Lecture Notes, UC Davis Philosophy 102, Theory of Knowledge

A Priori Knowledge

G. J. Mattey, Senior Lecturer

Philip Kitcher

Text: "A Priori Knowledge" (1980)

Since the beginning of Western philosophy, there has been a tendency in theory of knowledge to regard reason as superior to sense-perception as a means to knowledge. Some philosophers have gone so far as to claim that reason is the only human faculty whose operations can yield knowledge. We have seen that this was the view of Plato. At best, sense-perception can produce true opinion. Later philosophers would allow that beliefs that are based on sense-perception are probable but are not secure enough to be counted as knowledge.

Traditional Views

Traditionally, knowledge that is independent of sense-perception has been called a priori knowledge. Plato thought that mathematical knowledge is a priori, and many philosophers would add that knowledge of the truths of logic is a priori as well. What logic and mathematics have in common is that their objects are not accessible through sense-perception but only by reason alone. For Plato, it is through "recollection."

Descartes claimed that we have a priori knowledge not by recollection, but because we possess "innate ideas" that are implanted in our minds by God when we begin to exist. He resorted to metaphors such as the activity of "the light of nature" to explain how a priori knowledge arises.

For Descartes, the paradigm of a priori knowledge is knowledge of mathematical truths. One feature of mathematical truths is that they are necessary: there is no possibility of the falsehood of a true mathematical proposition. Descartes explained this necessity by reference to God: it is God's eternal will that these propositions be true. The same holds for the truths of logic.

Another item of a priori knowledge for Kant is the existence of God. In the fifth Meditation, Descartes tried to show that if one merely understands the meaning of the idea of God properly, one knows that God exists. Given that a priori knowledge is of necessary truths, the existence of God is a necessary truth as well. Further, this suggests another way to describe propositions that might be known a priori, namely those that are true by virtue of meaning, such as that a triangle has three sides.

Some of the more interesting debates regarding a priori knowledge are these. Is there a real distinction between a priori knowledge and knowledge through sense-perception ("a posteriori" or "empirical" knowledge)? Are human beings in possession of any such knowledge? Are all propositions which are known a priori (a priori truths) true by virtue of meaning? Are they all necessarily true? Is their any a priori knowledge of objects of the senses?

For some time, philosophers influenced by W. V. Quine had largely abandoned any claims to a priori knowledge. But interest was revived in 1970 by Saul Kripke, who argued that we have a priori knowledge of contingent (non-necessary) truths. (See Naming and Necessity. In the last ten years or so, Laurence BonJour (University of Washington) and George Bealer (University of Colorado) have defended claims of a priori knowledge.

In this class, we will be looking at a piece by Philip Kitcher (Columbia University) which tries to break the traditional mold in thinking about a priori knowledge. He seeks to clarify the definition of a priori knowledge given by Kant. More importantly, he tries to show that a priori knowledge can be understood from a standpoint according to which knowledge is a product of human beings in the world of physical nature. In this way, he rejects the view of Plato and Descartes that a priori knowledge is restricted to a soul that operates separately from the body.

Independence from Experience

Kant had defined a priori knowledge as that knowledge which is independent of all experience. According to this definition, knowledge obtained from recollection or innate ideas would qualify as a priori. But Kitcher points out that Kant thought that a priori knowledge is not obtained in this way. He gave experience some role.

What is experience? Kitcher suggests, following philosophers such as John Locke, that it can be divided into to sorts. "Outer experience" is the result of stimulus by external objects, while inner experience is the result of inner stimulus. If a priori knowledge is to be independent of all experience, then we might say that there is no particular stimulus, inner or outer, that is required for a particular item such knowledge.

But it may be that experience does play a role in a priori knowledge. For example, consider the claims that some propositions ("analytic truths") are true by virtue of their meaning. It may be that experience is required in order for one to obtain the concept which is analyzed in the proposition. This leads to a modification of Kant's definition. Knowledge is a priori when the only experience required to obtain it is that which is needed to obtain the concepts involved in the known proposition.

The next point Kitcher makes is to emphasize that a priori knowledge is independent of all of one's experience. If one engages in reasoning to draw a conclusion from information that was drawn from experience, one's resulting knowledge is not a priori. Even though the reasoning process itself may be independent of experience, it relies on premises that are derived from experience. With a different sequence of experience, the reasoner might lack the required premises. (See page 528 in the text for Kant's own statement.)

First Analysis of A Priori Knowledge

Having given an informal characterization of a priori knowledge, Kitcher puts it together in the form of an analysis. An analysis is a set of necessary and sufficient conditions for something to be of a certain kind. Here, the kind in question is a priori knowledge. So he will tell us what conditions must hold if someone X has a priori knowledge of some information p, and which conditions are such that if they hold, X has a priori knowledge of p.

The analysis is based on the notion of a person's experience at a time, a defintion of a person's "life" at that time, and a definition of a life's being "sufficient for" p. The two definitions are "stipulative," in that their author (Kitcher) is laying down or stipulating how it is that he will use "life" and "sufficient for" in the context of the analysis.

A person X's experience at time t consists in the sensory state of that person at that time. This state may either be "outer" or "inner" states, depending on the location of the receptors whose stimulation give rise to the sensory state. The state itself is a physiological state, or state of the body.

X's life at t is the sum of X's experiences. So it will be a collection of physiological sensory states. This is not how we ordinarily use the term "life," but as mentioned above, this is how Kitcher is using the term in this context.

The definition of X's life's being "sufficient for" p is more complicated. The idea is that some lives (perhaps lives that are only possible) contain experiences which allow X to form the concepts needed for X to believe that p. As was mentioned earlier, a person lacks a priori knowledge if that person lacks the concepts contained in the known proposition. If a life is one that X could have had, and if in that life X gains sufficient understanding to believe that p, then that life is sufficient for p.

Now we have the tools required for the analysis of a priori knowledge:

X knows a priori that p at t if and only if X knows that p at t and, given any life sufficient for p, X could have had that life at t and still known at t that p.
Suppose I know that q: the sum of the angles of a triangle is equal to two right angles. Do I know this a priori? According to the analysis, we have to look at any sequence of experiences which are enough to allow me to believe that q and ask whether it is possible for me both to have had that sequence of experience and at the same time know that q. The idea is that no sequence of experiences that give me enough conceptual material to understand q should block my knowledge that q. It does seem to be the case that any sensory stimulation I might encounter would never interfere with my knowledge that q.

The analysis seems to be strong enough that any cases of a priori knowledge will conform to its conditions. But Kitcher objects that it is too strong, i.e., that it counts as a priori knowledge cases that should not be so counted. Suppose that I learned that q on the basis of having been taught that it was the case. I never understood the proof, however, and just took its truth on the authority of my teacher, who is an expert in mathematics, whom I had every reason to trust, and who in fact was telling the truth on the basis of what she knew. Then I might be said to have knowledge that q. But by the analysis, I have a priori knowledge, because no experience sufficient for q is incompatible with my knowing that q. To weaken the analysis, we need to say something about how it is that X knows that p in the first place.

The Psychologistic Account of Knowledge

To describe how a person knows something, we should have in hand some notion of what it is to know. Kitcher adopts what he calles "the psychologistic account of knowledge." He allows that knowledge requires true belief, in addition to an appropriate explanation of the existence of this belief in terms of some process, which he calls the "warrant" for the belief. He does not give an account of what makes an explanation appropriate, because he does not have to do this for his present purposes.

The psychologistic account of knowledge is quite controversial. Most theories of knowledge do not require an explanation for the existence of the belief. They would hold that it does not matter where a belief originates. All that matters is whether or not the belief is "justified." Justification is usually understood in terms of having good reasons to think the belief is true.

Kitcher thinks that Kant's account of a priori knowledge is psychologistic. A Kitcher interprets Kant as saying that the psychological process involved in a priori knowledge of the elementary properties of a triangle is obtained by "using our grasp on the concept of triangle to construct a mental picture of a triangle and by inspecting this mental picture with our mind's eye." This explanation of our beliefs about these properties is psychological, as it involves nothing more than a description of some psychological processes.

Contrast this account with another one, in which a person measures, as best he can, the dimensions of roughly triangular figures he finds in the sensible world. Any knowledge one gains from this would be empirical, not a priori, because the process involves sensory states which rely on the external world for their input.

We can say that a life which allows the process of "pure intuition" that operates with mental pictures might be sufficient for beliefs about the elementary properties of triangles. Thus being sufficient is in this case independent of experience. The belief is warranted independently of experience if its generated in the appropriate way. So if the use of pure intuition is appropriate, belief is warranted independent of experience. Finally, if the warranted belief is true in other situations where experiences are different ("counterfactual situations"), then one has knowledge independent of experience--a priori knowledge.

Final Analysis of A Priori Knowledge

Putting this all together, we have the following analysis of a priori knowledge.

  1. X knows a priori that p if and only if X knows that p and X's belief that p was produced by a process which is an a priori warrant for it
  2. α is an a priori warrant for X's belief that p if and only if α is a process such that, given any life e, sufficient for X for p, then
    1. some process of the same type could produce in X a belief that p
    2. if a process of the same type were to produce in X a belief that p then it would warrant X in believing that p
    3. if a process of the same type were to produce in X a belief that p then p
This analysis is supposed to block cases of empirical knowledge from being counted as a priori. Kitcher does not tell us in any detail how the cases are blocked. He only says that the analysis implies that in all cases of a priori knowledge, "the knowledge be obtainable in the same way."

Take the case of the student being told by her teacher that the sum of the angles of a triangle is equal to two right angles. The idea is that in any life the student could know its truth using, say, pure intuition, but in some lives the student could not know its truth by the process of being in a sensory state and interpreting it as hearing a sentence uttered by her teacher. Perhaps this sensory state is wholly misleading, in which case she would lack warrant for her belief and hence lack knowledge.

Types of Processes

You might wonder why I described the process in terms of sensory states and their interpretation. The reason is that Kitcher wants in general to confine these processes "to those segments which consist solely of states and events internal to the believer." The reason he wishes to do this is that he wants to allow for a priori knowledge, which in turn requires him to allow for purely psychological states to confer warrant.

Kitcher does not explain exactly how it is that psychological processes should be divided into types. Instead, he leaves that to anyone who is trying to develop a substantive account of a priori knowledge. He further clarifies the meaning of "could" in the analysis, so that it supposes that X has the standard mental capacities human beings actually have. Otherwise, the conditions might be satified by a life in which X has enhanced mental powers.

A Priori Knowledge of Psychological States

As was noted above, the standard accounts of a priori knowledge take it to be about mathematical truths, logical truths, and the like. Kitcher's account allows knowledge of our own psychological states: a priori knowledge may be self-knowledge.

Because not all of one's own psychological states can be known a priori, there is less of a problem than it appears at first. If one can be mistaken about one's psychological state, the truth condition is violated. If the process requires certain experiences before it can exist, then in those lives lacking them, the belief cannot be produced. Thus, if the belief is that I feel pain, if I can have the belief without feeling pain, or if I must have the pain in order to have the belief, I lack a priori knowledge that I feel pain.

It might be thought that when we have the power to produce psychological states at will, and then exercise that power, we have a priori knowledge that we are in those states. So, if I will myself to imagine a red patch and by so doing produce a red patch in my imagination, perhaps I have a priori knowledge that I am imagining a red patch. Kitcher notes that we may be able to undermine this result by allowing that in some cases, I have enough neuro-physiological knowledge of myself that I find that I am unreliable in detecting my psychological states, so my belief will not be warranted.

But maybe we will have to acknowledge that we have a priori knowledge in cases such as that of imagining a red patch. This is not so bad, since there are clear cases (Kitcher contends) of a priori knowledge of one's own psychological states. One such case is my knowledge that I exist, when I am awake. The process that produces this may not be a "non-optical inner look," but there seem to be such processes. Perhaps the process of rational thought would work. It might also be the case that no experience is inconsistent with this belief.

A Priori Knowledge of Contingent Truths

One "doctrinaire" objection is that we can have a priori knowledge only of propositions that are necessarily true. This is a claim that Kant made, for example, on page 529 of our text. This claim rests on the fact that a priori knowledge is "ultra-reliable," as captured in the last condition: "if a process of the same type were to produce in X a belief that p then p." This codifies the idea that X is entitled to ignore all input of experience when X knows a priori.

Kitcher thinks this objection rests on a fallacious line of reasoning.

  1. Suppose X knows a priori that p, but p is not necessarily true.
  2. So, there are worlds w at which p is false.
  3. Suppose X inhabits a world w at which p is false.
  4. So, X would have an a priori warrant for a p that is false.
  5. But one can only have an a priori warrant for true beliefs.
  6. So, if X knows a priori that p, then p is necessarily true.
The invalid inference here is at step 4. The fact that X inhabits a world where X is false and knows a priori that p does not imply that X would have an a priori warrant at w for the truth of p. For the warrant at the actual world to "carry over" to w, the belief there would have to be produced by the same process at both worlds. But p might be a proposition such as that I exist, which if false could not be believed. So it may be that the worlds where p is false are worlds that are not relevant to a priori knowledge.

Degrees of A Priori Knowledge

It may be that the present analysis of a priori knowledge is too weak, in that it counts what Kitcher calls "universally empirical" knowledge. This would be knowledge that independent of any particular experience because it is about some feature shared by all experience. An example might be that objects have shapes.

One might oppose this kind of case on the grounds that, although this knowledge is "ultra-reliable" and cannot be refuted by experience, it still relies on experience to be generated. If so, then one could distinguish between Kitcher's analysis as capturing a weaker notion of the a priori. A stronger analysis could add the clause that the process producing a priori warrant may not involve any perceptual operations.

I may then be thought that adding such a condition would make the original analysis redundant. Why not just say that a priori knowledge is just knowledge produced by a process which does not involve any perceptual operations? Kitcher responds that such a process may fail to have the appropriate degree of reliability. That is, the belief may not hold up if one's experience is sufficiently unreliable. He does not give an example of how this might work.

The opposite objection is that the analysis is too strong. There may be almost nothing which meets the conditions Kitcher lays down. Maybe in some lives our experience is so bizarre as to be able to block our knowledge of practially anything in them. Kitcher's response is that he doubts this is the case, but he can accommodate it using the apparatus of his analysis. We can approximate a priori knowledge to a greater or lesser degree by restricting the range of experiences that are relevant. In this way, we could say that knowledge that holds up in all but the most far-out lives approximates very closely to a priori knowledge.

Non-Psychologistic Accounts

In the twentieth century, most analytic philosophers who have tried to develop accounts of a priori knowledge have rejected the psychologistic approach. A crude attempt is to hold that X has a priori knowledge that p just in case X believes that p and p is true by virtue of meaning. But X could believe p for any reason at all, and not because it is "analytic," so this account seems too weak. If I came to believe p because I was taught that p, and later proved that p, my knowledge in both cases would have to be counted as a priori. Kitcher suspects that philosophers have used the expression "a priori" to single out knowledge that they think is in some way privileged. His account gives a systematic reason for favoring one kind of knowledge over another: its independence from experience.

[ Course Home Page | Menu of Lectures ]