BonJour offers two main reasons for supposing that there is such a thing as rational insight into necessity. First, he says there are many examples in which it clearly seems that one has such insight. Second, he argues that any epistemology denying the existence of rational insight into necessity is committed to a narrow skepticism. After commenting about possible frameworks for epistemological justification, I argue against these two claims in reverse order.
Let us say that a belief or inferential procedure is "foundational" for a given person if and only if the person is prima facie justified in so believing or inferring in the absence of any appeal to further beliefs or procedures of inference. Then we can distinguish three broad kinds of theory as to which, if any, beliefs and inferential procedures are foundational.
A coherence theory holds that no beliefs or inferential procedures are foundational. A special foundations theory holds that only certain specified beliefs and inferential procedures are foundational. A general foundations theory holds that all of one's beliefs and inferential procedures at a given time are foundational at that time.
In his book, BonJour seeks to resist "the prevailing skepticism concerning a priori justification" (6). But his argument misfires through assuming a special foundations framework, while the "prevailing skepticism" he refers to arises within the general foundations framework, which I will now briefly describe.
Where special foundationalism says that justification must start from a small number of special foundational beliefs and inferential procedures, general foundationalism says that justification starts from wherever one is at the moment, that is, it starts from all of one's current beliefs and inferential practices.
General foundationalism is the epistemology suggested by Neurath's image: one's beliefs and inferential procedures are
like the planks making up a ship at sea; one must repair the ship as one goes.
It is the epistemology behind Rawls' idea
(borrowed from Goodman) that one seeks
"reflective equilibrium." One starts with one's current beliefs and
inferential practices, using some parts to criticize others, with an ideal goal
of arriving at a result in which all parts of one's view are in equilibrium
with each other. This is, of course, a very familiar idea, accepted by many
contemporary philosophers, even though it is rarely discussed by
It may be useful if I briefly mention standard replies to standard objections to general foundationalism.
Objection: "In this view, as soon as one randomly comes to believe P, one is automatically justified in believing P." Reply: If one believes not only P but also that one randomly came to believe P, the two beliefs are in tension and one has a reason to abandon at least one of them.
Objection: "General foundationalism cannot account for the special
epistemological relevance of perception." Reply: Perception causes beliefs,
which are therefore prima facie justified. Furthermore, if one's
inferential practices give special weight to certain perceptual beliefs, those
practices are prima facie justified.
Objection: "The sort of justification envisioned by general foundationalism is not epistemic justification with its special connection with truth." Reply: It is too strong to claim that epistemically justified beliefs have to be true; all that can be said is that they are likely to be true. But any theory of epistemic justification offers an account of the relevant sort of epistemic likelihood. General foundationalism, like special foundationalism, is a framework for an account of epistemic likelihood.
Objection: "General foundationalism is just another name for a coherence theory of justification, with all of its problems." Reply: They are opposite views. The coherence theory says that every belief and inferential practice requires a justification in terms of other beliefs and inferential practices; general foundationalism says that no belief requires such a justification in the absence of a specific well-motivated objection to the belief.
Objection: "General foundationalism is just Quine's view that we should stop talking about the justification of belief and restrict our attention to how beliefs are caused." Reply: General foundationalism agrees with Quine that we should give up trying to defend a special foundationalist approach to justification. But it offers a competing account of epistemic justification (and one I think it is clear Quine accepts, apart from terminology).
Special foundations theories tend to imply scepticism. Scepticism follows unless most of the various sorts of beliefs we normally take to be justified can be justified from special foundations using only foundationally acceptable modes of inference. So, special foundationalism faces the difficult problem of saying how anyone might be justified on that basis in believing in other people, in the external world, or in the conclusions of inductive reasoning.
These problems do not arise in any straightforward way within the framework of general foundationalism. Since one believes in other people and the external world and since one makes inductive inferences, these beliefs and inferences are prima facie justified. Conflicts may arise, in which case one may have to make changes, but one starts where one is.
My goal is to consider how BonJour's defense of pure reason and a priori justification looks to someone who starts out accepting general foundationalism. Here is BonJour's initial definition of "a priori justification"
I propose to count a proposition P as being justified a priori (for a particular person, at a particular time) if and only if that person has a reason for thinking P to be true that does not depend on any positive appeal to experience or other causally mediated quasi-perceptual contact with contingent features of the world, but only on pure thought or reason, even if the person's ability to understand P in question derives, in whole or in part, from experience. (11)
If "having a
reason for thinking" applies even to foundational beliefs and means something
like "is justified in thinking," then this definition could be interpreted to
imply that all of one's beliefs are justified a priori, according to
general foundationalism. Nevertheless, that is not the sort of a priori
justification with which BonJour is concerned. He is concerned with something
that is needed only in the special foundations framework.
Any theory within that framework has to say what distinguishes the special foundations from other beliefs. The foundations are not all of one's initial beliefs, as in general foundationalism. Something has to set them apart as special. Typically, it is suggested that certain perceptual beliefs are special in this way. But such beliefs are not by themselves enough to justify most of the things we believe. At the very least, some sorts of inferential practices have to be specially foundational, and they will have the relevant sort of special a priori status.
In particular, BonJour argues that what he calls "rational insight" can be a source of a priori justified belief. Such insight involves grasping or apprehending the necessity of a proposition (102). If we suppose that a priori justification requires apprehension of necessity, then it will not turn out that all beliefs count as a priori justified according to general foundationalism.
BonJour argues that a non-skeptical epistemology requires accepting rational insight as a genuine and autonomous source of epistemic justification and knowledge. That conclusion is obviously too strong. All that is required is that enough of one's beliefs be foundationally justified. It is not necessary that any beliefs be justified via rational insight, if rational insight involves grasping the necessity of a proposition. Someone who never grasps the necessity of anything can be as justified in his or her ordinary beliefs as someone else who (thinks he or she) grasps the necessity of certain beliefs or patterns of inference.
Although the notion of "rational insight" plays no role in general foundationalism, it does have a use in special foundationalism, namely, to help to specify part of the special foundations. In BonJour's version, any belief that arises from apparent rational insight counts as foundational.
BonJour offers as "intuitive examples" of rational insight coming to see (1) that nothing can be red all over and green all over at the same time, (2) that, if A is taller than B and B is taller than C, then A is taller than C, (3) that there are no round squares, (4) that 2 + 3 = 5, and (5) certain obvious logical implications.
But reflection on these examples reveals that there is no reason to suppose
that rational insight in BonJour's sense is involved in any of them.
How could one come to think that these claims are necessary? Presumably, by using one's imagination. For example, one tries to imagine something that is both red and green all over and finds that one cannot. From this concludes that red and green are incompatible. Or one imagines three people, A, B, and C, as described and sees in one's mind's eye that A is taller than C.
But all that can be concluded is that one has a limited imagination.
On reflection, one must allow that red-green incompatibility may be an illusion. Suppose an object looks red from one angle and green from a slightly different angle. In that case, might we not say that the object is red all over and also green all over, even though one cannot see both colors at the same time? In imagining cases, one forgot about that possibility; one tried to imagine something that looked both red all over and green all over at the same time. But something could be both red all over and green all over at the same time without looking both red all over and green all over at the same time.
As for whether something can look both red all over and green all over at
the same time, the fact that one hasn't yet been able to imagine something's
looking that way is clearly not decisive. Consider something that looks red to
one's right eye and at the same time looks green to one's left eye. How will it
look overall? Does rational insight alone reveal that the experienced object
will not look to be red all over and green all over? Crane and Piantanida
report somewhat a somewhat different case in which some subjects do report that
objects seem to be both red and green at the same place at the same time.
The source of apparent red green incompatibility probably has to do with the way color perception works. There are no ordinary color "chords" in the way there are musical chords. It would seem to be a contingent matter that this is so. Belief in red green incompatibility does not seem to be due to rational insight in BonJour's sense.
Similarly, one cannot tell simply through imagination that taller than is transitive, because one cannot know just by using one's imagination that one is imagining all relevant cases. Compare using one's imagination to assess this: "If A is to the right of B and B is to the right of C, then A is to the right of C." When one imagines the case, this claim may seem necessarily true, but that is only because (for example) one ignores the possibility that the three are spaced evenly around a circular table.
"If A is to the right of B, then B is to the left of
A." Again, that may seem necessarily true, until one considers the
possibility that A and B might be facing in opposite directions,
so that A is to B's right and B is to A's
"If A is taller than B and B is taller than C, then A is taller than C." Imagination may suggest that is true, but how can one know one isn't overlooking a possibility?
Again one cannot tell just by using one's imagination that it is impossible for there to be round squares. Suppose space is not Euclidean. Then maybe there are no small squares that are round, but what happens when the squares are large. Perhaps a right angle and a straight angle coincide under certain conditions? How can one know this is not so?
"2 + 3 = 5." Isn't this something that one memorized as part of learning addition? How does one use "rational insight" to assess it?
Reasoning about deductive implications. Example: "Either David ate the last piece or else Jennifer ate it. Jennifer did not eat it. So David ate it." There are two leading psychological theories about how people recognize such implications: the mental rule theory, as defended for example by Rips, and the mental models approach, as defended by Johnson-Laird and Byrne. Are these theories about how rational insight works? The mental models approach involves imagining cases and is subject to the worry that one has not considered all possible cases, since there is no way to tell just through imagination that one has indeed considered all such cases.
One reason why so many philosophers are doubtful about a priori knowledge is the seemingly clear cases are not clear after all. Many supposed paradigm cases of a priori knowledge may actually be false, examples like "All unmarried adult males are bachelors,"  "All women are females," or "Red and green are incompatible colors."
In this connection, it is interesting to consider "All cats are animals." Putnam argues that we can imagine discovering that cats are not actually animals but are clever robots controlled by space aliens. On the other hand, if cats are animals, perhaps it is necessary that cats are animals. "All cats are animals" seems to be an example of what Kripke counts as an a posteriori necessary truth. BonJour takes rational insight to involve an apprehension of necessity. So, might someone have rational insight into the truth of "All cats are animals" even though we can imagine discovering that it isn't true? I would guess that the relevant grasp of necessity must include grasping that the proposition in question cannot be discovered to be false and perhaps on that basis grasping that it is necessarily true.
In the end BonJour offers two arguments for the existence of rational insight into necessity. On the one hand, he claims that there are many examples where we seem to have such rational insight. On the other hand, he argues that denying such rational insight commits an epistemology to widespread skepticism.
Neither of these arguments is persuasive. The examples don't work and a general foundations approach in epistemology does not need to appeal to rational insight in order to avoid skepticism.
 Laurence BonJour, In Defense of Pure
Reason (Cambridge: Cambridge University Press, 1998).
 Alvin Goldman calls this approach "negative coherentism," Knowledge in a Social World (Oxford: Clarendon Press, 1999),p. 128. I called it "general conservatism," in "Rationality," Smith, E. E., and Osherson, D. N., eds., Thinking: Invitation to Cognitive Science, Volume 3, Cambridge, Massachusetts: MIT Press (1995), pp. 175-211, revised version in Harman, G., Reasoning, Meaning, and Mind (Oxford: Clarendon Press, 1999), pp. 9-45.
 Otto Neurath, "Protocol Sentences," in A. J. Ayer, Logical Positivism (NY: Free Press, 1959), p. 201.
 John Rawls, A Theory of Justice (Cambridge, MA: Harvard University Press, 1971), p. 20.
 Nelson Goodman, Fact, Fiction, and Forecast (Cambridge, MA: Harvard University Press, 1955), pp. 65-68.
 Hewitt D. Crane and Thomas P. Piantanida, "On Seeing Reddish Green and Yellowish Blue," Science 221 (1983) pp. 1078-9. (I am indebted to David Lewis for this reference.)
 Rips, L. J. The Psychology of Proof: Deductive Reasoning in Human Thinking (Cambridge, MA: MIT Press, 1994). Johnson-Laird, P. N., and R. M. J. Byrne, Deduction (Hillsdale, NJ: Erlbaum, 1991).
 The Pope is not a bachelor, nor is a man who lives with a woman for a long without getting married.
 Robert Schwartz pointed out to me many years ago a case in which a woman was disqualified from an Olympic event because she had the wrong chromosomes to count as female.
 Saul Kripke, Naming and Necessity [fix reference].