of certain characters, dispositions, and actions, it immediately feels the sentiment of approbation or blame. The characters, which engage our approbation, are chiefly such as contribute to the peace and security of human society; as the characters, which excite blame, are chiefly such as tend to public detriment and disturbance: Whence it may reasonably be presumed, that the moral sentiments arise, either mediately or immediately, from a reflection on these opposite interests. (E 8.35)
Crime “arises from a complication of circumstances, which, being presented to the spectator, excites the sentiment of blame, by the particular structure and fabric of his mind” (M App1.6). Hume takes the inference as basic and imagines a “judicious spectator” (T 3.3.1.14), but the spectator need not be ideal (Sayre-McCord 1994). Juries require no special training in reasoning, just evidence and law.
Verdicts & Verdictive Inference
Judgments of blame are verdicts. They enter philosophy by way of J.L. Austin’s speech act theory. For Austin, verdictives “consist of the delivering of a finding, official or unofficial, upon evidence or reasons as to value or fact” (1975, 153). However, the verdicts I discuss are conclusions of a particular inference type: from matters of fact (or evidence) and social norms (or laws) to normative judgment. There is no type of speech act that corresponds to verdicts in this sense. In fact, verdictives drop out of most contemporary speech act theories (Green 2021, §3). For John Searle (1975), a judge’s verdict is a declaration since it makes something true that was not before true. But the critic’s judgment is assertive, representing a state of affairs (the performance’s quality).
Verdictive inferences in Hume’s thought have specific characteristics. Their structure is always from matters of fact and norms to judgment. Since verdicts are adduced from norms, they have normative content. Like perceptual and causal inference, verdictive inference is “hard to be certain about” (Austin 1975, 151), and is non-demonstrative. Verdictive inference involves both matters of fact and the acceptance of norms, so it is similar to cause-effect prediction. Inferences from cause to effect, or effect to cause, are quasi-immediate when the causal relations are internalized, as with smoke and fire. Whenever norms are internalized, verdictive inference is quasi-immediate. The verdicts of ‘Guilty,’ ‘Safe,’ and ‘Beautiful’ come easily and are abundant in human conversation.
My main concern is verdicts or judgments of blame, but a similar theory applies to judgments of beauty or deformity (ST). Judgments of blame are not judgments of the understanding. They are not grounded in anything internal to the judgment, for it is possible to consistently deny such claims. Nor is there anything external to the judgment to play the role of a moral fact. A blameworthy act cannot correspond to an event like stealing since not all stealing is criminal (e.g. taking bread to save a starving child). Crimes are wrong doings without exemption or excuse (Strawson 1962), not mere doings corresponding to worldly events.
Crime is doubly complicated. It is a “complicated object” (M App1.8), arising from a ‘complication of circumstances’ (1.6). The ‘circumstances’ are what we usually associate with crime since they include a victim and the cloistered mental state of intent. However, the circumstances are not the blameworthy act. The crime ‘arises from’ the circumstances after ‘being presented to the spectator’ and exciting ‘the sentiment of blame.’ A crime is the kind of act that would produce the sentiment of blame in the average person, someone with sufficient evidence who accepted the relevant social norms. Provided the social norms are internalized – as in the case of horrendous crimes like murder – the judgment comes immediately given ample evidence. The structure of verdictive inference suggests norms work best when they have wide social acceptance, otherwise the relevant sentiment is not ensured to follow, and verdicts are impeded.
In a similar way, works of art are beautiful provided they cause sentiments of beauty in “a true judge in the finer arts,” one who is “united to delicate sentiment, improved by practice, perfected by comparison, and cleared of all prejudice” (ST 23). Of course, one might contest the standard of norms in Hume’s view of aesthetics, but I am not interested in those specifics. What interests me is that judgments of beauty, like judgments of blame, are a function of verdictive inference, a natural inference we all share, and one that can be better or worse.
In the end, verdicts have standards. Some judgments of blame are warranted, and some are not. Moreover, the criteria for warrant are clearer in the case of judgments of blame than of judgments of beauty. In the former, the evidence must be sufficient, and the norms must be accepted, at least provisionally otherwise the inference is impeded. Of course, warrant comes at the cost of an unavoidable connection to human sentiment, so even with a theory of moral responsibility in hand, skepticism remains. On one side, we have arguments for free will skepticism with no standard reply opposing them; on the other side, support rests in part on sentiment and invalid inference. From the point of view of the understanding, skepticism is the winner. Nonetheless, in the process, Hume transforms questions about the quality of normative judgment – e.g., What grounds moral truth? – into questions about our shared normative inference – e.g., Is the verdict supported by sufficient evidence and accepted social norms?
Concluding Remarks
I have tried to recast Hume’s epistemology in terms of inference-types instead of judgment-types while at the same time adding new judgment and inference types. There are now three prongs to Hume’s fork: relations of ideas, matters of fact, and verdicts.
Relations of Ideas
a. True depending on internal relations of ideas
b. Demonstrative inferences
Matters of Fact
a. True depending on external correspondence with objects
b. Causal inferences
Verdicts: Approbation or blame, beauty or deformity
a. Warranted or unwarranted relative to evidence and social norms
b. Verdictive inferences
This addition of a third prong is consistent with a variety of interpretations of Hume. For instance, it remains open whether normative sentiments are true or false. I stress that even if sentiments are never true (Mackie 1977), it doesn’t follow that they are never warranted, but that is a topic for another day.
References
Austin, J.L., 1975. How to Do Things with Words, 2nd edition, J.O. Urmson and M. Sbisá (eds.). Cambridge, MA: Harvard University Press.
Green, Mitchell. 2021. “Speech Acts.” The Stanford Encyclopedia of Philosophy (Fall 2021 Edition), E.N. Zalta (ed.), URL: <https://plato.stanford.edu/archives/fall2021/entries/speech-acts/>.
Hume, David. Hume Text Online, A. Merivale and P. Millican (eds.), URL: <https://davidhume.org>.
T = A Treatise of Human Nature (1739-40; referenced by book, part, section, paragraph)
E = An Enquiry Concerning Human Understanding (1748; referenced by section, paragraph)
M App1 = An Enquiry Concerning the Principles of Morals, “Appendix 1. Concerning Moral Sentiment” (1751; referenced by paragraph)
P = A Dissertation on the Passions (1757; referenced by section, paragraph)
ST = The Standard of Taste (1757; referenced by paragraph)
D = Dialogues Concerning Natural Religion (1779; referenced by part, paragraph)
Lewis, David. 1996. “Elusive Knowledge.” Australasian Journal of Philosophy 74: 549–567.
Locke, John. 1975. An Essay Concerning Human Understanding, Revised Edition, P.H. Nidditch (ed). Oxford, UK: Clarendon Press.
Lolordo, Antonia. 2021. “Hume on Power.” Powers, J. Jorati (ed.). Oxford, UK: Oxford University Press.
Mackie, J.L. 1977. Ethics: Inventing Right and Wrong, Harmondsworth, UK: Penguin Books.
Owen, David. 1999. Hume’s Reason. Oxford, UK: Oxford University Press.
Russell, Paul. 1995. Freedom and Moral Sentiment: Hume’s Way of Naturalizing Responsibility. New York: Oxford University Press.
———. 2014. “‘Hume’s Lengthy Digression’: Free Will in the Treatise.” The Cambridge Companion to Hume’s Treatise, D.C. Ainslie and A. Butler (eds.). Cambridge, UK: Cambridge University Press.
Sayre-McCord, Geoffrey. 1994. “On Why Hume’s ‘General Point of View’ isn’t Ideal – and Shouldn’t Be.” Social Philosophy and Policy 11: 202–228.
Schmitter, Amy M. 2021. “17th and 18th Century Theories of Emotions.” The Stanford Encyclopedia of Philosophy (Summer 2021 Edition), E.N. Zalta (ed.), URL: <https://plato.stanford.edu/archives/sum2021/entries/emotions-17th18th/>.
Searle, John. 1975. “A Taxonomy of Illocutionary Acts.” Language, Mind and Knowledge, K. Gunderson (ed.). Minneapolis, MN: University of Minnesota Press.
Strawson, Peter. 1962. “Freedom and Resentment.” Proceedings of the British Academy 48: 1–25.
Wittgenstein, Ludwig. 1969. On Certainty, G.E.M. Anscombe & G.H. von Wright (eds.). New York, NY: Harper & Row.
Leave a Reply