If you survey what analytic philosophers have said about God’s omniscience over the past 50 years, you’ll find that the vast majority of them define omniscience is strictly propositional terms. For example, William Lane Craig in vol. 2a of his Systematic Philosophical Theology (2025), defends the following definition of omniscience (p. 203):
(Oprop) S is omniscient =def. For all propositions, p, if p, then S knows that p and does not believe that ~p
In normal English, Oprop (for “propositional omniscience”) says that an omniscient being knows all truths and believes no falsehoods. Some proposed analyses of omniscience are more elaborate than Oprop, but in nearly all cases the objects of an omniscient being’s knowledge are said to be propositions.
I believe this is a bad approach to thinking about omniscience. A better approach, one that gets much closer to the intuitive (and Scriptural) idea that God has maximal knowledge, is to say that God knows everything as well as it can possibly be known. I’ll call this “full omniscience” (Ofull).
(Ofull) S is omniscient =def. S knows all of reality as well as it can possibly be known.
In what follows, I first (§1) step back from omniscience to the more basic concept of knowledge and begin with a brief history of attempts to define knowledge. I then (§2) offer three reasons for thinking that a strictly propositional approach to knowledge is inadequate. Next (§3), I discuss different strategies for defining a concept and argue that knowledge cannot be adequately defined in terms of necessary and sufficient conditions because it is a degreed perfection concept, one that points us toward the concept of full omniscience or maximal knowledge, i.e., the best conceivable instance or limit case of knowledge. In §4 I begin to unpack what maximal knowledge looks like by distinguishing between three tiers of knowledge (animal, casual human, and expert human). In §5 I then identify nine dimensions along which knowledge can vary in perfection. When the concept of knowledge is maximized along all nine dimensions, we arrive at a fairly good understanding of what full omniscience / maximal knowledge entails. Finally, in §6 I briefly explore some important implications of my analysis for thinking about God more generally.
1. What is knowledge? A brief history
Philosophers have typically sought to define the concept of knowledge (Greek, episteme) in terms of necessary and sufficient conditions. For most of over two thousand years the general consensus has followed Plato in thinking of knowledge as “justified true belief.” With some variations as to what counts as epistemic justification, medieval thinkers following Aristotle often supposed that knowledge properly speaking (Latin, scientia) required decisive justification, whether by direct rational apprehension or by demonstration via a connected series of such apprehensions. Early modern thinkers, writing in the wake of the nominalist revolution, the Copernican revolution, and the Protestant Reformation, were on the whole much more skeptical about the possibility of decisive justification. Even if, as rationalists like Descartes, Spinoza, and Leibniz supposed, certain rational principles could be known with infallible mathematical certainty, knowledge broadly speaking was increasingly regarded as inherently fallible and probabilistic, beyond the possibility of decisive justification.
Once knowledge became separated from decisive justification, the concept of knowledge as “justified true belief” became unstable, as there was no longer any way to ensure that the justification and truth conditions remain in sync. In 1963 a famous paper by Edmund Gettier exploited this and showed rather convincingly that it was possible to have justified true belief without having knowledge. This paper spurred philosophers into a three decade-long struggle to either identify a fourth condition in addition to justification and truth for defining knowledge or to reconceive and/or replace justification with other concepts like reliability and warrant. But philosophers couldn’t agree on a new set of necessary and sufficient conditions for knowledge and eventually sidelined that controversy to focus on more tractable topics like the nature of evidence, Bayesianism, social epistemology, and the like. Some of them even decried the whole project of defining knowledge and suggested taking it as an undefinable, basic concept.
I believe that much of this philosophical trajectory is mistaken in two main ways. First, as I argue in the next section (§2), knowledge isn’t exclusively or even fundamentally propositional and so its definition shouldn’t be centered on believing and justifying propositions. Second, as I argue in §3, knowledge is degreed perfection concept and so isn’t a suitable candidate for definition in terms of necessary and sufficient conditions.
2. Why propositional definitions of knowledge are inadequate
To be clear, I have no objection to the ideas that there is such a thing as propositional knowledge and that God, as omniscient, has all available knowledge of that sort. What I object to are the ideas that omniscience and knowledge generally should be defined as propositional. I object to this for three reasons.
First, there is much more to knowledge than propositional knowledge. Propositional knowledge, which Bertrand Russell called “knowledge by description,” is abstract. That is, it lies at least one step removed from concrete reality. In this respect, propositional knowledge contrasts with what Russell called “knowledge by acquaintance.” There is, for example, a clear difference between knowing the proposition <Tat the cat is on the mat> and having first-hand acquaintance with Tat and the concrete situation of Tat’s being on the mat. The proposition can be known, assuming it’s true, simply by overhearing a second-hand report from a reliable source, but the concrete reality that proposition represents can only be known by being there first-hand. Knowing a concrete individual like a person is not reducible to knowing true propositions about that person.
Second, by its very nature propositional knowledge is analytic. It breaks meaning down into discrete propositional chunks. It makes meaning digital, and thereby overlooks the analog, the holistic, the synthetic. But deep and practical knowledge of any subject matter is holistic and synthetic. It requires understanding connections between things, and not just abstract logical connections among propositions (i.e., more propositions), but the concrete real-world connections they represent. For example, knowing how to ride a bike involves much more than internalizing a list of “true bike-riding propositions.” It requires a performative ability to handle a bike in concrete circumstances. Contemplating a list of propositions while trying to ride a bike would only get in the way and make one more likely to fall over or crash.
Third, knowledge requires both a knower and a known, a knowing subject and known object. Propositional knowledge puts the emphasis almost entirely on the known, thereby sidelining the knower. This skews our understanding of knowledge. Thus, a propositionally-focused approach naturally lends itself to quantitative or extensional questions such as whether one knows this or that proposition or which propositions one knows. These are fine questions in themselves, but in many contexts, such as when a knowledge claim is contested, it’s at least as important to address the qualitative or intensional question of how well one knows, or is in a position to know, some claim. This requires that we keep both the knower and the known in focus. For example, first-hand testimony normally counts for more than second-hand testimony or “hearsay.” The first-hand closeness or proximity of the knower to the known is thus qualitatively relevant. If two people know the same proposition, but one knows it first-hand and the other only knows it second-hand by way of testimony from the first person, then there is a clear sense in which the first knower’s knowledge is better than the second knower’s knowledge.
3. Knowledge as a degreed perfection concept
There are several different ways one might try to define a concept. First, a lexical definition is what you get from a dictionary. A dictionary describes word usage. But we want to define a concept (knowledge), not a word. The two are related, of course, as words are typically used to express concepts, but words are also notoriously polysemous. That is, they often have multiple, distinct usages and therefore can express multiple, distinct concepts. Consequently, while consulting a dictionary is often a decent starting place for thinking about a concept, it’s rarely a good place to stop. By itself it can’t tell us which, if any, of the listed usages are what we need. Second, a stipulative definition is what we offer when we stipulate that “by term ‘X’ I mean” followed by a description of how one intends to use term ‘X’. This circumvents the polysemy problem, but at the expense of arbitrariness. Why should anyone else accept our stipulations? Third, an ostensive definition avoids arbitrariness by pointing at a clear example. This is how most basic words are introduced. Thus, to teach a child what the word “duck” means, you say the word while pointing at a duck or a clear picture of a duck. But not all concepts have clear examples that can be pointed at (e.g., God, infinity, nothing, etc.) and simply pointing at something doesn’t tell us explicitly what makes that thing a good example. Fourth, a conceptual definition helps us to understand a concept by reference to other concepts, ones we already understand. In standard form a conceptual definition specifies a genus (what kind of thing we’re interested in) and a specific difference (how it differs from other things of the same kind). For example, we can define a circle as a two-dimensional closed shape (its genus) the entire boundary of which is equidistant from a fixed point (its specific difference). A good conceptual definition articulates necessary and sufficient conditions for the application of a concept. Not all concepts can be defined in this way, however. This is especially true when the concept is vague, i.e., has a “fuzzy” boundary. In such cases we may be able to specify an inner boundary (sufficient but not necessary) and an outer boundary (necessary but not sufficient), but there remains a “grey area” between those boundaries where it’s not clear whether the concept applies or not. Most degreed or “spectrum” concepts fall into this problem category. For example, where does red begin and end on the color spectrum? There’s no precise answer. (Yes, we can stipulate precise boundaries for red, but then we’re no longer offering a conceptual definition but a stipulative one.) The same issue applies to most perfection concepts, for example, goodness. Goodness isn’t all or nothing; it comes in degrees. We have an outer (or lower) boundary—some things (e.g., murder) are clearly not good. But we don’t have an inner (or upper) boundary because there’s no point at which goodness becomes literally too good so as to be not good. Nor can we even clearly conceive of what the absolute limit case of goodness looks like. So, we can’t give necessary and sufficient conditions in a way that removes all vagueness.
To define a degreed perfection concept like goodness our only option is to combine aspects of ostensive and conceptual definitions and then project it toward “perfection” insofar as we can grasp it. First, we focus on the clearest, most paradigmatic examples of the concept we can find in our experience. That’s the ostensive component. But, however clear these examples may be, they may still be far from absolute perfection. This is the case when perfection lies beyond anything we can experience, such as with omniscience, or perfect / maximal knowledge. Second, we use our understanding of the cases we can experience to reflect on how even those cases could theoretically be improved. This helps us identify some of the concept’s necessary conditions. That’s the conceptual component. Necessary conditions define the concept’s outer (or lower) boundary. These are baseline features that never vary or drop out as we consider how our paradigmatic examples might be improved. Third, the variant features, those respects in which even our best examples of the concept can be improved define a “vector” that points us in the general direction of absolute perfection. The maximal or perfect degree of the concept is thus understood by analogy with the paradigmatic cases we started from. They provide an experiential anchor that keeps the concept grounded even as it points us toward a maximum that lies beyond us.
This is how I want to approach the concept of knowledge. But first I need to address some critical questions.
First, why should we think that knowledge is a degreed concept? Because it rather obviously is. All of the concepts chiefly used to define knowledge (justification, truth, reliability, belief, warrant, probability, credence, relevance, salience, grounds, evidence, etc.) as well as many closely related concepts (understanding, wisdom, know-how, etc.) all admit of degrees in one way or another. Thus, one can have more or less justification, warrant, grounds, or evidence for a claim. The evidence that one has for a claim can be more or less relevant or salient to it. It can have a greater or lesser effect on the probability of the claim. The processes by which one forms a belief or acquires evidence can be more or less reliable. One can have a stronger or weaker belief in the claim, or place more or less credence in it. A claim can be more or less true, that is, it can correspond more or less closely with reality. As for related concepts, one can have more or less understanding of a matter, have more or less wisdom, more or less know-how, and, finally, more or less knowledge. In short, if all of the concepts surrounding knowledge admit of degrees, it’s really hard to see how knowledge could fail to admit of degrees, and we already speak as if knowledge comes in degrees anyway. That knowledge is a degreed concept should not, therefore, be controversial.
Second, why should we think that knowledge is a perfection concept? Because it’s obvious that goodness is a perfection concept and that having knowledge of what is good is categorically better than lacking knowledge of what is good. Sure, we sometimes say “ignorance is bliss” because we think there are things that are not good that we’d be better off not knowing about. And, I’ll concede that if reality were fundamentally bad or neutral, perhaps knowledge wouldn’t be a perfection. But, as a Christian theist committed to the existence of a perfectly good and loving God as the singular foundation of reality, that knowledge is a perfection concept like goodness concept seems undeniable to me. For a perfect being like God, even bad things are worth knowing because (a) they can’t overwhelm Him, (b) nothing bad is all bad but is rather a corruption of something otherwise good, and (c) knowing bad things is a necessary first step toward turning them back toward the good.
Third, if knowledge is degreed perfection concept, then why shouldn’t we just identify knowledge with perfect knowledge and say that everything short of that just isn’t knowledge at all? Because this is a self-defeating skeptical proposal that, if adopted, would leave us no way to reason analogically from lower-level instances of knowledge (because there would be no such instances) and thereby make it impossible for us to develop any clear idea of perfect knowledge. Furthermore, we already establish that knowledge is a degreed concept. We can’t affirm that and also equate knowledge with only the highest degree. That would be inconsistent. Finally, a noteworthy feature of degreed concepts is that their applicability is normally a matter of pragmatic judgment. We say, for example, that this surface is “flat” knowing full well that if we were to zoom in closely with a powerful microscope that the surface is anything but flat at the micro level. We can, of course, conceive of a surface being strictly or perfectly flat, but if we were to restrict our usage of the word “flat” to refer only things that are perfectly flat, then the term would become nearly useless. When we say that a surface, like this table or Nebraska, is “flat” we mean that it is flat enough for the practical purposes we currently care about. Likewise, we can say that a certain proposition (e.g., “pi = 3.14”) is “true” even when we know that it is strictly false provided it is true enough in our current practical context. One who protested and said “Oh, but that’s not really flat, not really true, etc.” as though pointing out something we already know—namely, that in most contexts degreed concepts don’t strictly apply—would be an overly pedantic and socially inept jerk. Likewise with knowledge. If knowledge is a degreed concept, then the question of whether someone S knows (by description) proposition p or knows (by acquaintance) entity e, becomes a matter for pragmatic judgment. Does S know p or e well enough in the relevant context? The exception, of course, is the ideal or perfect case. Like a mathematical plane stipulated to be perfectly flat, if there is a coherent conception of perfect knowledge, then any instance of that sort of knowledge is knowledge strictly so-called, knowledge full stop. But this is no more reason to refuse attributing “knowledge” to less-than-perfect instances than my desk’s not being perfectly flat is a reason to deny that it is, in a relevantly approximate sense, flat.
4. Knowledge: animal, human, and divine
To the extent I can, I’m going to try to unpack the concept of perfect or maximal knowledge by considering clear-cut cases of animal- and human-level knowledge, noting what they reveal as necessary or invariant about knowledge and what they reveal as variable dimensions along which knowledge can theoretically be improved. The result will be an analogical conception of omniscience as perfect / maximal knowledge that gives us insight into the “as well as it can possibly be known” clause of Ofull.
Consider the following three levels or tiers of knowledge:
Tier 1: Animal knowledge. This kind of knowledge is unreflective and often instinctual. My dog Misa knows that when I say to her “Where’s your toy?” that I’m talking about one of her squeaky toys, and she knows that if she can’t find a toy in her immediate vicinity that she can get one from the doggie toy box downstairs. Likewise, many animals know instinctively what sort of food they need; when it’s time to mate, migrate, hunt, or hibernate; how to build a proper nest; and how to navigate long distances back to their breeding grounds. As far as we can tell, animals don’t encode knowledge propositionally and they don’t reflectively consider whether or how they know such things. But that they do know such things is demonstrated by their ability repeatedly to navigate the world successfully to obtain their goals.
Tier 2: Casual human knowledge. Most of our everyday knowledge is of this sort. We know who the President is; how to ride a bike; that we’ve got to do X, Y, and Z by next Thursday; how to state and apply the Pythagorean theorem; etc. This kind of knowledge is partly reflective and often at least partially propositionally articulated. It is acquired experientially and often linguistically through explicit instruction, reading, or conversation. It’s partly reflective because we sometimes wonder whether and how we know what we think we know, and we may sometimes consciously try to consider things from different perspectives.
Tier 3: Expert human knowledge. This kind of knowledge is highly reflective, highly articulate, and accompanied by deep understanding of some domain. The best way to understand this is by contrast with casual human knowledge. A causal math student, for example, can be said to know the Central Limit Theorem after having just read about it in an introductory statistics textbook, but the mathematician who wrote the textbook knows the theorem in a qualitatively deeper sense. He not only knows how to state the theorem with precision, but can rigorously prove it in half a dozen independent ways and apply it correctly and creatively across a diverse range of contexts. The casual math student can’t do any of that.
My first observation in reflecting on these tiers is that all of them qualify as knowledge. Knowledge at the animal and causal human level isn’t a difficult achievement. It’s rather commonplace. All it requires at a minimum is an ability to recognize patterns in experience as meaningful and to navigate the world successfully (for the most part) on that basis. It doesn’t require self-conscious reflection, deep understanding, or propositional articulation. Most of us, for example, know how to ride a bike, but how many of us can describe with any precision how to do so?
My second observation is that as we move up the tiers, more is required. We arrive at qualitatively better grades of knowledge. All other things equal, knowledge systems that are articulate, self-reflective, and rooted in deep understanding can withstand more scrutiny and track reality more closely over time. There is a clear trajectory of improvement as we move from animal knowledge to casual human knowledge to expert human knowledge.
My third observation is that expert human level knowledge is nowhere near the maximum. If we continue the same trajectory of improvement from tiers 1–3 we can at least begin to catch a glimpse of what perfect, maximal knowledge is like. Everything points toward a fourth tier of knowledge, divine knowledge, that preserves all of the core strengths of tiers 1–3 while surmounting their limitations. To make matters more precise, I will now argue that there are at least nine respects in which knowledge can be improved as we move from animal knowledge to divine knowledge.
5. Nine aspects of maximal knowledge
From our own experience, let’s think about ways in which knowledge can be improved or enhanced. Just as the expert human knows things in his area of expertise better than a casual human knower does
- Maximal scope – God’s knowledge extends to all of reality.
- Maximal understanding – God fully understands how everything is connected to everything else.
- Maximal accuracy – God knows reality precisely as it is.
- Maximal immediacy – God knows all of reality first-hand, without need for any mediation (e.g., photons, sound waves, testimony, etc.).
- Maximally reflexive – For everything God knows, God knows that He knows it.
- Maximal security – God’s knowledge is infallible.
- Maximal stability – God’s knowledge can’t be inadvertently lost (e.g., by forgetting).
- Maximal credence – God’s subjective confidence in His own knowledge is always 1.
- Maximally articulate – God has no background or “tacit” knowledge. God’s knowledge is fully explicit in God’s mind.