It is now common, though not uncontroversial, to regard facts as the normative reasons that justify our actions and attitudes. But in order for a fact to be a reason that justifies an action or attitude for a given agent it must be a fact that is possessed by that agent. There are a range of views of what it takes to possesses a fact in a way that allows it to function as a reason for an agent, and as has become increasingly common in recent epistemology some have sought to put knowledge in the theoretical forefront. Accordingly, one prominent class of views are centered on knowledge: we possess a fact P as a reason only if we know P (Williamson 2000; Littlejohn 2017) or we are in a position to know P (Lord 2018a; Neta 2017).

Discussions of knowledge-centric views of possession have largely overlooked the insight that knowledge is itself an instance of a kind. Quite roughly, if knowing that P is a form of non-accidental true belief that P, and belief is but one way representing P, then knowledge is but one kind of non-accidental true representation that P. I stipulatively call the general state of non-accidental true representation of which knowledge is an instance ‘awareness’. I explore the implications of different substantive characterizations of awareness in light of recent theories of knowledge. The upshot is an alternative way of thinking about possession: S possess a fact P as a reason only if S is in a position to be aware that P. This account of the epistemic condition on possession is logically weaker than knowledge-centric accounts, allowing for the possession of reasons in a range of cases where knowledge and the potential for knowledge are absent. I argue that this is just the result we need if we’re to explain the reasons we possess in cases of environmental luck and certain very ordinary cases involving explicit deductive updates.

It must be emphasized that what I’m primarily investigating in this paper is the epistemic relation agents must stand in to facts in order to possess them as reasons to respond in some way. But there is more to possessing facts as reasons than standing in a suitable epistemic relation to those facts. For example, in order to possess something as a reason to dance a jig I arguably have to be able to dance a jig as well as be capable of using P to dance a jig. These are potential practical constraints on possessing facts as reasons for a response and there is much to say about such constraints (cf. Whiting 2014; Sylvan 2015; Littlejohn 2017; Lord 2018a, Chap. 4). While the emphasis of my paper is on the epistemic condition involved in possessing facts as reasons, I will conclude with a brief reflection on how my arguments undermine Lord’s view of the practical condition involved in possession.

1 Background I: A quick primer on reasons-first epistemology

The reasons-first approach to justification is reductionistic. It purports to give us an illuminating and complete account of ultima facie justification (rationality) in terms of the possession of sufficient reasons. That is, what it is for an action or an attitude to be ultima facie justified just is for one to possess sufficient reason for performing that action or for having that attitude.Footnote 1 This generates three questions: what are reasons? when are they sufficient? and what does it take to possess them? In what follows I’ll summarize a prominent answer to the first two questions before introducing two knowledge-centric ways of answering the third.

What are reasons? When it comes to the ontology of reasons, the standard view in metaethics and a quickly growing view among epistemologists is that reasons are either facts or true propositions. For expressive convenience I will identify facts with true propositions and refer to the resulting view of reasons as broad factualism. The driving motivation behind broad factualism has to do with our justificatory practices: we reference (or attempt to reference) facts when engaged in the activity of justifying the actions and attitudes of ourselves and others. We say things like: the fact that there are elephants in Africa is a reason to believe that they’ve not yet gone extinct, and the fact that you’re hungry is a reason to get a snack, and the fact that a potential action would cause harm is a reason to refrain from that action, and so forth. When we recognize that our attempts to justify actions and attitudes reference falsehoods rather than facts we view the attempted justification as defective. Those who take reasons to be facts take this aspect of our justificatory practice as illuminating the sources of our justification for our attitudes and actions, i.e. facts.Footnote 2 If you think falsehoods as well as facts can be reasons that justify actions and attitudes you can think of this paper as a paper about the possession of objective reasons and a corresponding sort of objective justification (Whiting 2014; Sylvan 2015; Schroeder 2015).

When are reasons sufficient? Generally, sufficiency is regarded as a matter of weightiness: S has sufficient reason to believe P iff the reasons S possesses for believing P are at least as weighty as the reasons S possesses for not believing P. Talk of the weight of reasons is meant to capture the way in which reasons can “stack-up” in favor and in opposition to certain responses. For example, in the case of belief you can have all kinds of reasons to believe P and all kinds of reasons to refrain from believing P. But you only have sufficient reason to believe P when your reasons in favor of believing are not outweighed by your reasons to refrain (Lord and Maguire 2016). For the most part this way of thinking about sufficiency is independent of one’s stance on the nature of reasons.

What does it take to possess a reason? Broad factualists need to say something substantive about what it takes to possess (= to access, to have) a reason. For if reasons are facts or true propositions they are not automatically within one’s ken in a way that could justify one’s attitudes and actions (Williamson 2000; Whiting 2014; Schroeder 2015; Sylvan 2015, 2016, 2018; Littlejohn 2017; Neta 2017; Lord 2018a). Imagine that you are currently standing in front of an oak tree and you have no other information about the oak apart from your general knowledge of oaks and what you can visually “pick up” about this oak while looking at it. The fact that the oak you’re looking at is 613 inches tall is a conclusive reason to believe that the oak is between 612 and 614 inches tall. But if you are wholly ignorant of the fact that it’s 613 inches tall, then without further information about its height you cannot justifiedly believe that the oak is between 612 and 614 inches tall. This idea is generally agreed upon and so it’s generally assumed among broad factualists that an understanding of the nature of the epistemic aspect of the reasons-possession relation is essential to the integrity of the reasons-first approach to justification.

In what follows I assume one cannot possess a reason without possessing it as a reason for at least some kind of response. Accordingly, given broad factualism, ‘S possesses a reason’ is short for ‘S possesses a fact as a reason to respond in some way.’ When it matters I will specify some of what is included in the relevant range of responses.

2 Background II: Motivations for knowledge-centric views of possession

One prominent view of possession is centered on knowledge: we possess only those facts that we know (Williamson 2000; Littlejohn 2017):Footnote 3

Possession Requires Knowledge

\((\hbox {P}\rightarrow \hbox {K})\) S possesses a fact P as a reason to \(\phi \) only if S knows P.

Williamson (2000) and Littlejohn (2017) are both proponents of \((\hbox {P}\rightarrow \hbox {K})\) and their motivations for it largely overlap in that they appeal to cases where we seem not to possess a reason, P, because we fail to know P. I’ll focus attention on Littlejohn’s recent defense of \((\hbox {P}\rightarrow \hbox {K})\) in this section and return in the final section of this paper to discuss the lottery-like cases that Williamson (2000) has used in defense of \((\hbox {P}\rightarrow \hbox {K})\).

While Littlejohn (2017, pp. 26–27) points to various criteria a theory of possession must satisfy, the main argument that favor’s his knowledge-centric view of possession is driven by a kind of gettier case:

Consider Nozick’s experience machine. Agnes undergoes a series of experiences that dispose her to form false beliefs about her surroundings. It seems to her that she and everyone she cares about are flourishing. [While nearly all of her beliefs are mistaken] some of her beliefs happen to be true. ...it seems to her that her brother has just crossed the stage at graduation and a smile stretches across Agnes’ face because she believes he just graduated. What the lab technicians don’t realize is that precisely as Agnes undergoes this experience her brother crosses the stage and accepts his diploma.

Here’s Littlejohn’s (2017, pp. 26–27) judgement about Agnes and her reasons:

While she believes correctly that her brother is graduating and is happy because she believes this, her reason for being happy isn’t that her brother is graduating. She cannot be rationally guided by such a fact, not when she’s cut off from reality.

Littlejohn is right, Agnes is not nor can she be happy for the reason that her brother is graduating. Though she truly believes this, she’s too far separated from that fact for it to be a reason she could respond to in any way. As Littlejohn argues, it seems plausible that possessing a reason requires at least the possibility of responding to it. Accordingly, Agnes fails to possess the fact that her brother is graduating as a reason. And \((\hbox {P}\rightarrow \hbox {K})\) can easily explain this fact since Agnes is unable to know that her brother is graduating.

But not all knowledge-centric theorists have been so optimistic about \((\hbox {P}\rightarrow \hbox {K})\). Against \((\hbox {P}\rightarrow \hbox {K})\), Lord (2018a, 72ff) draws attention to counterexamples involving inattention. Here’s how I like to illustrate this type of counterexample:

Inattentive Perception

I have a multicolored carpet in my home that looks like this:

figure a

I love the way it looks, so does my family. Suppose you came over for dinner one evening. You’d have a visual experience not unlike the one you’re having now, and there are all kinds of facts that your visual experience would seem to put you in possession of even if you did not believe them. For example, (F1) that each determinate color in each square occurs exactly four times, (F2) that there are two determinate shades of blue, (F3) that there are four rows, etc. There is an exceptionally large number of facts that this perceptual experience could give you access to while having the perceptual experience, but many of these facts will not be converted into beliefs simply because you’re not paying attention to them.

Before coming into my home you had no beliefs about (F1)–(F3), and even while looking at my carpet you might not come to believe (F1)–(F3) simply because you’re attention is elsewhere, e.g. suppose we were having a rich and gripping philosophical conversation while you were absently looking directly at this carpet. But since you’re looking directly at this carpet it seems like you’re unjustified in failing to believe (MoreBlue): that there are more determinable blue squares than any other colored square.Footnote 4 What could explain this lack of justification? Not your knowledge of, say, (F1) and (F2). For you don’t believe (F1) or (F2) because you’re attention is swept up by our conversation. So \((\hbox {P}\rightarrow \hbox {K})\) doesn’t allow you to posses (F1) and (F2). But if your non-belief in (MoreBlue) is unjustified while you’re looking at my carpet this has to be explained by reasons that you have access to, and in the present case it seems like this will have to include (F1) and (F2). This is a problem for \((\hbox {P}\rightarrow \hbox {K})\).

If knowledge is to be central to our understanding of the possession of reasons and if we are to take cases of inattention as counterexamples to \((\hbox {P}\rightarrow \hbox {K})\), what we need is a theory of possession that can explain how I possess (F1) and (F2) in Inattentive Perception. But it must be a theory of possession that retains the explanatory power of \((\hbox {P}\rightarrow \hbox {K})\), which lies with it’s ability to explain why Agnes fails to possess the fact that her brother has just graduated.

Lord (2018a, pp. 67–96) and Neta (2017) offer just the thing. When it comes only to the epistemic constraints on possessing a fact the following holds:

Possession Requires the Position to Know

\((\hbox {P}\rightarrow \hbox {PK})\) S possesses a fact P as a reason to \(\phi \) only if S is in a position to know P.

Lord (2018a, p. 92) characterizes being in a position to know as follows:

You’re in a position to know P when all the impersonal conditions for knowledge are met. The personal conditions are (1) believing P and (2) believing P for the right reasons. The impersonal conditions are just whatever are left over. [Like the truth of P and standing in a non-accidental relation to it.]Footnote 5

While knowing is a logically stronger relation than being in a position to know, they both involve truth and non-accidentality requirements. Thus, for example, if knowledge requires safe belief, then being in a position to know requires being in a position to safely believe. Since Agnes neither safely believes that her brother is graduating nor is she in a position to safely believe it, it follows on a safety view of knowledge that Agnes neither knows nor is she in a position to know that her brother is graduating. And thus by \((\hbox {P}\rightarrow \hbox {PK})\) it follows that Agnes doesn’t possess that fact. Accordingly, the truth of \((\hbox {P}\rightarrow \hbox {PK})\) seems just as well placed as \((\hbox {P}\rightarrow \hbox {K})\) to help us understand why our intuitive judgement about Agnes is correct.

Furthermore, in Inattentive Perception while you do not know (F1) or (F2) you’re clearly in a position to know them despite your inattention. For as Lord (2018a, p. 92) characterizes it, if you don’t already know P you may yet be in a position to know P (and hence stand in a relevant non-accidental relation to P) if “in [such] uninferential cases...you have some experiences such that, if you could and did attend to certain features of those experiences and uninferentially form a belief that P in the right kind of way, then you would know that P.” Clearly, this condition can hold in cases like Inattentive Perception with regard to you and (F1) and (F2), thereby giving \((\hbox {P}\rightarrow \hbox {PK})\) an explanatory advantage over \((\hbox {P}\rightarrow \hbox {K})\).

3 Problems for knowledge-centric accounts of possession

In this section I lay out two problems with knowledge-first accounts of possession. The first concerns explicit deductive updates. The second concerns environmental luck.

3.1 Explicit deductive updates

Just as it was plausible to think that we could come to possess reasons even before we came to believe and know them when inattentive (as in Inattentive Perception), it’s even more plausible to think that we can come to possess new reasons by being attentive to entailment relations from existing knowledge.

Here is an example. I know (G): that North Rhine-Westphalia is the most populated state within Germany and that Germany has no more than 85 million residents. This entails (N): that North Rhine-Westphalia has no more than 85 million residents. And through a simple exercise of my deductive competences I can come to know: that (G) entails (N). Now deduction itself is not a belief-entailing activity. One can deduce P from Q without coming to believe Q. Think of all the deductive exercises you were required to do with obviously false premises in order to help you get a grip on the difference between validity and soundness. When the premises were obviously false you didn’t believe their conclusions despite deducing them. Similarly, coming to know that an argument’s premises entail its conclusion is also not an activity the ensures one believes the conclusion.

Now consider the fact that when it comes to agents like us, we sometimes update our beliefs in response to (and therefore after) recognizing entailments from our beliefs. That is:

Explicit Deductive Updating

It is possible for a thinker S to first come to know P and to know P entails Q, and then in response to begin her update process so that she later comes to believe Q.Footnote 6

Explicit Deductive Updating identifies not just a possible way to update, but an ordinary one. To see this let’s continue with the example above.

It was, embarrassingly, not so long ago that I learned that North-Rhine Westphalia was a German state. So it was not so long ago that I failed to know (G). But even before knowing (G) I could easily have known that (G) entails (N) since (N) is a rather obvious deductive consequence of (G). To know it I need only have been presented with (G) and (N) and reflected on their rather obvious logical relations. In such a case it is possible that I would first come to know (G) and know that (G) entails (N), and then in response update my beliefs so that I believe and come to know (N). This is an explicit deductive update.

Explicit deductive updates are a two-stage process. They involve first recognizing an entailment from one’s knowledge, and after that responding to that recognition by forming a new belief in the conclusion. I’m interested in what reasons I possess in between the recognition and the response. That is, after deducing (N) from (G) but before coming to believing (N), what epistemic relation do I stand in to (N)?

It seems implausible to say that I’m ignorant of or blind to that fact–surely ignorance of (N) is incompatible with consciously deducing (N) from my knowledge of (G). And if I’m not ignorant of (N), then don’t I possess it? It’s intuitive to think that I do have epistemic access to (N). At the very least, it’s just as intuitive to think that I possess (N) in virtue of deducing (N) from my knowledge of (G) as it is intuitive to think that I possess (F1) and (F2) in Inattentive Perception above. We can capture this idea with a kind of closure principle for possession:

Extended Possession

Other things being equal, if S knows that P is entailed by something she knows, then S possess the fact P as a reason for a response.

In the case of my deduction of (N) from my knowledge of (G), Extended Possession implies that I possess (N) unless something odd is going on such that “other things are not equal.” This is a wrinkle we needn’t worry about with regard to the case at issue involving (G) and (N).Footnote 7

Together Extended Possession and Explicit Deductive Updating cause problems for knowledge-centric theories of possession. \((\hbox {P}\rightarrow \hbox {K})\) is straightforwardly inconsistent with Extended Possession and Explicit Deductive Updating. This is because Extended Possession implies that in cases of Explicit Deductive Updating there are facts we possess before we believe them, and hence before we know them. Thus there are facts we can possess without knowing them. So we have one more reason to reject \((\hbox {P}\rightarrow \hbox {K})\).

Notice that Extended Possession and Explicit Deductive Updating don’t immediately cause problems for \((\hbox {P}\rightarrow \hbox {PK})\). Since \((\hbox {P}\rightarrow \hbox {PK})\) doesn’t imply that believing or knowing P is required for possessing P. \((\hbox {P}\rightarrow \hbox {PK})\) only says that being in a position to know is necessary. So the advocate of \((\hbox {P}\rightarrow \hbox {PK})\) need only explain how one can possess reasons in cases of Explicit Deductive Updating before coming to know them.

Lord’s (2018a) discussion of \((\hbox {P}\rightarrow \hbox {PK})\) on this point is admirably clear, and he explains it in terms of the following condition for being in a position to inferentially know (Lord 2018a, p. 92):

...it’s possible to be in a position to know P when you don’t believe [and thus don’t know] P. In cases like this, what is it in virtue of which you are in a position to know P? In the inferential cases, it’s the fact that you have some beliefs with contents such that, if you could and did attend to those contents and inferentially form a belief that P in the right way, you would know P.

Put differently:

Counterfactual Inferential Knowledge

When S does not already know P, S is in a position to inferentially know P iff S has some beliefs with contents such that, if S could and did attend to those contents and inferentially form a belief that P in the right way, S would know P.

Since I can come by such knowledge after properly forming a belief in (N) on the basis of my knowledge of (G) and that (G) entails (N), I count as being in position to know (N) even before I believe it. Thus, \((\hbox {P}\rightarrow \hbox {PK})\) allows me to possess (N) even before I come to know it. This is just what we need \((\hbox {P}\rightarrow \hbox {PK})\) to do if it’s to remain consistent with Extended Possession and Explicit Deductive Updating.

But there is trouble here. Counterfactual Inferential Knowledge yields too strict a theory of possession, one that turns out to be inconsistent with Extended Possession. Take the following possible scenario:

Kripkean Study

I’m part of a psychological experiment that is aimed at testing my responsiveness to my reasons as well as the extent to which I’m subject to the error of Kripkean dogmatism (though I do not know that this is what is being tested).Footnote 8 The psychologists test this by giving me new knowledge and then helping me realize new logical consequences of my new knowledge. So they give me some new, non-misleading evidence E that actually enables me to come to know (G). Then they explain why (G) deductively entails (N). (Assume this deductive consequence was unknown to me before the study.) Since I’m rational this would ordinarily bring about a new belief in (N) on the basis of my knowledge that (G) and that (G) entails (N). But ever so quickly, as they intended, before I form my belief in (N) the psychologists give me a defeater for my original knowledge that (G). Specifically, just after after helping me come to know that (G) entails (N) they very quickly and convincingly tell me a lie: that the original evidence I was relying on for (G) was all fabricated for the purposes of the experimental study and that (G) is indeed false. (This is not true. The evidence is real and (G) is true.) Since I’m not a crazed Kripkean dogmatist and since I don’t know they’re lying to me about their “fabrication”, I don’t ignore what they tell me. Rather, I give up my belief in (G) and so cease to know (G).

For me and other possible participants in this type of study it will be false that: were I to form a belief in (N) in response to an explicit deduction from my knowledge of (G) and knowledge that (G) entails (N), then I would know (N). This is due to the fact that the study is constructed in such a way that I’m supposed to be given a defeater for (G) before I have a chance to believe (N). So if the study is reliably executed, then it’s false that I would come to know (N). There are just too many nearby worlds where I get the defeater before I come to believe and know (G), and the actual world is one such world. So according to Counterfactual Inferential Knowledge I am not in a position to know (N).

The point here is that Kripkean Study is a possible situation. About this possible situation \((P\rightarrow { PK})\) and Counterfactual Inferential Knowledge together imply that I do not possess (N). But that’s inconsistent with Extended Possession. For there is some small amount of time where I know (G) and know that (G) entails (N)–it’s the short amount of time it takes for the psychologists to give me the defeater for my belief and knowledge of (G). So Extended Possession, \((\hbox {P}\rightarrow \hbox {PK})\), and Counterfactual Inferential Knowledge form an inconsistent set.

Some advocates of \((\hbox {P}\rightarrow \hbox {PK})\) might seek to undermine the objection arising from Kripkean Study by undermining the idea that (N), while a fact in Kripkean Study, is not a fact that can function as a reason for me. Recall that possessing a fact, P, as a reason to respond in some way requires that I be able to respond to P in some way. And it’s not clear just what responses are available to me if I’m going to be given a defeater so quickly. But it takes only a moment to appreciate that even in Kripkean Study I can possess (N) as a reason to update my beliefs or at least as a reason to begin to update my beliefs so that I give up or begin to give up my non-belief in (N).Footnote 9 This will involve me becoming increasingly confident in (N) and forming the dispositions to act and think on (N) in the ways characteristic of belief. Beginning to update my beliefs in these ways is clearly consistent with being given a defeater before that update process is completed.

Some advocates of \((\hbox {P}\rightarrow \hbox {PK})\) might seek to undermine the objection arising from Kripkean Study by seeking to refine Counterfactual Inferential Knowledge so that these problems don’t arise. For example, one could do away with the counterfactual aspect of Counterfactual Knowledge and thus do away with its implicit constraint that I possess only those reasons that I would know in all nearby worlds–i.e. in circumstances not importantly different from one’s actual circumstances. But this is a very attractive feature of Counterfactual Knowledge. For ordinary ways of thinking and talking about being in a position to know that are of normative significance for deliberation don’t seem to track the weaker notion of it being possible to know in circumstances that may be importantly different from one’s actual circumstances. In the present case, the important difference will involve the psychologists not giving me a defeater despite my participating in study where they intend to give me a defeater. The primary worry I have about weakening the target concept of being in a position to know in this way concerns the normative significance of what goes on in increasingly distant worlds. For, by definition, the reasons I possess are the facts that I have access to in my actual situation and they are facts that call for responses in my actual situation. But why should my knowledge in non-actual, distant worlds have an impact on what I should do or think or feel in the actual world?

3.2 Environmental luck

Cases of environmental luck are taken to be Gettier cases by many epistemologists, thus, by definition, these epistemologists take them to be cases of justified true belief that fail to be cases of knowledge. Consider the following case:

Hooligan

Bad Henry is a hooligan. He delights in wanton property destruction. Today he wants to destroy a barn. He knows what barns look like and he can reliably visually identify barns as barns when looking at a barn at a reasonable distance in normal (local) visual conditions. He drives out into the country to find a barn. He pulls over after an hour, retrieves his bazooka, and takes aim with unerring accuracy at the first roadside barn he looks at. He thinks to himself: “I’m looking at a barn.” Calm, cool, and collected he pulls the trigger and then thinks to himself: “My projectile will destroy the barn.” However, unbeknownst to Henry he is in an epistemically unfriendly environment when it comes to finding barns since most objects that look like barns in these parts are in fact barn facades.Footnote 10

When Henry stops in front of the barn is it plausible that Henry possesses the fact that the object he’s looking at is a barn? This is a question that can be hard to answer directly. When I, for example, consider the question I don’t have clear intuitions either way. But we can home in on an answer to this question by considering a related question.

Notice that the fact that the object Henry is looking at is a barn is a fact that gives (or could give) Henry a reason to respond in various ways. For example, it’s a reason to run in the direction of the barn if in need of shelter, it is a reason to place a substantial bet on the claim that the object he’s looking at is a barn, it is a reason to believe that barn-related supplies are inside the object he’s looking at, it is a reason to believe that a barn will be destroyed by his projectile, and so forth. In the story Henry actually responds in one of these ways: before destroying the barn Henry formed a belief in (Destruction): that a barn will be destroyed by the projectile.

Is Henry justified in holding that belief in (Destruction)? In answer to this question I do have a clear intuition: his belief is justified. I cannot see how Henry fails to have a justified belief. Indeed, in taking cases of environmental luck to be Gettier cases it is widely assumed that these are cases of justified belief. But if it’s justified he must possess reasons that justify the belief. What are his reasons? Doubtless if we were in Henry’s position and we were asked to cite our reasons that justify our belief we would respond by citing the fact that the object we’re looking at is a barn and some facts about bazookas.Footnote 11 But if Henry possesses the fact that that the object he’s looking at is a barn, it is a fact he possesses in circumstances of environmental luck.

Accordingly, on the assumption that Hooligan is a Gettier case it gives us evidence for two theses:

No Environmentally Lucky Knowledge: Usual cases of environmental luck are cases of non-accidental true belief that P, but the type of non-accidentality present in cases of environmental luck is insufficient for knowing or being in a position to know P.

Environmentally Lucky Possession: S can possess a fact P as a reason even if one’s non-accidental true belief that P is formed in circumstances involving environmental luck.

If these two claims are true, then \((\hbox {P}\rightarrow \hbox {PK})\) and \((\hbox {P}\rightarrow \hbox {K})\) are false.Footnote 12

4 Awareness: a generalization of knowledge

The arguments from Sect. 2 in defense of the idea that knowledge is central to possession turn on the way in which being accidentally related to a fact can prevent you from possessing it. Knowing and being in a position to know both rule out such an accidental relation, thereby giving knowledge-centric views of possession an advantage over other views of possession that permit one to be accidentally related to their reasons. But knowing P and being in a position to know P are not the only ways of being non-accidentally related to P.

On nearly all accounts of knowledge, knowledge is a type of non-accidentally true belief. The idea that knowledge requires true belief has been remarkably stable among philosophers despite occasional resistance, and I will not be defending it here.Footnote 13 The idea that knowledge is a kind of non-accidentally true belief has also been a mainstay among epistemologists. Of course, there are many intuitive senses in which a belief might count as “non-accidental” and epistemologists have been hard at work trying to identify the relevant non-accidentality relation that is constitutive of knowledge. The leading options here are familiar, including (but not limited to) process reliablism, proper functionalism, virtue reliabilism, safety theories, sensitivity theories, explanationist theories). Below I will mention where taking a stance on this can make a difference to one’s theory of awareness and possession.

Belief (outright belief) is a representational mental state with propositional content, i.e. it is a mental state that represents propositions as being true of the world. But believing P isn’t just a matter of representing the world as if P is true, believing P also involves dispositions to act and think in ways that presuppose the truth of P in a certain range of circumstances. For example, agents who believe P are agents who have the disposition to act on P, assert P, and draw inferences from P in a certain range of circumstances. But the idea that ‘belief’ in epistemically central uses refers to a representational state with propositional content that involves these minimal dispositions is generally agreed upon in both the philosophy of mind and epistemology.Footnote 14 Since beliefs are a kind of representational state it follows that true beliefs are a kind of true representation.

Thus it follows that knowledge is a kind of non-accidental true representation. For expressive convenience I will stipulatively use the term ‘awareness’ to refer to the target state of non-accidental true representation of which knowledge is an instance. Accordingly, the argument to follow is not being guided by ordinary usage of the term ‘awareness’.

To say knowledge is a kind of awareness is to leave much unsaid. For the account of awareness for safety theorists will be different than the account of awareness for virtue epistemologists, and this will in turn be different than the account of awareness for proper functionalists, and so on. This is just because they each take different stances on the nature of knowledge and thus will likely take different stances on awareness. These are differences that can make a difference as I will discuss below. But the overall argument against \((\hbox {P}\rightarrow \hbox {PK})\) and \((\hbox {P}\rightarrow \hbox {K})\) will stand independently of these issues since the counterexamples to \((\hbox {P}\rightarrow \hbox {PK})\) and \((\hbox {P}\rightarrow \hbox {K})\) stand independently of them.

Now, it is the standard view in epistemology and philosophy of mind that representational states other than belief exist (though I will not lean on this assumption in what follows). For example, it’s widely held that perceptual states, memorial states, introspective states, intuitive states, and seeming states are representational states that are not constitutively belief states.Footnote 15 And because these are representational states they are true if their contents enjoy the right kind of match with the world. If correct, these non-doxastic forms of representation are what we might think of as direct forms of representation, since they are mental states that represent P by having P as their propositional content.

There is also an indirect way of representing the world as being such that P. If Q entails P, then believing Q itself involves some interesting notion of indirect representation of P since Q rules out all the ¬P possibilities. But even if you believe Q this kind of indirect representation of P is not obviously a normatively significant kind of representation because the entailment might be too complicated for you to appreciate. But if Q entails P and you not only believe Q but also believe that Q entails P, then you are indirectly representing the world as being such that P in a way that has the potential to be normatively important. For part of what it means to believe that Q entails P, is that one believes that the world is structured in such a way that there is no possible way for P to be false when Q is true. Accordingly, when one believes Q and believes that Q entails P there in a perfectly legitimate and epistemologically important sense in which one indirectly represents the world as being such that P is true even if they fail to believe P.Footnote 16

Some might worry about indirect representation in cases when one’s belief that Q and belief that Q entails P is merely dispositional and thus in some way not present to one’s conscious reflections, or in cases when a given thinker is not putting these two beliefs together in such a way that they can “see” that P logically follows. Such cases are irrelevant. I’m only interested in cases where a competent deductive reasoner occurrently believes Q and occurrently believes that Q entails P. I think it clear that such thinkers genuinely represent the world as being such that P even if they fail to believe P (occurrently or otherwise). And I will limit my claims about indirect representation of P to such cases. While I’m open to and optimistic about more expansive notions of indirect representation being of epistemological importance, I limit myself to the specified case here since I think it obvious and uncontroversial that agents who are skilled at deduction can represent the world as being such that P when they occurrently believe Q and occurrently believe that Q entails P. (For convenience I’ll leave the ‘occurrently’ qualification implicit in what follows.)

It is important to note that these other non-doxastic ways of representing P can be non-accidental in the way that believing P can be. Basically, a true representation is non-accidental in the relevant sense whenever it stands in a generalization of the non-accidentality relation required for knowledge. Again, since there are different accounts of the non-accidentality constraint on knowledge there will be different options here. Since I cannot walk through all potential accounts of awareness I’ll just give two examples. First take a view on which knowledge is safe belief:

Knowledge is Safe Belief

S knows that P iff S could not have easily had a false belief that P (given the method or process that actually led S to believe P).Footnote 17

We can generalize this for awareness as follows:

Awareness is Safe Representation (SafeA)

S is aware that P iff S could not have easily had a false representation that P (given the method or process that actually led S to represent P).

For an alternative account of awareness take robust virtue reliabilism about knowledge. At its heart is the attempt to understand knowledge in terms of non-defective exercises of the competences of agents. More specifically, the unifying thesis of these reliabilist virtue epistemologists is that we can give an informative account of knowledge in terms of an agent’s cognitive success from a non-defective exercise of her cognitive competences (Sosa 2007, 2015; Greco 2010; cp. Pritchard 2012a; Kelp 2013; Miracchi 2015). While there are plenty of disagreements among virtue reliabilists about how to understand the relevant ‘from a non-defective exercise’ relation, there are two notable points of agreement. First, ‘cognitive competences’ are to be understood as circumstance-relative, reliable abilities to succeed. Second, when it comes to propositional knowledge the relevant type of cognitive success involves true belief.Footnote 18 Accordingly, the unifying thesis of the robust virtue theoretic approach to knowledge can be put more specifically:

A Virtue Theory of Knowledge

S knows P just in case S truly believes P from a non-defective exercise of her cognitive competences.

We get a virtue theoretic account of awareness from this by generalizing the virtue theory of knowledge to all forms of true representation:

A Virtue Theory of Awareness (VirtueA)

S is aware that P just in case S truly represents P from a non-defective exercise of her cognitive competences.

While (VirtueA) is a generalization of the virtue theory of knowledge, it’s important to note that (VirtueA)’s truth doesn’t depend on the truth of the virtue theory of knowledge. For example, one could think (as non-robust, anti-luck virtue epistemologists do) that knowledge requires some further anti-luck condition, e.g. safety or sensitivity (Pritchard 2012a; Kelp 2013, 2018).

Since true belief is not the only way of having a true representation, and since we can easily give theories of non-accidental true representation of which knowledge is an instance (=awareness) we arrive at the thesis that knowing P is not the only way of being aware that P.

5 Awareness and the possession of reasons

With substantive accounts of awareness on the table we can appreciate an alternative theory of the possession of facts construed in terms of awareness rather than knowledge:

Possession Requires Awareness

\((\hbox {P}\rightarrow \hbox {A})\) S possesses a fact P as a reason to \(\phi \) only if S is in a position to be aware that P.

\((\hbox {P}\rightarrow \hbox {PK})\) and \((\hbox {P}\rightarrow \hbox {A})\) are very much alike. Both are epistemically demanding views of possession, both are views that do not require belief, and both are factive views. So in this respect, the two views of possession are on a par. Similarly, both appeal to natural epistemic relations. After all, knowledge is a non-ad hoc explanatorily rich relation we can stand in to facts. Since the awareness relation is a very modest generalization from the knowledge relation it too is a non-ad hoc explanatorily rich relation we can stand in to facts.Footnote 19

Further \((\hbox {P}\rightarrow \hbox {A})\) is just as capable as \((\hbox {P}\rightarrow \hbox {PK})\) when it comes to explaining your possession of (F1) and (F2) in Inattentive Perception. For given that knowledge is a form of awareness two entailments follow: knowing P entails being aware that P, and being in a position to know P entails being in a position to be aware that P. So the fact that you’re in a position to know facts like (F1) and (F2) while inattentive–as Lord argues–ensures that you’re also in a position to be aware of (F1) and (F2) while inattentive. So the explanatory power of \((\hbox {P}\rightarrow \hbox {PK})\) to handle cases of inattention is completely carried over to \((\hbox {P}\rightarrow \hbox {A})\).

It is easy to see how \((\hbox {P}\rightarrow \hbox {A})\) can explain the fact that Agnes does not possess the fact that her brother is graduating. On (SafeA) this will be explained by the fact that Agnes cannot safely represent that fact because the experience machine scenario guarantees that she could easily have had a false belief. For there are nearby worlds where she falsely represents that fact, e.g. worlds where the machine feeds her that experience just before or just after her brother is graduating. On (VirtueA) this will be explained by the fact that Agnes has no cognitive competence she could non-defectively exercise to arrive at a representation of the fact that her brother is graduating. The is because her visual belief-forming competences are not competences (reliable abilities) relative to the situation that she is actually in within the experience machine. So on both views Agnes is not simply unaware, she’s also not in a position to be aware. So \((\hbox {P}\rightarrow \hbox {A})\) can easily be used to explain why Agnes fails to possess the target fact.

Accordingly, \((\hbox {P}\rightarrow \hbox {A})\) can explain why Agnes does not possess the fact that her brother is graduating, and \((\hbox {P}\rightarrow \hbox {A})\) can also explain why you do possess (F1) and (F2) in Inattentive Perception. So \((\hbox {P}\rightarrow \hbox {A})\) is on an explanatory par with both \((\hbox {P}\rightarrow \hbox {K})\) and \((\hbox {P}\rightarrow \hbox {PK})\).

We saw two objections to \((\hbox {P}\rightarrow \hbox {PK})\) and \((\hbox {P}\rightarrow \hbox {K})\) above: one from explicit deductive updates and one from environmental luck. They were cases where agents possess facts despite failing to be in a position to know them and hence failing to know them. \((\hbox {P}\rightarrow \hbox {A})\) has no such difficulty, as I’ll now explain.

5.1 Awareness and explicit deductive updates

\((\hbox {P}\rightarrow \hbox {A})\) nimbly avoids the problem raised by the application of Extended Possession to Kripkean Study because \((\hbox {P}\rightarrow \hbox {A})\) can easily explain the fact that I possess (N) before being given a defeater in the psychological study. This is because \((\hbox {P}\rightarrow \hbox {A})\) doesn’t make the potential for knowledge in all (most, many) nearby worlds a requirement for possession. Rather, all that is required is that one be in a position to be aware that (N). And in Kripkean Study I’m not merely in a position to be aware that (N) before being given the defeater, I’m actually aware that (N) before I’m given the defeater. Recall, awareness is a matter of non-accidental true representation. (N) is true, and while I don’t represent (N) by believing (N) there are at least two ways I might represent (N) without believing (N) in Kripkean Study. First, many epistemologists think that seeming states are sui generis representational propositional attitudes that are distinct from belief. Thus even if I do not believe (N), (N) can still be the content of a seeming state. Perhaps (N) seems true in virtue of (N) being a newly appreciated logical entailment of (G) in Kripkean Study. This would be a case of direct representation of (N) because (N) is the propositional content of the seeming state.

But we needn’t rely so much on seemings. For in Kripkean Study I occurrently believe (G) and occurrently believe that (G) entails (N) for some short amount of time. Through that short span of time I indirectly represent (N) as being true in virtue of the fact that I’m a competent deductive reasoner who occurrently believes (G) and occurrently believes that (G) entails (N). Accordingly, so long as my indirect representation of (N) is non-accidental in the way required for awareness, I count as being aware of the fact that (N). And in the present circumstances it is non-accidental in that way. For I know (G) prior to being given defeaters, and I also know that (G) entails (N), and I’m a competent deductive reasoner. Since we’re assuming things are otherwise normal prior to being given the defeater, I can be aware that (N) even before I believe it.

It is easy to see how (VirtueA) delivers awareness of (N) in this case. For knowing (G) and knowing that (G) entails (N) while being a competent deductive reasoner ensures that my indirect representation of (N) is from a conditionally reliable competence to indirectly represent facts–that is, it is reliable conditional on the truth of the premises. Perhaps it’s less obvious how (SafeA) could deliver awareness of (N). To see this notice that my indirect representation of (N) counts as safe in a sufficiently relevant sense. For before I’ve been given the defeater for (G) I know (G) and I know that (G) entails (N). Thus if knowledge requires safety, my beliefs in these premises will be safe. And so an indirect representation of (N) will count as derivatively safe when it’s constituted by safe belief in (G) and safe belief in (G) entails (N) when one is a competent deductive reasoner. So even on the modally more demanding view of awareness, (SafeA), we can explain the consistency of \((\hbox {P}\rightarrow \hbox {A})\) with cases like Kripkean Dogmatist. So not only is \((\hbox {P}\rightarrow \hbox {A})\) consistent with Extended Possession and Explicit Deductive Updating, it can also explain the fact that I possess (N) before being given a defeater for it in Kripkean Study.

\((\hbox {P}\rightarrow \hbox {A})\) neatly avoids the problem raised by Kripkean Study because I’m actually aware of (N), and thereby count as being in a position to be aware of (N). This raises the question: when one is not actually aware that P, when is one in a position to be aware that P? We need only take Lord’s (2018a, p. 92) proffered conditions for inferential and non-inferential cases of being in a position to know and generalize from knowledge to awareness:

Counterfactual Inferential Awareness

When S is not already aware that P, S is in a position to be inferentially aware that P iff S has some beliefs with contents such that, if S could and did attend to those contents and inferentially form a belief that P in the right way, S would be aware that P.

Counterfactual Non-Inferential Awareness

When S is not already aware that P, S is in a position to be non-inferentially aware that P iff S has some experiences such that, if S could and did attend to certain features of those experiences and non-inferentially form a belief that P in the right kind of way, then S would be aware that P.

Against these counterfactual conditions I can imagine someone trying to produce a version of Kripkean Study in cases like Inattentive Perception. This might be a case where, while you’re actually inattentive to (F1) and (F2) while looking at my carpet I stand poised to quickly deliver a defeater for (F1) and (F2) if I get even a whiff of you thinking too much about the carpet (e.g. perhaps I’ve even planted colored lights above the carpet and I’m prepared to tell you some nonsense about carpet squares being illuminated by blue lights to teach my family and friends about undercutting defeaters).

If so, and if you haven’t already become aware that (F1) and (F2) are true (thereby triggering the limiting condition), then it seems to me entirely intuitive to claim that you don’t possess (F1) or (F2). Doubtless they are facts that are “right there” for you to appreciate in some broadly modal sense. But the fact that you do not actually appreciate them and the fact that you’re only accidentally related to (F1) and (F2)–since the nearby worlds are ones where you’d likely be given a defeater were you to attempt to appreciate them–seem to me to make these facts normatively irrelevant when it comes to assessing whether or not your non-belief in (MoreBlue) is unjustified.

Those readers who have the lingering thought that you do possess (F1) and (F2) despite these circumstances are, I suspect, in the grip of a representationalist view of perceptual experience on which perception can give you (F1) and (F2) directly even if you don’t believe them. If so, that’s fine. For then you will count as already being aware that (F1) and (F2) are true since perception that (F1) and (F2) would typically put you in the knowledge-relevant non-accidental relation to (F1) and (F2). So the initial limiting condition of Counterfactual Non-Inferential Awareness will not be satisfied in that case because you count as already being aware of (F1) and (F2). Thus, according to \((\hbox {P}\rightarrow \hbox {A})\), you will already possess (F1) and (F2).

5.2 Awareness and environmental luck

Above we saw that treating cases of environmental luck as Gettier cases supports the following two theses:

No Environmentally Lucky Knowledge: Usual cases of environmental luck are cases of non-accidental true belief that P, but the type of non-accidentality present in cases of environmental luck is insufficient for knowing or being in a position to know P.

Environmentally Lucky Possession: S can possess the fact P as a reason even if one’s non-accidental true belief that P is formed in circumstances involving environmental luck.

This is a problem for advocates of \((\hbox {P}\rightarrow \hbox {K})\) and \((\hbox {P}\rightarrow \hbox {PK})\).

But it is easy to see how an advocate of \((\hbox {P}\rightarrow \hbox {A})\) could accommodate the thought that we possess reasons in cases of environmental luck. For example, one could adopt (VirtueA) while maintaining some kind of modal anti-luck condition that only applies to knowledge (safety, sensitivity). On this approach, knowledge will have a more demanding non-accidentality condition than awareness. This is something non-robust, anti-luck virtue epistemologists will find attractive (Pritchard 2012a; Kelp 2013). For on their view, cases of environmental luck involve a genuine cognitive achievement even though they fall short of knowledge, and the relevant cognitive achievement is explicitly identified as an instance of (VirtueA): true belief (and hence true representation) from a non-defective exercise of a relevant cognitive ability. But to know, they argue, one’s cognitive achievement must satisfy some further modal anti-luck condition (safety, sensitivity).

For those of us who think that knowledge is incompatible with environmental luck, one of the nice things about going on to endorse \((\hbox {P}\rightarrow \hbox {A})\) together with a condition like (VirtueA) is that it provides a way of seeing how cases of environmental luck form a middle-ground between knowledge and ignorance. For while Henry doesn’t know that the object he’s looking at is a barn, he’s not ignorant of that fact either. So what’s in between knowledge and ignorance? Intuitively, it’s something robustly epistemic that puts one in possession of facts. After all, if Henry were not in a robustly epistemic relation to the target fact there would presumably not be so much debate over whether or not environmental luck cases are cases of knowledge. Since awareness is robustly epistemic we have our middle ground as well as a novel explanation for why some epistemologists have been (wrongly) tempted to think there is knowledge in cases of environmental luck. For awareness is so much like knowledge (in being a certain generalization of it) that it’s presence in cases of environmental luck was doubtless going to mislead some reflective people.

To conclude, in this section we’ve seen that identifying possession in some way with the more general state of non-accidental true representation (awareness) can also explain the facts we possess in cases of environmental luck and in cases involving certain deductive belief-updating procedures. In contrast, knowledge-centric accounts of possession cannot explain the facts we possess in such cases. Accordingly, we have our main conclusion: accounts of possession framed in terms of the more general state of non-accidental true representation (awareness) are at least as well placed to explain our possession of facts as existing knowledge-centric accounts, and they are better placed to explain our possession of facts than existing knowledge-centric accounts.

6 Concluding reflections on Lord, lotteries, and other counterexamples to knowledge-centric theories of possession

While this paper has focused on the epistemic condition required for possessing reasons–i.e. possessing facts–it’s worth pausing to see how the arguments of this paper might impact other aspects involved in formulating an adequate theory of possession. For example, Lord (2018a, Chapter 4) argues for the following claims:

Lord’s Theory of Possessing Reasons (LPR). S possesses a fact P as a reason to \(\phi \) iff S is in a position to manifest knowledge about how to use P to \(\phi \).

Lord’s Entailment Thesis (LET). If S is in a position to manifest knowledge about how to use P to \(\phi \), then S is in a position to know P.

In the course of arguing for \((\hbox {P}\rightarrow \hbox {A})\) above I provided two arguments for thinking that one can possess a fact as a reason to respond in some way without being in a position to know it. In the case of environmental luck I argued that one can possess the fact that a barn is nearby as a reason to run it its direction, without being in a position to know that a barn is nearby. In the case of Kripkean Study I argued that one could possess the fact that (N) as a reason to update one’s beliefs without being in a position to know (N). Notice that this entails that (LP) and (LET) cannot both be true.Footnote 20 Therefore, either Lord’s theory of possessing facts as reasons to \(\phi \) is mistaken, or Lord’s entailment thesis is mistaken. I do not have space here to explore this further.

The next issue to discuss concerns lottery cases and their attendant form of modal luck. Above I argued that we could exploit the theory of awareness in such a way that we can be aware of and possess facts in cases of environmental luck while failing to know them. All that is required is that awareness be somewhat less demanding than knowledge. I illustrated this possibility for \((\hbox {P}\rightarrow \hbox {A})\) by appealing to (VirtueA) while maintaining some kind of modal anti-luck condition that only applies to knowledge (safety, sensitivity). But lottery-like cases generate worries about the stability of this position (cf. Williamson 2000, 200ff).

To see the problem take a standard lottery case:

Lottery

You have a ticket for a fair lottery with very long odds. The lottery has been drawn, although you have not heard the result yet. Reflecting on the odds involved you conclude (L) that your ticket is a loser. Besides your competent assessment of the odds, you have no other reason to think your ticket is a loser. As it turns out, your belief that you own a losing ticket is true.

A common judgement about (L) is that it cannot be known on statistical evidence alone and therefore you cannot be in a position to know (L) in Lottery (Hawthorne 2004; Williamson 2000). Together with \((\hbox {P}\rightarrow \hbox {PK})\) this has the intuitive consequence that you cannot possess (L) as a reason to believe or act. For instance, you cannot have (L) as a reason to sell your ticket or to believe that you will not have enough money to go on an expensive vacation later in the year. This raises problems for endorsing \((\hbox {P}\rightarrow \hbox {A})\) together with a theory of awareness, like (VirtueA), that allows you to be aware of facts in cases of environmental luck. I emphasize, this is not a problem for \((\hbox {P}\rightarrow \hbox {A})\) itself, rather it is a problem with theories of awareness like (VirtueA) that don’t rule out possession in cases like Lottery.

Fortunately, this problem can largely be avoided by appreciating a certain form of defeat that seems to apply in typical lottery cases. For it is plausible that you cannot have sufficient reason (=justification) to believe (L)–despite your statistical reasons–if you have further undefeated reason to think that you’re not in position to know (L). For being in a position to know (L) requires that your over-all circumstances are hospitable to knowledge. Thus having undefeated reason to think that your circumstances prohibit the possibility of knowing (L) seems like the kind of higher-order information that can defeat one’s justification for responding to (L) (Smithies 2012; Prado 2020).

How is it that you and other reflective agents come to have undefeated reason to think that you’re not in a position to know (L)? In my experience people usually flag the modal defect with lottery propositions like (L) by appreciating that their ticket might be a winner–i.e. (L) might be false–and by concluding from that fact that they don’t know (L). So provided one cannot be justified in believing (L) if one has sufficient reason to think one doesn’t know (L), it will follow that one must lack sufficient reason to believe (L) despite it’s statistical support. Notice that the conclusion is not that one lacks any reason to believe (L). Rather, the claim is that in Lottery cases where one has undefeated (=sufficient) reason to think they don’t know (L) one lacks sufficient reason to believe (L). So, in my view, while \((\hbox {P}\rightarrow \hbox {A})\) and (VirtueA) ensure that it’s possible for thinkers to possess and have sufficient reason to respond to (L), typically mature thinkers also possess further reasons that prevent them from having sufficient reason to respond to (L) in belief and action.

The final thing to discuss before concluding concerns further potential counterexamples to \((\hbox {P}\rightarrow \hbox {K})\) and \((\hbox {P}\rightarrow \hbox {PK})\). For as I mentioned above that many believe there are ways of representing a fact without belief. All of these ways can be used to generate instances of awareness without belief. For example, some have argued that one can see that P without believing and hence without knowing P (McDowell 1998; Pritchard 2012b; Turri 2010; Schroeder MS); others have argued that one can remember that P without believing that P (Bernecker 2010); similar claims can be made about perception in general (hearing that P, tasting that P, etc.), rational intuition that P (Bealer 2000; Pust 2000), and perhaps even the idea of having been informed by someone else that P. If there is a range of representational states wherein one can non-accidentally represent P without believing it, then the counterexamples to \((\hbox {P}\rightarrow \hbox {K})\) and \((\hbox {P}\rightarrow \hbox {PK})\) will be many. While I’m rather optimistic about this expansion of counterexamples, they tend to rely on controversial assumptions. So I’ve set discussion of them aside.

In conclusion, I’ve argued against putting knowledge-first within the reasons-first framework. The good news is that there is an alternative view, \((\hbox {P}\rightarrow \hbox {A})\). \((\hbox {P}\rightarrow \hbox {A})\) does at least as well as \((\hbox {P}\rightarrow \hbox {PK})\) when it comes to explaining the reasons we fail to possess in Littlejohn’s case of Agnes and the reasons we possess Lord’s cases of inattention. Moreover, \((\hbox {P}\rightarrow \hbox {A})\) does better than \((\hbox {P}\rightarrow \hbox {PK})\) when it comes to explaining the reasons we possess in cases of environmental luck and cases involving explicit deductive updates. Indeed, Extended Possession, Explicit Deductive Updates, and Kripkean Study generate no problems for \((\hbox {P}\rightarrow \hbox {A})\). At the very least, then, \((\hbox {P}\rightarrow \hbox {A})\) represents a unique and theoretically interesting theory of the epistemic demands of possessing reasons that has so far gone unconsidered in the expanding literature on possession.