Grounding and the Epistemic Regress Problem

Abstract

Modal metaphysics consumed much of the philosophical discussion at the turn of the century, yielding a number of epistemological insights. Modal analyses were applied within epistemology, yielding sensitivity and safety theories of knowledge as well as counterfactual accounts of the basing relation. The contemporary conversation has now turned to a new metaphysical notion – grounding – opening the way to a fresh wave of insights by bringing grounding into epistemology. In this paper, I attempt one such application, making sense of the epistemic regress problem in terms of grounding. I argue that the relation that generates the epistemic regress is a grounding relation, showing that grounding can make sense of proposals by epistemic foundationalists and charting the course for similar applications to epistemic coherentism and epistemic infinitism. If it is right that grounding is involved in the epistemic regress, this points the way forward both for epistemologists and metaphysicians, revealing the prospects of solutions to the epistemic regress problem while providing grounding advocates with yet another example of grounding with which to theorize.

1 Introduction

Innovations in metaphysics often give rise to insights within epistemology. One recent example is the wide-ranging influence of work on the metaphysics of modality. Pioneering work on modal logic made reasoning with modal notions more tractable, paving the way for progress on the metaphysics of possible worlds. Work on possible worlds, then, gave rise to novel applications ranging from analyses of meaning to the semantics of counterfactuals. The explosion of work on modality was not without its critics, most notably Quine, who thought that modal notions were too confused and obscure.¹ Despite these worries, the modal revolution gradually came to command a following due its theoretical fruitfulness.

A new revolution is now underway within metaphysics. A number of leading theorists have begun to argue for the notion of grounding, a relation they claim arises across philosophy in places as distinct as logic and normativity:²

• (1) Logic: The fact that snow is green or lying is wrong $\Lleftarrow$ the fact that lying is wrong³

• (2) Analysis: The fact that Jones is a bachelor$\Lleftarrow$ the fact that he is an unmarried male

• (3) Existentials: The fact that there is an object in the room $\Lleftarrow$ the fact that there is a chair in the room

• (4) Normative-Non-Normative: The fact that lying is wrong $\Lleftarrow$ the fact that lies in general produce worse consequences than truth-telling.⁴

• (5) Determinate-Determinables: The fact that the shirt is red $\Lleftarrow$ the fact that it is crimson

The development of grounding mirrors a number of the early movements in modal metaphysics. (Fine 2010, 2012a, b) has published extensively on the logic of ground. Rosen (2010) and Schaffer (2009) have written on the metaphysics of grounding, and numerous applications have already been found: Schaffer (2010a) employs grounding considerations in defending monism, Dasgupta (2014b) claims that grounding should be used to formulate physicalism, and Cameron (2018) takes the truthmaking relation to be an instance of grounding. Grounding is also not without its detractors. Like Quine on modality, Hofweber (2009) and Daly (2012) find the notion of grounding too obscure and ill-defined, while Wilson (2014) thinks that all the work grounding theorists want to accomplish can be done by other dependence relations. Like with modality, this has not prevented grounding theorizing from moving forward. It is partially this moving forward, in fact, that has given grounding further traction, just as it was a theoretical fruitfulness that cemented the impact and staying power of the modal revolution. What is left, then, is to apply grounding within epistemology. In doing so, I will not be answering the grounding skeptics directly,⁵ but nevertheless motivating the adoption of grounding idioms, as the more diverse uses to which grounding is put lend credence to the usefulness of thinking in terms of grounding more generally.

In this paper, I will argue that the relation that links justified beliefs in the epistemic regress can be characterized as an instance of grounding, applying this idea to epistemic foundationalism. I will then argue that these insights might also extend to epistemic coherentism and infinitism as well, though the full case for these additional applications will need to take place in future work. In Sect. 2, I examine some historical formulations of epistemic foundationalism, coherentism, and infinitism, noting that they centrally employ paradigmatic grounding talk, making it appear that determination factors in the epistemic regress problem. I then clarify in Sect. 3 that, even though basing is a relation between justified beliefs, I do not intend to be giving a theory of the basing relation. I then give an argument from transitivity in Sect. 4, contending that grounding makes sense of the transitive justificatory link between beliefs in the epistemic regress. For this argument, I show that the relation between the normative and the non-normative is transitive with the relation between justified beliefs, arguing that this gives strong reason to think that the epistemic regress problem is generated by grounding. I then conclude in Sect. 5 by suggesting that, if grounding is what is at issue in the epistemic regress problem, this has implications for what solutions to the regress are actually viable. In this way, it could well be that epistemologists have a lot to learn from metaphysicians, but it could also be that advocates of grounding themselves might gain new insights by taking the epistemic regress problem to feature the same sort of metaphysical priority.⁶

A note before we begin. In this paper, I will be approaching the epistemic regress problem by considering questions that have not always come to the forefront of discussions between epistemic foundationalists, coherentists, and infinitists. These debates are often focused on the nature of justification or what it takes to be a good epistemic reason, whereas I will primarily be concerned with the metaphysical link between the justified and their justification-makers. This approach mirrors the way that these linking questions been brought to bear on a number of other philosophical issues. Take, for instance, Dasgupta’s (2014b) application of grounding to physicalism. Instead of just directing his attention to the concept of physicalism itself, Dasgupta brings new insights to the physicalism debate by discussing the link that allows the physical to generate everything else. Just as focusing on these types of linking questions has brought work on grounding into conversations about monism, physicalism, and truthmaking, I hope that this approach will also shed light on discussions surrounding the epistemic regress problem.

2 The Regress Problem in Epistemology

The history of the epistemic regress problem traces back at least to Aristotle. This, in itself, is suggestive, as Aristotle is usually credited with founding the project of grounding metaphysics.⁷ The regress concern gets started with the plausible thought that, in order for a belief to be justified, it must be based on other justified beliefs:

• Regress Generator - For all beliefs b, b is doxastically justified for a person S only if S has another belief or beliefs $b_{1}$...$b_{n}$ such that (i) $b_{1}$...$b_{n}$ are doxastically justified for S, (ii) $b_{1}$...$b_{n}$ provide a good epistemic reason for S to believe b, and (iii) S appropriately bases their belief b on $b_{1}$...$b_{n}$

There are a couple things to note about Regress Generator. To begin with, if beliefs $b_{1}$...$b_{n}$ are going to play a role in providing justification for a belief b, those beliefs themselves must be justified. Suppose that I come to the conclusion that it rained earlier in the day. I infer this from my beliefs that the sidewalk is damp and that rain was in the forecast. If these further beliefs are justified, then I am also justified in believing that it rained. Justification for these beliefs can come in many forms – I could see the damp sidewalk or I could learn through testimony that there is a high chance of rain. However, if I have no evidence for these other beliefs, that the sidewalk is damp or that rain was in the forecast, then my belief that it rained earlier is also not justified.

It is not enough, however, for $b_{1}$...$b_{n}$ to be justified. They must also collectively provide a good epistemic reason or reasons to believe b. Even if I am justified in believing that the sidewalk is damp and that rain was in the forecast, if I infer from this that my favorite basketball team will win their game this evening, I am not justified in this further belief. Unless the game is played outdoors and my team is particularly good in the rain, there is simply no connection between the rain from this morning and my team winning the game this evening. The rain does not make it any more probable that my team will win this evening, and so these beliefs do not provide good epistemic reasons for believing that my team will win.

All of these points seem plausible. In order to be justified, a belief b should not only be supported by other justified beliefs, but those beliefs must provide good epistemic reasons to believe b. If the Regress Generator principle is accepted with full generality, however, the specter of a vicious regress looms. In order for b to be justified, it must depend on further justified beliefs $b_{1}$...$b_{n}$, and then $b_{1}$ must also depend on some justified $b_{2}$...$b_{n}$, and so on. But, if we also accept that human agents are incapable of entertaining an infinitely long series of inferences, then skepticism concerning justification, the thought that ultimately none of our beliefs are justified, is the natural result. If, for our beliefs to be justified, Regress Generator requires something of which we are not capable, then it turns out that none of our beliefs are justified. Another way to complete the chain of beliefs created by the Regress Generator would be to engage in circular reasoning. The problem, of course, is that circular reasoning seems like a paradigmatic example of unjustified inference, leaving us without justification for any beliefs that are adopted on its basis. Appealing to circular reasoning thus also leads to skepticism about justification.⁸

Three responses that have emerged to the challenge of skepticism are epistemic foundationalism, coherentism, and infinitism.⁹ Let’s look at these positions in terms of how they interact with the Regress Generator principle:

• Epistemic Foundationalism: The Regress Generator principle is false – not all beliefs depend on other beliefs for their justification. In particular, there are foundational beliefs that are justified even though they are not based on any other beliefs

• Epistemic Coherentism: The Regress Generator principle is true – all justified beliefs are justified by other beliefs. This does not necessarily result in skepticism, however, because beliefs can form a web of mutual support, making them all justified by other beliefs

• Epistemic Infinitism: The Regress Generator principle is true – all justified beliefs are justified by other beliefs. This does not necessarily result in skepticism, however, because it is possible to be justified via an infinite regress of beliefs

Epistemic foundationalists deny the Regress Generator, taking the regress problem to show that there are some beliefs that are justified apart from other justified beliefs. Epistemic Coherentism and Epistemic Infinitism both accept the Regress Generator. Coherentists affirm Regress Generator and take it that the structural features of a set of beliefs determines whether that set is in fact justified, allowing that those structural features can include aspects of circularity. Infinitists accept Regress Generator and contend that human cognitive limits are not problematic after all, that skepticism does not result from taking the structure of justified belief to be an ever-descending chain of justification-makers. Each of these positions, then, contra skepticism, captures the common sense thought that we are justified in at least some of our beliefs.

A question that arises for all of these positions is how, precisely, justified beliefs are linked with their justification-makers. Even though Epistemic Foundationalism does not accept the Regress Generator principle in full generality, foundationalists do accept the principle when it comes to non-foundational beliefs. Non-foundational beliefs depend for their justification on foundational justified beliefs, requiring that the epistemic foundationalist accept a modified version of Regress Generator. But what kind of relation holds between justified beliefs and all the things that go into justifying them? Close attention to historical formulations of epistemic foundationalism, coherentism, and infinitism reveal that they are rife with grounding talk. Consider, for example, Sosa’s (1980) characterization of epistemic foundationalism in his seminal paper “The Raft and the Pyramid”:

Classical foundationalism in epistemology is the view that

• (i) every infallible, indubitable belief is justified

• (ii) every belief deductively inferred from justified beliefs is itself justified

• (iii) every belief that is justified is so in virtue of (i) or (ii)¹⁰

And also his formulation of epistemic coherentism:

A belief X is justified for S in virtue of membership in a coherent set only if S is justified in believing

• (1) that most of his beliefs with the property of thus cohering are true, and

• (2) that X has that property¹¹

Sosa chooses to formulate both epistemic foundationalism and coherentism using ‘in virtue of,’ one of the standard natural language locutions used to express a grounding relationship. Further argumentation will be needed to show that Sosa’s uses of this phrase do in fact pick out the grounding relation, but there is indisputably some sort of dependence going on between justifiers and the justified within epistemic foundationalism and coherentism. This point, of course, also extends to epistemic infinitism, as can be seen in the following, even earlier passage from I.T. Oakley:

• Few seem to have held the view that a belief might be justified in virtue of its membership in an infinite series of beliefs, each depending on its successors¹²

Oakley characterizes epistemic infinitism using both the language of dependence as well as ‘in virtue of,’ making it clear that he takes infinitism about justification to be making some sort of dependence claim. Thus, all the responses to skepticism regarding the epistemic regress problem can be understood in terms of a dependence relation.

So we have dependence, but do we have grounding? Grounding theorists most often identify instances of grounding through the use of several natural language locutions, including ‘because,’ ‘makes it the case,’ ‘in virtue of,’ and ‘depends.’¹³ If the grounding thesis is correct, then, we can formulate the second clause of Sosa’s formulation of epistemic foundationalism in all of the following ways:

• (6) A non-basic belief b is justified in virtue of being inferred from justified basic beliefs $b_{1}$...$b_{n}$ that provide good epistemic reasons to believe b

• (7) A non-basic belief b is justified because b is inferred from justified basic beliefs $b_{1}$...$b_{n}$ that provide good epistemic reasons to believe b

• (8) A non-basic belief b’s being inferred from justified basic beliefs $b_{1}$...$b_{n}$ that provide good epistemic reasons to believe b makes it the case that b is justified

• (9) A non-basic belief b depends for its justification on being inferred from justified basic beliefs $b_{1}$...$b_{n}$ that provide good epistemic reasons to believe b

• (10) The fact a non-basic belief b is justified $\Lleftarrow$ the fact that b is inferred from justified basic beliefs $b_{1}$...$b_{n}$ that provide good epistemic reasons to believe b

(6)-(9) all seem to get epistemic foundationalism right. Just as for Sosa and Oakley, these ways of speaking come naturally when discussing the epistemic regress problem. Foundationalists hold that justified non-basic beliefs depend for their justification on the justification of basic beliefs. In this limited sense, epistemic foundationalists can sign on for Regress Generator – non-basic beliefs inherit their justification from basic beliefs. The question I’ll be interested in, then, is what relation brokers this inheritance of justification.

My positive thesis is that the inheritance that occurs in Regress Generator is best explained by (10), that the link between justification makers and the justified in the epistemic regress is an instance of grounding. The best way to understand the thesis is by comparing it to the positing of grounding relations across philosophy. The central issue in formulating physicalism is how to understand the dependence relation, call it the physicalist relation, that links the mental and the physical. As already mentioned, the most recent attempt at capturing physicalism has been to identify the physicalist relation as a grounding relation. Similarly, the truthmaking relation has received many treatments, being interpreted as an instance of the necessitation and supervenience relations, before receiving an interpretation on which truthmaking is an instance of grounding. My hope is to expand the purview of grounding to the epistemic regress problem – just as there are those who take the physicalist and the truthmaking relations to be instances of grounding, I will be defending that the relation in the justification regress is a grounding relation.

3 Basing, Grounding, and ‘Grounds’ in Epistemology

3.1 Basing is Not Grounding

Even though (6)-(9) provide accurate characterizations of epistemic foundationalism, this is not enough to conclude (10), that what is at issue in the epistemic regress argument is a grounding relation. There are a number of other notions, after all, that can be picked out by grounding theorists' pet phrases. One such notion is the basing relation. Suppose, for example, that I based my belief that Joe ate all the cookies on the fact that he was the last person in the kitchen. The idea can easily be expressed using ‘because’:

• (11) Joe ate all the cookies because he was the last person in the kitchen

Perhaps, then, what is at issue with formulations of epistemic foundationalism is the basing relation. This is not implausible – if I infer a non-basic belief from a basic belief, then I am basing my non-basic belief on a basic belief. Foundationalism could, then, be described in terms of basing, and no reference to grounding would be required.

If it is correct that basing can factor in the epistemic regress, one way to defend the thought that grounding is involved in the epistemic regress would be to argue that the basing relation is an instance of grounding. This proposal naturally flows from the recognition that many relationships that were once thought to be modal are now taken to be paradigmatic instances of grounding. Truthmaking has often been characterized in terms of necessitation, but recent developments now take truthmaking to be an instance of grounding. Similarly, the modal account of essence has recently been challenged by the thought that modal truths are instead grounded by essential truths.¹⁴ A similar progression might be tempting for the basing relation. To the extent that counterfactual accounts of the basing relation are unsatisfactory,¹⁵ perhaps it will also be superseded by a grounding account of basing.

Lending credibility to a grounding take on basing is the fact that ‘grounds’ in epistemology is often used precisely in a basing role. Many philosophers use ‘grounds’ as simply synonymous with ‘reasons for which a person believes,’ as in the following passages:

• Roger White – “In deliberating myself about whether to believe p, in seeking an answer to the question of whether p, I will naturally consider what reasons or grounds I have to suppose that p.”¹⁶

• Robert Audi – “Reasons for believing something are one or another kind of ground for believing it.”¹⁷

• Conor McHugh – “If KK is true, to know p, you must first have positive grounds to believe not only that p, but also that you have warrant to believe p.”¹⁸

In all of these examples, ‘grounds’ is used to pick out the reasons that one takes as supporting their belief, the reasons on which they base that belief. So long as we are discussing doxastic justification, these reasons play a role in justifying the belief that they support. There are thus a number of reasons to take the basing relation to be grounding.

Unfortunately, the seemingly plausible thought that basing is grounding runs into serious difficulties. Grounding is transitive, so if we suppose (1) and (4), then it is also the case that the fact that snow is green or lying is wrong $\Lleftarrow$ the fact that lies in general produce worse consequences than truth-telling.¹⁹ The basing relation, however, is not transitive with the grounding relation. Consider the following example. Suppose that (4) is true, that lying is wrong because lies in general produce worse consequences than truth-telling, making (4) able to be formulated using the grounding. Now take Jane, who believes that James’s lie was wrong because she believes that lying is wrong, an instance of the basing relation. Due to grounding’s transitivity, if the basing relation is the grounding relation, we would expect this instance of the grounding relation to be transitive with this instance of the basing relation. But this isn’t right - this is not enough to establish that Jane believes that James’s lie was wrong because lies in general produce worse consequences than truth-telling. Jane could very well believe that lies are wrong because they mislead the hearer, or because lies fail to respect others, or any other reason offered by normative theorists, disagreeing with the grounding claim that we are assuming in (4). Thus, given that the grounding is not transitive with the basing, it would be a mistake to take basing as grounding.

There are other reasons to think that basing is not a grounding relation. One rather cursory reason is that most theorists hold that grounding holds between two facts, whereas basing has belief as at least one relata.²⁰ Basing takes Jane’s beliefs and connects it to other beliefs and experiences, making basing unsuited as a connective glue between two facts. An even more compelling way basing differs from grounding is that evidential support can proceed in both directions across a grounding relation. Take (5), the determinate-determinable link between red and crimson. Learning that a shirt is crimson is a good reason (indeed, a necessitating reason) for believing that the shirt is red. This is in step with the original directionality inherent to the grounding link between determinates and determinables. But evidential support can also run the other way. Learning that a shirt is red is a reason to think it is crimson, as it makes it more likely that the shirt is crimson given that hues of green and blue are ruled out. Likewise, it makes sense to say that Jane believes that the shirt is crimson because it is red, just as it’s possible that Jane believes the shirt is red because it is crimson.²¹ Finally, basing is not grounding because metaphysical grounds necessitate the grounded, whereas epistemological grounds do not. Necessarily, if a shirt is crimson, then the shirt is red. Reasons, however, never necessitate that someone holds a particular belief. Beliefs, after all, can always be unjustified, so even if I have overwhelming reason to think that the shirt is red – I know that the shirt is crimson and that crimson is a hue of red – this does not necessitate that I believe that the shirt is red.

3.2 Basing Does Not Generate the Epistemic Regress

Because grounding is not identical to basing, those that want to defend that grounding generates the epistemic regress will also have to argue that basing is not what generates the regress of justification. One way to get this result is to see that the epistemic foundationalist position is not a claim about how we should base our beliefs – foundationalists are not claiming that we should base our belief that there are justified non-basic beliefs on the belief that there are justified basic beliefs, or that we should believe the latter because of the former. Instead, when epistemic foundationalists claim that non-basic beliefs are justified because they are inferred from basic beliefs, they are instead making a claim about the source of that justification, what gives rise to the fact that the inferred belief is justified.

Another way to argue that basing does not generate the epistemic regress problem is because basing does not always transmit justification. If basing always transmits justification from $b_{2}$ to $b_{1}$, it would not be possible to have unjustified beliefs that were based on justified beliefs. It is possible, though, that $b_{1}$ can be based on a justified $b_{2}$ without inheriting justification from $b_{2}$. Instead, something needs to be added to basing in order to secure that $b_{1}$ is justified. In this vein, the right way to think about basing is that it is one of the ingredients in the grounds of justified non-basic beliefs, not the relation itself between justification-makers and the justified. Recall the formulation of the foundationalist position in (6), that a non-basic belief is justified in virtue of being inferred from justified basic beliefs. Part of what grounds justified non-basic beliefs is that they are inferred from the basic, a way of appropriately basing one belief on another. The fact that appropriate basing obtains, then, is part of the grounds themselves, allowing us to reformulate (6) as (12):

• (12) A non-basic belief $b_{1}$ is justified in virtue of being appropriately based on justified basic beliefs $b_{2}$...$b_{n}$ that provide good epistemic reasons to believe b

Basing, then, is not the ‘in virtue of’ relation that connects the justification of basic and non-basic beliefs, but part of the full grounds for justified non-basic beliefs, making it one of the ingredients in Fig. 1 necessary for justified belief.

Fig. 1

Ingredients of Justified Belief

4 The Transitivity Argument

Now that we have done the necessary ground clearing, distinguishing between basing, epistemic grounds, and metaphysical grounds, we can now make the positive argument that the epistemic regress relation is an instance of metaphysical grounding. Grounding is transitive, and so the relation that links justified beliefs should be transitive with instances of grounding. Grounding theorists are unanimous in taking grounding to relate the normative and the non-normative, and so one way to make the positive case for the relation that generates the epistemic regress being an instance of grounding is to show that it is transitive with the instance of grounding that relates the normative and the non-normative. In this section, I do precisely that, exploiting the normative-non-normative connection between justified basic beliefs and their justification makers to show that the relation at issue in the epistemic regress is transitive with the grounding relation.

One of the most prominent instances of grounding is the connection between the natural and the normative. Though it has long been held that moral properties supervene on the natural, a position held as early as Sidgwick (1907), Moore (1922), and Ross (1930), it became clear even before grounding was explicitly on the scene that supervenience was not completely adequate for describing the link between natural properties and moral properties. Consider, for example, the following passage from Dancy (1981) contrasting supervenience and his “resultance” dependence relation:

• “Only resultance provides a genuine sense in which the non-moral properties fix the moral properties: Supervenience only tells us that once they are fixed, they cannot change without a change in the non-moral properties and that if all those properties recur in another action they would ‘fix’ the same moral property.”²²

Here Dancy emphasizes the point that, though supervenience can characterize how one set of properties might ebb and flow alongside another set of properties, this does not yet specify that the first set fixes the second. Supervenience is not the glue that holds together the natural and the normative – it just results from the relation that does. Dancy thus also points out that “supervenience is not really a relation between the moral and the non-moral properties of an action, but is...a consequence of whatever relation holds in the particular case between moral and non-moral properties.”²³

Grounding theorists take Dancy’s point, arguing that grounding is the resultance relation that makes it the case that supervenience holds between the moral and the non-moral. It is widely held that the relationship between the natural and the normative is a paradigmatic instance of grounding, and so this will be the first premise of the transitivity argument:²⁴

• (13) Normative-Non-Normative– Normative properties are (at least partially) grounded in non-normative properties²⁵

One example of (13) is our earlier grounding example (4), the rule consequentialist claim that the wrongness of lying is grounded in the outcomes to which lying typically leads. Here, we get a specific example of how such grounding might obtain. Even if rule consequentialism is incorrect, accounts of right action will be constrained by (13) in that fixing the non-normative properties will then fix the normative properties.

The Normative-Non-Normative link extends past moral normativity to include epistemic normativity as well. An important component of the epistemic foundationalist program is not just to argue that there are justified basic beliefs, but also to offer an account of how they are justified. To this end, epistemic foundationalists offer many candidate non-normative properties on which the justification of basic beliefs is thought to supervene. A short survey of some of the possibilities includes the following:

A basic belief b is justified due to its

• being clear and distinct²⁶

• being self-evident²⁷

• being produced by direct acquaintance²⁸

• being produced by a reliable process²⁹

All of these proposals attempt to explain why basic beliefs are justified.³⁰ They do so by offering a non-normative property on which immediate justification is thought to supervene – Descartes stumps for clarity and distinctness, Reid for self-evidence, and Russell for direct acquaintance.

If grounding theorists are correct about the Normative-Non-Normative link, then the supervenience in question with justified basic beliefs is secured because these non-normative properties ground the normative fact that b is justified. Given (13), we can generate grounding claims that propose to explain how it is that basic beliefs are justified. For example, Descartes’s doctrine of clear and distinct ideas can be understood as (14):

• (14) S’s basic belief $b_{2}$ is justified $\Lleftarrow$ $b_{2}$ is clear and distinct

Assuming Cartesian foundationalism is correct, (14) captures how it is that basic beliefs are justified, taking their justification to be grounded in the fact that they are clear and distinct. From (14), we can then make the grounding claim in (15), going from purely non-normative grounds to a normative groundee:

• (15) S’s non-basic belief $b_{1}$ is inferred from a justified basic belief $b_{2}$$\Lleftarrow$ $b_{1}$ is inferred from the clear and distinct basic belief $b_{2}$

Even with an account of the justification of basic beliefs, however, (15) is not yet complete. After all, S cannot simply infer $b_{1}$ from any old $b_{2}$ – the proposition expressed by $b_{2}$ must also qualify as a good epistemic reason to adopt $b_{1}$.

Now, just like there are non-normative accounts of what makes a basic belief justified, there are also non-normative accounts of when one proposition provides evidence for another proposition. Some take it that, in order for p to be evidence for q, p must raise the probability that q.³¹ Others take it that any proposition p that is entailed by q is evidence for q.³² We will not adjudicate between these accounts here, but will simply note that, in order for (15) to be complete, it must include an account of evidence as well as an account of the justification of basic beliefs:

• (16) S’s non-basic belief $b_{1}$ is inferred from a justified basic belief $b_{2}$ that provides a good epistemic reason to believe $b_{1}$$\Lleftarrow$ $b_{1}$ is inferred from the clear and distinct basic belief $b_{2}$ that raises the probability of $b_{1}$

With (16), we have non-normative grounds for both of our normative properties, being justified and being a good epistemic reason. The decisions we have made about which non-normative properties to choose, being clear and distinct and being probability-raising, are not necessary to demonstrate that the epistemic regress is a matter of grounding, but assuming these will suit our purposes of laying out the transitivity argument.

If the relation that generates the epistemic regress is grounding, then by the transitivity of grounding, it should be possible to chain (7) with (16). Recall that the epistemic foundationalist claim was the following:

• (7) S’s non-basic belief $b_{1}$ is justified because $b_{1}$ is inferred from the justified basic belief $b_{2}$ that provides a good epistemic reason to believe $b_{1}$

If (7) actually is an instance of grounding, then the relation that links the two facts should be transitive with the grounding relation in (16), which we can see is the correct prediction with (17):

• (17) S’s non-basic belief $b_{1}$ is justified $\Lleftarrow$ $b_{1}$ is inferred from the clear and distinct basic belief $b_{2}$ that raises the probability of $b_{1}$

Here, we have the left-hand side of (7) paired with the right-hand side of (16), creating a grounding claim that seems intuitively correct. Apart from considering how we derived (17), it simply seems plausible on its face, making it reasonable to think that the relation from the epistemic regress and the grounding relation are transitive, thus providing a positive case that the relation in the regress of justification is grounding.

Beyond just providing this positive argument that the regress of justification features the grounding relation, we can also show that the relation found in the epistemic regress problem does not suffer from any of the same deficiencies as the basing relation did in Sect. 3. For starters, the relation that figures in the epistemic regress relates facts instead of just beliefs. In (7), for instance, the relation that features in the epistemic regress is flanked by facts, the fact that S’s non-basic belief is justified and the fact that S’s non-basic belief was inferred from a basic belief. And, unlike the basing relation, we have seen that the relation of the epistemic regress is transitive with the grounding relation. Another way in which the relation from the epistemic regress differs from the basing relation is that it can necessitate. With basing, we pointed out that being based on a particular epistemic reason never necessitates what someone believes. With the epistemic regress relation, however, if S appropriately bases their belief $b_{1}$ on a justified belief $b_{2}$ that provides a good epistemic reason to believe $b_{1}$, then this necessitates that S’s belief $b_{1}$ is justified. Thus, not only do we have a positive reason to think that the relation from the epistemic regress is grounding, we can also see that it does not suffer from the same defects as the basing relation.

5 Future Work

If it is correct that the epistemic regress problem is a matter of grounding, this is by itself an interesting discovery. The observation is made more intriguing, though, by its possible upshots both for proposed solutions to the epistemic regress problem and theorizing about grounding. One way of exploring the connection would just be to read results off of the grounding literature onto solutions for the epistemic regress problem. It might be, for instance, that if we assume certain characteristics of the grounding relation, Epistemic Infinitism and Epistemic Coherentism are untenable. It might also be that this is the wrong direction to argue – perhaps, in generalizing about the grounding relation, advocates of grounding have not extended their list to a wide enough range of examples. If the regress of justification is taken into account, then the independent plausibility of epistemic infinitism and coherentism might suggest that what needs to be revised is our understanding of grounding. Either way, the upshot of realizing that the epistemic regress problem is a grounding problem is that it opens new avenues for discussion between epistemologists and metaphysicians. In what follows, I briefly chart some of implications moving forward.

5.1 Well-Foundedness and Epistemic Infinitism

The viability of solutions to the epistemic regress problem depend on what relation generates the regress of justification, a point that Peter Klein was privy to even before grounding made an appearance. In the following response to Carl Gillett, Klein observes that getting clear about a solution to the epistemic regress problem requires getting clear about what kind of dependence is involved in the chain of justification:

• Gillett is relying on our understanding of what “in virtue of” means in [the regress problem] and, as Aristotle remarked, that phrase is multiply ambiguous. Aristotle did point out that it is used in the same senses as “cause.” That isn’t much help, given the many senses of “cause,” but it does seem that in this context “in virtue of” is, at base, probably a phrase meant to be roughly synonymous with “because of” and I take the dependence here to at least involve explanatory dependence.”³³

Here, we see Klein sorting through some of the very same issues that we have been considering in this paper. What is the ‘in virtue of’ relation at issue in the epistemic regress? Klein is intent on answering this question as he takes the viability of his infinitist project to rest on what kind of relation is involved in the epistemic regress. Klein continues his discussion, arguing that the regress of justification does not involve explanatory dependence.³⁴ Thus far, we have seen reason to think that Klein is mistaken – the epistemic regress problem is generated by the grounding relation, and as such, also coincides with an explanatory regress.

We can, of course, take on the point that Klein was incorrect in thinking that epistemic infinitism does not involve an explanatory regress while not also thinking that Klein was right to think that an explanatory regress is problematic for infinitism. Once we accept that grounding generates the epistemic regress problem, there is a further question about whether grounds themselves must come to an end or whether they can reach infinitely downward. This contemporary discussion is over whether or not grounding is well-founded, that is, whether there are foundational, ungrounded grounds. Some grounding theorists think it obvious that grounding must be well-founded. Schaffer, for instance, takes it that if grounding were not well-founded, “being would be infinitely deferred, never achieved.”³⁵ It is not difficult to find many similar statements from critics of epistemic infinitism. Henry Johnstone argues that infinitely deferred justification is no justification at all, saying that “X infinitely postponed is not an X.”³⁶ If grounding is well-founded, then Johnstone is right – epistemic infinitism cannot confer justification on beliefs at all.

Not all grounding theorists, however, accept that grounding is well-founded. Rosen (2010), p. 116, and Raven (2015b) do not take it as obvious that grounding must be well-founded, and Bliss (2013) and 2014), Morganti (2009), (2014), and (2015), and Tahko (2014) argue against it. To the extent that it is not obvious that chains of grounding must have a foundation, then epistemic infinitism cannot simply be ruled out due to the epistemic regress’s connection to grounding. More work will need to be done, then, on whether grounding is well-founded in order to discern whether infinitism is a viable solution to the epistemic regress problem.

The way that I have been discussing the problem might give the impression that the only import of the epistemic regress relation being an instance of grounding is that grounding theorists can now dictate plausible solutions to the epistemic regress. Even though it is true that the question of grounding’s well-foundedness is a metaphysical question, the lessons might easily go the other way as well. To the extent that infinitism about justification is independently plausible as a solution to the epistemic regress, epistemologists could argue that this is a consideration in favor of taking grounding itself to not be well-founded. If this is right, then my thesis, combined with the plausibility of epistemic infinitism, is itself an argument against grounding being well-founded. It is not obvious at the outset which way things should go – perhaps the lesson is that the prospects of infinitism about justification hang on whether grounding is well-founded, or maybe the takeaway is that epistemic infinitism demonstrates that grounding is not well-founded.³⁷ Either way, the connection between justification and grounding provides another link in logical space around which infinitists and grounding advocates must theorize.

5.2 Asymmetry and Epistemic Coherentism

Even though the jury might still be out on whether grounding is well-founded, it is nearly unanimously thought that grounding is asymmetrical. A primary motivation for positing grounding in the first place was that it was capable of capturing directionality in a way that supervenience cannot. Even though it is true that the existence of $\{$Socrates$\}$ supervenes on the existence of Socrates and vice versa, it is not also true that the dependence between the two entities goes in both directions, as $\{$Socrates$\}$ exists in virtue of Socrates existing.³⁸ If it is true that the relation in the regress of justification is a grounding relation and that grounding is asymmetrical, then the same is true in the case of justification-makers and the things they justify – the dependence between the two goes in only one direction. This creates trouble, however, for certain understandings of epistemic coherentism, as one characterization of the coherentist position is that justification proceeds in a circle, that $b_{1}$ inherits its justification from $b_{2}$, that $b_{2}$ inherits its justification from $b_{3}$, and so on until $b_{n}$ inherits its justification from $b_{1}$.

Such a structure, however, violates asymmetry. Even though the difficulty is not as glaring as taking $b_{1}$ to directly inherit justification from $b_{2}$ and $b_{2}$ to directly inherit justification from $b_{1}$, a circular ordering is ultimately problematic. Due to the transitivity of grounding, we can straightforwardly generate the consequence that $b_{1}$ both inherits justification from and grants justification to $b_{2}$.³⁹

Even though circular justification is a popular way of construing epistemic coherentism, it is not the only way, or even the most plausible way, to make sense of the coherentist proposal. We have already seen that circular reasoning is suspect apart from its relationship to grounding, and so some versions of epistemic coherentism are holistic rather than linear. Perhaps the lesson that coherentists about justification can learn from grounding theorists is that structures of holistic support must be conceived in a way that does not force them to violate asymmetry.⁴⁰ On the other hand, perhaps there are strong reasons to think that justification can proceed in a circle. If that is the case, then perhaps pressure pushes in the other direction and suggests a revision to how metaphysicians think of grounding itself. If the epistemic regress relation is an instance of grounding, and justification can plausibly be circular, then epistemic coherentism would instead provide a reason to think that metaphysicians have instead been wrong about the asymmetry of grounding.

6 Conclusion

I began this paper by pointing out that work in metaphysics often has much that can be appropriated within epistemology. Theorizing about grounding is currently taking hold in metaphysics despite a number of criticisms about the theoretical usefulness of positing a grounding relation. In this paper, I hope to have provided reason to think that such work has promise by showing that grounding can both help make sense of the epistemic regress problem as well as give a characterization of foundationalism. Given how well-worn the regress problem is in epistemology, it is surprising that grounding talk can shed new light on how to move the debate forward, and suggests that grounding might have further promise in terms of being applied to other positions within epistemology. In particular, grounding might lead to fruitful discussions of the possibility of epistemic coherentism and epistemic infinitism. It is also important to note that, along with applying new insights in metaphysics to epistemology, sometimes the influence can flow in the other direction. By adding the epistemic regress to the cases that metaphysicians consider when theorizing about grounding, epistemologists can also provide insights for the metaphysicians, opening the door to promising future work through dialogue between the two camps.