For those who haven't caught up, Stanley Fish has some more of his friendly God-talk, over on his New York Times blog.
Brian Leiter has a scathing response, followed up by an interesting discussion thread.
The gist of Fish's argument is that all knowledge claims must start somewhere, e.g. with assumptions about what counts as evidence, what counts as logically cogent argument, and so on. Hence, there is a sense in which all claims to knowledge are ultimately grounded in something that we might as well call "faith". Therefore, scientific claims such as the claim that the Earth revolves around the Sun are ultimately no better grounded than, say, typical evangelical Christian claims, such as the claim that, by his death on the cross and subsequent resurrection, Jesus of Nazareth atoned for our sins and broke the hold of sin and death over us. Both categories of claim rely on faith sooner or later, or so Fish wants us to think. Neither is epistemically superior to the other.
It is, of course, true that the grounding of any knowledge claim will eventually run out. If somebody does not accept our basic assumptions about what forms of argument are cogent and what counts as evidence, we can not convince her of anything that she does not want to be convinced of. For that reason, it's true to say that there is no argument about anything that is effective in persuading all comers, no matter how fanatical or even insane.
You can, of course, demonstrate the havoc that ensues if you violate ordinary logic by embracing a contradiction, such as "P & ~P". But even the demonstration will require reliance on some basic logical rules. Someone who is prepared to embrace a contradiction probably won't mind denying these, and in any event may not care if they are committed to accepting any proposition at all.
It's easy enough to demonstrate this, by the way. Assume any proposition you like, "Q", which might, for example, mean "Stanley Fish is the Great Beast of Revelation". Let's assume "P & ~P" and try to derive this directly by using some logical moves that are pretty standard.
1. P & ~P (assumed)
2. P (from 1. by & Elimination)
3. P v Q (from 2. by v Introduction)
4. ~P (from 1. by & Elimination)
5. Q (from 3. and 4. by Disjunctive Syllogism)
If your pet logical system doesn't have Disjunctive Syllogism as a fundamental rule, this might take a bit longer, but no plausible system of first-order logic fails to provide for Disjunctive Syllogism somehow. So you get the idea. I've just demonstrated that if any contradiction is true then Stanley Fish is the Great Beast of Revelation.
As I said, that might not worry someone who is not worried about accepting a contradiction in the first place. Still ...such a person must also either abandon some very basic rules of reasoning and accept that (as well as being the Great Beast of Revelation) Stanley Fish is the very same Great Fish that swallowed Jonah, not to mention that fact that he is one of the evil spirits cast into the Gadarene swine. And he's also the bottle of cough mixture that I've just been sipping from to try to loosen up some congestion in my lungs.
Likewise, chaos arises if I reason poorly about empirical matters. Suppose I often find that I bump into walls when I attempt to walk through them. In fact, it turns out that every time I've tried to walk through a wall (let's say I've attempted this 473 times), I've failed, and instead bumped into it. I now reason along the following lines: "I've failed 473 times in a row in my attempts to walk through walls. That's suggests I'm now long overdue for success; it's evidence that I'll succeed on my next attempt. I'll now walk through the wall that's directly ahead of me." Don't try that at home, folks.
Of course, someone who reasons like that can hang on stubbornly to their approach, with each failure providing stronger evidence that the next attempt will succeed. If someone is sufficiently stubborn, she will always be convinced that the evidence favours success on the next attempt.
But, while our various chains of inference cannot be justified all the way down to all comers, it does not follow that none are better than others. Chains of inference don't need to be justified all the way down. In fact, the very idea is incoherent. But some can be justified down into claims that no sane person would deny.
In fact, that's more typical than not of honest truth claims made in daily life. Scientific claims often move beyond what we can observe directly and involve things that are very small, or very far away, or no longer exist (but have left traces). In some cases, however, they are supported by consilient chains of inference that are strong at each step and that ultimately rely on claims that no sane person denies. Accordingly, science is often able to make real progress. Scientific investigations can converge on very robust propositions that are acceded to by properly trained people from many different personal and cultural backgrounds. Science's methods are continuous with the ordinary methods of reasoning that we use in day to day life, but made more rigorous in various well-known ways, to make up for the ubiquity of circumstantial evidence and heavily theory-laden reasoning.
Religion is simply not in this position. When we say that it relies on faith, we don't just mean that it eventually depends on assumptions about what counts as evidence and what counts as cogent reasoning - assumptions that can't be proved without relying upon them, because they count as our standards for what can be proved or evidenced. One problem is that religious statements are often in contradiction to statements that can be justified as far as it reasonable to try to justify anything. For example, the claim by Christian fundamentalists that the Earth is 6,000 to 10,000 years old contradicts the well-evidenced scientific finding that the Earth is well over 4 billion years old. It's true that the latter can't be justified all the way down - not in the incoherent sense that the most basic assumptions about evidence and logic on which the claim rests can themselves be justified in a non-circular way. But, in any coherent sense that you wish to specify, the claim is justified and the Christian fundamentalist claim is unjustified.
Similarly, the claim that the Nazis killed 5 to 6 million Jews during the Holocaust perpetrated in the World War Two years is justified in any coherent sense that you wish to specify. Claims to the contrary by Holocaust deniers are not justified. They are false.
In other cases, religious claims may not formally contradict other claims that are well-justified by evidence and cogent reasoning; it is just that they are not justified by our normal standards of what counts as justification. There is no good reason why someone who does not already believe these claims should do so. They are believed on the basis of socialisation or emotional experience, or wishful thinking, or some such thing, but not on the basis of evidence. A person who does believe these claims can be given good reason to doubt them, but she can continue to believe by relying on an inner feeling of calm or something similar. It is not just that these claims can't (incoherently) be justified all the way down. They can't even be justified as far down as is necessary for them to be well-evidenced. If our ordinary standards of evidence and reasoning were applied, then they would be assessed as probably false.
These distinctions are not all that difficult to make. It is easy to establish that some claims are better evidenced than others, by ordinary standards that all sane people accept, or by the use of reasoning and investigative techniques that are continuous with those standards. It's true that we can't justify our ultimate standards of justification themselves - to demand that we do so is incoherent. However, it's also true that some claims are well-supported by evidence, as assessed by standards of evidence that it is perfectly rational to acknowledge and that no sane person fails to apply in day to day life. Other claims, notably religious ones, well ... not so much.