Just finished teaching a section on epistemic principles or principles of evidence, as Chisholm sometimes calls them, using Jim’s paper “Foundationalism, Epistemic Principles, and the Cartesian Circle.” Jim rightly criticizes certain coherentists for first accepting the truth of certain epistemic principles and then insisting that they must also be known to be true or rational to believe in order to play a role in any story about why a given target belief is justified. So, consider the following representation of a ground for belief:
It appears to me that there is an electron path in that Wilson cloud chamber.
So, there is an electron path in that Wilson cloud chamber.
Complaints about this representation are manifold: the conclusion doesn’t follow (irrelevant, if “follow” means “follow deductively”), you forgot to rule out potential defeaters (hmmm…this raises issues I’ll bypass here), your description of your appearance is too theory-laden (good luck drawing the relevant distinction), you should have first concluded that it is reasonable to believe something and then detached the epistemic operator (no, I shouldn’t have, though perhaps I might have done so). The complaint I’m interested in, though, is the one that Jim criticizes. It’s the one that grants the truth of the epistemic principle “if you’re appeared to in a certain way, then it is reasonable to believe that things are as they appear (in the absence of grounds for doubt),” but then criticizes the representation in question for not including anything about “the epistemic status” of the principle in question. It is this criticism that I think goes wrong.
At first pass, we could put this point as follows. The line separating support from supported encodes the rule of evidence, and if we insist on putting the rule in as an additional premise, or if we insist on putting an epistemic operator on the conclusion, we make one of two mistakes. The latter mistake is the one underlying the modal fallacy of confusing the necessity of the consequence with the necessity of the consequent. The former mistake is the one Lewis Carroll parodies, when the unwitting agrees to put the corresponding conditional to modus ponens in as an additional premise of the a modus ponens argument, thereby launching an infinite regress.
But now, the kicker: if the relevant principles are, let us say, encoded in the line we draw to separate premises or grounds from that which they support, then what can we say about Plantinga’s many theistic proofs, one of which is:
think of all the numbers there are.
So, God exists.
OK, one response, which I won’t allow here because it doesn’t get to the heart of the matter, is that the conclusion is false. If that’s your whine, then replace the conclusion with “2+2=4”. Or if you are now worried about necessary truths in the conclusion, try your favorite examples of familiar and unfamiliar correlations:
It is raining.
So, it’s cloudy.
The tides are high.
So, traffic on the Brooklyn Bridge is up.
On these inferences, we will rightly insist on the linking premise, and question it’s truth (haven’t we all seen cases where rain comes out of an apparently cloudless sky?). But, of course, this point doesn’t get to the heart of the matter either, since if we want linking conditionals here, why not in the cloud chamber example? And the linking conditional will be as false there as here. So, to be ridiculously inexact, let’s say that the linking claim is true enough so that failure of truth value isn’t the issue.
So, the first point is that I don’t think we have much of a clue as to when a representation of a reason for belief is an enthymeme and when it isn’t, except the following: when there is a true principle of evidence linking the kind of information in the premise with the kind in the conclusion, then we don’t have an enthymeme. OK, but still mere truth isn’t going to be enough. There is a true principle of evidence tying the presence of rain with the presence of clouds, isn’t there?
I anticipate the reply that it won’t be a necessary truth, and that the correct principles of evidence must be necessary truths; only then can they be encoded in the line separating support from supported, rather than being required as an additional premise.
But why think that? If our epistemic theory is complete, there will be at least one necessary truth characterizing the conditions under which a belief is rational. But the necessary truth of principles of evidence is a further claim that goes beyond this claim. So if an evidential argument is an enthymeme unless the principle of evidence is a necessary truth, perhaps every non-deductive argument has to have the principle of evidence as a premise.
But even if the principles of evidence are necessary, it isn’t clear why that is sufficient. In deducing q from p, we wouldn’t count the deduction as imparting knowledge of the conclusion from knowledge of the premise if q followed from p only in virtue of some quite complicated logical fact, such as one of the incompleteness results. So something more than mere necessary truth is needed, even if we grant the claim that there are necessary principles of evidence.
What more is needed? Chisholm added something. He holds that principles of evidence are synthetic, a priori, necessary truths. Adding that they are synthetic probably makes the issue harder rather than easier, but a prioricity might be thought to help. Well, it won’t help with the incompleteness result example above, and we certainly want to be able to explain how Hilbert was justified in believing various aspects of his formalist program that entail the falsity of the incompleteness results.
The bottom line, I think, is that we have no good reason to think that the only issue about linking principles is whether they are true, or necessarily true, or all of that plus knowable a priori as well. Holistic coherentism deny that all of these are sufficient for prima facie evidential support, and though the specific ways they raise this complaint that are discussed by van Cleve are inadequate, the general point remains untouched.