Open Book Publishers logo Open Access logo
  • button
  • button
  • button
GO TO...
Contents
Copyright
book cover
BUY THE BOOK

4. Evidence and Transmission Failure

© 2017 Mark McBride, CC BY 4.0 https://doi.org/10.11647/OBP.0104.06

Against the claim that the impression that there are instances of transmission failure is the product of a flawed conception of evidence, I argue that transmission failure (conceived as related to epistemic circularity) is possible on most plausible accounts of evidence. The importance of this (exploratory) chapter is primarily theoretical, rather than dialectical.

0.1 Consider the following argument:

(MOORE)

(1) I have hands.

(2) If I have hands an external world exists.

(3) An external world exists.

Our focus in this chapter is on (MOORE), but points will generalise to (reasoning by means of) arguments with a similar structure (e.g. (TABLE)/(EK)). (MOORE) is clearly valid and its premises are true. Yet, to many, reasoning by means of this argument — which we are calling (MOORE)-reasoning — seems defective. On one view (e.g. Wright’s), it is an example of transmission failure. In section 2 we’ll come to precisify this concept. For now let’s make do with a very rough first pass at transmission failure. Philosophers locating the defectiveness of (MOORE)-reasoning in transmission failure, typically take (MOORE)-reasoning to suffer from a form of epistemic circularity: it’s only if one already has warrant for (3), that one can have a warrant for (1). If (MOORE)-reasoning suffers from transmission failure in this way, we can say that propounding the argument cannot offer one a new route to a warranted belief in its conclusion.

0.2 What’s one’s warrant to believe (1)? Put differently: what’s one’s evidence for (1)? (Suppose the conditional premise (2) is knowable a priori.) These questions are badly formed. There are various different warrants one might have for (1). Lots of things could be one’s evidence for (1). Suppose I’m sitting blindfolded, unable to see my hands, and Edna whispers gently in my ear: ‘You have hands’. Edna’s utterance — on one plausible view — counts as good testimonial evidence for me that I have hands. That’s one form of evidence I might have for (1), and there are many others.

Let’s focus, though, on visual-perceptual evidence that I have hands, and specifically the mere having of a visual experience as of having hands — the having of which is compatible with my failing to have hands.1 (Indeed, unless we focus on a particular form of evidence it will be indeterminate whether transmission failure occurs.) With this assumption about the nature of my evidence for (1), the idea that (MOORE)-reasoning involves epistemic circularity has some initial plausibility. For, it might be said, it’s only if I antecedently and independently have a warrant for (3) that an experience as of having hands constitutes a warrant for (1). If, by contrast, I had good reason to think that (3) was false and that I was a handless brain-in-a-vat (BIV) with experiences that were being generated by an evil scientist, then merely having a visual experience as of having hands would not be such good evidence for (1). (But, whether or not you think that you are a BIV, it can be maintained that an experience as of having hands raises the probability of (1) [I have hands — rather than flippers, or bloody stumps] and also of (1)´ [I am a BIV and the evil scientist is producing an experience as of having hands — rather than producing an experience as of having flippers or as of having bloody stumps].)

If (MOORE)-reasoning does involve epistemic circularity, then following through the reasoning could not provide a new route to a warranted belief in (3). Even if one were to allow that it would provide a new, gratuitously circuitous, route, still it would not be a way of arriving, for the first time (cf. Wright 2002: 333), at a position in which one had a warrant to believe (3). (Recall, in Chapter One I showed that, even if it would not be a way of arriving, for the first time, at a position in which one had some warrant or other to believe (3), it could still be a way of arriving, for the first time, at a position in which one had a (partly) perceptual warrant to believe (3).)

0.3 Some philosophers — viz. Timothy Williamson (2000: ch.9), John Hawthorne (2004: 34, n. 86),2 and Nicholas Silins (2005)3 — (can be taken to) claim that there are no genuine instances of transmission failure, provided we operate with the right account of the sources of warrant-or-evidence4 for future reasoning.5 This novel terminology requires some preliminary exposition. As to sources of warrant-or-evidence, pre-theoretically, or before commitment to any particular theory, warrant or evidence can come in myriad forms: blood and DNA samples, hearsay, facts, states of affairs, objects etc. Our principal focus, though, will be on two particular possible sources of warrant-or-evidence: an experience and a premise (i.e. a proposition).6 As to future reasoning, we may take inference to be our paradigm case thereof (as contrasted, for example, with coming to know that from which one infers).7 My aim in this chapter is to (begin to) clear the way for instances of transmission failure regardless of the account of the sources of warrant-or-evidence for future reasoning with which one operates. My aim is not to claim there are in fact genuine instances of transmission failure; merely to render it possible on all — or most — plausible accounts of the sources of warrant-or-evidence for future reasoning.8

0.4 Silins (2005: 84–89) claims instances of transmission failure only arise on a particular — putatively flawed — account of the sources of warrant-or-evidence for future reasoning.9 Call this Silins’s claim. Here’s a key commitment of this account. (Silins is going to reject the commitment and reject the account of which it is a commitment. It may help to keep (MOORE) in mind when reading this passage.):

[Y]ou might know a certain premise, or indeed be absolutely certain that the premise is true, yet still not possess the premise as a warrant for a conclusion you have competently deduced from the premise. On this view, if your inference provides you with any warrant for believing10 the conclusion at all, the warrant provided is the prior warrant for which you believe the premise. Thus even if you know that p, or are absolutely certain that p, it is still not the case that your warrant for believing the conclusion is that p. (Silins 2005: 88)

Sticking with (MOORE), suppose, as before, my warrant to believe (1) is my having an experience as of having hands. Nonetheless, suppose I can come to know, on the basis of this warrant, that I have hands. Suppose also, that I know (1), and am certain that (1) is true. Finally suppose I competently deduce — using the conditional, (2) — (3) from (1). Silins now cuts in that, on this (to-be-rejected) account of the sources of warrant-or-evidence for future reasoning, if my inference provides me with any warrant to believe (3) at all, the warrant provided is my experience as of having hands. Crucially, Silins points out, on this account, even though I know that I have hands, and am certain that I have hands, it’s not the case that I can use the fact that I have hands as a source of warrant-or-evidence for future reasoning. Specifically, I can’t use the premise that I have hands — a premise, note, whose truth is incompatible with my being in an idealistic world — as a source of warrant-or-evidence to believe there to be an external world. Rather, on this account, all I can use is my experience as of having hands, combined with inference — an experience, note, the having of which is not incompatible with my being in an idealistic world.11

Silins’s claim that instances of transmission failure arise only on one (to-be-rejected) account is resisted in this chapter. One cannot confine the possibility of transmission failure to this one account of the sources of warrant-or-evidence for future reasoning. So, at least, I shall argue.

1. Conceptions of Evidential Warrant and Accounts of Sources of Warrant-or-Evidence for Future Reasoning

1.1 Consider (where, throughout, ‘S’ is a placeholder for a subject and ‘p’ for a proposition):

Fallibilism: S can know p when S’s evidence for p is compatible with ~p.

Infallibilism: S cannot know p when S’s evidence for p is compatible with ~p.

Focusing on (MOORE), fallibilists can suppose my having an experience as of having hands is — or can be — (part of) my way of coming to know that I have hands.12 Infallibilists, by contrast, typically suppose my seeing that I have hands is my way of coming to know that I have hands, where my seeing that I have hands is incompatible with my not having hands.13 Think of fallibilism and infallibilism as particular conceptions of evidential warrant: they’re particular views of what it takes for a piece of evidence to warrant a belief in a proposition for a subject such that that subject can acquire knowledge. A philosopher’s conception of evidential warrant is typically associated with a particular account of the sources of warrant-or-evidence for future reasoning. A fallibilist typically takes an experience as of having hands as his source of warrant-or-evidence for future reasoning (and indeed sometimes, perhaps often, for (1) itself). An infallibilist typically takes (knowledge of) (1) itself — a premise — as his source of warrant-or-evidence for future reasoning (and indeed sometimes, perhaps often, as his source of warrant-or-evidence for (1) itself).

1.2 Philosophers who’ve explored the possibility of transmission failure in (MOORE)-reasoning (and in like arguments) — viz. Martin Davies (2004, 2009), James Pryor (2004), and Crispin Wright (1985, 2002) — operate with a fallibilist conception of evidential warrant. Let’s — exhaustively and exclusively — partition fallibilism as follows:

Fallibilism—: S can know p when S’s evidence for p is compatible with ~p, and when S knows p, p itself is not part of S’s evidence-base for future reasoning.14

Fallibilism+: S can know p when S’s evidence for p is compatible with ~p, and when S knows p, p itself is part of S’s evidence-base for future reasoning.

These theses are constructed by combining a conception of evidential warrant (first conjunct: fallibilism) with an account of the sources of warrant-or-evidence for future reasoning (second conjunct: ‘evidence-base for future reasoning’).15 Precisely what the second conjunct of fallibilism forbids is a subtle matter (cf. n. 11 supra). I take it, somewhat suggestively, to be the following: the fallibilist— denies that a known premise, p, can itself be used in future reasoning, but this leaves open that future reasoning from a known premise, p, can lead to, for example, further knowledge.16 Fallibilists are typically fallibilists—;17 but nothing in fallibilism rules out being a fallibilist+.

1.3 Now consider:

Infallibilism+: S cannot know p when S’s evidence for p is compatible with ~p, and when S knows p, p itself is part of S’s evidence-base for future reasoning.18

As with our previous two theses, this thesis is constructed by combining a conception of evidential warrant (first conjunct: infallibilism) with an account of the sources of warrant-or-evidence for future reasoning (second conjunct: ‘evidence-base for future reasoning’). To conclude: how do the fallibilist+ and infallibilist+ differ? Each, after all, allows p itself into one’s evidence-base for future reasoning. Two features distinguish them. First, most obviously, they operate with different conceptions of evidential warrant: for the former — but not the latter — one can know p when one’s evidence for p is compatible with ~p. Second, and less obviously, the latter’s allowance of p itself into one’s evidence-base may well not be confined to future reasoning (p itself, on versions of this view, is part of one’s evidence-base for knowing p), while the former’s allowance may well be so confined (on versions of this view) — cf. nn. 13 and 14 supra. In part because our chief focus in this chapter is on future reasoning, and in part because we only seek to secure the possibility — rather than genuine instances — of transmission failure, neither difference, though important, turns out to be fundamental for our purposes.19

We now have a (novel) taxonomy of conceptions of evidential warrant and accounts of the sources of warrant-or-evidence for future reasoning to hand. Now let’s try to precisify the concept of transmission failure and locate it within this taxonomy.

2. Transmission Failure and (In)fallibilism

2.1 Consider this (sufficient) condition for non-transmission:

(NT) Non-transmission of warrant

Epistemic warrant is not transmitted from the premises of a valid argument to its conclusion if the putative support offered for one of the premises is conditional on its being antecedently and independently reasonable to accept the conclusion. (Davies 2004: 221)

Let’s spell out one view of how (MOORE)-reasoning might suffer from transmission failure. Let’s take the fallibilist—. Suppose the fallibilist— claims (cf. Wright 1985, 2002, 2004) that:

(AW) In order for one’s visual experience as of having hands to act as warrant for (1) one must, antecedently, have warrant for (3).20

If (AW) is true, one’s warrant for (1) will depend on an antecedent warrant for (3). So (MOORE)-reasoning will suffer from epistemic circularity and (NT) will be triggered. Suppose that one in fact has antecedent warrant for (3), so that one’s visual experience does count as warrant for (1). Then, according to (NT), the warrant for (1) that is constituted by one’s experience as of having hands is not transmitted across the conditional — (2) — to serve as warrant for (3).21

2.2 What about our fallibilist+ (cf. Silins 2005)? Consider him debating with our fallibilist—, and arguing thus: if you, fallibilist—, are allowing, as a good fallibilist should, that I can come to know p notwithstanding my evidence for p is compatible with ~p, and allowing that I do indeed come to know (1) on the basis of such non-entailing evidence, why are you preventing me from using (1) itself as evidence in my future reasoning? (I leave open whether the non-entailing evidence for (1) is just the experience, or is additionally the antecedent warrant for (3).) If you were a sceptic about our knowledge of quotidian propositions, I’d vehemently disagree with you, being a good fallibilist myself, but I could at least make sense of your prohibition on my use of (1) as evidence in future reasoning: I couldn’t so use (1) as I wouldn’t, ex hypothesi, know (1). But given, as we can happily agree as fallibilists, I know (1), your exclusion of it that premise — from my evidence-base for future reasoning is unmotivated.22

Our fallibilist+ makes a good case. So let’s grant him, pro tem, victory over his fallibilist— foe. Our fallibilist+ may have won this particular battle, but not yet the war. Fallibilist— is a dogged soul. Suppose he accepts defeat over the inclusion of (1) itself in one’s evidence-base for future reasoning, but then marshals (AW) and claims that fallibilist+’s (MOORE)-reasoning thereby suffers transmission failure.

If, however, fallibilism+ is true, it’s no longer clear that (AW) leads to (MOORE)-reasoning suffering (complete)23 transmission failure. Why not? After all, isn’t it still the case, thanks to (AW), that (NT) is triggered? Recall, (NT) distinguishes between premises (which are propositions like (1), in the (MOORE) example) and support for premises (which is an experience as of having hands, in the (MOORE) example). (NT) talks about a valid argument, and the only valid argument in view is from (1) via (2) to (3).

The wrinkle, though, is that it’s no longer clear that (NT) is applicable, given the truth of fallibilism+. This is because of (NT)’s mere talk — when it comes to matters of transmission — of “support […] for one of the premises”, and not, additionally, talk of premises themselves.24 (NT) seems, pace fallibilism+, to preclude (1) itself from counting amongst one’s evidence-base for future reasoning. In other words, (NT) doesn’t seem to capture adequately the account of the sources of warrant-or-evidence for future reasoning with which a fallibilist+ operates. Take this to be what I’m driving at in suggesting (NT) isn’t applicable to fallibilism+ (and, mutatis mutandis, in subsequent uses of ‘applicable’). Put differently: the non-transmission of one’s warrant for (1) — which, by (NT), is one’s support for (1) — leaves open that (1) itself (which one ex hypothesi knows) transmits across the conditional ((2)) — perhaps, better, to account for the optionality of (2): across the relevant inference — to provide warrant for (3). In sum, I think, for all we’ve said, our fallibilist+ has won another (interim) battle.

Might it be objected that, if (1) itself is here being considered as a warrant that might be transmitted across the valid argument from (1) to (3), then presumably (1) itself, qua warrant, starts out as a warrant for (1), and is then transmitted to (3)? But, continues the objector, qua warrant for (1), (1) itself is not a fallibilist warrant but an infallibilist warrant. Such an objection misunderstands things. We noted before (cf. n. 13) that fallibilism— is compatible with the premise itself being part of one’s evidence-base for knowing the premise (it just precluded the premise from forming part of one’s evidence-base for future reasoning); a fortiori this is so for fallibilism+ (which has no such preclusion with respect to future reasoning). Fallibilism, abstracting from its species, is only a claim about what can, not what must, be the case when coming to know a proposition.

2.3 Before responding to our fallibilist+, let’s tie together the fallibilist+ and the infallibilist (i.e. the infallibilist+, cf. n. 18 supra). To be sure each philosopher, both the fallibilist+ and the infallibilist, operates with a different conception of evidential warrant. However, for the purposes of this transmission failure debate, this difference is less important than one’s account of the sources of warrant-or-evidence for future reasoning.25 Of more importance, then, is whether or not, when a subject comes to know (1) (or a like proposition), (1) itself becomes part of that subject’s evidence-base for future reasoning. Their response unites the fallibilist+ and the infallibilist: it does. We thus must alter standard taxonomies. Typically, in debates pertaining to evidence, fallibilists and infallibilists are taken to be the two camps ripe for comparison.26 However, in this transmission failure debate, it’s best to group fallibilists— in one corner (by dint of their refusal to allow (1)-type propositions into one’s evidence-base for future reasoning), and fallibilists+ and infallibilists together in the other corner (by dint of their allowing (1)-type propositions into one’s evidence-base for future reasoning). To be clear, the fallibilist— does allow that (1)-type propositions are premises in future reasoning, in the sense that conclusions can be (competently) deduced from them; the prohibition is simply that (1)-type propositions can’t be sources of warrant-or-evidence for said conclusions. We thus might hope that a good response, in the case at hand, to the fallibilist+ will also count as a good response to the infallibilist.

3. Modifying (NT) (and other germane theses)

3.1 The challenge is to see if we can construct a motivated (sufficient) condition for non-transmission of warrant which, unlike (NT), is applicable to fallibilism+ and infallibilism, and which, if the condition is satisfied, precludes (1) itself from transmitting across the conditional ((2)) to provide warrant for (3).27 A natural first response to putting the challenge this way would be to ask: what does it mean for (1) itself to fail to transmit across the conditional ((2)) to provide warrant for (3)? After all, (MOORE) is an instance of modus ponens. Let’s assume a plausible knowledge-closure principle, say: If S knows p and S knows p entails q, then S knows (or at least is in a position to know) q. Given this (and given (2) is knowable a priori), if I know (1), I know (or am in a position to know) (3). What room is there for a motivated principle — a modification of (NT) — granting this, yet allowing that (1) itself can fail to transmit across the conditional ((2)) to provide warrant for (3)?

3.2 First, let’s distinguish between a closure question and a transmission question. The germane closure question asks: is it possible to know (1), know that (1) entails (3), yet not be in a position to know (3)? The germane transmission question asks: is it possible to know (1), know that (1) entails (3), yet not thereby — that is, in virtue of knowing (1) and recognising the entailment from (1) to (3) — be in a position to know (3)? Or, putting the transmission question differently: is it possible to know (1), know that (1) entails (3), yet not be in a first time position to know (3)? Let’s answer the closure question ‘no’ (closure is exceptionless), yet remain agnostic about the transmission question.28 I take it that, even answering the closure question ‘no’, conceptual room is left open to deny that (1) itself transmits.

Second, let’s expand what I’ll call the standard account of warrant. Standardly, one’s warrant is one’s warrant for this or that proposition, and does not include the proposition itself; but we’re not bound by standard usage. Indeed we’re currently granting the fallibilist+ and infallibilist their claims that premise (1) becomes part of one’s evidence-base for future reasoning, once one knows (1). We might therefore modify (NT) by making explicit that warrant includes any known premises.29 Here’s a first pass:

(NT*) Non-transmission of warrant

Epistemic warrant — where warrant includes any known premises — is not transmitted from the premises of a valid argument to its conclusion if the putative support offered for one of the premises — or knowledge of any of the premises themselves — is conditional on its being antecedently and independently reasonable to accept the conclusion.30

Recall that, to show that (MOORE)-reasoning suffers transmission failure by (NT), our fallibilist— marshalled thesis (AW). And we can construct a thesis paralleling (AW) (call it (AW*)) which makes room for (complete) transmission failure in (MOORE)-reasoning by (NT*):

(AW*) In order for one’s warrant for (1), where warrant includes any known premises, to so act as warrant, one must, antecedently, have warrant for (3).

If (AW*) is true, and (as a central case) if one in fact has antecedent warrant for (3), (MOORE)-reasoning suffers transmission failure by (NT*). Is the pair of theses (NT*) and (AW*) motivated?31

4. Motivating our Modifications

4.1 Let’s start with (AW*). As with (AW), I don’t want to marshal arguments purporting to secure (AW*)’s truth. Instead, I simply want to present a way of formulating or expressing a defense of these (modified) theses. Recall, our overall strategy is not to secure genuine instances of transmission failure; it’s merely to render it possible, on all — or most — plausible accounts of the sources of warrant-or-evidence for future reasoning. So, how might one go about expressing a defense of (AW*)? Let’s approach this question indirectly. We’ll first consider how one might express a defense of (AW) and see if that helps us in discerning a way to express a defense of (AW*). Recall (AW):

(AW) In order for one’s visual experience as of having hands to act as warrant for (1) one must, antecedently, have warrant for (3).

One (likely) way of formulating or expressing a defense of this thesis would be by making use of notions contained within an account of a relation of epistemic antecedence (or priority):

(EA) One’s warrant for accepting p is antecedent to one’s warrant for accepting q iff one’s reasons for accepting p do not presuppose or rest on one’s reasons for accepting q.32

Someone defending (AW) would likely claim that one’s visual experience acting as warrant for (1) presupposes, or rests on, one’s reasons for accepting an external world exists. We don’t need to agree with that bald statement offered in defense of (AW). Indeed we here offer no reasons for (AW)’s (or (AW*)’s) truth. We just need to understand the claim so formulated or expressed.33

4.2 That’s (AW). Now what about (AW*)? We might formulate or express a defense of (AW*) by making use of notions contained within an account of the following relation of epistemic antecedence:

(EA*) One’s warrant for accepting p is antecedent to one’s warrant for accepting q — where warrant includes any known premises — iff one’s reasons for accepting p do not presuppose or rest on one’s reasons for accepting q — where one’s reasons include any known premises.

We’re supposing, recall, that (1) itself — which we ex hypothesi know — is part of our evidence-base for future reasoning. (We’re leaving open whether or not (1) itself is part of our warrant for accepting (1). If it is, one might argue that my experience as of having hands isn’t also part of my warrant for accepting (1), because (1) itself is an entailing (logically sufficient) warrant for (1), and so (1) does not need any additional support from the experience. Be this as it may, our chief focus is on one’s evidence-base for future reasoning, and, as we’ll come to see, this argument will not carry over to future reasoning in a straightforward manner.) In assessing (MOORE)-reasoning armed with notions from (EA*), we’re (in part) asking: does my knowing that (1) presuppose or rest on my reasons for accepting (3)?

Let’s make an assumption — an assumption favourable to fallibilists+ in particular. To understand why we’re making this assumption we need to understand a claim some (prominent) infallibilists make. Some infallibilists claim that all and only what one knows is one’s evidence (see Williamson 2000: ch.9). So for this type of infallibilist our supposition that (1) itself is part of one’s evidence-base for future reasoning, and that one’s evidence-base for future reasoning includes (1) itself, is a slightly misleading supposition: for them, (knowledge of) (1) itself exhaustively constitutes one’s evidence-base for future reasoning (prior to an act of inferring (3)).34 Any pragmatic implication that there must or may be more to one’s evidence-base for future reasoning than (1) itself should be cancelled. (We can call this the pragmatic implication objection, with the cancellation operating as a reply thereto.) I hereby cancel any such implication by explicitly noting that ‘part’ does not mean ‘proper part’, and ‘includes’ does not mean there need be any more evidence. For our fallibilist+, meanwhile, all that one knows is in one’s evidence-base for future reasoning, but not only what one knows: one still, in addition to (knowledge of) (1) itself, has one’s visual experience as of having hands as part of one’s evidence-base for future reasoning. This, at least, is one plausible way of carving up evidence for a fallibilist+.35

What’s the purpose of this detour? Recall, the strategy is to explore using notions from (EA*) to formulate or express a defense of (AW*), and to claim that (AW*) opens up the possibility of transmission failure for fallibilists+ and infallibilists by (NT*). Given this, we don’t want the fallibilist+36 to be inexorably lumbered with (complete) transmission failure in (MOORE)-reasoning because part of his warrant for (1), the part he shares with the fallibilist—, viz. a visual experience as of having hands, fails to transmit to (3).37 To allow for this would be to slur over the important difference between the fallibilist— and the fallibilist+. To preclude this from happening, we make our assumption, viz. some of one’s warrant can fail to transmit in an argument without all of one’s warrant failing to transmit.38 Maybe this assumption seems too obvious to need stating. But note that, without this assumption, if we suppose that (AW*)’s truth can be secured (perhaps with use of notions from (EA*)), this inevitably leads to a fallibilist+ being committed to (MOORE)-reasoning suffering transmission failure by (NT*). That’s because our trio of theses, (EA*), (AW*), and (NT*), don’t themselves discriminate between different warrants one might have for (1) (and for (3)). Given the two types of warrant on which we’re focusing, viz. an experience and a premise, are lumped together, if one type fails to transmit in (MOORE)-reasoning (viz. the experience) (NT*) is sufficiently coarse-grained (a bad thing) that it delivers the result that (MOORE)-reasoning suffers from transmission failure, period. (We can call this the coarse-grained objection, with the ensuing deflection operating as a reply thereto.) This is an unwelcome result: our modified (transmission failure) theses — (EA*), (AW*), and (NT*) — should be able to discriminate on the basis of the salient differences between the fallibilist— and the fallibilist+. We need theses which leave room, in line with our assumption, for one type of warrant to fail to transmit in (MOORE)-reasoning (e.g. here, the experience), and another type of warrant to transmit (e.g. here, the premise). Rather than further refining, and complicating, our modified theses themselves, I leave the task of avoiding the foregoing unwelcome result to sensitive application of the modified theses.

(Let’s pause briefly to think back to Chapter One. If one adopts Wright’s view (which Silins would locate as fallibilism— in the present taxonomy), then (MOORE)-reasoning does not put one into a position of having warrant for (3) for the first time, but it does, I argued in Chapter One, put one into the position of having perceptual warrant for (3) for the first time. The antecedent warrant for (3) does not transmit, because one already had that warrant; but the perceptual warrant provided by the experience as of having hands (given the antecedent warrant for (3)) does transmit. Is this the opposite of what I am now saying in Chapter Four? No. Here I am simply countenancing the experience failing to transmit, compatibly with which the premise could transmit. In engaging with Silins, that was the apposite compatibility relation to consider. Whether the foregoing possibility actually obtains will depend on what the correct (non-)transmission theses decree (which, in turn, I’ve argued here, will depend on what relations of epistemic antecedence obtain). In Chapter One I made an attempt, given certain background assumptions, at sketching some plausible (non-)transmission theses on which antecedent warrant does not, but perceptual warrant does, transmit.)

4.3 Let’s take stock. We might reconstruct the foregoing dialectic thus: we set out fallibilism—. We then set up three theses, (EA), (AW), and (NT), which seemed motivated, and captured adequately the fallibilist—’s account of the sources of warrant-or-evidence (for future reasoning). We used notions from (EA) to descry one way to express a defense of (AW). We then noted that, assuming (AW) and fallibilism—, we had transmission failure in (MOORE)-reasoning by (NT). Fine. We then noted that it wasn’t clear the account of the sources of warrant-or-evidence (for future reasoning) presupposed by these three theses was the same account with which fallibilists+ and infallibilists operate. Next we set up three modified theses, (EA*), (AW*), and (NT*), to see if we could construct a motivated transmission failure thesis that operated with an account of the sources of warrant-or-evidence for future reasoning that was friendly to fallibilists+ and infallibilists. But while our three modified theses operated with a (disjunctive) account of the sources of warrant-or-evidence (for future reasoning) not unacceptable to fallibilists+ and infallibilists, they were vulnerable to two objections from the two camps: a pragmatic implication objection (from infallibilists, which we cancelled) and a coarse-grained objection (from fallibilists+, which we deflected).

In sum, our original three theses, (EA), (AW), and (NT), are serviceable for analysis of fallibilism— (but not fallibilism+ and infallibilism). Our three modified theses, (EA*), (AW*), and (NT*), are serviceable for analysis of fallibilism+ and infallibilism (but not fallibilism—). We therefore have two sets of three theses, which jointly cover all, or most, plausible accounts of the sources of warrant-or-evidence (for future reasoning).

5. Might Introducing (EA*) be More Difficult than Introducing (EA)?

5.1 There may be especial problems involved with introducing our modified theses. Let’s focus on a putative problem arising over introducing (EA*) (which doesn’t arise when introducing (EA)). How exactly can a premise, a source of warrant-or-evidence for future reasoning by fallibilism+ and infallibilism, fail to transmit? I accept this explanatory challenge. Consider a fallibilist+ or infallibilist debating with a proponent of our modified theses, and reasoning thus: I can partially understand support for a premise — say, an experience — failing to transmit across entailments or inferences. I’ve heard talk of an abstract space of warrants and I can see that some warrants might be epistemically antecedent to others in that abstract space of warrants. So if I’m trying, without begging the question, to engage the external world sceptic, I can see that my warrant for (1) could fail to be epistemically antecedent to my warrant for (3). However, I can’t make sense of relations of epistemic antecedence between a premise ((1)) and a warrant for a conclusion ((3)). The former is a proposition I, ex hypothesi, know; the latter is warrant for a proposition I fail to know prior to inference. To talk of a relation of epistemic antecedence between a premise and warrant for a premise is close to a category error: it’s introducing an epistemic relation between two radically different things. While, for example, knowledge is an epistemic relation between a (knowing) subject and a proposition (at a time) — two very different relata if ever there were — there’s something distinctively problematic about introducing a relation of epistemic antecedence between two relata of radically different kinds. To introduce a relation of epistemic antecedence, don’t the relata have to be of the same kind?39 This, I take it, is a forceful objection that could be made by a fallibilist+ or infallibilist to (EA*), and any resultant use of (NT*) via (AW*).40

5.2 We must distinguish between a premise and knowledge of a premise. Once we do so the objection dissolves. So far I’ve played fast and loose with what form evidence for future reasoning takes (better: how we should describe such evidence) for a fallibilist+ or infallibilist. I’ve switched between (1) itself and knowledge of (1). I take it these are, for our purposes, two different descriptions of the same piece of evidence given we suppose I know (1). We can note that epistemologists who operate with the former description take (1) itself to be in one’s evidence-base for future reasoning because one (ex hypothesi) knows (1). I propose we describe (part of) the fallibilist+’s and the infallibilist’s evidence-base (source of warrant-or-evidence) for future reasoning as knowledge of (1) (rather than as (1) itself).41 Once this is done there’s no especial problem about seeing how a relation of epistemic antecedence might hold. We simply ask: does my knowing (1) presuppose or rest on my warrant for (3)? To be sure, different epistemologists will give different answers to this question, but that’s no slur on its intelligibility: indeed, it counts in its favour.42 A caveat: to require that the fallibilist+ and infallibilist describe this evidence in this way in this transmission failure debate, isn’t to require that he always so describes his warrant. Nor, relatedly, is it to say that the alternative description of the evidence is deficient.

In sum, one might have difficulty with the intelligibility of notions of epistemic antecedence or epistemic dependence, period. But I’ve shown there’s no especial problem of applying the notion of epistemic antecedence to a fallibilist+’s and an infallibilist’s account of the sources of warrant-or-evidence for future reasoning. Put differently: if one hasn’t a problem with (EA), then one shouldn’t have a problem with (EA*).

6. Conclusion

6.1 I take myself to have shown that motivated transmission failure conditions can be constructed which are applicable to — capture adequately — all, or most, plausible accounts of the sources of warrant-or-evidence for future reasoning. I haven’t shown that there are in fact genuine cases of transmission failure. I have, though, demonstrated that the fallibilist+ and infallibilist cannot evade questions of transmission failure as easily as they might have hoped. Williamson’s, Hawthorne’s, and Silins’s hard line against genuine instances of transmission failure is not warranted.43


1 An experience as of x is an experience representing x, the having of which is compatible with ~x (and, indeed, compatible with x). This may need to be finessed, depending on one’s view on object-dependent experience, to the following: an experience as of x is an experience representing x, and the having of an experience with the same conscious character is compatible with ~x (and, indeed, compatible with x). Finally, note that even when we settle on a specific possible warrant for (1), the questions may still be alleged to be badly formed on account of the notion of evidence not having been explicitly introduced (e.g. the Bayesian might ask is evidence simply a proposition on which credences can be updated by conditionalisation?).

2 “I find most alleged examples of [transmission failure] unconvincing.”

3 But cf. Silins (2007: 128, n. 27) for an evolution in his views. These three philosophers — viz. Williamson, Hawthorne, and Silins — say very different things of relevance to the issues presently under consideration; moreover, they do not always explicitly mention such issues when saying things of relevance thereto (hence my hedging device of ‘can be taken to’ in the main text). For example, here are two distinct, but closely related claims (and it is not always clear which is best ascribed to these philosophers): (1) there are no genuine instances of transmission failure (full stop); (2) it can seem as if there are cases of transmission failure, but when you look more closely you will see that these seemings are the product of a flawed view about sources of warrant-or-evidence for future reasoning.

4 Cf. Klein’s (1995) notion of ‘sources of justification’.

5 We can certainly grant to these epistemologists that there is an exceptionless notion of what we might call warrant transmission. We would then say that when other epistemologists talk about there being instances of transmission failure, this suggests that they are using a more demanding notion of warrant transmission than the exceptionless notion. And one possible more demanding requirement — our focus — would be to exclude epistemic circularity.

6 Regarding experience: cf. Conee and Feldman (2008), for whom all ultimate evidence is experiental. Regarding premises: some readers may find it hard to think of a premise in an argument as warrant-or-evidence. My parenthetical reference to a proposition may make it less hard. Regardless, I ask for patience on the part of the reader, and trust the idea will be clearer by the chapter’s end.

7 And so ‘future’ is not the temporal notion where the contrast is with ‘past’ and ‘present’ reasoning. Instead, in our paradigm case, we suppose a reasoner knows a proposition and ask what sources of warrant-or-evidence such a reasoner has at his disposal should he go on (in the future; subsequently) to perform an inference. Finally, our focus in this chapter is on future deductive reasoning.

8 And, indeed, on all — or most — plausible conceptions of evidential warrant (a term explained in 1.1 infra).

9 I call the account fallibilism— (explained in 1.2 infra). Strictly it is an account of the sources of warrant-or-evidence for future reasoning and a conception of evidential warrant. Silins’s claim receives a sharpening later (see n. 37 infra), but this version will do in the meantime. Finally, I take it this claim can be fairly ascribed to Silins. Nonetheless, should he resist it, the claim itself is worth independent exploration.

10 For Silins (2005: 74), “when [he] say[s] that someone’s warrant for believing that p is that w…[he] means that the person believes that p, and that the person believes that p on the basis of the relevant [warrant, w]”. (And so Silins appears to use ‘warrant for believing’ for doxastic, rather than propositional, warrant.) I do not take this to be orthodox usage, but we can grant it for present purposes without harm.

11 Silins pins this account on Wright (2002) and Davies (2003). It might be worth quickly pointing out what else a philosopher operating with this account can do (all assuming he meets the epistemic standards for knowing (1)). First, he can competently deduce (3) and come to believe (3) on that basis: he can allow that, in the exceptionless sense of what we might call warrant transmission (cf. n. 5 supra), the experiential warrant for (1) is transmitted to (3). Second, if such a philosopher is a dogmatist (see Pryor 2000), he can allow that, even in a more demanding sense of warrant transmission that excludes epistemic circularity, the experiential warrant for (1) is transmitted to (3). Finally — whether dogmatist of not — he can allow that, if asked why he believes (3), he can legitimately reply: well, I have hands — I know that much — and (3) follows from what I know.

12 Such fallibilists will be dogmatists. All dogmatists are fallibilists; but not all fallibilists are dogmatists. For non-dogmatist fallibilists (e.g. Wright) we must rely on the parenthetical ‘part of’: this experience is supplemented with my having an unearned warrant to accept/trust/believe that I am not a BIV etc.

13 Cf. McDowell (1982, 1986, 2008), a fact disjunctivist, and Williamson (2000: ch.1) for whom seeing that p entails knowing that p. Note that a fallibilist may also, for example, allow that sometimes, perhaps often, we see that p — that is, we enjoy evidence incompatible with ~p.

14 Note that this leaves open that, though p itself is not part of S’s evidence base for future reasoning, it is part of S’s evidence base for knowing p.

15 A note about the exhaustivity of these two theses: I take it that p either necessarily is or necessarily is not part of S’s evidence base for future reasoning. Consistently with this, contingent matters could be in play rendering p, though part of S’s evidence base for future reasoning, not exploitable (e.g. on account of epistemic circularity). (Alternatively, one might recast the second conjunct of fallibilism— as ‘p itself necessarily is not part of S’s evidence base for future reasoning’ and the second conjunct of fallibilism+ as ‘p itself is or can be part of S’s evidence base for future reasoning’.)

16 Where ‘used’ is, perhaps, elliptical shorthand for ‘used as a source of warrant-or-evidence’. And where ‘future reasoning from a known premise’ is, perhaps, elliptical shorthand for ‘future reasoning from warrant for a known premise’. Finally, such a fallibilist can allow that, if the premise, p, is not known, perhaps because it is not true, then reasoning from p to a conclusion may be above reproach and yet will not normally lead to knowledge.

17 E.g. Fred Dretske, notwithstanding that he’s classifiable as an infallibilist on a different way — present in the literature — of carving things up. Finally, I take it that dogmatists can be either fallibilists— or fallibilists+.

18 There is close to being merely logical space for infallibilism— (modulo ‘evidence base’ being used in the way I describe). At any rate, it’s difficult to discern the motivation for such a thesis. Given this, in the main text I quickly revert to simply talking of infallibilists.

19 Many issues I’ve opened up in this section require more detailed treatment. For example: What does it take — knowledge? warrant(ed belief)? truth? — for a proposition to be included in one’s evidence? Is (possession of) propositional evidence for p compatible with ~p? On some views your evidence will include the proposition that p when you know p, but you could have had that evidence even if you warrantedly, yet falsely, believed that p. Moreover, my discussion of the presented theses has proceeded in hedged terms with, for example, ‘often’, ‘typically’, and ‘may well’ in place of ‘necessarily’ at many junctures (here and in what follows). I don’t see that this is avoidable, however.

20 Rejection of (AW) is a defining feature of dogmatism. In this section we focus on fallibilists not flatly rejecting (AW): that is, fallibilists allowing that (MOORE)-reasoning might involve epistemic circularity, and so might suffer from transmission failure. Indeed given the chapter’s overall focus is not to claim there are in fact genuine instances of transmission failure, but merely to render it possible on all, or most, plausible accounts of the sources of warrant-or-evidence for future reasoning, we do not definitively adjudicate on (AW) (or its cognate, (AW*) — see 3.2 infra) at all in this chapter.

21 This failure of transmission is, trivially, not a failure of transmission in the exceptionless sense (cf. n. 5 supra). It is failure of transmission in the more demanding sense that requires that there not be epistemic circularity. And it seems that this remains the situation even if we regard the warrant for (1) as being constituted by both the experience as of having hands and the antecedent warrant for (3).

22 Both of these fallibilists should agree that, provided that one knows (1), one will do well to believe (3), since it follows from (1) by a palpably valid inference (cf. n. 11 supra). (Of course, if one gets good evidence against (3), then the palpably valid inference may have to be run in the opposite direction.) But, because of possible epistemic circularity, there is more to be said about why one does well to believe (3). The argument via (1) and (2) appears to involve a detour and, in order to understand one’s warrant to believe (3) after the inference, one needs to articulate the nature of the antecedent warrant — that is, one’s warrant to believe (3) even before the inference.

23 As will become clear, here (and in what follows), when I talk of ‘complete’ transmission failure I am bracketing any exceptionless notion of warrant transmission (cf. n. 5 supra) and focusing on a more demanding notion. If there is, in this sense, ‘complete’ transmission failure, it means all, and not just some, of one’s sources of warrant-or-evidence for future reasoning fail to transmit. I explicitly expand what I call the standard account of warrant in 3.2 infra.

24 This is evidence that philosophers operating with (NT) are (closet) fallibilists—.

25 Our excursus into conceptions of evidential warrant was still useful. First, standard taxonomies foreground conceptions of evidential warrant: so this excursus helps locate our taxonomy in relation to standard ones. Second, while there are no entailments in play from one’s conception of evidential warrant to one’s account of the sources of warrant-or-evidence for future reasoning (or conversely), we might think that the former helps explain the latter.

26 Or a divide is effected between forms of internalism (typically associated with fallibilism) and externalism (typically associated with infallibilism). Effecting such a divide will not serve present purposes.

27 The notion of non-transmission of warrant in play here is still based on epistemic circularity. Ideally this condition would also be applicable to fallibilists—. But this seems unrealistic. Nonetheless, a desideratum of parsimony (viz. ceteris paribus, construct as few theses as possible) operates in what follows.

28 For philosophers answering the closure question: ‘yes’, cf. Dretske (1970, 1971, 2005a, 2005b), Nozick (1981), and Heller (1999). For remarks on how one’s operative account of the sources of warrant-or-evidence for future reasoning affects one’s answer to the closure question, see Klein (1995).

29 ‘Warrant’ is hereinafter used in this broad sense unless indicated otherwise.

30 Objection: if knowledge of a premise is conditional on antecedent warrant for the conclusion then this will be because epistemic warrant for the premise is conditional on antecedent warrant for the conclusion. It is therefore unclear why we need to introduce (NT*). Reply: insofar as this objector is adopting my expanded account of warrant, I agree with the premise but not the conclusion. Only (NT*) allows us to make explicit the pivotal relation of epistemic antecedence between a (known) premise and a standard warrant (see sections 4 and 5 infra). Finally, insofar as warrant now includes known premises, it also a fortiori includes doxastically justified premises (cf. Tucker (2010), sections 2.1 and 3.1).

31 My focus in this section (and those to come) is on known premises. What happens when, say, the premise is false but warrantedly believed (cf. n. 19 supra)? I bracket this question, but see, inter alia, Klein (2008: 25–61).

32 Cf. Pryor (2000: 525). Should the left-hand-side of (EA) strictly be: it’s not the case that one’s warrant for accepting q is epistemically antecedent (or prior) to one’s warrant for accepting p? Couldn’t the right-hand-side be satisfied and one’s warrant for p be epistemically neither-prior-to-nor-posterior-to one’s warrant for q? Finally, note that we can also, I take it, say: one’s warrant for p1 is antecedent to one’s warrant for p2 iff one’s warrant for p2 is dependent on one’s warrant for p1 and one’s warrant for p1 is not dependent on one’s warrant for p2.

33 Two related points. First, to see (EA)’s relevance to (AW), it is best to revert to n. 32’s left-hand-side: if warrant for (1) presupposes or rests on one’s warrant for (3), then it follows from (EA) that one’s warrant for accepting (3) is epistemically antecedent (or prior) to one’s warrant for accepting (1). Second, we need not find this notion of one reason presupposing or resting on another perfectly perspicuous; it’s left intuitive.

34 An infallibilist might claim that one’s evidence for (1) consists of knowledge of (1) and knowledge of one’s having an experience as of having hands. If he does so, the points I make in relation to the fallibilist+ here can be marshalled, mutatis mutandis, with respect to this infallibilist. But this is — for good reason — currently an unpopular way to be an infallibilist, so I’ll ignore it in what follows. An infallibilist will typically only claim that one’s evidence for (1) consists of knowledge of one’s having an experience as of having hands in the bad case — that is, when ~(1) — but our principal focus in this chapter is not on bad cases.

35 There’s logical space for a fallibilist+ to claim that his knowledge that (1) trumps or wipes out any non-entailing evidence he has for (1). But I find this a peculiar way of carving things up. To complete the picture, for our fallibilist— neither all that one knows, nor only what one knows, is in one’s evidence base for future reasoning.

36 Or the infallibilist (cf. n. 34), but the fallibilist+ is the philosopher driving our present concern.

37 Recall that the fallibilists+ allows that when S knows p, p itself is part of S’s evidence base for future reasoning. It is thus not a flawed account of the sources of warrant-or-evidence for future reasoning by the lights of Silins’s claim (cf. 0.4 supra). It should, however, be clear that we would not have answered Silins’s claim by showing that the warrant the fallibilist+ can share with the fallibilist— may fail to transmit. To make the dialectic clearer we may wish to sharpen Silins’s claim thus: instances of complete transmission failure only arise on a particular, and putatively flawed, account of the sources of warrant-or-evidence for future reasoning (cf. n. 23 supra).

38 Mutatis mutandis for reasons: some of one’s reasons for (1) may presuppose or rest on one’s reasons for (3) without other of one’s reasons for (1) presupposing or resting on one’s reasons for (3).

39 Cross-categorial relations are routinely posited in philosophy. An example from metaphysics: the truthmaker relation has been posited between a (non-propositional) truthmaker and a (true) proposition (see Armstrong 2004). So the objection in the main text is a specific objection focused on the relation of epistemic antecedence.

40 A trio of theses such as these is a (prominent) way to argue for transmission failure; it may not be the only such way (cf. 2.1 supra).

41 Why, then, did I play fast and loose with my description of the evidence? First, I want to leave open (cf. n. 19 supra) that something short of knowledge of a proposition can suffice for a proposition to be in one’s evidence base for future reasoning. Second, I want to keep vivid, with Silins, the thought that, upon competent inference from a known premise, the premise itself can become one’s warrant to believe the conclusion.

42 Indeed, even the following extreme response, I take to count in its favour: once I know (1), never mind how, I’ll be able to exploit (1) as evidence for (3).

43 It is a further task — one not attempted in this chapter — to show why, for example, an infallibilist might plausibly take (MOORE)-reasoning to suffer from some kind of epistemic circularity. Finally, see Tucker (2010: esp. 517ff.) for another hard line against transmission failure, although it is different in form from those targeted in this chapter, in not (as I understand it) being reliant on any particular conception of evidence.