Knowledge-that is knowledge-of

I. Two Kinds of Knowledge?

The word ‘know’ is…used in two different senses. (1) In its first use it is applicable to the sort of knowledge which is opposed to error, the sense in which what we know is true…In this sense of the word we know that something is the case. This sort of knowledge may be described as knowledge of truths. (2) In the second use…the word applies to our knowledge of things…The distinction involved is roughly that between savoir and connaître in French, or between wissen and kennen in German. (B. Russell, Problems of Philosophy, 1912/2001, 33)

Thus Russell enshrined in modern philosophy the idea that there are two genuinely distinct kinds of knowledge. In subsequent discussions these show up variously as:

• Knowledge of truths vs. knowledge of things

• Factual or propositional knowledge vs. objectual knowledge

• Knowledge-that vs. knowledge-of or knowledge NP [noun phrase]

Although Russell introduced these on equal footing, their fates in epistemology have widely diverged.

The first kind has been the star. It is the topic of much theorizing. Some epistemologists explicitly claim that it is the fundamental or central kind of knowledge; many imply as much by focusing on it exclusively, as when for example asserting that knowledge is a kind of belief.

The second kind, by contrast, has gone largely undertheorized. Contemporary epistemologists sometimes have a say about it, but it is clearly second fiddle—or even third, after knowledge-how.¹ Some construe it as a special phenomenon occurring in only a narrow range of circumstances: knowledge of people or places based on personal interaction, or the “knowledge by acquaintance” we have in direct, unmediated experience.² Others argue that it reduces to knowledge-that, so that knowing New York, for example, is just a matter of knowing facts about New York.³

In this paper I propose a very different view—one that I find in the works of Aristotle (although I do not present much evidence for the Aristotelian origin here), and that I think deserves to be put back on the table. On this view, the division between knowledge of facts and knowledge of things does not carve nature at the joints. There is just one generic kind of knowledge, and it can be had either of facts or of things. It does however have many of the hallmarks of Russell’s second kind: it is what we pick out with uses of ‘knows-NP’; it comes in degrees; it entails recognitional capacities; it is characterized by something roughly like familiarity. Indeed, arguably it just is Russell’s second kind, construed so as to remove the restriction to things in contrast to facts. To avoid confusion however I give it a new label: the Ancient Greek term ‘gnosis’. I will argue that gnosis is generally best understood as the condition in which some part or aspect of reality is revealed to us, rather than being opaque or obscure.⁴

Thus, the central kind of knowledge is knowledge-NP, gnosis. As for knowledge-that, it turns out not to be a distinct kind of knowledge but rather just a special kind of gnosis: gnosis of facts.⁵ All knowing is gno-ing.

My aim here is twofold: first, to draw attention to and begin to develop an account of gnosis as a neglected and misunderstood but important phenomenon; and second, to present and begin to defend an account of knowledge-that as gnosis of facts. I will argue that this account shows promise of succeeding where its rivals do not, both in respecting our core intuitions about knowledge-that and in explaining them.

II. Gnosis

We speak of knowing-NP all kinds of things: persons, God, Namibia, tree-frogs, quantum mechanics, Beatles songs, the taste of Marmite, the pain of loss, Cretaceous dinosaurs, the works of Aristotle, and so on. Of course, there are significant differences among these: some rely on direct awareness (“I know that taste”), while others rely on causal interactions (“I know your mother”), and others instead on conceptual and propositional abilities (“She knows her Cretaceous dinosaurs/the works of Aristotle”). These differences are stark enough that one might suspect that there is no unified phenomenon here, but instead several quite different things lumped together under the same grammatical construction. Knowing an object would be either having direct acquaintance with it, or having personal experience with it, or knowing facts about it. If this is so, then knowledge-NP is not a natural kind, and certainly not a plausible candidate to be the primary, central, or sole kind of knowledge. Perhaps direct awareness is a natural kind, and perhaps so too is personal experience—but both of these, while interesting in their own right, are clearly quite limited in their domains.⁶

I want to show that there are in fact significant commonalities across all these cases—enough, in fact, to suggest that they are all really instances of a single kind, what I’ll call gnosis. I will not attempt an analysis of gnosis here, partly because I am not sure that one can be provided. I will, though, attempt to show that there is a distinctive and unified phenomenon, and to identify its typical features. My strategy is to look at the process by which we come to have such knowledge.

Take knowledge of persons as a paradigm. At first, a person is unknown to you, a stranger: they are opaque or mysterious, or perhaps you are completely unaware of them. Then you get what I will call a window onto them: something that gives you cognitive access to them. The window could be direct, such as perception or personal interaction, or indirect, such as trusted, reliable testimony (you read a biography, or hear a description from someone who knows them). If things go well, you use this window to get a good view, literal or metaphorical, of the person. You pay attention when you are looking at them or hearing about them, and your view is not clouded by serious confusions (you are not falsely convinced that everything you hear about them is a lie; you are not under a delusion that they are someone else in disguise). If these conditions are met, then you are having what I will call a revealing encounter with the person.

Now you have crossed a threshold in your cognitive relation to this person: to you they are now a known person. That is, they are now revealed to you, at least to some small extent—they are no longer wholly opaque or hidden. You have registered them in your mind. And in virtue of this registration, you now have a certain cognitive facility with them: at a minimum you might be able to recognize them by sight when you see them again, or be able to “place” them mentally when you hear about them in some other context.

Moreover, if the encounter continues, or if you have more such encounters, you can get to know them better—more fully and deeply. Now that you have a window onto them you can take a closer look. Perhaps at first you know them just barely, but with a better view you can become an expert about them, or even come to understand them.

So a known person, in contrast to an unknown person, is one who is revealed to you to some extent, and a better-known person is one who is revealed more fully and deeply.

It may seem a stretch to say that when you merely notice someone in such a way that you can now recognize them by sight, this counts as them being “revealed” to you. I say it nonetheless, to draw attention to the sharp contrast between even this minimal knowledge of someone and total ignorance of them, and to the continuum between this minimal knowledge and the knowledge you have of someone who is very revealed to you, well-known. What I want to insist on is that the cognitive relation you have to someone you can recognize by sight is the same relation, in a much lower degree, as the cognitive relation you have to someone you know well. This is why we can refer to both with uses of ‘knows-NP’.

I have not here offered an analysis of knowing a person. But I do hope to have captured some core intuitions about the difference between a person being unknown, known, and well-known.

What I want now to emphasize is that this same account also applies to other cases of knowledge-NP. Paris, dinosaurs, the Pythagorean Theorem, the works of Aristotle, or a particular shade of blue can, just like persons, be unknown, barely known, or well-known. And while the details vary, the basic pattern is the same. For any of these things to be unknown to you is for that thing to be mysterious, or obscure, or totally outside your awareness. Then, just as you can get a window onto a person, so you can get a window onto any of these things, although what counts as a window will depend on what the thing is. (For a color, the best window is vision; for a city, a visit or a documentary; for dinosaurs, books or bones or a lecture series; for philosophical works, a good read; for a theorem, a statement or a proof.) In these cases too, if you pay attention, and do not have your mental view blocked by serious confusions, you can have a revealing encounter with the thing, although what gets revealed in the encounter will also vary (the color itself; some distinctive features or parts of a city; the skeletal structures or eating habits of dinosaurs; the content of Aristotle’s works, or of the theorem.) Now you have crossed a threshold: the thing is known to you, at least to some extent. As a result, just as with persons, you will have some cognitive facility with it, where this too will vary depending on the kind of thing. (You can recognize the color on another sighting, or discriminate it from similar ones; you can recognize or navigate the city; you can identify some dinosaur skeletons or Aristotelian doctrines; you can state the theorem.) Over the course of prolonged or repeated encounters, moreover, these things can become better known, where just as with knowing persons, the metrics are fullness and depth.

These parallels suggest that it is not just a grammatical accident that we speak of knowing-NP various kinds of things, and that we speak of doing so not only through direct acquaintance or causal interaction but also through testimony or inference. It is plausible that there is a real, unified phenomenon here.

Of course, some core uses of ‘knows-NP’ do very strongly imply personal experience. But while “She knows Obama” or “She knows Kansas” have this implication—she has met him, she has been there—near-synonyms do not. “Obama is known to all Americans” is true not because we have all met him, but because he is familiar to us: we can recognize him in pictures and make predictions about his behavior. “Ancient Troy is well-known to her/She is well-acquainted with it” is true of an expert historian because she knows lots of facts about it. I surmise that, because knowing a person or place by direct experience is a particularly salient and important way of knowing them, the verb ‘know’ used in connection with people and places tends to have this narrower sense: know-by-direct-experience, know-personally. But—and this is the crucial claim for my purposes—knowing-personally is just one way of knowing in the broader sense. Interacting with a person is one (salient, important) way of putting yourself into the very same general relation you can get into by reading a biography of them, or by hearing about them from their friends: a relation that is the opposite of their being opaque to you, that is gained through a revealing encounter, and that comes in degrees, where the lowest level is superficial familiarity and the highest level is expertise or understanding.

Colloquially we sometimes refer to this relation as familiarity, or being well-acquainted with something, but both these terms have misleading connotations.⁷ We also refer to it as knowledge-of (“She has deep/superficial knowledge of Cretaceous dinosaurs/Aristotle’s corpus”)—hence this paper’s title—but this too can be misleading: “I know of her” implies slight awareness by contrast with real knowledge. Thus, I choose as a neutral label the Ancient Greek term ‘gnosis’.

Here is a rough attempt to characterize it, summing up the above discussion:

Gnosis: You gno a thing to the extent that:

(a) it is revealed to you in a way appropriate to the kind of thing it is, enough so that as a result

(b) you have some cognitive facility with it, appropriate to the kind of thing it is.

I submit that gnosis thus understood is a familiar notion, one to which we appeal regularly in our uses of knows-NP. It does not, however, correspond precisely to any category prominent in contemporary epistemology. Gnosis is obviously akin to Russell’s second kind of knowledge, “knowledge of things,” but it does not quite fit with the ways philosophers often characterize this. It is not Russellian acquaintance:⁸ although this often amounts to gnosis, one can see something without getting to gno it at all, and one can gno something through testimony without direct acquaintance; moreover gnosis comes in degrees, while direct acquaintance does not. It is not just knowledge based on personal experience or interaction, although this is one instance of it. It is not just de re knowledge or awareness of an object,⁹ for although this is necessary for gnosis—you cannot get to gno a thing without that very thing somehow “getting into your thoughts”—it is not sufficient: arguably one can think de re of something without gno-ing it at all (“I’ve heard you mention Ana but I don’t know her”). It is not identical to understanding, for one can gno something without understanding it. Instead, understanding something is gno-ing it to a high degree.

Finally, and very importantly: counter to the way that objectual knowledge is often construed, gnosis is not essentially non-propositional—and thus not essentially opposed to knowledge-that.¹⁰ Everything I have said is compatible with the view that all gno-ing is propositional. Moreover, even if there is non-propositional objectual knowledge through direct acquaintance, one can gno a person or a place by learning facts about them through testimony, and there are other things one can only get to gno in this way. In addition, as I will argue below, we can have gnosis of facts or propositions just as much as we can have it of people or places. What is essential to gnosis is not the kind of object it can take, but instead its nature: roughly, familiarity, or good-acquaintance—a relation where something is to some degree revealed to one, rather than remaining obscure.

Recent philosophical attention seems, then, to have passed gnosis by. Nonetheless, I have argued, it is a very familiar notion. It is also a very important one. Gnosis is something we strive for, and deeply value. We want reality to be open and revealed to us, rather than hidden and obscure; we want to be well-acquainted with reality, rather than have it be a stranger. Here is a statement of just that view from one of the first theorists of knowledge, Aristotle:

The fact that most people avoid death also displays the soul’s love of learning; for it flees the things it does not know (gignōskei), what is obscure and not clear, and by nature it pursues what is evident and what is known (gnōston)…It is for this same reason that we also enjoy what we are accustomed to, both things and people, and call known/familiar (gnōrimoi)¹¹ people ‘friends’. These things, then, show clearly that what is known and evident and clear is loved; and if what is known and clear [is loved], then it’s clear that necessarily so too are knowing and being wise (gignōskein, phronein).¹² (Aristotle, Protrepticus B102)

Aristotle holds that we love things that are known to us in the way that familiar people are known—indeed, that we love familiar people just because we know them, and hate death just because it is obscure.¹³ And he takes this love of the familiar as a sign of our love of knowledge.¹⁴ We want and love the kind of knowledge that one has of a known person, and that one lacks of a mystery. That is, we want and love gnosis.

Aristotle is onto something here: we do tend to want reality to be revealed to us. This desire can manifest as simple curiosity, as when we want to know a secret or solve a puzzle, no matter how inconsequential. It can also manifest as the “thirst for knowledge,” as when we want to learn all about history or biology or human nature. Or—because gnosis comes in degrees, and to gno a complex, structured object well just is to understand it—it can also manifest as the desire for understanding, as when we desire to get to the bottom of something, to see how it hangs together, to grasp not just the facts but their explanations.

I have made a case that gnosis is a real, unified phenomenon, one both very familiar and highly valued. One might, however, accept all this, and still think that there is also another and at least equally important kind of knowledge: knowledge-that—factual knowledge, knowledge of truths.

It may seem that this kind of knowledge is obviously different from gnosis—that knowing that p is not a matter of being familiar or well-acquainted with something in the way one is with a person. It may also seem that it is more fundamental than gnosis: that to gno a thing just is to know some important facts about it.

I want to show that, on a promising view, knowledge-that is neither more fundamental than gnosis nor even distinct from it. Instead it is just one kind of gnosis: gnosis of facts. More precisely, it is gnosis of facts, constituted by true belief. And thus, although knowing a person or city usually entails knowing facts about them, the factual knowledge is not something distinct from and more fundamental than the gnosis; rather, there is gnosis all the way down.

I begin by showing the roots of this account of factual knowledge in Aristotle, and then defend the account in its own right.

One caveat: there is an important question I will not try to answer here about the relation between gno-ing an object and gno-ing facts about it. In many cases, gnosis of an object is arguably constituted wholly by gnosis of facts about it: I know you by being well-acquainted with your sense of humor, your love of polka, etc., where this entails knowing (gno-ing) the fact that you have this property, the fact that you stand in that relation, and so on. On the other hand, perhaps one’s awareness of the properties and relations can be strong enough to support gnosis of the object without being so strong or salient as to constitute gnosis of these facts. In addition, perhaps one can gno a simple phenomenal object like a color or sound without needing to focus on connections between it and other things, or connections between its properties, and thus can gno it without gno-ing facts about it. I leave these questions for another occasion.

III. Knowledge-that as gnosis in Aristotle

Aristotle is arguably the first to introduce into Western philosophy a general notion of knowledge.¹⁵ He applies the noun ‘gnōsis’ and related verbs to perception, experience-based knowledge, craft-knowledge, practical wisdom, scientific understanding, intellectual intuition, and philosophical wisdom. What he means, as I argue elsewhere, is that these are all ways of being well-acquainted with reality.¹⁶ Recall his claim in the Protrepticus: we love learning and knowledge because we love things that are known to us in the way that friends are known. The implication is that all knowing is gno-ing.

If we think of gnosis as knowledge exclusively of things, by contrast with knowledge of facts—Russell’s distinction—we might think that in construing knowledge as gnosis, Aristotle is ignoring knowledge-that. But many of the phenomena he characterizes as gnosis clearly consist in knowledge of facts: for example, scientific understanding (epistēmē) of the conclusion of a demonstration, or the judgment of practical wisdom (phronēsis) that a particular action is the right one. In calling these gnosis, is Aristotle conflating knowledge-that with knowledge-NP?

It is striking that he does not seem to notice any need to track the distinction. Indeed, he switches back and forth so casually between talk of knowing objects and talk of knowing facts or propositions that there is real indeterminacy about whether, for example, the objects of perception in De Anima, or the objects of nous in Posterior Analytics, are propositional or objectual.¹⁷

Moreover, when we look at his treatment of known “thats”—facts or propositions—we find strong evidence that they are known in the same way that people or places are known. For example:¹⁸

Practical wisdom (phronēsis) is about the particulars, which become familiar/known (gnōrima) through experience…[Such knowledge is important because] mistakes are either about the universal in deliberation or about the particular: either that (hoti) all heavy water is bad, or that this [water] is heavy (Nicomachean Ethics 1142a14—23)

“The that” is a fact: the fact that heavy water is bad, or that this is water is heavy. Aristotle is saying that practical wisdom, which he elsewhere calls gnosis,¹⁹ involves having such facts be known to one on the basis of experience—that is, on the basis of frequent exposure. The strong implication, bolstered by the connotations of familiarity in ‘gnōrimon’ (known), is that, to the practically wise person, such facts are well-known in broadly the same way that familiar people are well-known.²⁰

As for the facts known “by nature” rather than “to us,” such as the facts about essences which constitute the first principles of scientific demonstrations, these too are objects of gnosis: not now because we are familiar with them on the basis of repeated exposure, but because to the educated they are fully revealed, understood.²¹

Thus, for Aristotle, knowledge-that is not distinct from gnosis, let alone more fundamental. It is instead one kind of gnosis: gnosis of facts (either superficial gnosis (familiarity), or deep gnosis (insight and understanding)).

Of course, Aristotle might just be confused. But I want to consider another possibility: perhaps he was onto an important truth, and his innocence of the heavy weather made in later times of the difference between factual and objectual knowledge left him free to recognize their unity. Perhaps Aristotle is offering us an uncorrupted view to which we should return. Knowledge is one unified kind: gnosis. Knowing that something is the case is the main way we stand in the gnosis-relation to facts, just as seeing is the main way we stand in this relation to colors, or personal acquaintance is the main way we stand in it to people.

To be clear, I am not merely proposing the general view that knowledge-that is an objectual relation to a fact.²² I am proposing the specific view that it is the relation of gno-ing a fact—being cognitively intimate with it in broadly the same way that one can be with a person. That, with some refinements, is the view I wish to defend in the rest of this paper.

IV. Knowledge-that as Gnosis: First Pass

The view may seem a simple non-starter. Here is a stark rejection:

Certainly one may speak with equal propriety in English of knowing individuals and knowing facts, but the word ‘know’ is used in different senses in the two uses. To know an individual is to recognise or identify it, or at least to have the capacity to do so; or to know it in the sense of being familiar or acquainted with it, or of having personal experience of it. Knowledge of a fact lacks these connotations.²³ (Hossack 2007, 8, emphases mine)

I want to show that, on the contrary, an analysis of knowledge-that as gnosis fits well with our core intuitions about knowledge-that. I will make a brief, broad-strokes case for this claim in this section and then defend it more carefully in the rest of the paper.

As a way into the idea, consider some suggestive albeit inconclusive linguistic data. We frequently speak of people being familiar with or acquainted with facts. Sometimes these locutions have connotations of frequent exposure (“We’re so familiar with the fact that the Wright brothers invented the aeroplane that the miraculous nature of their achievement goes unheralded”),²⁴ but often they seem simply paraphrases for “S knows that p” (quite frequently, for whatever reason, in questioning during trials or hearings: “Are you familiar/acquainted with the fact that…?”).²⁵

There are parallels in the very languages that have been thought to show a gulf between the two kinds of knowledge. In languages with a dedicated word for know-that (e.g., saber), when a fact is nominalized (“the fact that p”), it takes the word used for knowing people or places (e.g., conocer)—what I will call the gnosis-word. Moreover, in these languages, knowledge-that at least sometimes entails the other kind of knowledge: if you saber that the sum of the hypotenuse is equal to the sums of the squares of the other two sides, you conocer the Pythagorean Theorem, at least to some degree.²⁶ In the same vein, one can call a widely known fact bien connu, or wohl bekannt, just like a well-known person. Most strikingly, in many of these languages the general word for knowledge is the gnosis-word: connaissance, conocimiento, conoscenza. And the name for epistemology is often derived from this word (Craig 1990 mentions Théorie de la Connaissance, Erkenntnisstheorie, and equivalents in Polish and Hungarian; Greek sometimes uses gnosiologia, and Italian gnoseologia)—as if gnosis were the general phenomenon under investigation when we investigate knowledge.

All this suggests that in these languages, the knowledge-that verb does not refer to something distinct from gnosis, but is instead used to signify gnosis in a specific context. Perhaps these languages have two words not because there are two distinct kinds of knowing, but instead because knowing-that is a special, salient kind of gno-ing: perhaps saber is to conocer as ‘walk’ is to ‘move’.

(Some alleged counter-evidence: sentences such as “She knows that he is cruel, and him” are thought to be instances of zeugma, suggesting that ‘knows’ has two distinct meanings. I am dubious that these are really zeugmatic, rather than merely odd—indeed, with the order reversed [“She knows him and that he is cruel”] they sound pretty good—but the question deserves more attention.)

Why can we paraphrase “S knows that p” as “S is familiar/acquainted with the fact that p,” or do the equivalent in other languages? The practice makes sense if knowing a fact is similar to knowing a person: having it be revealed to you, rather than hidden or obscure—that is, gno-ing it.

Certainly facts, just like people or cities or species, are things one can gno—as we can see by drawing on the account of gnosis-acquisition developed above. Construe facts with minimal metaphysical commitment as ways things are. Clearly a fact just as much as a person can be unknown in the sense of being opaque, mysterious, or totally outside your awareness: you are wondering what the weather is like, or what the crisis is in Ukraine, or you have no idea that there is any crisis there at all. Then you can get a window onto a fact: you see or are told that it is raining, you hear or read that Russia invaded. If you pay attention, and your mind is not clouded by serious confusions, you get a mental view of it: you have a revealing encounter with it, in which you register it. As a result, in the future you will have a certain facility with it: at a minimum, you will be able to place it mentally when you encounter it in another context or under a slightly different guise (“Did you know that there’s war in Ukraine?”, “Ah yes, I did know that”). Now you have crossed a threshold in your cognitive relation to it: for you it is now a known (gnon) fact—revealed at least to some extent; no longer wholly obscure. Moreover, just as with persons or places, from here you can also get to gno the fact better – more fully or deeply. You go outside and feel that it is raining; you study the situation in Ukraine and come to understand better how things are, and why.²⁷

Thus, just like a person, a fact can be known (gnon) instead of unknown, and a known fact can fall anywhere on the scale from just barely known to very well-known, where the lowest level is superficial awareness, and the highest level is insight or understanding.

My proposal is that knowledge-that is not some other cognitive relation we bear to facts, but this very same one. Knowing that it is raining is being in a condition such that (a) the fact is revealed to you to some extent, enough so that as a result (b) you have some cognitive facility with it. Knowing a fact is gno-ing it.

For this proposal to have a chance at succeeding, the most urgent task is to show that it can respect our core intuitions about knowledge-that. In the rest of the paper I argue that the proposal—with one refinement, made in the next section—shows promise of not only accounting for these intuitions but also explaining them, and thus of offering that elusive thing: a satisfying account of knowledge-that.

V. Gnosis and True Belief

Although epistemologists disagree about how (and whether) to analyze knowledge-that, most agree on the following: if S knows that p, then S believes that p, S’s belief is true, and her believing truly is not a matter of accident. Any plausible account of knowledge-that should fit these constraints.

But it may seem that the account I have proposed roundly fails, for gno-ing a fact may seem very different from non-accidentally truly believing that it obtains (thus Hossack, quoted above). Indeed, it might seem that the two are not even mutually entailing. Couldn’t one non-accidentally truly believe that a fact obtains, without having enough of a handle on it to gno it? And couldn’t one gno a fact while not believing that it obtains, or while believing but only as a result of luck? These worries disappear, I want to show, once we spell out the notion of gno-ing a fact.

For the sake both of simplicity and of making the challenge most sharp, I proceed on two widespread though controversial assumptions: that part of knowing that p is believing that p, and that belief is an attitude toward propositions, where true propositions correspond to but are distinct from the facts underlying them. For those who reject these assumptions, the view will need to be refined, but the results will be variants on the same general picture. My aim is to defend a view on which knowledge-that, just like knowledge of a person or planet, is gnosis of some aspect of reality. Precisely how we construe the aspect of reality (as a fact? an obtaining state of affairs? objects and their properties and relations?), and precisely how we construe the relation between gnosis and propositional attitudes (as identity? entailment?), are questions I am happy to leave in principle open.²⁸

Operating with these assumptions, I aim to show that if knowledge-that is gnosis, knowing that p does entail non-accidentally truly believing it, because:

1. Under the proper conditions, true belief that a fact obtains constitutes gnosis of that fact, and

2. The requisite proper conditions ensure that gnosis-constituting beliefs are non-accidentally true.

In this section I explain and defend (1); in the rest of the paper I make a start on (2), although a fuller defense must wait for another time.

Gno-ing anything, I have argued, requires registering it during a revealing encounter, but what it takes to register something depends on its kind: registering a color is different from registering a song, pattern, person, or proof. What does it take to register a fact? Whatever your view of the metaphysics of facts, they are very plausibly the kinds of things well-represented by propositions. Therefore, one has an important kind of mental facility with a fact when one can represent it propositionally. This means that one registers a fact best by representing it to oneself propositionally.

Compare: neighborhoods are the kinds of things well-represented by maps, and so one registers the layout of a neighborhood best by drawing a mental map of it. If you get a window onto a neighborhood by walking around it, and you pay attention, absorbing and registering its layout, this registering consists in making a mental map of the neighborhood. You now have gnosis of the neighborhood, and the gnosis is constituted by your having the mental map. Likewise, I am suggesting, if you get a window onto a fact through perception or testimony or inference, and you pay attention, absorbing and registering the way things are, this registering consists in having a propositional representation of the fact. You now have gnosis of the fact, and the gnosis is constituted by your having this propositional representation.

Moreover, your attitude to this proposition is not one of doubt, or provisional acceptance, or mere entertaining. Registering is taking in and absorbing, and the proposition is your tool for doing this. So just like your attitude toward the mental map by which you register a neighborhood, your attitude toward the proposition by which you register a fact is one of what we might call reliance, or endorsement, or assent, or taking-to-be-true. And when held toward a proposition, this very plausibly just is the attitude of belief.

Thus we can best register a fact via a process which amounts to forming the belief that it obtains. Moreover, since the thing registered is, by stipulation, a fact, the belief will be true.

Consider a paradigm revealing encounter with a fact: you look outside and you register the fact that it is raining. My claim is that you do this by forming the belief that it is raining. Most picturesquely, you register the fact by saying to yourself “Ah, so that’s how things are: it’s raining.” More neutrally, you undergo some mental event that disposes you to say that it is raining if you are asked, or to look for an umbrella if you are heading out. Doing this counts as registering the fact, for now (a) the fact is revealed to you, at least to some extent—you are familiar with it under an important, revealing description, captured by the proposition that it is raining; and (b) this gives you a certain facility with the fact: you can recognize it under other guises (“The weather is bad today”—“Yes, I know”), or make inferences from it, or use it to guide your behavior.

A more vivid illustration, from a case involving more mental work: you sift carefully through various reliable news websites, trying to figure out what is really happening in Ukraine. The websites give you a window, and if you pay attention with an unclouded mind, you have a revealing encounter—you come to gno how things are. And you do this by forming a propositional representation of how things are, and settling on that representation as the right one—that is, by forming a belief (e.g., the belief that the Russians are mounting a full-scale invasion).²⁹

Believing a corresponding true proposition is not the only way to gno a fact. There may be worse ways. Arguably non-human animals, or humans using “System 1,” can be quite well-acquainted with facts without forming any propositional representations at all. Arguably, if q is a false proposition that is helpfully illuminating of a fact (e.g., a simplified model of a complex truth), one can get to gno the fact a bit by believing falsely that q.³⁰

There are also better ways: on the plausible assumption that a single fact can be represented by various propositions, sometimes truly believing one proposition will amount to only very superficial gnosis of a fact. Suppose you overhear people discussing the news, and you register just enough to form the belief “Something bad is happening in Ukraine.” Now you know that something bad is happening in Ukraine, but this is very low-level gnosis of how things are. You can get to gno the fact better by coming to believe other, more revealing descriptions: that there is war in Ukraine, that Russia has invaded Ukraine, and so on; ultimately, you can get to gno it so well that you not only know that it obtains, but also understand it.³¹ Nonetheless, just as registering a person’s physical appearance can be a superficial but genuine way of gno-ing them, so believing that something is happening in Ukraine can be a superficial but genuine way of gno-ing how things are in that portion of reality.

Note that this point allows the analysis to accommodate a widely cited difference between gnosis and knowledge-that: that the former comes in degrees (you can know a place or person better or worse), while the latter is absolute (either you know that p or you do not). (See, e.g., Dretske 1981, 363.) Once you register a fact by forming the belief that it obtains, in the course of a revealing encounter, you know that it is the case; past that point you can get to gno the fact better and better, even gno it so well that you understand it, but you will not thereby more know that it is the case.

My claim, then, is that believing even just one corresponding true proposition is a salient and important way to gno a fact. It is one that allows us to do a lot both theoretically and practically: to use our gnosis of the fact in reasoning, to reflect on it consciously, to express it verbally, to transmit it to others, and so on. Thus it stands to reason that this way of gno-ing merits a special locution of its own: knowing-that, or saber, kennen, etc.

We now have a refined and clarified version of the proposed account:

The Gnosis Account of Knowledge-that: Knowledge that p is gnosis of the fact that p constituted by true belief that p.

And so we see that the account meets one major constraint on any plausible account of knowledge-that: it preserves the intuition that knowledge-that entails true belief.

VI. Gnosis and Non-accidental Truth

But, of course, not every true belief amounts to knowledge. On a widely shared view, to count as knowledge a belief must be non-accidentally true³²—where non-accidentality is ensured by some special feature of the belief, although the identity of this feature is of course subject to much dispute.

So, one might object, the Gnosis Account is incomplete: at best, knowing that p is truly believing that p when (i) one’s belief constitutes gnosis of the underlying fact and (ii) the belief also meets some further condition which renders it non-accidentally true (such as being formed by a reliable process, or caused appropriately by the fact, or being sensitive and safe, or whatever the right condition may be).

My first response is that if this were so, we would still have made an important advance in our understanding of knowledge-that. The connection between knowledge-that and gnosis, although not providing an analysis, would illuminate knowledge-that by showing it to be one variety of a more general phenomenon: not radically different from knowledge of people and places, but instead, like such knowledge, a species of gnosis.³³ In particular, as my arguments in the final section will show, we would have made progress in understanding knowledge’s value.

I want, however, to show that a more ambitious project is promising. There is a good case to be made that there is no need for any further condition beyond what the Gnosis Account provides—that if a true belief meets the conditions for constituting gnosis, its truth will ipso facto be non-accidental in the way required for knowledge-that. If this is right, then the Gnosis Account is complete, and the connection with gnosis offers not just an illumination of knowledge-that, but an analysis. I cannot hope here to show decisively that this move will work: there are too many different kinds of case where true belief is thought to fall short of knowledge because its truth is accidental, and I cannot cover them all. But I aim to make a substantial start on the project by covering some classic cases.

Recall that, to constitute gnosis, a belief must be formed in special conditions (claim (1) above), namely as part of a revealing encounter with a fact. We form a belief only if we think that what we are encountering is a fact, but sometimes we are wrong. If we believe on the basis of misleading testimony, for example, or misperception, faulty inference, or tea-leaf reading, we are not having a revealing encounter with a fact, and so the belief will not constitute gnosis. This is obviously the case if the belief is false—if there is no corresponding fact out there—but it is also the case even if there happens to be a corresponding fact. Misleading testimony and the like afford no window onto facts; therefore beliefs formed on their basis are not formed as part of revealing encounters with facts, and therefore cannot constitute gnosis of facts. (Likewise, there are ways to draw an accurate mental map of a neighborhood that do not constitute gno-ing the neighborhood: one might make up a map out of thin air, in a way that happens to correspond to an actual neighborhood. But when one draws an accurate map on the basis of a revealing encounter with the neighborhood, the result is a special kind of mental map, analogous to the special true beliefs that amount to knowledge-that.)

I now make a case for claim (2): the special conditions required for revealing encounters turn out to ensure that the beliefs we form in such encounters—that is, gnosis-constituting beliefs—are non-accidentally true.

The basic argument for claim (2) is simple. Revealing encounters presuppose a special connection between mind and world, and it is this connection that is missing in true beliefs that fall short of knowledge-that. Such beliefs are true only because they happen to coincide with reality, while beliefs that constitute knowledge are true because they are connected to reality in the right way—namely, via a window allowing for a revealing encounter.

It will follow that non-accidentality is characteristic of knowledge-that not because non-accidentality has epistemic value beyond gnosis, but because the features securing it are entailed by gnosis. Because knowledge-that is gnosis, like all gnosis it requires a special connection to reality. This special connection renders the beliefs constituting it non-accidentally true.

A note: this discussion will bring out a similarity between the Gnosis Account and the causal theory of knowledge, on which a person knows just in case her belief is caused by the facts in the right way. On both views, knowledge requires a special connection between fact and belief, a connection which rules out a certain kind of accidental truth. It is worth emphasizing two differences, however. First, while it may be promising in many cases to spell out the notion of a window in causal terms (perception gives you a window onto the things causing it, for example), there may be non-causal windows too: perhaps informed prediction gives us a window onto the future, calculation gives us a window onto probabilities, abstract reasoning gives us a window onto abstract entities, or empathic imagination gives us a window onto others’ feelings. Second, and crucially, on the Gnosis Account a special belief-fact connection is required not because knowledge-that just is true belief connected appropriately to the facts, but because knowledge-that is gnosis, and gnosis requires that connection.

VII. Gnosis and Classic Gettier Cases

To show that gnosis entails the kind of non-accidentality required for knowledge-that, my strategy will be to show that this kind of non-accidentality is also present in gnosis of things other than facts, such as persons.

To begin with a case analogous to those that have played such a prominent role in debates about the analysis of knowledge-that, consider a (partly true!) story about gnosis-failure:

CATFISHED: Manti has been texting with someone he met on the net. The texts say that her name is Lennay, and that she is 22 and a student at Stanford. A distinctive voice comes through in the texts, as well as the impression of someone who is “gifted in music, multi-lingual, has dreams grounded in reality and the talent to catch up to them.”³⁴ Unbeknownst to Manti, however, there is no Lennay behind the texts. The author is a catfisher: Manti’s former fellow football-player, Ronaiah. But by an impressive coincidence, somewhere at Stanford there really is a 22-year-old woman named Lennay who is musical, multilingual, and has the same kind of writer’s voice and the same dreams and promise as “Lennay.” Neither Manti nor Ronaiah have ever encountered this person in any way. Is Manti getting to know the real Lennay?

The answer seems very obvious: No! The problem is not just that Manti does not know Lennay personally; as I argued above, people can be known to us without personal interaction. The problem is that he doesn’t know her at all—she herself is not in any way known to him.³⁵ (Indeed, he is not even aware of her existence!) Why doesn’t he know her? Because he lacks a window onto her. The texts he is reading are not from the real Lennay—she herself is not, as it were, there on the other side of those words to be known—nor does he have any other window onto her. She is totally obscured from his view. He therefore has no way to have a revealing encounter with her, and so cannot get to gno her.

Like the subjects in Gettier cases, Manti lacks knowledge. Moreover, like Gettier subjects’ beliefs, Manti’s cognitions (beliefs, impressions) match reality, but only accidentally so. They have this status because they correspond to a real person but were not formed via an encounter with that person. Metaphorically we might say that they were formed via a pseudo-encounter with a pseudo-person who happens to resemble a real person. If Manti’s impressions had come about through a genuine revealing encounter with the real Lennay—the very condition that would have afforded gnosis—then their correspondence to reality would not have been accidental. Gnosis of a person requires appropriate cognitive contact with the person, and this entails that gnosis-constituting cognitions conform non-accidentality to the person.

I want now to show that the parallel holds for gnosis of facts. Consider then the familiar:

COINS: Smith believes falsely although with justification (testimony from the boss) that Jones has been tapped for the job. He also believes truly that Jones has 10 coins in his pocket. He infers:

C The man who has been tapped for the job has 10 coins in his pocket.

As it happens, C is true: in reality Smith himself has been tapped for the job, and has 10 coins in his pocket. Thus Smith’s belief that C is true, but only accidentally so. Does Smith know that the man who has been tapped for the job has 10 coins in his pocket?

The classic and compelling answer is: No! Smith has a justified true belief, but he doesn’t know. Why? If knowledge-that is gnosis, Smith’s problem is just like Manti’s. He too thinks that a part of reality is being revealed to him: in his case, the fact described by C. And as it happens there really is a fact matching that description: call it RC (for “real coins”). But Smith has never encountered RC. If the boss had told the truth, that testimony would have given Smith a window onto the piece of reality that consists of he himself having been tapped for the job, and from there, with the help of observation and inference, he could have had a window onto RC. But that is not the situation here. RC itself is not there behind the boss’ testimony, being revealed through it, nor does Smith have any other window onto RC; he has therefore never had a revealing encounter with RC, and he therefore does not gno it.³⁶

Smith’s belief is true, meanwhile, only because (metaphorically) it was formed via a pseudo-encounter with a pseudo-fact happening to resemble a real one. If his belief had come about through a revealing encounter with RC—the very condition that would have afforded gnosis—its truth would not have been accidental. As with gnosis of persons, then, so here too: gnosis of a fact requires a revealing encounter with the fact, and this precludes the kind of accidental truth that makes us deny knowledge in Gettier cases.

The same strategy will work well in other classic Gettier cases.³⁷ When Smith looks at a stopped clock that happens to show the right time, the clock fails to serve as a window onto the fact that it is 3 pm, so he is not encountering that fact and so does not gno it. When he remembers parking his car on a corner but the car has in fact been stolen and left at another corner, his memory is not giving him a window onto the fact that his car is parked on a corner, so he is not encountering that fact and so does not gno it.

A caveat: of course there are important questions about what counts as a window, and if we need to appeal to intuitions about whether someone has knowledge in order to determine whether he has a window, the Gnosis Account risks becoming trivial and ad hoc. I am not here aiming to present necessary and sufficient conditions for windows, or for revealing encounters. Perhaps that could be done, but we would need different sets of conditions for different types of fact—empirically observable facts, mathematical facts, statistical facts, and so on. I do think, however, that we have intuitions about what constitutes a window in various cases: for example, it seems clear that the fact that Smith’s car is parked on the corner where the thief left it just is not there on the other side of his memory of parking the car himself, getting revealed through it. Moreover, such intuitions can be guided and anchored by analogies with gnosis-enabling windows onto persons. So while the notion of a window needs elaboration, we have enough to go on here to see not only that the Gnosis Account delivers the right verdicts in these cases, but also that it does so without being trivial.

To take stock: I have argued in this section that classic Gettier cases present no need to add a non-accidentality clause to the Gnosis Account of knowledge-that. When our true beliefs constitute knowledge, it is because we have them in virtue of a gnosis-enabling connection with the facts; this connection precludes the kind of accidentality characteristic of Gettier cases. It follows that the Gnosis Account not only accommodates our intuitions about these cases, but can also explain them. Why do we shrink from counting Gettierized beliefs as knowledge, given all their merits? It is not that we fetishize a certain hard-to-define lack of accident for its own sake. It is that we are distinguishing beliefs that just happen to get things right from beliefs that allow us to do something very important: to know—gno—reality.

VIII. Other True Beliefs That Fall Short of Knowledge

The Gnosis Account does well with classic Gettier accidents, but what about other cases of knowledge-undermining accident? Although I cannot hope to cover all such cases, here is a general prediction. Since, like the causal theory, the Gnosis Account makes knowledge depend on a successful connection between mind and fact, it too will perform well with cases where such a connection is lacking (like Gettier cases, and perhaps others too, e.g., lottery cases);³⁸ but it too will have trouble with cases where such a connection seems to hold but other factors count against attributing knowledge. In such cases, proponents of the Gnosis Account will either have to deny that the subject really has gnosis of the fact, perhaps by drawing on intuitions about what counts as a revealing encounter, or they will have to bite the bullet and insist that the subject does have knowledge. I will show briefly how this plays out for some classic cases. To begin with:

BARN:³⁹ Henry is looking at a barn and believes truly that there is a barn over there; unbeknownst to him, however, every other thing that looks like a barn in the area is a convincing fake, and he just lucked out by looking at the one real barn. Does Henry know that there is a barn over there?

The Gnosis Account seems to say that he does, for his visual experience gives him a window onto the fact that there is a barn over there, and thus allows for a revealing encounter with it—the fact is right there on the other side of his visual experience, being revealed through it. But, the standard line goes, Henry does not know, because his belief is true by luck. Is the Gnosis Account giving the wrong answer?

Intuitions diverge about BARN,⁴⁰ and rather than accepting as given that Henry lacks knowledge, we can use our account to help us figure out what to think. Following the strategy of understanding knowledge-that by comparison with gnosis of persons, compare the following:

NOT CATFISHED: Every other “woman” tweeting at Manti these days is a catfisher, but he lucks out by following the one thread that can lead him to a real woman, Lana. Through their ensuing texts he comes to have lots of impressions that are true of Lana. Is Manti getting to know Lana?

Our picture of gnosis says that he is, since the texts provide a window onto her. But his impressions are true only by luck. Are we getting the wrong answer? No: Manti very clearly does know Lana. He is lucky to know her, in that he could easily have failed to be getting to know anyone, but nonetheless he really is getting to know her, because their correspondence really is a window, with Lana herself right there on the other side being revealed. And the analogy suggests that, in BARN, we should just bite the bullet—as others have done on other grounds. Henry is lucky to have knowledge, in that he could easily have failed to have it, but he nonetheless really does have it, because he really is having a revealing encounter with the fact.

A similar approach works for another kind of putatively knowledge-undermining luck, the kind caused by unpossessed defeaters:

ASSASSIN:⁴¹ Jill reads a true news report of a political assassination, from a respectable source, and then misses a convincing but false broadcast denying the first report. Does Jill know that the assassination took place?

Again, the Gnosis Account seems committed to saying yes, since Jill had a revealing encounter with the fact. Likewise, if after texting with Lana (the real woman), Manti missed a lying voicemail from Ronaiah claiming Manti had been catfished, he would nevertheless still know Lana. So here too the Gnosis Account encourages us to bite the bullet (again in company with others):⁴² Jill is lucky to know, in that she narrowly escaped having her window onto the fact clouded by a confusion—but she really does know.⁴³

In this section and the previous one I have reviewed prominent cases in which it seems that true belief falls short of knowledge, and I have argued that the Gnosis Account handles them well. In these cases, the account not only respects but indeed explains the view that knowledge-constituting belief must be non-accidentally true.

Of course, I have not been able to review every kind of problem case, and if induction is to be trusted, the history of epistemology shows that there will be apparent counterexamples! What I hope to have shown is that the Gnosis Account is promising enough to make it worth trying to meet the counterexamples, rather than retreating to the weaker position on which the connection with gnosis does not define knowledge-that, but merely illuminates or elucidates it.

The most straightforward strategy for dealing with apparent counterexamples would be to refine the notions of “window” and “revealing encounter,” showing that, in each case in which our intuitions tell us that a subject lacks knowledge-that, she also lacks a genuine window onto the fact, or is not genuinely registering it, and hence does not gno it. I am optimistic about this project—although wary of the danger of letting intuitions about when a subject has knowledge-that dictate theories about when she has a window or a revealing encounter, such that the account loses its explanatory power.

The other strategy would be to refine the account itself, adding further conditions after all. The extent to which this expansion would depart from the spirit of the Gnosis Account would depend on the rationale for those further conditions.

Recall that not every way of gno-ing a fact counts as knowing that it obtains: we have knowledge-that only when our gnosis is constituted by a true belief. This condition, I argued above, is not arbitrary or ad hoc, for this way of gno-ing a fact is a very important one—it is useful for reasoning, reflection, expression, and other purposes. Hence it is sufficiently salient to merit its own special locution: ‘knowing-that’. The hope would be that if we need further conditions to rule out accidentality, those conditions would turn out to be similarly non-arbitrary. We would wind up with a more refined account—to know that p is to gno the underlying fact when (a) the gnosis is constituted by belief in a corresponding true proposition, and (b) XYZ—where the conditions expressed in (b) are not jerry-rigged to stave off counterexamples, but instead mark out an important way of gno-ing a fact beyond just condition (a): a way that makes gnosis more useful, or salient, or distinctive, and thereby renders gnosis that meets both conditions worthy of special classification under its own title, ‘knowing-that’.

IX. The Value of Knowledge

I have argued for the Gnosis Account by showing its potential to account for our intuitions about the nature of knowledge. In closing, I make a further and very important claim for the view: that it can also account for our intuitions about the value of knowledge.

Knowledge, as Aristotle famously declares, is something humans naturally desire. But the focus on knowledge-that and its relation to true belief has led epistemologists to struggle to explain the value of knowledge. It is clear enough that we value truth, but if knowledge is valuable for the sake of truth, why is it any better than true belief (the “Meno problem”)? Standard analyses of knowledge as true belief plus various other factors (justification and more) offer hope of identifying some further value beyond that of truth; in the end, however, they may only exacerbate the worry.⁴⁴ These problems are sometimes even thought to show that knowledge is not more valuable than true belief—that what we really care about beyond true belief is not knowledge but something else, understanding.⁴⁵

If knowledge-that is gnosis, however, the problem dissolves. For as I claimed above, the value of gnosis is clear. I am not sure how to defend the claim that we value gnosis beyond simply stating it: just as Aristotle says in the Protrepticus, we obviously do value having the world be revealed to us rather than opaque, clear rather than obscure, familiar rather than strange. And therefore, if knowing facts is gno-ing them, we obviously do value knowledge-that.

Indeed, if knowledge-that is gnosis, we can account for the intuitions behind the proposal that understanding is more valuable than knowledge, for gno-ing things well often amounts to understanding them. The desire to know can be satisfied to a low degree by a low degree of gnosis—curiosity about a trivial fact can be somewhat satisfied by a yes or no answer, just as curiosity about a person can be somewhat satisfied by learning their name. But it is much better satisfied by a high degree of gnosis: understanding. That is, we value understanding because we value knowing (gno-ing), and understanding is gno-ing to a high degree.

Most generally, if knowledge-that is gnosis, we can say why it is more valuable than mere true belief. When people say “I want the truth,” what they mean is usually not in the first instance that they want their beliefs to be true (although this will follow), but that they want answers: they want obscure things to become clear, hidden things to be revealed. What we want is not just to get things right, but to be intimately acquainted with how things are, in something like the way we can be acquainted with a person. If knowledge-that is gnosis of facts, then the true beliefs that give us what we want are the very ones that count as knowledge.⁴⁶