1 Introduction

This paper is about two seemingly separate questions in epistemology and about whether giving a particular answer to one commits us a corresponding answer to the other. These questions concern two paradigm epistemic sources: perception and testimony. Roughly, do we need justification to believe that perception is reliable in order to acquire justification via perception, and, correspondingly, do we need justification to believe that testimony is reliable in order to acquire justification via testimony. The goal of this paper is not to advance particular answers to these questions. Rather, our focus is going to be on whether an answer to one commits us to a corresponding answer to the other. The kind of view that I will be challenging holds that we do indeed need some prior reasons or justification for the reliability of testimony but that this in no way commits us to the corresponding claim about perception. This view is common in the literature on testimony. So-called reductionists about the nature of testimonial justification are committed to the prior justification claim while attempting to distance themselves from the equivalent claim about perception. Such views attempt to establish important theoretical differences between testimony and perception that would allow for the stricter epistemological policing of testimony. In the language of this paper, these views attempt to establish a theoretical asymmetry between testimony and perception. The goal of the paper is to show that the arguments for asymmetry fail, the upshot of which is that the kind of strict requirements on testimony that are popularly endorsed do, in fact, entail a commitment to corresponding strict requirements on perception which are widely seen as implausible.

2 Theoretical symmetry

One important question in the epistemology of perception is whether the acquisition of perceptual knowledge or justification requires some prior knowledge or justification for the reliability of perception. For example, in order to know that, say, there is a laptop on the table in front of me based on the relevant visual experience, do I first need to know that my perceptual faculties are a generally reliable guide to the truth about my immediate environment? If I do, then it looks like I cannot rely on perception in order to first establish its reliability. I would need to establish the reliability of perception on some non-perceptual basis, such as perhaps introspection or a priori reasoning. The opposing positions in this debate have come to be referred to as liberalism and conservatism.Footnote 1 Conservatism is the view that the acquisition of justification via a method M requires prior, M-independent knowledge or justification of M’s reliability. In contrast, liberalism denies that prior, M-independent knowledge or justification is necessary in order to acquire justification via M. According to the liberal, absence of reasons to take M to be unreliable are sufficient to justify the use of M in forming beliefs. For example, absent defeaters for the reliability of perception, my perceptual experience as of a laptop on the table is enough to justify me in believing that there is. Perceptual experiences provide a prima facie kind of justification that requires no auxiliary justification to take it that the experience is veridical.

The debate between liberals and conservatives has for the most part focused on the specific case of perception as a method of belief formation. It is however curious to note that elsewhere in epistemology a structurally similar debate has taken place independently of but concurrently to the liberalism-conservatism debate on perceptual justification. In the epistemology of testimony, one question of interest is whether hearers in testimonial exchanges can acquire knowledge or justification for testimonial beliefs independently of having some prior justification for the reliability of testimony or the reliability of individual testifiers. Discussion of this point has tended to be subsumed under a related discussion over whether testimony is a fundamental epistemic source or whether its epistemic status can be reduced to other more fundamental sources such as perception, memory, and induction. If the epistemic status of testimony can, as reductionists hold, be reduced to these other sources, then a form of testimonial conservatism is true: testimonial knowledge and justification are acquired via a kind of inference from the speaker’s utterance that p, together with justification to trust that the speaker is a reliable, trustworthy source, to the conclusion that p. As with perceptual conservatism, the prior justification for the reliability of testimony must be entirely independent of testimony. Those who reject the reduction of testimony to more basic epistemic sources (anti-reductionists) seem committed to a kind of testimonial liberalism, according to which no independent reasons or evidence to take a speaker to be a reliable, trustworthy source of testimony are strictly necessary. Liberalism about testimony takes there to be a prima facie kind of justification that subjects can enjoy merely by virtue of being a hearer in a testimonial exchange. This justification can be easily defeated if, for example, there are reasons to take it that a testifier is being insincere.

I will use the term theoretical symmetry to talk about the kind of view that says our answers to these questions about perception and testimony ought to match. Two epistemic domains A and B are symmetrical if defending conservatism (liberalism) in A commits us to conservatism (liberalism) in B, and vice versa. Several authors seem to go along with something like symmetry. For example, Wright (2012, 2014), Coliva (2015), Coady (1973, 1992), Graham (2000, 2004, 2006), and Burge (1995, 1996). Notably, there are symmetrists in both liberal and conservative camps. Some notable defenders of asymmetry are Elizabeth Fricker (1987, 1995) and Jennifer Lackey (2006). The following sections of this paper will deal with a variety of arguments from Fricker and Lackey in defence of asymmetry (though neither use this term it is clear from how I define it that both are committed to it). I will show that their arguments either fail in attempting to establish differences between perception and testimony or that the differences they do establish are not epistemologically significant.

If symmetry is correct, then testimonial conservatives are committed to perceptual conservatism. But conservatism about perception is a deeply problematic view. Indeed, on a certain interpretation whereby one needs to antecedently know or to have prior evidential justification for the reliability of perception, the conservative requirement seems to induce a vicious form of scepticism. After all, where could such knowledge or evidence for the reliability of perception come from if not via perception itself? Those who defend conservatism about perception thus tend to endorse the weaker claim that some prior warrant or justification for the reliability of perception is required, though this is not per se evidential but rather a kind of default justification or entitlement.Footnote 2 We are warranted by default (entitled) in presupposing that our perceptual faculties are a reliable guide to reality. This watered-down version of conservatism need not concern us, however, given that what testimonial conservatives of the above kind (i.e. Lackey and Fricker) seem to be committed to—conditional on the assumption of theoretical symmetry—is an inferentialist, evidentialist conservatism about perception, which is a view all anti-sceptical epistemologists would do well to reject. Testimonial conservatives recognise the worry. Fricker defends a conservative account of testimony according to which knowledge can be gained through testimony only inferentially via knowledge of speakers’ reliability (1995, p. 399). Fricker notes that corresponding views in other domains are fraught with problems: “It is familiar that it is hopeless to treat knowledge through either memory or perception as inferential rather than direct” (1995, p. 400). What Fricker needs then is to reject symmetry—she needs argument for why we ought to treat testimony differently to memory and perception, requiring stronger epistemological standards than do these other sources. Echoing this concern, Jennifer Lackey nicely summarizes the situation:

“If [symmetry is true], there seems to be a problem of overgeneralization here. For now it looks as though positive reasons are needed to justifiedly hold, not just testimonial beliefs, but any beliefs. And this, in turn, leads us into all of the problems facing traditional internalist theories of epistemic justification, such as infinite regresses, circularity, foundations, and so on.” – (2006, p. 176).

Both Lackey and Fricker defend versions of views which posit a default requirement on hearers to have positive reasons to trust that speakers are reliable testifiers. The corresponding view about perception is practically a non-starter, which is what the above quotes are gesturing at. If we first need to establish that our perceptual faculties are reliable prior to using them to form beliefs, we cannot rely on them in doing this. But if we cannot rely on perception to establish its reliability, then it is very hard to see what other options are available—after all, it is arguably a contingent, empirical matter whether our perceptual faculties are reliable so a priori reasoning cannot help here. It is therefore crucial for Lackey’s and Fricker’s projects, and indeed all reductionist, inferentialist accounts of testimony, that they can establish that there are reasons for us to apply different epistemic standards to perception and testimony—i.e. to establish what I am calling asymmetry.

3 Neutrality and non-neutrality

In the following sections of the paper I will discuss four arguments that attempt to establish the asymmetry of perception and testimony. Each argument attempts to show that we can reasonably treat testimony conservatively while treating perception liberally. These arguments amount to the following four claims: an assertion is neutral with respect to the truth of the asserted content; testimony is de facto less reliable than perception; testimonial unreliability has more dimensions to it than perceptual unreliability; the sources of testimony are heterogenous. We begin with the first and take each in turn.

Fricker’s case for applying stronger standards to testimony begins with the observation that it is part of the nature of an experience that a state of affairs is represented to one as true (1987, p. 74). Or perhaps more cautiously, where an experience is an experience as of a certain state of affairs, the state of affairs is represented as true.Footnote 3 So a perceptual experience as of, say, a cup on the table in front of one is an experience in which it is represented to one as true that there is a cup on the table. This direct, immediate link between experience and perceptual representation means that such experiences cause a direct impulse to believe: “the very nature of an experience means that it is a state with an internal, unmediated impulse to believe in what is experienced” (ibid). And thus beliefs based on perceptual experiences need not be based on any kind of inference. However, Fricker thinks this is not the case with testimony. “There is nothing”, she argues, “in a hearer’s perception of an utterance as, say, an assertion that P which is, from either a phenomenological or an epistemological point of view, his experiencing P itself as instantiated” (ibid). This is the crucial difference we were looking for. Fricker locates the asymmetry between perception and testimony in the thought that perceptual experiences are representations of states of affairs as true, while assertions are not. And thus, hearers need to perform a kind of inductive or abductive inference, from the speaker’s utterance that P, to the conclusion that P, via some additional premise to the effect that the speaker’s testimony can be relied upon.

Granting Fricker’s point that there is something in the nature of a perceptual experience that the state of affairs represented in the experience is represented as true, why should we think this is not also the case with testimony? To help get some clarity on the issue, let’s first consider a kind of representation of a state of affairs in which it is trivially not the case that the state of affairs is represented as true. Consider a painting of a particular scene in which some event is taking place. The event is represented in the painting by the artist, but is it represented as having happened? We surely can imagine a kind of case in which it is—perhaps an artist working for the police has painted a representation of an event based on witness testimony. But this is beside the point. The question is whether it is part of the very nature of representational painting in general that this the case. Clearly the answer here seems to be no. Knowing nothing else about a picture, about why it was painted and by whom, we would feel no compulsion to believe that the events depicted had actually taken place, nor would there be any expectation of believing. All else equal, a painting of an event taking place is neutral with respect to the issue of whether the event did in fact take place. With this contrast case in mind, we are better positioned to ask the question of whether testimony is more similar to perception, in which the represented state of affairs is represented as true, or more like a painting, in which the state of affairs is not represented as true. I think it should be relatively clear that an assertion is more similar to a perception than a painting in this regard. Consider the fact that seeing or hearing about something surprising might cause us to update our picture of the world or else respond with incredulity. If someone tells me that the president of France has been awarded the Olympic gold medal in the one hundred metre sprint, I will probably respond with surprise and incredulity. If someone shows me a painting of the same, my response will be more likely one of mild amusement at the artwork. It would not normally be proper of me to respond to the painting as if my friend had just made the corresponding assertion about the president by asking for evidence to back it up. An assertion that P is not neutral with respect to the truth of P in the same way that something like a painting is.

Reflection on the nature of the assertoric speech act reveals some interesting characteristics that go some way to explain why we are inclined to take assertions to be representations of states of affairs as true. Analyses of assertion usually point to two essential characteristics: content and force. Assertions have content in the sense that they represent propositions. But this is true of other types of speech acts as well: questions, suggestions, guesses. What distinguishes assertions from these other speech acts is assertoric force. Assertoric force is more than simply representing a proposition as true via a particular speech act. It is also to represent oneself as having the authority to make an assertion, and thus as having met the requisite norms of assertion, with the intention that one’s audience themselves form the corresponding belief. Far from being neutral with respect to the truth of the represented content, an assertion is a speech act that is defined by its non-neutrality.

One further point against the thesis that assertions are truth-neutral is that comprehension of assertoric content cannot occur in a truth-neutral manner. That is, the inferentialist picture according to which we first comprehend the content of an assertion in a neutral manner before then choosing whether to infer its truth is neither theoretically nor empirically well founded. On the theoretical side, Patrick Rysiew argues that hearers cannot comprehend the content of assertions unless they presume the speaker is being truthful (2012). This entails that arriving at comprehension of assertion in a truth-neutral manner in the way proposed by the inferentialist is impossible. I cannot interpret your assertion as an assertion without a presumption that you are attempting to be truthful, for otherwise the verbal act underdetermines whether you are asserting as opposed to guessing, joking, acting, and so on. Furthermore, even within the scope of assertoric speech, given loose ways of talking, the verbal act underdetermines which of a number of semantic interpretations is the correct one. The presumption of truthfulness is necessary in order for me to understand that when you say everyone is coming to dinner tonight, you do not mean this to be taken literally. In order to get to the correct interpretation of your assertion, I must presume that you are adhering to the proper rules and conventions of linguistic communication such as the Gricean Maxim of Quality—‘try to make your contribution one that is true’ (1989, p. 27). On the basis of considerations such as these, Rysiew concludes that “if the presumption of truthfulness is required just to arrive at what a speaker is telling, questions about whether/why one should believe what one’s told can only arise against the background of the presumed reliability of testimony” (2012, p. 293). A presumption of truthfulness is thus the default, which means that neutral comprehension of testimony that is prior to truth assessment is an inferentialist illusion. This presumption should not be conflated with the conservative claim that we need prior justification to presuppose that speakers are reliable. The former is a claim about the psychology of understanding, the latter is a claim about the structure of justification. The takeaway here is simply that assertions are not truth neutral because understanding them requires the presumption of truth. That is compatible with the liberal claim that justification to believe them requires no prior justification to take speakers to be reliable. Taking speakers at their word may be a matter of basic entitlement.Footnote 4

On the empirical side of things, the psychological literature on truth bias, sometimes also called truth default theory, corroborates the foregoing theoretical insights. For example, Daniel T. Gilbert and his team provide strong evidence for thinking that acceptance of communication coincides with comprehension (Gilbert, Malone, & Krull, (1990), Gilbert (1991), and Gilbert, Tafarodi & Malone (1993)) . That is, subjects do not first comprehend an idea that is presented in communication before then deciding whether to accept or reject it. Rather, acceptance is the default setting and only after having accepted a communication as true does the possibility of rejection as false arise. Gilbert labels this the Spinoza hypothesis, on the grounds that Spinoza held the view that “all ideas are accepted (i.e., represented in the mind as true) prior to a rational analysis of their veracity, and that some ideas are subsequently unaccepted (i.e., represented as false)” (Gilbert, Malone, & Krull 1990, p. 601). The evidence for this Spinozan view is that processing a proposition as false requires more time and cognitive capacity than processing as true. Moreover, they find that this is no mere social convention but is part of the very nature of the cognitive mechanisms in play. Incredibly, even if subjects are primed to expect false information, still it looks as though there is an initial representation-as-true which precedes a subsequent judgement of falsity:

“In short, knowing ahead of time that information would be false apparently did not enable subjects to adopt a skeptic's set and represent the information initially as false. This suggests that the initial coding of ideas as true may (as Spinoza suggested) be an operation that is not readily amenable to voluntary control.” (Gilbert, Malone, & Krull 1990, p. 607).

What the Spinozan hypothesis means for the current discussion is that assertions, by default, are experienced as true by hearers, according to the literature from psychology on truth bias. In conjunction with the claim above that force as well as content is a distinguishing feature of assertions, this provides a very strong case for rejecting the claim that assertions are not representations of states of affairs as true. The truth-neutrality account of assertions is neither a plausible account of the nature of the content of assertions, nor a plausible account of the actual cognitive mechanisms underpinning our receipt of testimony. We will need to look elsewhere for a theoretically important difference between perception and testimony.

4 Degrees of reliability

A thought that both Lackey and Fricker pick up on is the idea that testimony just is de facto less reliable than perception and thus more scrutiny is required of us when forming testimonial beliefs as compared to perceptual. In considering the liberal policy of believing in accordance with speaker’s utterances, absent positive reasons to take the speaker to be unreliable, Fricker writes that “the proportion of utterances which are made by speakers who are either insincere or incompetent is far too high for this to be an attractive policy” (1987, p. 76). Let us grant the point that testimonial unreliability is significantly more unreliable than perception. Is this difference epistemologically significant? Suppose that it is. Peter Graham points out that this kind of argument presupposes a problematic principle according to which differences in degree (of reliability) determine differences in epistemic kind (Graham 2006, p. 101). Arguably, perception is less reliable than introspection, memory less reliable than perception and testimony less reliable than memory. Degrees of reliability exist between each of these epistemic sources. Thus merely pointing out a difference in degree of reliability is not enough to establish asymmetry because there is a whole spectrum of degrees of reliability along which to place a variety of epistemic sources but only a single line to be drawn between conservatism and liberalism. What testimonial conservatives need to argue for in order to establish the asymmetry claim on the basis of a difference in degrees of reliability is that the difference in reliability between perception and testimony is theoretically significant. Simply pointing out that testimony is generally less reliable than perception as a way to establish asymmetry begs the question since it remains to be established that the theoretically significant threshold the asymmetrist needs is to be located somewhere along the spectrum between the precise loci of perception and testimony. Suppose we grant the thought that there is some threshold along a spectrum of degrees of reliability of various epistemological sources and methods. This is enough to establish that there is asymmetry between some methods. It nonetheless remains to be established just where exactly the threshold lies and between which methods. For consider the possibility that perception falls below the threshold. Perception is somewhat unreliable, testimony more so, but both fall below the threshold and thus symmetry is true of these two methods. What possible argument could the asymmetrist who wants to locate the threshold between testimony and perception give for excluding this option? A third possibility yet is that the threshold is in fact much lower, to the extent that both perception and testimony pass the bar needed to be treated liberally. Once again, the asymmetrist who wishes to ground their assymetrism in differences in degrees of reliability needs some independently plausible argument for excluding this version of the threshold view, and yet it is very difficult to see on what grounds one could attempt such an argument.

Lackey offers a related but nonetheless distinct argument for symmetry on the grounds of a difference in reliability. Her argument is a modal one, appealing to the familiar thought that the possible worlds in which most of our perceptual beliefs are false are quite distant from the actual world—the brain-in-a-vat scenario, for example. This means that perception, though not infallible, is relatively modally robust. Compare this with testimonial unreliability. Lackey thinks that worlds in which most of our testimonial beliefs are false are much closer: “indeed, for many people this is true in the actual world” (2006, p. 177). The kinds of examples she has in mind here are scenarios in which one is raised by parents who are in a cult, or in societies that are highly superstitious or under governments that are highly corrupt. What she takes this to show is that the chance of testimonial error is much higher than perceptual error and for this reason “the rational acceptance of the reports of others requires positive reasons in a way that is not paralleled with other cognitive faculties” (ibid.). Clearly this argument again appeals to the differences-of-degree-equal-differences-in-kind principle that we have rejected. However, there are some further worries for Lackey’s argument that, upon reflection, seem to point us away from the kind of asymmetry she is trying to motivate.

Radical sceptical scenarios are easy to posit. The most familiar ones involve radical perceptual deception such as the brain-in-a-vat, evil demon or Matrix scenarios. Though we ordinarily take such scenarios to be modally very distant from the actual world, the sceptical bite comes from the acknowledgement that they are perfectly logically consistent with our subjective experiences. Furthermore, we might easily imagine equivalent sceptical scenarios for other sources of knowledge besides perception. Consider Bertrand Russell’s argument that it is logically compatible with all of our apparent memories of the past that, in fact, the world simply sprang into being mere moments ago (2012, p. 94). Just like the perceptual-deception scenarios, Russell’s memorial sceptical scenario would seem to be perfectly internally consistent and consistent with our current subjective experiences. It is important to bear in mind that in these radical sceptical scenarios, virtually all of our perceptual (or memorial) beliefs would be false. With this in mind, it is far less clear that there are real-world scenarios of radical testimonial unreliability. Consider the examples Lackey appeals to, cults and oppressive regimes. Do these examples merit the label of radical deception scenarios, on a par with the brain-in-a-vat and Russellian scenarios? It would seem that they do not. After all, even living in a cult, while one could easily have many false testimonially-based beliefs (beliefs about the world outside the cult, beliefs about the origin of humanity, moral beliefs) one presumably has lots of mundane, ordinary, true testimonially-based beliefs also (beliefs about what time it is, about what is being cooked for dinner, about the location of objects in one’s environment). While it is perhaps part of the very essence of cults that they entail some kind of testimonial unreliability, the beliefs in question are typically restricted to a certain domain (religious, political, moral). It is therefore not clear that these scenarios are at all on a par with the familiar radical (perceptual, memorial) sceptical scenarios. What Lackey needs to substantiate her claim are cases in which practically all of one’s testimonially-based beliefs are false and then some reasons for thinking such scenarios are modally closer to the actual world than the perceptual cases. Far from this being an achievable goal, there are reasons for thinking that radical sceptical testimonial scenarios are not even coherent possibilities.

What would it mean for all of one’s testimonial beliefs to be false? Conceiving of such a scenario is not as simple as in the perceptual case. Firstly, in the perceptual scenario there is no need to think that radical perceptual unreliability need prevent one’s perceptual beliefs being generally coherent. However, in the testimonial case things are not so obvious. Consider a subject—let’s call him Truman—who is in the unfortunate scenario of being lied to by those around him. It is perfectly conceivable that those in Truman’s life seek to deceive him through false testimony. Truman is told all sorts of lies about the name of the small town he lives in, the world beyond the town, and the lives of those he lives with. It is easy to conceive of such a scenario of massive orchestrated testimonial deception. But this does not put testimony on a par with perception. For that to work, we need to conceive of a scenario of not just massive deception but total, radical deception. And the obstacle we encounter in attempting to do so is that, while in the radical perceptual case, perceptual unreliability need not affect reliability in other domains—the envatted subject may well have perfectly excellent powers of reasoning and she may also have impeccable memory. Things are different for testimony because, unlike for perception and memory, it is very easy to compare testimonial evidence with other kinds of evidence. Consider what will happen when Truman asks simple questions about objects in the immediate environment that can easily be checked by other means such as ‘What’s that you’ve got in your hand?’ or ‘Does the cheeseburger come with fries?’ Or consider what will happen when Truman asks questions about the recent past or near future the answers to which will soon become apparent to him anyway such as ‘What are we having for dinner?’ or ‘Did it rain last night?’ The point is that in order for the massive, orchestrated deception to function, a certain amount of the testimony that Truman receives needs to be reliable. If it is not, this will very quickly become obvious to him and if he is a normal functioning person he will presumably stop relying on testimony.

A case in which testimony is globally unreliable is much harder to conceive of than our Truman case. In fact, there are reasons for thinking that such cases could not possibly exist. The reason for this is that the practice of testimony itself will not survive wholesale unreliability. Suppose that we found ourselves in a world in which the propositions that others express when they seem to be making assertions bore no relation to the truth. The unreliability of these utterances will quickly become obvious: when it is raining, people will declare that it is not; when it is night, people will declare that it is day. One consequence of this is that no one would believe such barefaced lies. Moreover, as C. A. J. Coady argues, in a world in which there were no correlation between apparent reports and the facts, it is hard to see why we should think there are reports at all (1992, p. 84). Coady imagines encountering a Martian community whose language we could translate and who seem to systematically say things that we and they can observe to be false. In such a situation, he wonders, “what reason would there be for believing that they even had the practice of reporting?” (1992, p. 84). Peter Graham concurs with Coady that if all assertions are falsehoods, then the practice of asserting will come to an end because hearers will have no reason to accept assertions, giving speakers no reason to make them (2000, p. 698).Footnote 5 This shows why the radical sceptical testimonial cases, far from being modally closer to the actual world as Lackey claims, are arguably not even coherent scenarios. One might wonder whether we can nonetheless conceive of sceptical testimonial scenarios in which, if not all, then most of the assertions made are false. But Graham goes on to deny even the coherence of this less radically unreliable scenario. Even if, say, the imagined Martian community make equal amounts of true and false reports, still the practice of assertion would not survive because relying on the word of others would be no better than flipping a coin (2000, p. 699). Unless speakers mostly utter truths, the economy of testimony will thus break down.

The sustained practice of testimony conceptually implies a certain degree of testimonial reliability without which members of a community would have no reason for engaging in the practice. This is not yet to say that radical sceptical testimonial scenarios are impossible. Rather, it is an empirical claim about the institution of testimony, that wholesale unreliability would mean hearers have no reason to believe what speakers say, which in turn would mean speakers have no reason to give testimony. This argument is closely connected to Rysiew’s argument, discussed in the previous section, that the practice of testimony requires a presumption of truthfulness in order to function (2012). And since wholesale unreliability would lead hearers to forgo the presumption of truthfulness, future attempts to assert would fall flat, no longer interpreted as truth-apt declarative utterances. The consequences of this are far more radical than that the economy of testimony would likely break down. The point here is that the utterances such speakers make would no longer count as assertions because they would carry no assertoric force. If we accept the plausible claim that assertoric force is a constitutive component of assertions, then any speech act that ceases completely to bear any degree of force will thereby cease to even count as an assertion. That is why radical sceptical testimonial scenarios, far from being modally close, are not even coherent possibilities.

Reflection on differences in degrees of reliability does not therefore seem to offer a way to establish the kind of asymmetry claim that the testimonial conservative seeks. We must again look elsewhere for our theoretically significant differences between perception and testimony.

5 Multidimensionality of testimony

There are two dimensions to testimonial unreliability: incompetence and insincerity. A testifier can provide unreliable testimony if they are themselves incompetent believers. But a competent believer can still provide unreliable testimony by being insincere. These two dimensions to testimonial unreliability make testimony more unpredictable than perception, according to Lackey, because perceptual unreliability, by comparison, is one-dimensional (2006, p. 176).

This argument is somewhat similar to the differences-in-degree-equal-differences-in-kind argument from above—both arguments derive higher epistemic standards for testimony on the grounds of greater testimonial unreliability. But whereas above we were not able to find principled grounds to set an unreliability threshold—beyond which the higher epistemic standards kick in—this argument promises to provide such grounds. Contrary to the earlier argument, it is not the degree of unreliability that matters, but the complexity, or number of dimensions that are important. Where there is more than one dimension of unreliability, subjects need to be on guard in a way that they do not if there is but one dimension.

Once again, this argument fails to establish the asymmetry. To see why, assume that the principle it relies on is correct, namely the following. Where there is just one dimension to cognitive unreliability liberalism is true, but where there is more than one dimension to cognitive unreliability conservatism is true. It is not clear what the argument for this principle is meant to be, but it at least feels relatively less arbitrary than identifying some particular degree of reliability as the threshold beyond which conservatism kicks in. However, even with this principle in play, it still does not force the wedge between testimony and perception that the asymmetrist needs.

The first thing to note is that to suggest that perception is one-dimensional in this way is an oversimplification. Perception is multi-modal in that it may represent the size of an object, its shape, distance, motion, relation to oneself, and so on. These multi-modal aspects of perceptual representation may come together to give a subject a cohesive picture, but they rely on a variety of perceptual modalities, or sensory streams.Footnote 6 A reliable perceptual system depends on each of these many modalities functioning reliably and on their coming together in a reliable way. Visual agnosia occurs when one of these sensory streams breaks down, giving rise to an inability to process sensory information such as distinguishing objects from their backgrounds or motion detection. Perceptual representation is thus multimodal and the many dimensions to perceptual unreliability correspond to the many forms of visual agnosia.

Despite this recognition of perceptual multimodality, we might nonetheless wonder whether there is something unique about the agential nature of testimony that would support asymmetry. The thought would be that since speakers are free to choose to deceive us in a way that our perceptual systems are not, we ought to be on our guard against this possibility and that is why conservatism is true of testimony but liberalism is true of non-agential sources like perception. There are two assumptions here that we might reject. Firstly, Graham argues that the agential nature of testimony does not give rise to the kinds of strict epistemic standards that conservatives say it does (2000, 2004). If Graham is right about this, then differences in whether sources involve the possibility of deception or not will have no theoretical significance for the conservatism-liberalism discussion. I think Graham is indeed right about this. However, I want to push back against the second assumption. This assumption is that deception is unique to testimony. Assuming for the sake of argument that the possibility of deception does give rise to conservative justificatory standards for testimony, why should we assume does this not translate into a corresponding argument for perception? The obvious answer is that speakers have agency and this enables them to deceive us while our perceptual faculties do not have agency and so are unable to deceive us. Perhaps the multimodality of perception is beside the point if the dimension of insincerity in testimony has no perceptual counterpart for perception. This does not, however, seem to be the case. Consider the fact that con-artists, pranksters, illusionists and other practitioners of the deceptive arts may seek to cause us to form false beliefs based on perception in all sorts of ways. I visit a Broadway magic show and watch the magician enter a cupboard door only to have vanished into thin air a moment later. It is all an illusion of course, and the magician is preying on my credulity, but it is nonetheless an act of deception on a par with insincere testimony. In cases such as this, even though my perceptual faculties themselves are not malfunctioning, illusionists and other tricksters can lead me to form false perceptual beliefs through deceptive means. Deception is not unique to testimony.

Just like perception and testimony, we can also think about memorial unreliability along multiple dimensions. In particular, we can perfectly well think about different dimensions of memorial unreliability that mirror those discussed by Lackey. The standard way for memory to mislead is a simple act of misremembering. I seem to remember taking my keys with me when I left the house, but it turns out I left them at home and have misremembered. In such a case, my memory is simply letting me down, but it is not being ‘deceptive’ in the agential sense of intending to lead me astray. In this sense, it is on a par with the dimension of ‘incompetence’ that Lackey discusses in relation to testimony. Contrast this with a case involving the intriguing phenomenon of memory implantation.Footnote 7 In memory implantation cases, subjects in controlled experiments have been led to form false memorial beliefs by experimenters manipulating them using a variety of different means. In one experiment, subjects are shown photographs of their childhood that have been manipulated in such a way to depict an event that did not happen as having happened. In another experiment, subjects are given a narrative of events that happened to them as a child with one false event mixed up among several other true events. Unaware that a false event has been slipped into the narrative, a significant number of subjects adopt the false event as part of their memory of the series of events. What cases like these show is that it is in fact relatively easy for agents to manipulate us into seeming to remember things that did not occur. There is thus no reason to think that testimony is special in the sense that it is only through testimony that we can be deceived by agents who seek to deceive us. Deception can take many forms. If it is the multidimensionality of testimonial unreliability that gives rise to testimonial conservatism, then the fact that there are multiple analogous dimensions to perceptual and memorial unreliability also seems to commit conservatives to perceptual and memorial conservatism.

6 Perceptual homogeneity, testimonial heterogeneity

The next argument for asymmetry that we will consider comes again from Lackey. Her thought is that an epistemologically important difference between perception and testimony is that perceptual sources are homogenous while testimonial sources are heterogenous (2006, p. 177). To see what she means by this, it is best to run through the two contrast cases she presents.

Perceptual amnesia

Suppose that after her involvement in a car accident, Olivia has complete amnesia with respect to her perceptual faculties, that is, she remembers nothing about either the workings or the deliverances of such faculties. After leaving the hospital, she stops at the store to buy some groceries, bumps into some acquaintances on her way home, watches an episode of Seinfeld on TV while eating dinner, and spends some time on the internet before going to bed. Along the way, Olivia forms perceptual beliefs about all sorts of things, including beliefs about the vegetarian items that Trader Joe’s carries, the kinds of trees losing their leaves, the number of children her acquaintance now has, which Seinfeld episode is on, and the colour of the background of the MSN website. Now, because of her perceptual amnesia, Olivia’s acquisition of these perceptual beliefs is not governed by any acquired principles of perceptual belief formation. But even in the absence of such principles, it seems reasonable to conclude that the overall status of Olivia’s daily perceptual beliefs would be very high epistemically.

Testimonial amnesia

Edna, Olivia’s best friend, was in the same car accident that caused Olivia’s perceptual amnesia. In Edna, however, the accident caused testimonial amnesia: she remembers nothing about either the workings or the deliverances of testimony. After leaving the hospital, Edna’s day was nearly identical to Olivia’s. For instance, she stopped at the same grocery store, bumped into the same acquaintances on the way home, watched the same episode of Seinfeld, and visited the same internet sites before going to bed. Now, because of her testimonial amnesia, Edna’s acquisition of testimonial beliefs along the way was not governed by any principles of testimonial belief formation. As a result, Edna trusted to the same extent all of the testimonial sources she encountered throughout the day—which included a copy of the National Enquirer that she read at the grocery store, her acquaintance’s 3- year-old daughter, the characters of Jerry and George on Seinfeld, and an extremist, evangelical Christian internet site she stumbled upon while surfing the web—and she believed everything that she was either told or read along the way—which included testimony that a woman from Georgia was abducted by aliens, that there are real princes and princesses at Disneyland, that licking the envelopes of cheap wedding invitations can lead to one’s death, and that those who are gay will be sent to eternal damnation.

The intuition we are supposed to have towards these cases is that Oliva’s perceptually based beliefs are justified while Edna’s testimonially based beliefs are not justified. The explanation that Lackey offers for this difference is that perception is homogenous—there is not much difference between Olivia’s various perceptual beliefs and the ways they were formed. Her belief that there are leaves on the tree was formed by looking at the leaves on the tree. Her belief about the kinds of vegetarian items on offer in the supermarket was formed by looking, and so on. Because of this homogeneity, Lackey argues that “subjects do not need to be very discriminating in order to be reliably in touch with the truth” (2006, p. 177). On the other hand, the sources of testimony are a heterogenous group, with some sources being less reliable than others. According to Lackey it is the fact that Edna’s amnesia renders her insensitive to the differences between testimony’s various sources that lead her astray. She trusted sources that she was unaware she should not have trusted, such as the New York Post and the 3-year old’s depiction of Disneyland. Importantly, in both cases, were we to stipulate that our subjects did not suffer from amnesia, Edna’s beliefs would have been significantly different while Olivia’s would not. According to Lackey, what this demonstrates is that possessing acquired principles governing the acceptance of testimony is necessary for testimonial justification, while the same is not true for perception.

Has Lackey identified an epistemically significant difference between perception and testimony here? There are good reasons to think not. Consider what testimonial conservatives like Lackey want to say about young children who are incapable of meeting the conservative requirement on testimonial justification. It is plausible that even children who lack concepts such as sincerity, lying, reliability and so on can know things via testimony. Indeed, young children would seem to rely on testimony as a source of knowledge about the world to a greater degree than do adults. And yet, conservatism threatens to make it very hard for such children to properly acquire knowledge and justification via testimony given their supposed lack of the relevant concepts and lack of sophisticated abilities to make judgements about sincerity, competency, and so on. Reductionists of course have things to say about this objection. For example, Fricker attempts to sidestep this issue by making an exception for young children, arguing that they do not need justification to trust in the reliability of testimony while they are still in the developmental stage (1995, p. 403). But as they grow up and they acquire discriminative capacities and knowledge of what people and institutions are like, the conservative constraints kick in and they need to engage those discriminative capacities and make use of that background knowledge. Suppose that, as is highly plausible, young children can acquire knowledge via testimony even though they lack the kind of background supporting reasons that reductionism generally requires of hearers in testimonial exchanges.Footnote 8 This raises the question of why we cannot simply treat Edna like a child. Edna has lost the necessary concepts such as testimonial unreliability, lying, and so on, rendering her unable to discriminate between trustworthy and untrustworthy sources of testimony, much like a young child. Edna’s testimonial beliefs are therefore unjustified only if young children have no justified testimonially based beliefs. But young children plausibly do have some justified testimonially based beliefs and therefore so does Edna.

A further issue with these cases is that some of the differences that Lackey is picking up on seem to be merely a result of differences in the two narratives she tells. It is just a feature of the stories that Olivia forms lots of true beliefs while Edna forms lot of false beliefs, but we can easily imagine different stories, equally as plausible, that give us different results. For example, suppose that, as she is leaving the hospital Olivia believes her path ahead is clear but as she steps out she walks straight into a glass door. She then wanders past a pond and notices a pole that is sticking up out of the water. She believes the pole is bent, as it appears to be bent just at the point where it hits the water, though in fact the stick is straight, and this is a common illusion caused by refraction of light. Olivia then comes across a skilfully drawn pavement sketch artist drawing a perspectival illusion, intended to appear as a hole in the ground when viewed from the correct angle, and forms the corresponding false belief that she is in danger of falling through. Olivia next sees a billboard poster which displays the Müller-Lyer illusion, forming the corresponding belief that the two lines are off different lengths. The point here is that Lackey simply builds into her narrative that Olivia forms lots of true beliefs via perception despite her amnesia and concludes that the epistemic status of those beliefs is unaffected by the amnesia. But what these examples illustrate is that Olivia will be led to form false beliefs given that she is not on guard against these ordinary, everyday visual illusions.

In addition to leading her to form false beliefs based on perception, Olivia’s amnesia will also greatly hinder her ability to form any judgements at all for a wide class of perceptually-based beliefs. Having no understanding about the workings and deliverances of perception, no memory of past perceptual experiences, no awareness of how information or conceptual content can be encoded through visual representation, and no understanding of how different perceptual modalities represent information, Olivia will be incapable of taking in lots of the information that would otherwise be available to her. She may in some sense ‘see’ a sign telling her not to cross the road, but she cannot see that the sign is telling her not to cross the road. For that would require her to understand that perceptual experiences can encode instructional information, to remember which shapes correspond to which words and concepts, and so on. She may see an oak tree in the garden outside the hospital, but she does not see that it is an oak tree. She may appreciate the structure of the tree but she does not recognise it as an oak tree. The content of perceptual experience is in part a function of what we expect to see, of our background beliefs, and of the concepts we possess (Coady 1992, p. 147). Robbed of these concepts, beliefs and expectations, Olivia would not be able to judge that the items on the shelf in Trader Joe’s are vegetarian or that there is a particular episode of Seinfeld on the television. Of course, the same point applies to Edna about whom we ought to say she would not, in fact, be able to form beliefs based on testimony since a complete and total lack of understanding of the practice of testimony and absence of memories of testimonial interactions would render her incapable of recognising an assertion as an assertion. Moreover, we might reasonably think that Edna’s testimonial amnesia will affect not just her ability to form testimonially based beliefs but many ordinary perceptual beliefs too. The vast array of concepts that we use to make sense of experience are learned via testimony, meaning that Edna’s testimonial amnesia will rob her of these concepts, rendering her unable to make sense of her experiences. Consider that in Lackey’s description of the case, Edna is able to recognise and make use of a newspaper. But the concept of a newspaper is not represented by her visual system. What is represented by the visual system is the size, shape, location, colour, etc., of the object. In order for her to recognise that what she sees is a newspaper—which is necessary in order for her to form beliefs based on reading the newspaper—Edna needs to deploy the concept of a newspaper. However, since we acquire such concepts via testimony, and since Edna’s amnesia means she has lost all memory of the past deliverances of testimony, she will presumably no longer have the concept of a newspaper. Likewise for the concepts needed to form beliefs about trees, chairs, tables, and so on, each of which is acquired via testimony. This suggests that Edna’s unfortunate situation is indeed, as suggested above, much more like that of a child; a very young child who has yet to acquire the concepts needed to interpret her experiences.

The general lesson here is that these amnesia cases are vastly under-described. Exactly which beliefs a subject would and would not be able to form via perception or via testimony in the absence of any knowledge about the workings and deliverances of those sources or memories of past executions of them is a question that would require a much more detailed and empirically-informed discussion.

7 Conclusion

We have seen a number of arguments that seek to establish an epistemically significant difference between perception and testimony. Those arguments either fail to establish a difference or they establish some difference but not one that has the kind of epistemological significance needed to support asymmetry. A negative conclusion can be drawn in support of symmetry. None of the asymmetry arguments succeed, therefore, the default position is that we ought to put perception and testimony on a par. Why should this be the default position? Because the very general question we can ask of any belief forming method M is whether positive reasons to trust in the reliability of M are necessary for acquiring knowledge and justification via M. Absent reasons to think that we ought to give different answers to this question depending on which kind of method or source we are talking about, the default position should be to give the same answer across the board. Either we take the conservative line and demand that positive reasons to trust are necessary for perception, testimony, memory and so on, or we take the liberal line and deny such reasons are necessary, or we take a mixed approach. But in the latter case we need argument for why we should treat different sources or methods differently.Footnote 9