1 February 16,1987,96

2 Thus the class of idiosyncratic beliefs contains the false beliefs. There are, of course, false beliefs that one might hold without having in any plausible sense epistemically failed, and we therefore do not mean to suggest that a belief’s being idiosyncratic necessarily constitutes an epistemic failing. But it may, and in the central case we will be considering, it clearly will.

3 We say ‘virtually’ because we have (loosely) defined the teflonicity of a belief in terms of a believer’s resistance to belief change given the receipt of information concerning the content of that given belief. Thus, we leave open the possibility of a teflon belief’s changing or even being given up upon the receipt of some information unrelated to the content of the belief.

4 We have not, of course, said what it is to receive information; rather, we have cited what we take to be a sure sign that information has been received — a mark, one might say, of its having been received. (Other signs that the information that p has been received by a subjectS include: S’s (immediately) coming to believe that p; S’s (immediately) coming to form (further) evidential beliefs relative to his antecedently held belief that p.) Similarly, in section ll1 below, we do not require that Dretske give a full account of what causal impacting is. We argue, rather, that he is unable to cite a sign or mark of causal impacting (or, to put the point a little differently: he is unable to cite a difference that the causal impact of information on a subject’s belief makes), present in the meddler case and absent in the Reagan case, such that on the basis of its presence (or absence) the two cases, and by implication the two important classes of distinct cases they typify, can be distinguished.

5 Some might object to our suggestion that Reagan would have the very belief that p* that he actually does have, if (1) the peculiar actual causes of his belief that p* had preceded his receipt of the information that p*, and (2) the receipt of the information that p* had, instead, caused Reagan to believe that p*. Rather, it might be claimed, in cases in which the ‘other contributory causes’ precede a causal sustainer and thus where the relevant belief was in place before the sustainer occurred, to imagine that the contributory causes had never been is to imagine out of existence the original belief and to imagine into existence another belief that p*.

Now it is true that those who insist on this objection will be able to say, against the argument immediately below in the text, that Reagan’s belief that p* is not causally sustained by the information that p*, and hence that Reagan does not come out knowing that p* on Dretske’s view. This defense is not available to Dretske, however, for in cases like that of the meddler (which we discuss below) he demonstrates the need to count some causes as causally sustaining a belief even though the belief had been in place at the time of the occurrence of the sustainers. He must then take a sufficiently broad view of the identity conditions for beliefs to be able to count, against the objection, one belief as having had a different originating cause. As we shall see, other philosophers in the tradition have implicitly adopted such broad identity conditions for beliefs.

6 We can make sense of these notions only if we countenance sufficiently wide identity conditions for beliefs. Seen. 5, above.

7 It should be plain, however, that if the objector insists that only genuine overdeterrniners can play a causal role in beliefs, our claim can be recast accordingly: the difficulty for which we argue would be that Dretske’ s theory is not able satisfactorily to make out the distinction between genuine and pseudo overdeterminers in a way that will separate the problem cases typified by the Reagan case from cases like the meddler (or the Yankees) case. It is, in particular, unable to articulate what it is about Reagan such that it could be said on that basis that the information that p* is not a genuine overdeterrniner of his belief that p*.

In the end then, the issue between us and the objector and us is merely verbal; nothing of importance hangs on how it is settled.

8 Perhaps this is an appropriate place to raise and answer an objection that might occur to close readers of Knowledge and The Flow of Information.

Dretske is explicit that his concern is with perceptual knowledge. He writes ‘It should be emphasized at the outset that this is intended to be a characterization of what might be called perceptual knowledge, knowledge about an items, that is picked out or determined by factors other than what K happens to know (or believe) about it’ (86). Now, it might be held — especially given the suggestion that we are presently considering in the text — that the cases we’ve focussed on crucially involve inference, and hence are not perceptual in the relevant sense. If that were correct, then it could be suspected that the problem we have raised is irrelevant to the view that Dretske is expounding.

Our answer is that Dretske’s restriction to perceptual knowledge is not intended to exclude inference, as his discussions of informational content and nesting in Chapter 3 and of the communication of knowledge in Chapter 4 clearly indicate. We note, moreover, that some of his own crucial illustrations include: (1) knowing that the voltage drop across a resistor has a certain value, by seeing the position of the needle in a voltmeter (111-23), (2) knowing that a peanut is under a given shell, by seeing that it isn’t under the others (not, that is, by one’s seeing the peanut under that given shell) (79), and (3) learning that a person died, by reading a story to that effect in a newspaper (185). Quite clearly in the latter two, and arguably in the first, inferences are involved.

9 Nor is it helpful to retreat to the idea that I must at least have implicitly drawn the relevant inference. What is important is that I believe what the authority said and that his having said it had an impact on me. But in cases in which there is no independent evidence that any implicit inference has taken place, to claim nevertheless that some such inference must have taken place is just to claim that the evidence has impacted on me. It is not to specify, even broadly, how that can be so.

10 Furthermore, receipt of some information might well alter my disposition to affirm or deny a proposition in some deviant way — say by causing me to have a heart attack — without in any intuitive way impacting causally on the relevant belief

11 This idea can be found in Nozick’s, R. Philosophical Explanations (Cambridge, MA: Harvard University Press 1981)Google Scholar, in the discussion of outweighing (see 179-85).

12 This is not to say that there is no notion of causal sustaining that can be a component of any theory in any field. It may well be that the particular puzzles that populate the field in the analysis of knowledge, and the related field of epistemological skepticism, require a division of cases quite different from the puzzles that populate the fields of the various sciences. If that is right, a notion of causal sustaining that would be suitable for the latter could very well turn out not to be suitable for the analysis of knowledge (perhaps because it is not fine grained enough, or for some other reason).

This, we want to point out, would not be especially unusual. The rough semantics we presently possess for counterfactuals, it might be argued, are sufficient to permit their intelligible use generally in the sciences. This, even though the attempt to use any such semantics in a counterfactual account of knowledge — especially if it is to be invoked against the epistemological skeptic — is very likely to be, at the least, uncompelling. (See Lipson, M. ‘Nozick and the Skeptic,’ Australasian Journal of Philosophy 65 [1987)327-34CrossRefGoogle Scholar on this latter point.)

13 See Bach, K. (‘A Rationale for Reliabilism,’ The Monist 68 [1985] 246-63)CrossRefGoogle Scholar for this distinction.

14 ‘The Concept of Knowledge,’ in Midwest Studies in Philosophy IX, P. French, eta., eds. (Minneapolis: University of Minnesota Press 1984)

15 As another aside, we point out that there is an important vagueness in McGinn’s global reliability condition. Just which class of propositions is it which is the relevant class of propositions with respect to which global reliability is to be assessed? There may be no unique answer to this question. For instance, if the proposition putatively to be known has the form’s is F,’ must the process which led to this belief be globally reliable with respect to beliefs about S (or S’s) or with respect to beliefs about F’s?

This openness in the theory, of course, might be a virtue rather than a flaw. (The following test was suggested by the ‘framing effect’ for decisions discussed in Tversky, A. and Kahneman’s, D. ‘The Framing of Decisions and the Psychology of Choice,’ Science 211 [1981) 453-8.)CrossRefGoogle ScholarPubMed Suppose that one were able to describe in two different ways certain types of situations in which it is questionable whether or not R knows that p, one way suggesting that a certain class of propositions Cis relevant and the other suggesting some other class of propositions C’ is relevant. Suppose further that R was pretty clearly reliable over the class C and not reliable over the class C’. If one’s intuitions as to whether or not R knew that p in that situation varied with the mode of description, then one would have empirical evidence that the indeterminacy in McGinn’s theory matched up with an indeterminacy in our commonsense concept of knowledge.

16 We wish to thank Bruce Freed, Peter Loptson, and Peter Vallentyne for particularly helpful comments on and criticisms of earlier drafts of this paper, and John Collier and Ali Kazmi for help and encouragement along the way. Hector Casteftada suggested the term ‘teflon’ to us.