Orthodox Reliabilism and the Epistemic Significance of Testimony
Orthodox Reliabilism and the Epistemic Significance of Testimony
Abstract and Keywords
This chapter explores a position – “Orthodox Reliabilism” – that combines Process Reliabilism with Process Individualism, the thesis that belief‐forming individualistic assumption about the extent of belief‐forming processes. In particular, it examines the options available to this position on the epistemology of testimonial beliefs. It argues that the Orthodox Reliabilist is committed to a very traditional conception of the epistemological significance of testimony, according to which particular testimonies and their properties (including their reliability properties) are to be regarded as features of the hearer's local environment. On such a view, unreliable testimony can constitute a Gettier condition, and so can undermine a subject's claim to knowledge; but this exhausts the epistemic significance of the testimony.
Chapter 1 was the first of five chapters I will be devoting to the nature of epistemic reliance in testimonial belief‐fixation. I concluded by highlighting the reliability of the cognitive processes implicated in the production of the testimony. My thesis was ERTK: whether a hearer's testimonial belief amounts to testimonial knowledge depends on the reliability of cognitive processing taking place in the mind/brain of her source. Such a conclusion entails the falsity of the doctrine of Knowledge Individualism.
In this chapter, I explore how these results bear on another individualistic doctrine—one that is explicitly endorsed by prominent reliabilists. According to this doctrine, which I will label “Process Individualism,” the cognitive processes implicated in the formation or sustainment of a subject's beliefs all take place within that subject's own mind/brain. (I will use “Orthodox Reliabilism” to designate the combination of Process Reliabilism and Process Individualism.) My thesis is that the Orthodox Reliabilist can and should embrace ERTK, but that the combination of these views requires a commitment to a very traditional conception of testimony's epistemic significance, according to which particular testimonies and their properties (including their reliability properties) are to be regarded as features of the hearer's local environment. On such a view, testimony that is apparently reliable but de facto unreliable (p.37) (in the manner of T2 from the previous chapter) can constitute a Gettier condition, and so can undermine a subject's claim to knowledge; but this exhausts the epistemic significance of the testimony.
Chapters 3 and 4 are devoted to arguing against this traditional conception of testimony's epistemic significance. However, it will be helpful to appreciate why the Orthodox Reliabilist is committed to this conception, and what form this conception takes within the Orthodox Reliabilist framework. That is the aim of this chapter.
Since the pressure on the Orthodox Reliabilist to endorse this conception comes from ERTK itself, I begin, in Section 1, by noting that most epistemologists will regard ERTK itself as a humdrum claim—one which reflects an already widely acknowledged point about the role of local factors in ascriptions of knowledge. In Section 2, I argue that such a reaction to ERTK is mandatory for any proponent of Orthodox Reliabilism. Finally, in Section 3, I suggest how the Orthodox Reliabilist will employ the distinction between global and local reliability, in order to capture the traditional conception's idea that testimony and its reliability properties are features of the subject's local environment. The result, I argue, is that the Orthodox Reliabilist is committed to the Local Reliability Hypothesis (LRH), according to which facts about the reliability of a piece of testimony, by themselves, can affect only the local reliability, not the global reliability, of beliefs based on that testimony.1
This chapter concludes, then, by discussing how the Orthodox Reliabilist will employ the distinction between global and local reliability in her factorization of the reliability of testimony‐based (p.38) belief. This topic, which will be the focus of the next two chapters, can seem somewhat esoteric. Not only does it presuppose the significance of a distinction—that between local and global reliability—that is not without its own problems (see Gendler and Hawthorne 2005), and which in any case is not universally endorsed by Process Reliabilists (see Greco 2010). What is more, the issue regarding how to factor the reliability of testimonial belief would appear at first blush to be a minor “accounting” matter—an in‐house dispute among a certain subset of reliabilists, to be sure, but a dispute that does not tell us very much about the nature of our epistemic reliance on others. An underlying aim of this chapter is to show that these appearances are misleading. It is for this reason that I stress that the Local Reliability Hypothesis is the Orthodox Reliabilist's version of a very traditional conception of the epistemic significance of testimony—a conception that is prevalent among epistemologists generally (not just among reliabilists). The main conclusion of this chapter—that Orthodox Reliabilism is committed to LRH—is meant to pave the way for the subsequent two chapters, where I will be targeting the Local Reliability Hypothesis, in an indirect attempt to weaken the traditional conception's grip on our epistemological imagination.
Chapter 1 concluded that the acquisition of testimonial knowledge depends on facts regarding cognitive processing in the mind/brain of one's source. In particular, my claim was that:
ERTK Whether a testimonial belief amounts to testimonial knowledge depends on the reliability of cognitive processes implicated in the production of the testimony.
In fact, it can seem that ERTK is a special case of an already very familiar point having to do with the role of local features in ascriptions of knowledge. To see this, consider a doppelgänger scenario involving the familiar ‘stopped clock’ case. Let S and S* be doppelgängers, each of whom is presently observing a different clock: S is observing clock C1, while S* is observing clock C2. Both clock faces read 2 o'clock, so both S and S* come to believe that it is 2 o'clock. In fact it is 2 o'clock; but it turns out that, while C1 was functioning properly, C2 has been stopped at 2 o'clock for weeks. The following verdict then seems apt: S's true belief that it is 2 o'clock amounts to knowledge, while S*'s does not. We would thus appear to have a clock‐related analogue to ERTK: whether a belief formed through trusting a clock amounts to knowledge depends on the reliability of the clock at the time in question. And, since the stopped clock case of S and S* appears to mirror the testimonial case of H and H* from Chapter 1, it would appear further that there is nothing particularly special about testimony: like stopped clock cases, cases involving unreliable testimony merely illustrate a general point, already widely acknowledged, regarding the potential role of local features in undermining ascriptions of knowledge.
Before examining why epistemologists might analyze these cases in this way, it is worth acknowledging that the analogy between unreliable testimony and a stopped clock is not perfect. There is, after all, one obvious difference between getting testimony from a person and reading a clock face. In the former case the stimulus (= the testimony) is itself the ‘output’ of a cognitive process, whereas in the latter case the stimulus (= the arrangement of the hands on the clock's face) is not. Consider in this light the (p.40) following argument, which might be used to call into question the legitimacy of the analogy:
The crucial question, then, is whether facts about the reliability of a piece of testimony are suitably ‘external’ to the process by which, or the basis on which, testimonial beliefs are formed.
In the case where the stopped clock happens to give the correct time at the moment S* observes it, the proper verdict is clear: S*'s true belief is justified, but fails to amount to knowledge. Here, the bearing of the scenario on knowledge is clear: the fact that the clock was not properly functioning makes the truth of S*'s belief too much a matter of luck to count as knowledge. And the irrelevance of the scenario to justification is also clear: facts regarding the functioning of the clock are irrelevant to the issue of justification, since these facts are ‘external’ to to the process through which the belief is formed, to the grounds on which the belief is based, and so forth—and so these facts are ‘external’ to whatever it is that we assess when we assess the justification of S*'s belief. But—moving over to the case where H* forms a belief through accepting T2 (an unreliable piece of testimony which happens to be true)—it is not clear that we should support a verdict of ‘true and justified but not knowledge’ regarding H*'s testimonial belief. The bearing of the scenario on knowledge is clear: the fact that the testimony was unreliable makes the truth of H*'s belief too much a matter of luck to count as knowledge. What is not clear is the irrelevance of the scenario to justification: it is not clear that facts regarding the production of a piece of testimony are ‘external’ to the process through which the belief is formed, to the grounds on which the belief is based, and so forth. And if the facts regarding the production of testimony are not suitably ‘external’ in the relevant sense, then the analogy between stopped clocks and unreliable testimony is untenable.
At this point, the defender of the analogy between stopped clocks and unreliable testimonies can defend the viability of the analogy by appeal to a very traditional conception of the epistemic significance of testimony. According to this conception, a piece of testimony and its features (including its reliability features) are no different from any other external feature(s) of a subject's local environment. From this vantage point the analogy between (p.41) stopped clocks and unreliable testimonies remains as strong as can be: facts regarding the production (and so the reliability) of the testimony, like facts regarding the proper functioning of a clock, are relevantly ‘external’ to the materials that make for the doxastic justification of corresponding (testimony‐ or clock‐based) beliefs. The epistemic significance of such features (the functioning of the clock, the reliability of the testimony) is that of a potential Gettier condition: such features might conspire in a given case to undermine the subject's claim to know, but this is the extent of their epistemic significance.
To see why one might propose to treat testimony and its properties (including its reliability properties) in this way, we do well to review why it is natural to treat non‐testimonial Gettier cases in this way. Why are the familiar examples of Gettier‐type luck—stopped clock cases, Gettier's own case of NoGot and Havit, and so forth—standardly taken to show something about knowledge and knowledge alone? Although the standard treatment is perhaps not as clearly articulated and defended as one might like, something like the following picture appears to be in play.
To be clear, this picture does not tell us that, for purposes of assessing a belief's doxastic justification, we can bracket the “external” world (p.42) entirely. What the standard treatment of Gettier cases assumes, rather, is that, for the purpose of assessing the doxastic justification of a belief, we can ignore those highly contingent features of our local environment: the proper functioning of a clock, the actual distribution of Fords among members of the office, the prevalence of fake barns in the local vicinity, and so forth.3 The idea is that these features are irrelevant to an assessment of the goodness of the subject's grounds, or her belief‐forming process, or what‐have‐you—irrelevant, that is, to the materials that make for doxastic justification.
Knowledge requires compliance with the relevant portion of the world: assessing whether a belief amounts to knowledge involves assessing whether the truth of the belief is a matter of luck (given how or on what basis the belief was formed). But doxastic justification is primarily a matter of how or on what basis the belief was formed, and in assessing this we can bracket the highly contingent antics of the “external” world: these antics are irrelevant to how or on what basis the belief was formed.2
What I am calling the “traditional” conception of testimony's epistemic significance would treat the reliability‐making features of particular testimonies in precisely the same way. That is, the de facto reliability of a piece of testimony—whether it is actually reliable—is to be regarded as a candidate Gettier condition. It is something that can defeat a subject's claim to know, but it leaves considerations regarding doxastic justification untouched. To be sure, a testimonial belief formed through accepting unreliable testimony can fail to be doxastically justified, as, for example, when the hearer was not sufficiently sensitive to signs of mendacity or incompetence in the speaker. The point is merely that, in those cases, the testimonial belief's status as unjustified reflects an assessment of goings‐on in the mind/brain of the hearer—and this assessment ignores the de facto unreliability of the testimony itself.
To those who think about testimony and its features (including its reliability features) in this way, the analogy between unreliable testimony and a stopped clock remains an excellent one: a testimony's de facto reliability, like a clock's de facto reliability, are features of the scenario that are relevantly “external” to the way or the grounds on which the belief was formed. It is for this reason that the de facto reliability of testimony, like the de (p.43) facto proper functioning of a clock, can be safely ignored when it comes time to assessing the doxastic justification of beliefs formed through accepting that testimony. This is according to what I am calling the traditional conception of testimony's epistemic significance.
As I say, I think that this conception is broadly popular: it is endorsed by epistemologists of all stripes, no matter what their background ideology is regarding the nature of doxastic justification. To those epistemologists who endorse this conception, ERTK is a humdrum claim that illustrates a familiar point about external factors playing the role of Gettier conditions.
How should one think of these matters if one is a Process Reliabilist? I suspect that, like epistemologists generally, most Process Reliabilists will find themselves attracted to the traditional conception of testimony's epistemic significance, and so will regard ERTK in the manner described above. In this section I explain why. To anticipate: it turns out that there is a traditional assumption that Process Reliabilists make, which, once made, renders the traditional conception (and its understanding of ERTK) all but mandatory.
The assumption in question was announced, without much fanfare, in Goldman's (1979) paper “What is Justified Belief?”4 The key passage is one in which Goldman introduced his pre‐theoretical understanding of justified belief. He wrote:
For present purposes, it is the second sentence that is the crucial one. In it, Goldman is giving voice to a view that I will label Process Individualism:
A justified belief is, roughly speaking, one that results from cognitive operations that are, generally speaking, good or successful. But “cognitive” operations are most plausibly construed as operations of the cognitive faculties, i.e., “information‐processing” equipment internal to the organism. (Goldman 1979/2000: 346–7; second italics added) (p.44)
PI For every subject S, all of the cognitive processes implicated in the formation or sustainment of S's beliefs are cognitive processes that take place within S's own mind/brain.
It is easy to appreciate why those Process Reliabilists who endorse PI have a strong motive to embrace the traditional conception of the epistemic significance of testimony. In Chapter 1, I argued that the reliability of a piece of testimony is a matter of the reliability of the cognitive processes through which the testimony was produced. These processes take place in the mind/brain of the source speaker, and so are “external” to the ‘information‐processing equipment’ of the hearer. The result is that, by the lights of PI, they are ‘external’ to the cognitive processes through which the testimonial belief is produced. But, according to Process Reliabilism, doxastic justification is a matter of the reliability of the cognitive processes that are responsible for the production of the belief. It follows that facts about the reliability of testimony, including facts about the reliability of the processes that produced the testimony, are irrelevant to the doxastic justification of the resulting testimonial belief. And herein lies the pressure on the Orthodox Reliabilists to embrace the traditional conception of testimony's epistemic significance: insofar as the Orthodox Reliabilist admits that facts about a testimony's reliability have some epistemic significance—and surely this much must be admitted—she must regard that significance as restricted to be that of a feature of the subject's local environment. This is precisely as the traditional conception would have it. (p.45)
It should be clear, then, that the Process Reliabilist's move to endorse Orthodoxy is a significant one. For what we have just seen is that the reliabilist who endorses PI in effect restricts the epistemic significance she can ascribe to facts regarding the de facto reliability of testimony. It is worth underscoring that the Process Reliabilist is by no means forced to take on this restrictive view. To see this, suppose that we endorse the point behind Goldman's remark that “A justified belief is, roughly speaking, one that results from cognitive operations that are, generally speaking, good or successful.” Even so, we might go on to think that, at least in testimony cases, the relevant cognitive operations are distributed: while some of those processes take place in the mind/brain of the subject who forms the testimonial belief (i.e., in her comprehension and endorsement of the testimony), some of those processes do not. In particular, there are cognitive processes that take place in the mind/brain of the speaker (i.e., in the production of the testimony that was consumed) that are also relevant. A Process Reliabilist who approaches things this way will reject the analogy with the stopped clock case, for the simple reason that the processes involved in the proper functioning of a clock are not themselves cognitive processes. Such a position may or may not be an attractive one; my present point is merely that such a position is available to Process Reliabilists. Nothing in the Process Reliabilist position per se forecloses on this option.
Nevertheless, it is clear from Goldman's characterization of the relevant cognitive operations, as “ ‘information‐process equipment’ internal to the organism,” that he himself would reject this option out of hand. What is more, he does so for reasons that are highly programmatic in nature: they reflect views regarding how we ought to think about the processes that are relevant to epistemic assessment. To get a better appreciation for Goldman's reasons on this score, we should expand our focus a bit, and consider the extended passage from which the above quote is taken. In this passage, Goldman is offering what to my mind is the first statement and defense of an (p.46) individualistic conception of belief‐forming processes ever offered by a reliabilist:
Here it appears that Goldman's commitment to Process Individualism derives from his more general commitment to excluding any causal process that takes place “outside the organism,” where this is understood to be any process that is not “within the organism's nervous system.” Although matters are not entirely clear, Goldman appears to endorse this restricted focus on the grounds of a programmatic conception of justification, on which “Justifiedness seems to be a function of how a cognizer deals with his environmental input, i.e., with the goodness or badness of the operations that register and transform the stimulation that reaches him.”6 (p.47)
Clearly, the causal ancestry of beliefs often includes events outside the organism. Are such events to be included among the “inputs” of belief‐forming processes? Or should we restrict the extent of belief‐forming processes to “cognitive” events, i.e., events within the organism's nervous system? I shall choose the latter course, though with some hesitation. My general grounds for this decision are roughly as follows. Justifiedness seems to be a function of how a cognizer deals with his environmental input, i.e., with the goodness or badness of the operations that register and transform the stimulation that reaches him. . . . A justified belief is, roughly speaking, one that results from cognitive operations that are, generally speaking, good or successful. But “cognitive” operations are most plausibly construed as operations of the cognitive faculties, i.e., “information‐processing” equipment internal to the organism. (Goldman 1979/2000: 346–7; second italics added)5
It is worth noting that in the (1979a) quote just cited, Goldman endorses PI “with some hesitation.” By the time of his ground‐breaking (1986), however, Goldman is no longer hesitant in his commitment. Setting up his response to the generality problem, Goldman asks: “But how is it determined, in each specific case, which process type is critical?” His response to this question is instructive:
Here Goldman's reasoning appears to be that belief‐forming processes are to be individuated “internally” (I would say ‘individualistically’) because (i) they are psychological processes that (ii) are causally operative in producing the belief token.
One thing we do not want to do is invoke factors external to the cognizer's psychology. The sorts of processes we're discussing are purely internal processes. (1986: 51; italics added)
Goldman is not unique among prominent epistemologists who endorse Process Individualism on broadly programmatic grounds. Another epistemologist to follow Goldman in this regard is William Alston. Speaking of the extended quote from Goldman (1979) cited above, Alston (1995) wrote:
And, in a paper in which he is addressing himself to the relevance of “the social” to assessments of reliability, Alston reiterates that:
This seems to me just the right thing for a reliabilist to say on this point. If the epistemic status of a belief is a function of the reliability of the process that generates the belief, it is the reliability of the psychological process that is crucial. (1995/2000, 360; italics in original)
Alston's insistence on PI is all the more striking given that he goes on to acknowledge the “essentially social processes” (48) involved in the sort of cooperative inquiry and transmission of information found in, for example, the scientific enterprise (1994: 45).8 This gives us a clear indication that Alston, like Goldman, endorses PI on the grounds that belief‐forming processes are psychological processes, and that psychological processes are processes that do not extend beyond the (physical) boundaries of individual subjects.
So long as we are describing and analyzing doxastic mechanisms we are confined to individual psychology; we are studying the internal cognitive structure and processes of individual human beings. (Alston 1994: 30; italics added)7 (p.48)
The position being articulated in the quotes by Goldman and Alston advocates a combination of Process Reliabilism and Process Individualism. Although few other Process Reliabilists are as explicit as Goldman and Alston have been with respect to PI, I suspect that the reasons Goldman and Alston have offered are widely endorsed, and that this combination of Process Reliabilism and Process Individualism is widely taken to be “the” Process Reliabilist position. It is for this reason that I use the label “Orthodox Reliabilism” to designate any Process Reliabilist position that endorses PI.
We can now return to the main issue before us, namely, how those Process Reliabilists who are Orthodox Reliabilists will react to ERTK. I submit that such reliabilists will view ERTK in the manner described above, as a special case of the general point that external “local” features of one's environment can be relevant to ascriptions of knowledge, in the manner of a Gettier (p.49) condition. In effect, the commitment to PI forces the Orthodox Reliabilist's hand on the matter: she has no choice but to endorse the traditional conception of testimony's epistemic significance. The claim I want to bring out on this score is that, by the lights of Orthodox Reliabilism, two testimony cases that differ only in the de facto reliability of the testimony each hearer consumed will be regarded by the Orthodox Reliabilist as exhibiting a merely local environmental difference. To establish this, take any case in which a subject forms a testimonial belief through having accepted a particular piece of testimony. Now construct a variant case involving a hearer who is an intrinsic duplicate of the hearer in the original case, but where the testimony in play, though subjectively indistinguishable to the hearer, differs in its de facto reliability from the reliability of the testimony in the original case. We will imagine that this difference in testimonial reliability is the only difference between the two cases: in all other respects they are relevantly alike. In particular, the two hearers reside in the same community, and so encounter roughly the same proportion of reliable testimonies, employ the same processes for discerning reliable testimony, and so forth. Then we can say the following: whatever epistemic difference there is between the hearers' testimonial beliefs in such cases cannot be traced to a difference in the doxastic justification of these beliefs. This is because, by hypothesis, the hearers are doppelgängers, and so they employ the same belief‐forming process‐type; and they live in the same community, and so are in scenarios that share the relevant “background conditions.” Process Reliabilism thus requires that they be treated as alike, justification‐wise. Given this, it would seem that the only relevant difference that can be acknowledged, once one has assumed both Process Reliabilism and Process Individualism, is a relevant difference in the subjects' respective local environments. The clearest illustration of this is in a case like that of H and H* from Chapter 1, where one subject knows, and the other doesn't, merely in virtue of differences in the de facto reliability of the testimonies each consumed. But the point in question is perfectly general: the Orthodox Reliabilist (p.50) must regard any two cases that differ merely in the de facto reliability of the testimonies consumed as differing merely with respect to features of the relevant local environments. This is the basis for my contention that Orthodox Reliabilists must view the epistemic effects of (apparently reliable but actually) unreliable testimony as a special case of the knowledge‐undermining effects of “local environmental” factors.
I doubt whether any of this will come as a surprise to Process Reliabilists. On the contrary, I suspect that those Process Reliabilists who embrace Orthodoxy will be happy to endorse the idea that a testimony's de facto reliability is to be regarded as a feature of the hearer's local environment. In this final section, I want to acknowledge that the resulting Orthodox Reliabilist position is not without its virtues. In particular, the resulting position can exploit the distinction between local and global reliability, to good effect.
It is easy to appreciate why an Orthodox Reliabilist should want to exploit this distinction. Return to the case of Wilma in GOOD and in BAD from Chapter 1. In both, she accepts testimony from Fred, under conditions in which the testimonies appear exactly the same to her, where the cognitive processes that take place in her mind/brain in the two cases are type‐identical. The only difference between the cases resides in the testimonies: Fred's testimony in GOOD was knowledgeable, his testimony in BAD was not (and so was true only as a matter of accident). Regarding cases like these, I submit that all reliabilists—stronger, all epistemologists, regardless of their favored theory of knowledge—should want to acknowledge the following point:
(*) From the point of view of their reliability profiles, Wilma's testimonial belief in BAD compares unfavorably to her testimonial belief in GOOD. (p.51)
First, some background on the distinction between local and global reliability. The distinction itself was introduced for reasons having nothing to with the testimony cases. Following Goldman's seminal work in epistemology in the 1970s and 1980s,10 Process Reliabilists were quick to acknowledge that local environmental factors can conspire to render an otherwise reliable belief‐forming process less reliable in a particular context. The subject in Fake Barn County is perhaps the paradigmatic example of this phenomenon. (p.52) As an ordinary subject who has been exposed to a wide variety of barns and who can recognize them by their visual appearances, such a subject can be said to possess a highly reliable perceptual process of barn‐recognition. But this process will be much less reliable when she is traveling through the countryside of Fake Barn County, where virtually every barn‐looking structure is a mere facade, a well‐constructed “fake barn.” In cases like this, a process‐type that in more ordinary circumstances is highly reliable would yield a much higher proportion of false belief in local circumstances.
The distinction between global and local reliability was introduced to enable the Process Reliabilist to represent cases like this. The claim is that the mature subject's barn‐beliefs in Fake Barn County is formed through a process‐type—forming barn‐beliefs on the basis of barn‐like visual appearances—that is globally reliable but (when employed in Fake Barn County) locally unreliable. The global reliability of a process‐type is a matter of the preponderance of true belief produced by this process‐type across the actual range of situations in which tokens of that process‐type are ordinarily used; whereas the local reliability of a process‐type is a matter of the preponderance of true belief produced by this process‐type in environments relevantly like the present one.11 Beliefs formed in Fake Barn County are globally reliable, since they are formed through a process that would produce a preponderance of true beliefs in normal contexts; yet these beliefs are locally unreliable since the process in question would not produce a preponderance of true beliefs in contexts relevantly like the present one.
Here, then, we have one advantage that accrues to any Process Reliabilist who employs the distinction between global and local reliability: she can handle a certain range of problem cases in a natural way. But there is a second advantage that accrues to the Process Reliabilist who exploits the distinction between global and local reliability, which is that it provides her with a way (p.53) to understand epistemic (doxastic) justification in reliabilist terms. Once again, the intuitive point can be made in connection with fake barn cases. The prevalence (or not) of fake barns in the local landscape of an unsuspecting subject does not seem, intuitively, to bear on whether her barn‐belief, formed on the basis of the perceptual experience as of a barn, was justified. In particular, facts about the prevalence of fakes in the vicinity are “external” to the considerations pertaining to the process through which the subject's barn‐belief was formed, and so are irrelevant to doxastic justification as the Process Reliabilist conceives of this. At the same time, the prevalence of fakes does seem to bear on whether this subject's barn‐belief counts as knowledge, since it bears on the issue of how lucky it was that the subject's belief was true, given that she formed it as she did (through the process‐type in question). This motivates the addition of a ‘local reliability’ condition on knowledge: if a belief is to amount to knowledge, then in addition to being true and formed and sustained through a globally reliable process, the belief must also be locally reliable—formed and sustained through a process‐type that would produce a preponderance of truth in situations relevantly like the one in which the belief was formed.12
Return now to how this distinction might be exploited by the Orthodox Reliabilist in her account of the epistemology of testimony. We already saw in Section 2 that, having embraced PI, the Orthodox Reliabilist is forced to endorse the traditional conception of testimony's epistemic significance, and so is forced to regard the reliability of a piece of testimony as a feature of the hearer's local environment. For this reason, the Orthodox Reliabilist cannot regard unreliable testimony as such to have any effect on the sort of reliability that makes for a doxastically justified testimonial belief. Still, it would seem that a testimonial belief formed through accepting unreliable testimony is less well‐off, reliabilistically speaking, (p.54) than is the testimonial belief of a doppelgänger which is formed through accepting reliable testimony. ((*) is a special case of this claim.) If the Orthodox Reliabilist wants to acknowledge this, as I said she should, she must distinguish the sort of reliability that makes for the doxastic justification of testimonial belief, from the sort of reliability on which unreliable testimony can affect the reliability of the corresponding testimonial belief. The distinction between local and global reliability recommends itself for precisely this reason: the Orthodox Reliabilist can say that the reliability of testimony is relevant to the local reliability, but not to the global reliability, of the resulting testimonial belief.13 Seen from this perspective, the case of the hearer who accepts true, apparently reliable but actually unreliable testimony is a testimonial version of that familiar phenomenon whereby a globally reliable process yields a true belief which nevertheless fails to amount to knowledge owing to local conditions. A verdict of ‘globally reliable but locally unreliable’ can seem as apt here, as it is in the case of the unsuspecting subject in Fake Barn County, and as it is in the ‘stopped clock’ case.
I have little doubt but that this is the standard way to think about the relevance of a testimony's reliability to the reliability of those beliefs that are formed through accepting that testimony. At the same time, I believe that this standard view is the expression of a fundamentally flawed conception of the epistemic significance of testimony. Since this standard view derives from the particular hold that Process Individualism has on our epistemological imagination, it will take some work to dislodge this view. In the two chapters to follow, I attempt just this.
In this chapter I have argued that the Orthodox Reliabilist is committed to a very traditional conception of testimony's epistemic (p.55) significance, and that as a result she is committed to the Local Reliability Hypothesis:
LRH A testimony's degree of reliability affects the local reliability, not the global reliability, of the belief formed through accepting that testimony.
The time has now come for me to begin to push back against the traditional conception. To do so, I will be taking aim at the idea that testimony and its reliability properties are epistemically relevant only as a potential Gettier condition. While LRH gives expression to the Orthodox Reliabilist's version of this idea, my claim will be that LRH does not square with the nature of our epistemic reliance in testimony cases. In Chapter 3, I will argue for this on the grounds that testimonial belief‐formation is best seen as a belief‐dependent process, where testimony is the input into this process. As we will see, the reliability verdicts sanctioned by such a view are different from those required by LRH.
(1) LRH does not imply that a testimonial belief based on unreliable testimony is always merely locally unreliable. What it implies, rather, is this: insofar as a testimonial belief is globally unreliable, this is not in virtue of the unreliability of the testimony. (By the lights of Orthodox Reliabilism, the degree of global reliability of a testimonial belief is always a matter of the degree of reliability of the process‐type by which the testimonial belief was formed—a process‐type that is presumed to be individualistic, and hence having nothing to do with the reliability of the testimony.)
(2) To say that we can ignore such features is compatible with saying that we cannot ignore other “external” features of the environment, such as those invariant, or in any case more or less enduring, features of our environment—e.g., those features that ensure that our perceptual processes regularly deliver true representations of our surroundings. (One thinks here of the various assumptions about object permanence, the rigidity of object boundaries, illumination from above, etc., that might well be hard‐wired into our perceptual system.)
(5) Goldman's use of ‘internal’ here is unfortunate, since it is not meant in the sense in which people speak of ‘internalist’ positions in epistemology these days. But his idea is clear enough.
(6) I note in passing the point I made above: in testimony cases, at least, Goldman's programmatic point here is in some tension with his point that “A justified belief is, roughly speaking, one that results from cognitive operations that are, generally speaking, good or successful.” After all, in testimony cases one might well regard the cognitive processes that produced the testimony to be among those whose “goodness” ought to be assessed in assessing the goodness of the cognitive processes that produced the testimonial belief. I will return to this point in Chapters 3 and 4, where I will be arguing that Goldman's delimiting of the ‘cognitive’ domain as that domain ‘within the organism's nervous system’ begs an important question in the current discussion.
(7) It is noteworthy that Alston's actual quote here is susceptible to two very different readings. On one reading, the claim that “we are confined to individual psychology” amounts to the claim that every belief‐forming process is a process the entirety of which takes place in a single individual subject's mind/brain. On the other reading, it amounts to the claim that only that which is part of the psychology of some individual or other can be part of the reliabilist assessment of a belief‐forming process. As I will go on to argue below, the latter reading is compatible with the thesis for which I will be arguing in the latter half of this chapter, and then throughout Chapters 3–5. But it is clear from the context that Alston has the former, stronger, reading in mind.
(8) It is also noteworthy that Alston's 1994 defense of the epistemic relevance of “the social” aims to establish the relevance of social considerations to the case for thinking that sense perception is reliable. It is no part of his argument that social considerations themselves are relevant to the reliability of some of our belief‐forming and ‐sustaining processes (except insofar as the reliability of social processes itself is assumed by the case for thinking that sense perception is reliable).
(9) The idea is that acknowledging (*) is a desideratum on any reliabilist account of the epistemology of testimonial belief: a reliabilist account that fails to be able to acknowledge (*) may not be doomed, but such an account would have a strike against it.
(12) One interesting issue regards the relation between the local reliability requirement on knowledge, and the anti‐Gettier condition on knowledge: does the satisfaction of the former ensure the satisfaction of the latter? I think not, but I do not pursue the matter further here (although it comes up again, briefly, in Chapter 4).
(13) Some Orthodox Reliabilists may endorse this because they think that unreliability in testimony has no effect on the reliability profile of the testimonial belief.