1. Introduction

It is no secret that Lady Mary Shepherd’s philosophy has David Hume as its target. Shepherd’s treatise (1824: 2), partially entitled An Essay upon the Relation of Cause and Effect, Controverting the Doctrine of Mr. Hume, begins by stating that the Humeans’ arguments and conclusions are ‘illogical’ and ‘untrue’, respectively. Shepherd is particularly concerned with Hume’s skepticism about necessary causation and external existence. Shepherd, then, is standardly presented as Hume’s foil: her project seeks to restore reason as grounding beliefs in necessary causal connection and external objects’ existence. Shepherd thus hopes to free us from this skeptical ‘malady’ introduced by Hume, which he claims we have no ‘cur[e]’ for (2007a: 144).¹

While quite right, this presentation belies her account’s fundamental modesty. What Shepherd thinks we cannot know can be lost amid her discussion about all we do know through reason. For example, the secondary literature contains ‘only brief acknowledgements that Shepherd affirms the unknowability of essences of material things’ (Lindblom 2022). This neglect of Shepherd’s epistemic humility is accompanied by a more general tendency to focus on her metaphysics rather than her epistemology (Tanner 2022: 71–72). Though it is true Shepherd’s work is thoroughly metaphysical—Shepherd thinks refuting Hume requires uprooting his metaphysical system—skepticism prevails if the metaphysical tenets are not epistemically secure. It is one thing to say the world is constituted in such a way and another to prove we know it is such. To rebuke Hume, Shepherd must respond not only to his metaphysical challenge but also to his epistemological challenge: to establish knowledge of nature’s uniformity, such that we can then make rational inferences about the future from prior experience.² Indeed, even if we grant Shepherd her metaphysics (a generous allowance, given its critiques, e.g., her causal principle’s purported circularity and question-begging),³ she could still struggle to address Humean skepticism’s epistemological concerns, especially given her aforementioned epistemic humility about external objects.⁴ Accordingly, attending to Shepherd’s epistemology lets us better assess if her anti-Humean project succeeds.

Before beginning, it is important to set our scope. Hume’s skepticism is sprawling. The existence of external objects, other minds, the self, God, and necessary causal connection are all challenged, and Shepherd’s response, in turn, tries to restore belief in each as grounded in reason. The present discussion, however, cannot treat Shepherd’s entire epistemology. Instead, I discuss her epistemology as it concerns natural science, which seeks to prove ‘“an universal necessity of connection” between any given cause and its effect’ in the empirical world (Shepherd 2020: 150).⁵ Such scientific inquiry is characterized by its use of physical induction (Shepherd 2020: 145).⁶

This paper’s focus on science is motivated by the original text—Shepherd clearly thinks it is important that science is saved from skepticism—and my belief that science affords the best opportunity to bring out Shepherd’s epistemic humility, as it investigates the very objects Shepherd is epistemically humble about.⁷ By examining Shepherd’s science, we also see how her epistemology and metaphysics interact: what we can know about objects is partially a function of how our metaphysics understands such objects’ nature. Analyzing science highlights how Shepherd’s epistemic humility, when imposed on her metaphysics, requires her to make concessions to the Humeans about reason’s limits.

I begin by glossing Shepherd’s science and its foundations: the causal principle and the inductive principle derived from it. I then problematize this model of scientific inquiry by introducing Shepherd’s view of objects. For Shepherd, scientific claims must be probabilistic because of her account of objects. Specifically, her epistemic humility about external objects’ essences and strict conception of natural kinds mean we can never be certain that any given object before us is of a certain kind and thus are unable to know what all its causal powers will be, even after experimentation. I then examine what grounds scientific claims’ probability and contend that it rests on an inference according to which objects with the same sensible qualities probably have the same essences, i.e., that they are of the same kind and, as such, have the same causal powers: this object, because it looks like a previously observed object x (and is thus probably of the same kind as x), will probably have the same causal powers x did. Crucially, this inference—from sensible to essential likeness, and from past observation to future prediction—depends on a supposition about nature’s regularity that is not rationally grounded. I submit that Shepherd’s science, to get off the ground and make probabilistic inductive claims despite her epistemic humility and conception of kinds, must assume some proposition about nature that, to the chagrin of her response to Hume, cannot be grounded in reason.

2. The Causal Principle and Shepherd’s Science

Often, Shepherd’s anti-Humeanism is said to be theologically motivated. Shepherd (1824: 4), an avowed Christian, describes Humeanism as ‘lead[ing] directly to a scepticism of an atheistical tendency, whose dangerous nature can require no comment, nor any apology for its refutation’. Yet it is also true that Shepherd was deeply concerned about Humeanism’s consequences for science.

For Hume, we are mistaken to think that necessary causal causation is out there: an objective relation we can discover in the empirical world. This confused account of causation arises from repeated observations of conjoined objects—e.g., that the sun has risen each morning—and the associations we make thereby. From these past experiences, we infer that the sun will necessarily continue to rise each morning (all other things supposed equal). However, we cannot use reason to make this inference that the future will resemble the past: it is neither intuitively nor empirically demonstrable.⁸ On the one hand, a priori reasoning will not do, for this inference concerns the external, empirical world, which can only be apprehended through sensory experience, and not pure reason.⁹ But appealing to experience results in circularity. It only follows that the sun will rise tomorrow since it has in the past, because, in the past, the future has mirrored the past. We do not have any good reason to think this past state of affairs must continue, and it is thus just as well to suppose that the sun will not rise tomorrow. The upshot is that what we mean when we talk about causation cannot be a necessary and universal connection between certain objects. On Hume’s view, neither pure reason nor experience gives us the license to posit such a thing. Rather, what we actually have is a ‘connexion . . . which we feel in the mind’, such that we expect a certain object to be attended by another (Hume 2007b: 69). This subjective feeling, and the inference from past to future that underwrites it, is caused not by reason but by custom, which arises from ‘natural instincts’ and ‘alone determines the mind, in all instances, to suppose the future conformable to the past’ (Hume 2007b: 46; 2007c: 139). What first seemed to be necessary connection is thus, in fact, constant conjunction: ‘Beyond the constant conjunction of similar objects, and the consequent inference from one to the other [through custom], we have no notion of any [causal] necessity or connexion’ (Hume 2007b: 75).

Shepherd worries that this thesis of causal connection as constant conjunction—a subjective psychological construct—means scientific claims hold no water. If one thinks science makes claims about necessary causal connections in the external world, and we cannot in principle know of such causal relations, as Hume thinks, then scientific claims cannot be genuinely posited. Shepherd (1824: 186) writes that ‘as long as the notions of Mr. Hume shall prevail, inquiries of this [scientific] nature will be instituted in vain; nor indeed is there any received doctrine upon the relation of Cause and Effect, which can be securely used, as an efficient instrument in the advancement of science’. Refuting Hume and reinstalling reason as ‘the foundation of the whole principle of causation’ is how we rescue science (Shepherd 1824: 81). Indeed, Shepherd (1824: 192) says her causal principle, which establishes knowledge of necessary causal connection, is ‘the governing proposition in every science’, without which we would look ‘in vain’ for ‘improvement in any [science]’. We should now sketch how this causal principle (and the inductive principle following from it) are the ‘only true foundations of scientific research’ for Shepherd (1824: 194).

As the ‘primary tenet of her metaphysics’, the causal principle grounds Shepherd’s philosophy (Bolton 2021). From it, we derive a rational basis for belief in the existence of external objects, other minds, God, the self, and necessary causal connection: the last being most relevant for our present purposes. Shepherd (1824: 34–35) motivates the causal principle as follows:

Let the object which we suppose to begin its existence of itself be imagined, abstracted from the nature of all objects we are acquainted with, saving in its capacity for existence; let us suppose it to be no effect; there shall be no prevening circumstances whatever that affect it, nor any existence in the universe: let it be so; let there be nought but a blank; and a mass of whatsoever can be supposed not to require a cause start forth into existence, and make the first breach on the wide nonentity around;—now, what is this starting forth, beginning, coming into existence, but an action, which is a quality of an object not yet in being, and so not possible to have its qualities determined, nevertheless exhibiting its qualities?

The idea is that no object can come into existence of itself, that is, uncaused. An object which spontaneously ‘start[s] forth into existence’ would demonstrate a quality, or effect—that of the action of coming into existence—which could not, by definition, belong to that object when it was non-existent.¹⁰ It would be a contradiction to say a non-existent object could cause its own existence, as an object could not have the power to perform the action of bringing itself into existence if there did not first exist a capacity to actualize that power. This capacity cannot inhere in a non-existent object, given that an object cannot exist before it is actualized. An object thus can only get its existence (or any other quality) by virtue of other objects with the capacity and potential power to cause that quality to come into being. Such causal powers only belong to extant objects, different from the object they cause to have the new quality (in this case, the quality of coming into existence). It is thus that ‘this beginning to exist cannot appear but as a capacity some nature [other objects] hath to alter the presupposed nonentity, and to act for itself, whilst itself is not in being’ (Shepherd 1824: 36).¹¹

To see how this works, we can think of flint and steel producing fire. Fire and the qualities that constitute it—like its heat and orangish color—cannot exist absent flint striking steel. Rather, fire only has potential existence in the capacities of flint and steel to cause fire. When combined, flint and steel actualize their causal powers to create the qualities of fire. Fire, as the collection of these qualities, now exists as a new object, caused by two other objects combining. Though we need not dwell on the metaphysics here, we should note that the relata of causation are now objects, rather than a sequence of events, as Hume has it. It is in the union of flint and steel that fire simultaneously arises as ‘a new nature, capable of exhibiting qualities varying from those of either of the objects unconjoined’ (Shepherd 1824: 63). It is thus that flint and steel, upon joining with one another, are ‘really to be a producer of [a] new being’ (Shepherd 1824: 63).

From this causal principle, Shepherd (1824: 27) derives an inductive principle: ‘Similar causes must necessarily produce similar effects’. If the causal principle—that everything that begins to exist must have a cause other than itself—is true, it necessarily follows that like causes produce like effects. If there are two objects differing in qualities, we cannot explain each object’s coming to be as the effect of the same cause(s). Otherwise, without positing a different cause to explain the different qualities, we would have qualities, or effects, which were without a cause, something the causal principle rules out. In other words, a difference in effects (a new quality) cannot arise uncaused: ‘A difference of qualities could not arise of itself, could not begin its own existence’ (Shepherd 1824: 56). Given that everything that begins to exist has a cause, we must explain what caused this difference to come into being, and the causal principle precludes this difference from occurring uncaused, whereby different effects would have to arise spontaneously from the same cause. Thus, different effects must be the result of different causes. Conversely, same causes must give rise to same effects. If the causes are the same, there is nothing to cause a difference in the effects, and the causal principle guarantees that a difference in effects cannot come from nothing, i.e., happen uncaused.

Hume’s anti-realism about necessary causal connection, whereby custom and constant conjunction are ‘utterly incapable of affording an universal definition [of nature’s causal regularity]’, is thus vanquished (Shepherd 1824: 66). Shepherd (1824: 66–67) grounds our belief about causal regularity in reason, not custom, and establishes nature’s ‘absolute invariableness’ as a ‘strict necessity’. Not only is it thus rational, given nature’s uniformity, for ‘the mind to predicate for the future as for the past’, but we also have certitude about this (Shepherd 1824: 67). Like causes will, necessarily and universally, produce like effects.

Here, Shepherd thinks she has saved science from skepticism. The scientist now has a framework within which to make their claims and acquire knowledge of the external world and the causal relations among its objects. By ‘reasoning on experiment’, we can take the ‘abstract and demonstrative reason’ guaranteeing nature’s regularity and apply it to gain a posteriori knowledge of the empirical world (Shepherd 2020: 150). We can, in principle, conclude with ‘equally demonstrative evidence’ that there also ‘must exist “an universal necessity of connection” between any given cause and its effect’ (Shepherd 2020: 150). If we observe that certain objects give rise to others—e.g., flint and steel producing fire—and hypothesize that there was not an unknown variable contributing to the fire’s production, then the flint and steel must be the causes of this instance of fire. The causal principle and the inductive principle guarantee this relation will obtain necessarily and universally and allow us to project our finding onto all future occasions where flint and steel meet (in like circumstances, where it is assumed there is no confounding variable). We can thus say with certainty that the concussion of flint and steel will cause fire, now and always.

3. Shepherd’s Epistemic Humility

I want to throw some water on Shepherd’s fire now. The above sketch of Shepherd’s science is, to me, a touch rosy. Things become muddied when we appreciate Shepherd’s epistemology and understand her as a not ‘especially optimistic philosopher’ (LoLordo 2020: 23). Indeed, I am uncertain that Shepherd entirely delivers science from the specter of skepticism. To be sure, that she has secured rational justification for the causal principle and physical induction (if we grant Shepherd her metaphysics) is no small feat. Shepherd meaningfully challenges Hume’s attack on reason: we are no longer ‘epistemically shut off from a grasp of causation’s intrinsic nature’ (Fantl 2016: 104). We can rationally speak of necessary causal connection, and the inductive principle gives the conditions by which we can, in principle, know of particular causal relations.

Yet, this inductive principle is just a general law of nature (and, with the concomitant causal principle, the ‘one only law’ of nature there is for Shepherd) (Shepherd 2020: 149). Employing empirical observation to demonstrate specific causal relations among objects on the basis of this principle, as science attempts to do, is a different, thornier matter: ‘A priori, we know not what particular effect may arise as the results of any given cause’ (Shepherd 2020: 164). Knowing that like causes produce like effects need not mean we cannot be removed from knowing ‘what effects will follow from the events observed thus far’ (Fantl 2016: 104). I contend that when it comes to the beliefs we have about particular causal relations—e.g., that bread, if combined with the body, will always and necessarily have the effect of nourishing us (some ailment or spoiled bread notwithstanding)—we are forced to make extra-rational inferences, going beyond what the causal and inductive principles guarantee, because of a necessarily incomplete understanding of the external world and the essences of the objects therein. This is a direct consequence of Shepherd’s epistemic humility about external objects and her strict conception of natural kinds.

As suggested in our earlier discussion of the causal principle, Shepherd (1824: 46) takes objects to be just what their qualities are, such that ‘objects in relation to us, are nothing but masses of certain qualities, affecting certain of our senses; and which, when independent of our senses, are unknown powers or qualities in nature’. Likewise, she says ‘we know nothing of objects but the enumeration of qualities’ (Shepherd 2020: 145). Crucially, however, we cannot know all of an object’s qualities. Some qualities are necessarily beyond our grasp, for we cannot know in principle the qualities that external, mind-independent objects have independent of our sensations:

The union of the three following things are required to form the proximate cause for that great effect, the formation and combination of those aggregates of sensible qualities usually called objects; namely, first, the unknown, unnamed circumstances in nature, which are unperceived by the senses; secondly, the organs of sense, whose qualities mix with these; and thirdly, the living, conscious powers necessary to sensation in general. In this union, and with it, is the creation and production of all sensible complex qualities called objects, such as we know them. (Shepherd 2020: 60)

These sensible qualities—i.e., the effects of external objects causally interacting with our sensory apparatus—are ‘the only, the original and immediate materials of our knowledge’ (Shepherd 2020: 122). External objects ‘can never be contemplated, excepting under the forms of those unions [between the senses and external objects]’ (Shepherd 2020: 40). We thus only have acquaintance with objects indirectly via sensations. The ‘given data’ is ‘nothing but our sensations . . . and their relations’ (Shepherd 2020: 121).

These sensations, however, cannot be like the objects that caused them. Shepherd (2020: 112) thinks ‘nothing can be like a sensation, or idea, or perception, but a sensation, idea, and perception’. Our perception of an external object is like any other causal occasion: the combination of an external object, the self, and sense organs mix to produce something new, whose qualities vary from those of its causes before they were conjoined. In sensation, ‘there must be a modification of the objects, which must . . . in some way alter them from the state in which they were, when existing unperceived’ (Shepherd 2020: 41). We are then at a remove from the cause of our sensations (the external object) and its essence, which is ‘entirely different from the essence of the sensation [the external object] causes’ (Boyle 2018: 12). We thus know not of objects’ secret powers: ‘the real external unknown Causes in Nature, which determine the sensible qualities, as well as every other Effect’ (Shepherd 1824: 60).¹² It is these unknown and unperceived essential qualities of objects that cause them to appear as they do in sensation, i.e., in their ‘inward existence’ (Shepherd 2020: 48).

What we have at our disposal is then a necessarily ‘incomplete definition’ of objects (Shepherd 2020: 59). An object’s sensible qualities enumerate only ‘one class of [its] effects’, leaving us ‘wanting [qualities] for the proper definition of the objects’ (Shepherd 2020: 81, 45). And while Shepherd (2020: 137) thinks we make do with sensible qualities, likening them to ‘a language, which must be translated, before it can explain the actions of nature’, some things are necessarily lost in translation. It ‘never can be too much insisted on, (in order to maintain an exact philosophy,) that the positive nature and essence of unperceived beings cannot be known’ (Shepherd 2020: 130). Shepherd (2020: 220), recognizing these epistemic limits inherent to humanity’s lot, resigns that ‘although we be philosophers enough to know it is impossible to do so, we are for ever endeavouring to catch at, and yet for ever disappointed at not meeting with, those essences’.

This epistemic humility has significant implications when considering Shepherd’s account of kinds. For Shepherd (1824: 53), an object, to be of a certain kind, ‘must comprehend all [that kind’s] qualities tried and untried; observed and unobserved; determined and undetermined’. One object is not of the same kind as another ‘unless it hath all its qualities, and no other than its qualities’ (Shepherd 1824: 54). This all-or-nothing view is, as others have remarked, a ‘radical’ and ‘extremis[t]’ account with ‘troublesome’ and ‘unpalatable’ consequences (Tanner 2022: 76; Fantl 2016: 95; Tanner 2022: 87, 78). Here, it is essential to stress that an object’s having certain sensible qualities is not sufficient for kind membership (though it is necessary, as ‘truly similar objects, must necessarily appear the same’) (Shepherd 1824: 105). Even if two objects apparently share all sensible qualities, these are mere ‘signs’ of their secret powers, which Shepherd’s epistemic humility dictates we can never know (Shepherd 1824: 123). Unbeknownst to us, these apparently alike objects might have different essential, unperceivable qualities and would thus be different kinds of objects. Shepherd (2020: 140) explicitly endorses this possibility, writing that ‘apparently like objects may in every sensible quality be similar, and yet they may essentially differ in their remote causes’. Such objects would, despite their sensible alikeness, be ‘truly other’ objects of different kinds (Shepherd 1824: 71). Thus, such a strict conception of kinds—when considered with Shepherd’s epistemic humility about objects—rules out our ever being certain that an object in a present trial is of the same kind as one in a past trial.

This is a problem for causation, given that objects are the causal relata for Shepherd (1824: 55–56): ‘Like Causes, that is, like objects . . . must have like effects, or qualities’. If I do not know two objects are of the same kind, I cannot guarantee the object before me now will, in like circumstances, have the same causal powers as another putatively identical object I tested earlier. This present bread-looking object is not necessarily like that hour-old bread (which nourished me then) in kind, even if they appear alike. I therefore do not know if the present bread-like object will, when combined with my body, cause the effect of nourishment.¹³

The upshot is that the scientist must qualify claims about specific causal relations: it is highly probable that bread (precisely, an object deemed to bear all the sensible qualities we assign to the kind ‘bread’) will, if consumed, cause the effect of nourishment. This, to be sure, is not anything Shepherd herself would reject. Scientific claims are, in the best case, ‘but a probability, although a high one’ just because of our uncertainty about the essences of the objects these claims concern (Shepherd 1824: 119).¹⁴ The epistemic divide in science is no longer between us and necessary causal connection as such. We know nature is uniform, insofar as like causes necessarily and universally produce like effects. Rather, a gulf stands between us and the natures of objects, that is, of particular causes: ‘The only difficulty the mind has to do with, in forming a right judgment concerning its expectations of the qualities of objects, is the probability, or the contrary, whether the circumstances which formed them, are the same as heretofore or not’ (Shepherd 1824: 112).

The skeptical question raised by Hume has thus ‘shift[ed]’ to ‘the consideration of the method whereby [like causes’] presence may be detected’ (Shepherd 1824: 60). When I claim that this bread-like object will probably nourish me, what is probable is not the causal relation itself. We now know that necessary causation is a law of nature, so the bread will necessarily nourish me in like circumstances. Instead, the claim is probable insofar as I cannot be sure I have bread before me. Though if this object is, in fact, of the kind ‘bread’, it must nourish me.

So, despite our epistemological limitations, we persist and make probabilistic scientific claims. Now, it seems we need an account of what underwrites these claims’ probability. If we cannot know an object is of a particular kind, what gives us license to think that it is probably of that kind? Relatedly, why might we reasonably think one claim about an object’s expected effects is more probable than another? It seems much more probable (and reasonable to believe) that a bread-looking object will nourish me when consumed, rather than make me hungrier. Pushing further on what grounds this probability invites trouble for Shepherd.

4. The Problematic Supposition

Shepherd (1824: 118) writes that ‘the senses . . . are considered capable of nearly detecting the similarity of internal constitutions’. Elsewhere, we are told that ‘sensible qualities being similar is a presumption in favor of similar secret powers’, and, furthermore, that ‘the greater number of qualities which are exhibited as similar to the senses, the higher does the proof become, of the secret powers being also similar’ (Shepherd 1824: 128, 123). Thus, Shepherd thinks that when we are presented with objects that appear sensibly alike, this feature leads us to infer that their imperceptible essences are the same: ‘Similarity of appearance proves the presence of like proximate cause; other things therefore being equal, it proves the presence of a really similar object’ (2020: 161, n. 6).¹⁵ Together, these two shared features—sensible and essential likeness—fulfill the necessary and sufficient conditions required to deem these objects members of the same kind (though her ‘nearly’ qualification above reminds us we cannot be certain these conditions obtain, since objects’ secret powers are always unknown to us).

When we make probabilistic scientific claims, this is precisely the inference grounding their probability. The more confident we are that we have enumerated an object’s sensible qualities, the more probable it is that the object has the same secret powers (and, thus, is of the same kind) as another object with those same sensible qualities (and, hence, the more probable it is that the present object will cause those effects which we know its putative kind necessarily demonstrates). This inference allows for our observations of objects’ sensible likeness, which show that objects’ secret powers are similar ‘thus far’, to then ‘beget a proportional belief’ that leads us to ‘expect again the same, under similar circumstances’ in future instances concerning these apparently alike objects (Shepherd 1824: 61). For example, it is thought that a present object seemingly identical to the bread that nourished me yesterday will have this same effect today, just because the two objects appear alike and, as such, are thought to have the same causal powers. In this way, a reliable connection is established from the observable to the unobservable, as well as from past experience to present and future trials.

We can see this inference at work in Shepherd’s oft-cited ‘candle passage’ (2020: 160), whereby she describes ordinary inductive practices:

To the question which inquires, ‘Whence it is, the child supposes a candle will burn his finger upon a second trial, as upon a previous occasion?’ I answer, that the child considers, upon the second appearance of a candle, that the candle is a candle. He knows nothing about ‘secret powers, ‘methods of formation’, &c. but owing to the sensible qualities being precisely alike, he considers the object presented to him to be a similar one to that, which he formerly observed of the same appearance; he therefore expects it will prove itself the same in all its qualities [e.g., that it will burn him if he touches it].

The child implicitly employs the inductive principle—like causes create like effects—to reason that this candle, too, will burn him. But this glosses over an important and necessary antecedent step. The child’s reasoning must presuppose the second object is, in fact, a candle, just like the first object he encountered (Shepherd 2020: 162). We now know he cannot be sure of this. He must first infer that, if the objects are alike so far as appearances go, then they are likely of the same kind: this thing before him is probably a candle. Only then can he infer, using the inductive principle, that the putative candle will, like the first candle, probably burn him if touched. But what grounds this first inference from the sensible to the essential?

Shepherd (1824: 120, 118) says this inference is founded on a causal relation: an observed ‘frequency of repetition’, whereby objects with like sensible qualities are repeatedly shown to demonstrate like effects, leads us to think there is a ‘regularity in fact of the course of nature, which must itself be looked upon as a general Effect, from a general Cause’.¹⁶ We thereby come to form a ‘trust in the regularity of nature in forming her compound objects alike’ (Shepherd 2020: 87). This observed regularity among particular objects thereby licenses the general inference from appearance to essence sketched above.¹⁷ We ‘regularly depend on’ this apparent regularity to make inductive inferences (Shepherd 1824: 121).

The presumption that objects with like sensible qualities are of like kinds (and, thus, have like causal powers) therefore issues from an empirical claim about a causal relation observed among particular external objects. But such claims, as we have shown, are probable and cannot be necessarily demonstrated. Indeed, Shepherd (2020: 88) says we are ‘ignorant’ of this regularity’s cause. Because of Shepherd’s account of objects and her epistemic humility about them, we have no way of ascertaining whether nature’s regular state is such that sensibly alike objects are of like kinds: ‘Nature, whatever her apparent course may be, still keeps [us] “at a great distance from all her secrets”; . . . and therefore, . . . there is nothing contrary to her real course [i.e., like causes creating like effects] . . . that singular varieties should take place [by means of some secret efficient cause]’ (Shepherd 2020: 89). While it may seem that, in our experience, ‘the same sensible qualities have been regularly exhibited along with like secret powers’, for all we know, it could be the case that sensibly alike objects in fact never have the same set of causal powers (Shepherd 1824: 61). It is just that our experimentation hitherto has not found the circumstances that would reveal a difference in the objects’ powers. At a specific temperature and climate, perhaps this bread-like object will nourish, while that seemingly identical bread-like object will not. We cannot be certain that these objects are alike ‘unless we first know how they would behave in all possible circumstances’ (LoLordo 2019: 10).

But such further experimentation would in fact prove futile in resolving this worry. Should we grant (quite improbably) that we could exhaustively test two apparently alike objects in an experimentum crucis and observe that their effects never differed in any possible set of circumstances, there could remain a difference in the objects’ secret powers whose nature is unobservable to us. It is always possible that ‘some unobserved cause might creep in to alter the object, whilst appearing the same’ (Shepherd 1824: 61). These would be different objects, despite being, by all appearances, identical. There is then, no matter what, always a possibility that sensibly alike objects are essentially different.

Even if we stipulated that it was impossible for there to be such unobservable differences, meaning that we could, in principle, ascertain that the two objects have the same secret powers if they demonstrated no differing effects in all possible sets of like circumstances, this still would not be enough to guarantee the objects will necessarily have the same causal powers in the future (Shepherd 2020: 88). Indeed, nature can intervene at any time after such an experiment and alter the essences of the objects, such that they would then have different causal powers and, as such, bring about different effects: ‘Our past experience [can never] acquaint us, what latent influences, what new unseen events, what “secret powers” might be drawn from the mysterious storehouse of unperceived nature to alter our experience in future’ (Shepherd 2020: 166).¹⁸

So, once again, because of Shepherd’s epistemic humility, we can only say that nature’s regular course is such that it is highly probable that objects with like sensible qualities will be of like kinds, i.e., have like causal powers and demonstrate like effects. To be sure, this claim’s uncertainty itself need not be a problem for Shepherd. We know Shepherd accepts that scientific claims must be probabilistic because of our epistemic limits. However, this claim in question—about nature’s regularity—is what grounds those scientific claims’ probability. It therefore needs a distinct ground to justify its own probability. When we probe further here, things become epistemologically treacherous, at least by Shepherd’s lights. Indeed, this claim (the regular course of nature is such that objects with like sensible qualities will probably be of like kinds, and, as such, will probably have like causal powers and effects) is not rationally grounded.

Nothing about the causal principle—specifically, the inductive principle derived from it that like causes necessarily create like effects—entails this claim about nature’s regularity. This says something more about nature than the causal principle guarantees. We can easily picture a state of affairs in nature where sensible qualities do not serve as generally reliable signs of essences. That an object looks like bread would, in this world, not be a good reason to think it will probably have the effects expected of bread, e.g., nourishment when eaten. On this model of nature, objects’ secret powers would have to be altered more frequently to account for the unreliability of sensible qualities as a sign of objects’ kind membership. It could even be the case that apparent likeness among objects is a good reason to expect that those objects are probably not the same kind. In such a world, I would see a bread-like object and suspect that it probably will not nourish me, just because that other bread-like object I consumed an hour ago nourished me.

Either of these hypothetical regular states of affairs in nature would be just as compatible with the causal principle as the state of affairs Shepherd has us assume to support our scientific claims’ probability. Such worlds seem dramatically different from how we take ours to be, but neither scenario sketched above would violate the laws that like causes create like effects or that nothing can arise uncaused. Moreover, these two principles alone do not give us reason to endorse any of these models of nature over another, or to think that one is more probable. An appeal to experience cannot be definitive here either because of our epistemic limitations. We have no way of knowing for certain that this ‘regularity’ we perceive in nature’s operation is not in fact illusory. And, even granting that we could affirm that there is such a regularity, it would still potentially be subject to future change. Thus, that the regular course of nature is such that objects with like sensible qualities will probably be of like kinds is, at best, a contingent, and not necessary, feature of nature.¹⁹

The result is that the inference on which the probability of scientific claims is founded—that we can reliably infer from an object’s sensible qualities what kind of object it is, and, as such, what causal powers it has—bottoms out in a supposition about the state of nature that is extra-rational, i.e., not grounded by the causal principle. Hume’s epistemological concern thus looms large. Shepherd struggles to demonstrate how, in our inductive reasoning about particular causal relations among external objects, we rationally move from the proposition that ‘such an object has always been attended with such an effect’ to the separate claim that ‘other objects, which are, in appearance, similar, will be attended with similar effects’ (Hume 2007b: 35).

Shepherd’s problem is that induction, to be of any practical use, requires that we know (or, at least, reasonably suspect) an object is of a certain kind. While the causal principle guarantees induction’s basis—like causes create like effects—we can only make an inductive inference if we think that a given object, or cause, is of the same kind as one in the past. If I do not know that I have two like objects, then I do not know if this bread-like object before me will behave like yesterday’s bread. Here, her hand is forced. Because of our ‘inability’ to judge objects’ kind from ‘the mere appearance of their sensible qualities’ (Shepherd 1824: 117), we must suppose that nature is regular in a way that makes sensible likeness a reliable sign of essential likeness. Otherwise, past observations of objects would have no probative weight. Without it, the child’s first observation of the candle would tell him nothing about the causal powers of the second candle-like object he sees.

To get induction off the ground, then, we must make a supposition about nature’s regularity that licenses an inference from an object’s appearance to its kind: from what I know to what I cannot know. My belief that this present bread-like object is, in fact, bread and will thereby nourish me is based on something I cannot know through reason and is thus extra-rational. As stated before, Shepherd (2020: 88) admits that we are ‘ignorant of the cause for the regularity’ that underwrites this inference. However, I wish to say that our ignorance here in fact goes a step further. We cannot be certain that there is such a regularity in the first place—or even confident that there probably is such a regularity—even though we presuppose that there is one when reasoning inductively, as we do when making scientific claims. If my discussion has been successful, it should be clear how this extra-rational leap in scientific inquiry arises as a direct consequence of Shepherd’s epistemic humility and her account of kinds.

5. Conclusion

What to call this supposition and how to explain how we come to have it—and what might ground it, if not reason—I leave open. One possibility unmentioned thus far is that it is reason, albeit a conception of reason distinct from the Humean one presupposed by this discussion (see note 8). For example, Landy (2024: 81) argues that Shepherd can be interpreted as giving a generally Kantian account of reason: reason is in part constitutive of our representations of objects, rather than just a ‘process which operates over these perceptions’ after the fact (Hume’s view, according to Landy).²⁰

What Shepherd believes reason to be is important for present purposes, insofar as one’s account of reason has potential implications for what might then be sufficient to justify reason’s conclusions. Maybe Shepherd fails by Hume’s lights, but what about her own? Perhaps these are distinct assessments, and the shortcomings I charge her with are, in fact, misplaced. Landy thinks Shepherd’s Kantian view can explain why she says ‘an argument . . . in fact is never made’ in reply to Hume’s challenge that we explain the reasoning underlying induction (Shepherd 1824: 122). This ‘conclusion’ that certain sensible qualities will follow from the experience of other sensible qualities is ‘already contained in the experience of the object’ (Landy 2024: 93; emphasis mine). The proof sought is not found in a chain of ex post facto reasoning but is instead ‘nothing more than the clear perception of a universal relation’ (Landy 2024: 93). If Shepherd restores reason to science, then it is not the sort of reason susceptible to demonstration through argument. We have a ‘process of reasoning’, whereby ‘neither before nor after experience, the particular kind of Effects from given Causes should be discovered’ (Shepherd 1824: 139). One might then wonder if my work here—evaluating Shepherd as having to provide a chain of reasoning in support of induction—has been a category error of sorts.

Such a conclusion, however, is far from entailed by the above. For one, there is no critical consensus on what Shepherd thinks reason is. Landy (2024: 80) notes that ‘Shepherd writes very little that makes explicit what she takes reason to be’ and says his Kantian reading of her on reason is merely ‘plausible’. And other scholars, in contrast, hesitate to read Kantianism into Shepherd.²¹ But more importantly, even if Shepherd is a Kantian about reason (or otherwise non-Humean), that need not preclude her from giving the kind of justificatory account sought by Hume: to ‘produce’ a ‘chain of reasoning’, or a ‘demonstrative argument’ (Hume 2007b: 35; 2007a: 62).²² A descriptive account of our mental processes is distinct from an analysis of the associative reasoning that may issue from those processes.²³ The causal principle could be a ‘structural aspect of experience itself’ beyond demonstration (Landy 2024: 93), but the reasoning which facilitates that principle’s application to scientific inquiry is a different matter. Landy (2024: 93), glossing Shepherd, rightly observes that the conclusion that a given object must produce certain effects is demonstrative only when we have first ‘confirmed the causal powers of an object . . . and we experience the sensible qualities that we know are the effects of that object’. Justifying inductive reasoning requires explaining how we get such ‘confirmation’, and Shepherd is not entitled to it by the causal principle alone. She needs some kind of argument: a chain of reasoning.

Shepherd’s above remark about the absence of an ‘argument’ should also be considered in context. Shepherd cites Hume as seeking an argument supporting the ‘reasoning . . . [that leads us] from the mere sensible qualities of things to expect their future Effects’ (Shepherd 1824: 122; emphasis mine). But Shepherd thinks that argument would be a fool’s errand. We do not reason from sensible qualities per se but rather take them as ‘signs of the secret powers’ (Shepherd 1824: 123), a distinction Shepherd is at pains to draw throughout her work (see, e.g., 1824: 102, 109, 113; 2020: 44, 69, 77; see also note 15). Thus, while Shepherd rebukes any attempt to make this argument to Hume, using his particular framing of induction, that need not preclude an argument justifying her account of inductive reasoning. Textual evidence suggests this is just what she seeks to do.²⁴

So, I think it is reasonable to evaluate Shepherd as I have, even if Landy’s work makes it plausible that my framing of Shepherd’s response to Hume is misguided. And if I am wrong about all this, that still need not disturb my conclusion that Shepherd fails to silence Hume. If Shepherd is in fact uninterested in providing a chain of demonstrative reasoning, then it would seem she is no longer playing the game by Hume’s rules. If she succeeds, then she succeeds at a different game, perhaps resisting or sidestepping but not directly responding to Hume’s challenge as such. Regardless of whatever Shepherd thinks she is trying to do—or actually does—she cannot give a satisfactory reply to Hume, at least by his lights. This is my central point.²⁵

Besides reason, other options remain for explaining what might underpin the supposition at induction’s heart. Perhaps it is akin to Hume’s account of how we are naturally disposed to believe in nature’s regularity. We cannot help but think this is the way things are.²⁶ Alternatively, it could be parsimony. One may think nature is essentially simple and that we need not posit more natural kinds than those that objects’ observed qualities render necessary.²⁷ A third possibility, somewhat like Landy suggests, is that it is a foundational self-evident principle known via intuition but not justifiable through reason (at least in the Humean sense of the word).²⁸ In any case, however we understand this supposition’s ground, it is neither the a priori causal principle nor a posteriori knowledge found on its basis. Constrained by her epistemic humility and account of natural kinds, Shepherd’s science assumes a proposition about nature that cannot be demonstrated through reason.

However, it seems science could not accomplish much without this supposition as a ‘guiding circumstance by which to form a judgment of the future’ (Shepherd 2020: 77). Lacking reason to believe that objects before us are the same kind as those prior, past observations of objects would amount to isolated, one-off causal relations that give no future insights: something that is antithetical to scientific inquiry, which Shepherd (2020: 156) thinks of as ‘enabl[ing] us to understand and imitate nature’. Nevertheless, if my reading of Shepherd is correct, Hume’s problem of induction still lurks, albeit in an attenuated capacity. Though this concession to Humean skepticism is subtle, it shows that Shepherd has not entirely extricated science from Hume’s grasp. Her account of natural kinds and epistemic humility make this an unavoidable conclusion.

Competing Interests

The author has no competing interests to declare.