for our beliefs most strongly when we recognize our own epistemic failings. Consider this passage from Descartes' Meditations:
Some years ago I was struck by how many false things I had believed, and by how doubtful was the structure of beliefs that I had based on them. I realized that if I wanted to establish anything in the sciences that was stable and likely to last, I needed – just once in my life – to demolish everything completely and start again from the foundations … My reason tells me that as well as withholding assent from propositions that are obviously false, I should also withhold it from ones that are not completely certain and indubitable.
(1641, p. 1)
On this way of looking at things, we see the need for justificatory support when we recognize that our own beliefs are sometimes mistaken. Because of these mistakes, we can now see that the beliefs we've formed in the past are doubtful. The epistemic ailments, then, are mistaken beliefs and beliefs that are doubtful to some degree. The prescription? At least according to Descartes, it is to believe nothing that is doubtful to even the slightest degree2 – viz. to withhold assent when we don't have something that makes it certain that what we believe to be true is indeed true.
2.7 On this way of looking at things, if something stops a justificatory regress, it must provide us with certainty. Call this position Cartesian foundationalism:
Cartesian foundationalism (regress‐stopping claim): something stops the regress of justification by constituting a justifier for our foundationally justified beliefs iff this is something distinct from a belief that provides the thinker with certainty that her foundational belief is true.
If this is the right way to think about the termination of the epistemic regress, then the question as to whether experience can stop the epistemic regress can be restated as a question about whether experience can provide us with the right kind of certainty.
2.8 Many philosophers now believe that this is the wrong way to think about the epistemic regress. A standard complaint is that the Cartesian standard is too demanding. Let's quickly consider two reasons why we might think this. First, think about the times when you're trying to help those near and dear to you by providing them with information. Maybe a friend wants to know whether Venus Williams advanced to the quarter‐finals. You check the paper: it says that she has, and so that's what you tell your friend. You recognize that there's something good about telling the truth and something bad about telling them something false. Although there's a downside to giving them bad information, you don't think the risk is too great. (You don't say, “I couldn't say” or “I don't know.”) You also recognize, implicitly, that newspapers are fallible. And yet you think that the risk of error doesn't compel you to say “I couldn't say.” You think of your testimony as a regress stopper, albeit a fallible one. Reflecting on this, you might think that someone who had the paper and said that they couldn't say because they know that papers have made mistakes in the past isn't just showing an aversion to error; they seem to show a pathological aversion to error, one that seems far too concerned with the disvalue of getting things wrong. Whether it's rational to tell your friend that Venus advanced to the next round seems to turn on whether you want to help them discover the truth (you do), the evidence you have (the information you can extract from the paper), and the value and disvalue associated with getting things right and getting things wrong. If getting things wrong was a disastrously bad outcome, you might need exceptionally strong evidence to reasonably decide to tell your friend that something is so. That we seem to reasonably decide to tell our friends that certain things are true with only fallible evidence to rely on suggests we don't normally think that it's an utter disaster to acquire a false belief. Because of this, it's not clear why we should think that a small risk of saying or believing something false should mean that it's not fully rational to believe. It's hard to square this assessment of the importance of avoiding error with the stringent standards of the Cartesian view.3
2.9 In addition to this point about tolerating risks, it's worth highlighting a related point about our sources themselves. Just step back and think about paradigmatic cases of knowledge and of justified belief. Think about the knowledge we can gain by looking online. You might look online and find numbers for your favorite restaurants. It's quite natural to think that in committing the number to memory, you come to know that the number to your favorite takeout place is 07963‐123 456. It's natural because you know that the book was carefully researched. Imagine that you also discover that the book contains this line in the opening pages of the book: “Although we went to great lengths to ensure the accuracy of each entry, we are sure that mistakes will eventually be discovered and we will need to make changes accordingly.” It's tempting to think that you know the book contains at least one error. These knowledge ascriptions suggest that the fallibility of the book is compatible with the further idea that you can stop the regress of justification and come to know or justifiably believe things about phone numbers by consulting the book. If some paradigmatic cases of knowledge are cases where we gain knowledge from fallible sources, it's not at all clear why the fallibility of the senses should be taken to be a reason to think that the senses do not provide adequate justificatory support for our perceptual beliefs. If we do not need certainty to justifiably draw an inference, why should certainty be required for non‐inferential justification?
2.10 Most foundationalists now accept this less stringent view:
Modest foundationalism (regress‐stopping claim): something stops the regress of justification by constituting a justifier for our foundationally justified beliefs iff this is something distinct from a belief that provides the thinker with adequate albeit fallible support for her foundational belief, and this support is not defeated.
If we don't think that the possibility of error is what creates the need for justificatory support, we probably wouldn't think that the regress stoppers have to provide certainty. We might think of the justified beliefs as the beliefs that it makes sense to form given the end of forming true beliefs and avoiding false ones, say. On this way of looking at things, the adequacy of evidence doesn't turn on whether it's impossible for this evidence to lead you astray. If it makes truth sufficiently probable or shows that your beliefs are not held arbitrarily, this might be sufficient. The Cartesian might complain that nothing can provide adequate support if it's fallible, but this is one of those issues that we'll have to try to sort out as we work through a range of other discussions in the book. At any rate, the question for the modest foundationalist is whether experience can provide adequate (albeit fallible) support for our beliefs about the external world.
2.11 Now that we have two views concerning regress stoppers on the table, we can turn our attention to perception and perceptual experience. We need to know more about perceptual experience to determine whether it could stop the regress of justification on either of the foundationalist views just introduced.
2.3 How to Talk about Experience
2.12 We need to talk about talking about perception and perceptual experience. You see the page and the words on it. You see the black marks on the page and you see them as words, words that you're familiar with because you're an English speaker. In speaking about perception in the pages that follow, it will be helpful to remember the distinction between seeing something and seeing something as something:
1 Agnes sees the marks on the page.
2 Agnes sees the marks as words.
2.13 If Agnes could read, both (1) and (2) might be true. If Agnes were a dog, she could see the marks but would not see the marks
