cannot be revised. (The given entries are simply there, in sudoku, and it's built into the game that there's no room to doubt them.) Does the foundationalist really think that the foundation for all our thinking about the world rests on foundations that we have no room to doubt? Thinking that such beliefs are in short supply, critics of at least one version of strong foundationalism could argue as follows:
Argument from Defeasibility
P1. If there are non‐inferentially justified beliefs that are like the numbers given at the start of the puzzle, then you cannot form correct beliefs about the world by reasoning well from your justified beliefs in such a way that you're led to suspend judgment about whether one of these properly basic beliefs is true, and you cannot form the correct beliefs about the world by reasoning from your justified beliefs to the conclusion that one of these beliefs is false.
P2. However, none of our beliefs are immune to this sort of process of rational revision.
C. Thus, it isn't true that there are non‐inferentially justified beliefs that are like the numbers given at the start of the puzzle.
1.76 To understand the Argument from Defeasibility, we need to understand what it means to say that justification is defeasible. In short, “defeasible” literally means “able to be defeated,” or diminished. The kind of thing that can potentially defeat the justification you have for believing something is new information you might acquire. More specifically, this happens when the rational force of whatever evidence served to justify you holding a belief is undermined and overridden by you learning new relevant information.36 The terms “undermining defeater” and “overriding defeater” are used to pick out two different ways that you can “lose” the justification you have for believing something.
1.77 To understand the difference between undermining and overriding defeat, think about promises. If you promise to meet a friend, that's a good reason to meet them. If you have no reason to do anything else, that reason might be decisive in that it might determine what you should do, all things considered. If, however, you make a promise to meet a friend and you then encounter a child who needs your help, the reason you have to meet a friend could be overridden by a weightier reason, to render aid. Alternatively, if your friend calls the meeting off, the reason you had could lose its rational force because the ground of the reason has been cut away. In this case, the reason is undermined.37
1.78 Something similar happens with belief. The testimony of a friend might give you a good reason to believe that, say, they plan on staying around to have coffee with you. Seeing them slipping out a back entrance to the parking lot might override that, as it's strong evidence that they're sneaking away for some reason. In cases of overriding defeat, the justification provided by evidence for p is defeated or lost because you acquire strong evidence for ~p. In tasting the orange juice, you might judge that there's something wrong with it because of a funny taste. You remember that you've just brushed your teeth and remember that the minty toothpaste affects how things taste to you. Here, the justification provided by your evidence is undermined. The funny taste is typically some reason to think that there's something wrong with the juice, but there's some additional evidence in the form of your knowledge that you've just brushed your teeth that you need to take account of. The combined evidence doesn't support the hypothesis that there's something wrong with the juice, not even if a part of that evidence (i.e. the way it tastes) could have supported that hypothesis that there's something wrong with the juice if only you didn't have the additional evidence that brushing your teeth affects the way things taste to you.
1.79 The Argument from Defeasibility challenges the idea that there are non‐inferentially justified beliefs by challenging the idea that there are indefeasibly justified beliefs. Beliefs that are defeasibly justified are things that you can rationally reject or suspend judgment on as new evidence comes in. In the case of sudoku, there's no process that should lead you to believe that the rational way to solve the puzzle is to change the numbers given at the start of the puzzle. If non‐inferential beliefs are akin to the numbers given at the start of the puzzle, shouldn't we think of these beliefs as rationally unrevisable? If so, since every belief is rationally revisable, shouldn't we think that there aren't any non‐inferential beliefs?
1.80 What can the foundationalists say in response? They have two salient options which track different ways of characterizing the relationship between foundational beliefs and defeasibility:
Option 1: the foundationalist might agree that foundational beliefs have to be indefeasibly justified, and they might defend their view by trying to show that there are such beliefs and that there is a sufficient stock of such beliefs to serve as the foundation for all of our knowledge.
Option 2: the foundationalist might deny a foundational belief has to be indefeasibly justified. As they see it, so long as a belief is supported by something that isn't a belief and the support this belief receives isn't defeated, this could be good enough to justify a belief. Much as a promise, say, can justify an action in the absence of defeating reasons, perhaps an experience – or perhaps a certain causal source – can justify a belief in the absence of defeaters.
1.81 We may call foundationalists who opt for Option 1 classical foundationalists and those who opt for Option 2 modest foundationalists. The classical foundationalist simply embraces the idea that our foundationally justified beliefs have the two features that the given numbers in sudoku have, which is that they are (i) beyond rational revision as well as (ii) the foundation for all justified belief. The modest foundationalist, by contrast, rejects the idea that these two features come together. As they see it, the sudoku analogy is useful, but only within reasonable limits. As they see things, an important difference between, say, our beliefs about the external world and our beliefs about the grid is that the former can be justifiably held on the basis of reasons that provide only defeasible support.
1.82 We'll have more to say about the respective virtues of the classical and modest foundationalism in Chapter 2 (in connection with perception) and in Chapter 3 (in connection with the particular and vexed case of a priori justification). The important point to take from the present discussion is that there are two ways for the foundationalist to respond to the Argument from Defeasibility and thus two ways to flesh out the details of that view.
1.83 Whereas the Argument from Defeasibility invites us to take a side in a disagreement between classical and modest foundationalists without forcing us to abandon foundationalism altogether, a further line of argument poses a serious threat to foundationalism in any form.
1.84 What distinguishes the foundationalists from the critics of foundationalism is a commitment to the possibility of non‐inferentially justified belief. It should be possible, on the foundationalist model, for a belief, p, to be justified in virtue of possessing some feature, F, even if the subject doesn't have any opinions about whether her belief that p even has this feature F.38
1.85 And here, according to BonJour (1978), is the source of the problem. For a belief to be justified, it is supposed to be properly connected to the truth. Presumably, this is what F does. The possession of F ensures that beliefs that are F are properly connected to the truth. Here's the problem. BonJour thinks that it's important to the very idea of a justified belief that such a belief is responsibly held. To be responsibly held, however, he thinks that the believer would have to be cognizant of the fact that her beliefs have this feature F. This, however, creates trouble for the idea that the relevant belief is immediately justified. To be cognizant of the fact that the belief has this feature F, the subject would have to have an opinion – some further belief – about whether her belief had this feature. But