A Credibility-Backed Norm for Testimony

Abstract

I propose that testimony is subject to a norm that is backed by a credibility sanction: whenever the norm is violated, it is appropriate for the testifier to lose some credibility for their future testimony. This is one of a family of sanction-based norms, where violation of the norm makes it appropriate to lose some power; in this case, the power to induce belief through testimony. The applicability of the credibility norm to testimony follows from the epistemology of testimony, in that false or unjustified testimony weakens the reason for belief that is provided by the speaker's future testimony.

1. Non-standard arguments and bite

Here is one kind of argument concerning norms on a type of act A. We consider judgments about when particular As are proper. By and large, we notice, the As we judge to be proper are those with property F. We show that F is the property whose presence or absence best explains these judgments, and conclude that the norm of A is F. Call this form of argument a standard argument about norms.

Standard arguments, I claim, cannot answer one important question. Even if a standard argument succeeds in telling us that F is a norm on As, it would not tell us what kind of norm it is. Is an A without F morally wrong, prudentially wrong, wrong according to some social convention, or wrong in some other sense?¹ An argument based merely on intuitive data cannot answer this question.

For example, suppose a certain community has a robust practice of classifying apples, intuitively judging some apples good and some bad. A standard argument concerning norms on apples would be to find what the intuitively good apples have in common – say a certain size, shape, color, and texture – and conclude that these properties constitute a norm on apples. But this argument would not tell us what a good apple is good for. Is a good apple one that is tasty, nutritious, attractive in fruit arrangements, effective for throwing at inept vaudeville performers, or filled with viable seeds that will grow into new trees? The community's judgments about apples might be directed toward any of these ideals; knowing which apples pass the test will not tell us what the apples are being tested for. In contrast, if we know what kind of norm we are concerned with, we can work backwards to figure out what conforms to the norm. If we know that the community's norm is that a good apple is nutritious, we can use nutrition science to determine the characteristics of nutritious apples; mutatis mutandis for other sorts of norm.

In general, a complete account of a norm on A will specify not only which As conform to the norm, but also what type of norm it is. To say that it is a norm not to wear white after the first Monday in September is only to describe which sartorial acts conform to this norm.² It leaves open whether the norm is moral, legal, a norm of etiquette, or a norm of fashion. If it is a legal norm, violations might be punished with a five-dollar fine or a five-year prison sentence. These are important things to know about the norm. Call the specification of acts that conform to or violate a norm the norm's content, and the type of norm and the penalties for violating it (if applicable) the norm's bite. Standard arguments may determine the content of a norm of assertion, but they will not prove anything about its bite.

Another objection to standard arguments is that the intuitions may underdetermine the content of the norm. Pagin (2015) and Johnson (2018) have put forth such objections against the idea that assertion can be explained in terms of its norms. Johnson argues that intuitions concerning impropriety of assertions cannot provide evidence that a certain norm is constitutive of assertion unless we are certain that the impropriety is distinctively assertoric. Consider Williamson's lottery case (2000), in which someone tells someone else that their lottery ticket did not win, based only on the low probability of the ticket's winning. Johnson suggests that the assertion is rude, since the hearer is already aware of the low probability; and that perhaps our judgment that the assertion is improper stems from this rudeness rather than from a distinctively assertoric norm. Since “Don't be rude” is a norm that applies generally, our intuitive judgment about this assertion cannot yield a distinctively assertoric norm without a proof that the judgment does not come from a non-assertoric norm.

Pagin points out that theories of the norm of assertion often explain clashes between the theory and intuitive judgment by adverting to auxiliary principles. For instance, proponents of a knowledge account of assertion acknowledge that when a speaker reasonably but falsely believes that they know what they assert, we do not condemn them. Proponents might explain this apparent clash by saying that the belief excuses the speaker's norm violation (Williamson 2000) or gives the assertion a sort of secondary propriety (DeRose 2002). Pagin argues that this “goes a long way towards severing the empirical relation itself between norm theories and intuitive support” (Pagin 2015: 197); for incompatible theories may explain our intuitive judgments equally well when supplemented with different auxiliary principles, and so the judgments would provide no basis for choosing between them.

To avoid these problems for standard arguments, we need an argument that invokes more than intuitive judgments about propriety. Call such an argument non-standard. If a non-standard argument establishes that an action of type A must be subject to a norm with a certain bite and then determines the content of that norm as the property of As that is best suited to that bite (as when we use nutrition science to determine which apples satisfy a norm with a nutritional bite), the account it provides will be more complete than an account that merely determines the norm's content. As for the possibility that other norms or auxiliary principles might confound our intuitive judgments of propriety, non-standard arguments do not rely on intuitive judgments of propriety, and so there is nothing to confound.

The challenge is to formulate an argument that a norm applies to a certain kind of act A without adverting to judgments of propriety. One possibility is to argue from the nature of A to the fact that a norm of a certain bite must apply to A. I will propose such an argument for a specific subclass of assertion, testimony. The nature of testimony, I claim, is such that it must be subject to a certain sanction-based norm, whose bite is that a violation of the norm makes it appropriate to issue a certain sanction against the violator.

In outline, I will argue that when a speaker issues false (or perhaps unjustified) testimony, this will make it appropriate for hearers to be less likely to believe their future testimony. (Not that speakers necessarily will be less likely to believe that testimony; as we will see, sanctions-based norms can be misapplied.) The mechanism is epistemic; false testimony should make future hearers less likely to believe future testimony because it provides evidence against the truth of the future testimony. The nature of testimony itself will be shown to entail both that this credibility loss should apply to those who testify falsely and that this credibility loss counts as a sanction.

To be clear about the scope of the argument: I argue that the credibility-based norm is a norm of testimony, not that it is the norm of testimony. When a non-standard argument shows that a norm with a certain bite must apply to a certain type of action, this does not mean that other norms with different bites and different contents cannot apply to the same type of action. For instance, there is almost certainly a norm against lying with moral bite, and in many jurisdictions there are norms with legal bites against slander and perjury. My argument will proceed by first setting forth a series of sufficient conditions for a norm to apply to a certain kind of act and then showing that the credibility norm on testimony fulfills these conditions. Since the conditions are sufficient but not necessary, they will not rule out the existence of any other norm of testimony.

2. Sanction-based norms

The idea behind norms with a sanction-based bite is that violations of some norms make it appropriate to apply a specific concrete sanction. To spell it out, it is a sufficient condition for the existence of a norm on A with content F that anyone who performs an act of type A that lacks property F is appropriately subject to a normative sanction S.³ One paradigmatic example is a legal norm: a lawbreaker is appropriately subject to civil or criminal punishment. Another example is a rule of a sport that establishes a concrete penalty for a violation, such as when a hockey player is suspended for two minutes for an illegal check.⁴ Breaking a rule of a private club may result in the suspension of one's membership; and so on.

To show that testimony is subject to a sanction-based norm, I will establish some sufficient (but not necessary) conditions on what counts as a sanction. First, we must distinguish sanctions from run-of-the-mill bad consequences of actions. If I crash my car and make it undriveable, my inability to drive the car is a bad consequence of my past poor driving, but it is not a sanction that was appropriately applied to it. The application of the sanction itself must be normative. Whoever applies a sanction is themselves subject to norms (not necessarily themselves sanction-based) concerning whether the sanction has been applied appropriately. To a first approximation, applying a sanction is appropriate when and only when a sanction-based norm has in fact been violated; which need not mean the sanction will, or will usually, be applied when and only when the norm has been violated. For instance, when a batter fails to bat in their proper turn in baseball, the umpire ought to call them out on appeal. This does not mean that the umpire always will call them out, but that an umpire who fails to do so has violated a norm of umpiring. (In the next section we will consider further complications concerning applying a sanction when the applier is not aware of the violation of the norm.) Thus the sufficient conditions on the presence of a sanction should include the existence of norms governing its application.

We also must capture the idea that the sanction is in some sense bad for the norm-violator. Since the aim is to provide sufficient but not necessary conditions for the existence of a sanction, we can focus on one common kind of sanction: the loss of some power, privilege or entitlement. (I'll use “power” generically to refer to all of these.) The sanction for repeated violation of traffic laws may be the loss of driving privileges. The sanction for violating the rules of the Hasenpfeffer Club may be expulsion from the club and consequent loss of the privilege of entering the clubhouse. There are sanctions that are not losses of powers; for instance, being docked points in a game for violating a rule is plausibly a sanction but not obviously a loss of a power. But loss of a power is prima facie bad in a way that makes it a plausible candidate to serve as a sanction.

Still, not all normatively governed losses of power will be sanctions. Consider this case: Ordinarily movie tickets can be sold to any customer. However, if the movie contains certain swear words, then in the United States it will receive an R rating, and theater owners that abide by the rating system ought to prevent sales of tickets to unaccompanied children under age 17. Thus the presence of the swear words should cost the filmmakers the power to have tickets sold to unaccompanied children. Yet this loss of power is not a sanction that enforces a norm that movies ought not to contain swearing. Rather it reflects a common sort of conscious tradeoff, in which filmmakers make movies with swearing, foreseeing that they will be unable to have tickets sold to unaccompanied minors.

Such tradeoffs need not even be foreseen. A cook who heats milk in a pot loses the power to use that pot to cook kosher meat. This loss is normatively governed in that the laws of kashrut dictate that people should not treat meat cooked in this pot as kosher. This loss of power need not be foreseen or intended as a conscious tradeoff, for the cook may have no idea what the laws of kashrut are. Still, this is not a sanction enforcing some norm that one should not heat milk in a pot, for the cook may never have intended to use the pot to cook kosher meat in the future. The loss of a power that the actor does not intend to retain need not be a sanction.

Basing the distinction between a sanction and a tradeoff on the individual actor's intentions may make the existence of a norm unacceptably dependent on that actor's mental states. Instead, we may advert to the type of action to which the norm applies: a loss of a power P is a sanction on a type of act A when people who perform A acts typically intend to continue to exercise P. So the loss of the power to cook kosher meat in a pot is not a sanction for heating milk in the pot considered qua act of cooking, since it is not the case that cooks typically intend to cook kosher meat in particular pots in the future. However, if the pot in question is in a kosher fleishig restaurant (which serves meat), then the loss of this power will be a sanction on the act of heating the milk qua act of cooking in a fleishig restaurant, since cooks in fleishig restaurants typically do intend to use their pots to prepare kosher meat in the future. Hence there is a norm on acts of cooking in fleishig restaurants that one should not heat milk in pots, enforced by the sanction of losing the power to cook kosher meat in those pots.

Thus, our sufficient conditions on a sanction-based norm will be:

(SC) Suppose that

1. (i) whenever a person X performs an act of type A that lacks property F, it is appropriate for X to be deprived of some power P; and

2. (ii) people who perform acts of type A typically intend to exercise power P in the future.

Then there is a norm on A whose content is that acts of type A should have F, and whose bite is the sanction represented by the loss of power P.

Again, (SC) are sufficient conditions only; plausibly there are sanction-based norms that do not meet these conditions. I am putting forth relatively restrictive criteria for a sanction-based norm, so as to make the strongest case that whenever these criteria are fulfilled, a sanction-based norm is in fact present.

(SC) refers essentially to types of act. A single act may fall under different types and thus be subject to all the norms that apply to these different types. So an act of heating milk in a pot in a fleishig kitchen is both an act of cooking in general and an act of cooking in a fleishig kitchen. It can thus violate a norm on cooking in fleishig kitchens even if it does not violate a norm on cooking simpliciter. Similarly, when a spy says something, their act may fall under the types “testimony” and “utterance by a spy,” and may violate a norm on testimony without violating any norms on utterances by spies, or vice versa. Spies who testify truly about the identity of their superiors may obey general norms of testimony even as they violate norms imposed by their spy organization. So it is natural for an analysis of a norm on acts of type A to depend on A, and for our analysis of a norm on testimony to depend on what people who offer testimony typically intend to do in the future.⁵

3. Elaborations and clarifications

(SC) requires several clarifications and elaborations. The first elaboration is that the loss of a power need not be all or nothing. Sometimes it takes multiple violations of a norm to make it appropriate to deprive someone of a power. When a soccer player commits a certain type of foul and is shown a yellow card, it is not yet appropriate to send them off; on the second foul and second yellow card, it is appropriate to send them off, depriving them of the power to continue playing in the match. The first yellow card is a sanction for the foul, as it moves the player closer to the loss of this power. Weakening of a power may similarly count as a sanction. A monetary fine diminishes someone's power to pay for goods and services without necessarily eliminating it completely. For simplicity's sake I will continue to speak of all these cases as loss of power.

A further clarification concerns the idea that people who perform acts of type A typically intend to exercise power P in the future. Many different concepts of typicality might be invoked here: a statistical generalization, some sort of normative characterization, or something else. As we shall see, the argument that there is a sanction-based norm on testimony will not depend on the details of what counts as typical. The power that ought to be diminished when one issues false testimony is a power that speakers must intend to keep exercising, if they intend to continue offering testimony. This should be sufficient to meet clause (ii) on any plausible definition of typicality.

Another clarification concerns the application of the sanction specified in clause (i). Clause (i) states that, whenever X violates the norm, it is appropriate for X to be deprived of some power P; where “appropriate” means that the deprivation of P is itself governed by norms and can be done properly or improperly. Call the norms governing deprivation of P meta-norms on the application of the sanction-based norms. Someone will be responsible for depriving X of P, and thus subject to the meta-norms. In some cases this will be a designated person, group, or institution, as when referees apply penalties in sports, or when legal institutions apply legal sanctions. But a sanction can exist without such a special authority, as long as whoever applies the sanction is subject to meta-norms. Take a pickup game of basketball in which players call their own fouls; it is appropriate for a player to call a foul and exact the resulting penalty when and only when the foul has actually been committed.

Let J be the potential applier of the sanction. In some cases J will not be aware of a violation of the sanction-based norm or will have misleading evidence that the norm has been violated. Whether it is appropriate for J to apply the sanction in this case depends on whether we take an internalist or externalist view of the meta-norms. On an internalist view, whether it is appropriate for J to deprive X of P depends on whether J has evidence that X has violated the norm. For instance, in one sense it is not appropriate for judicial officials to impose a penalty unless they have sufficient evidence that the defendant has broken the law, even in cases where the defendant did break the law.⁶ On an externalist view of the meta-norm, it is appropriate for J to deprive X of P if and only if X has violated the norm; but when J's evidence is misleading, J is blameless for violating the meta-norm. In law, for an innocent person to be convicted is a miscarriage of justice even when the evidence points to their guilt, while for a guilty person to go unpunished can be what Forst calls an “error of impunity” (Forst 2004: 23). There is then a sense in which it is appropriate for the apparently innocent criminal to be punished and for the apparently guilty non-criminal to be let off, even though the judicial actors should not be blamed for following the evidence. I will not commit to either the externalist or internalist view of meta-norms; as we shall see, each view supports the existence of the credibility-based norm of testimony, so long as it is paired with a fitting view of the epistemology of testimony.

Note that J need not see themselves as enforcing a sanction when they deprive X of power P. Suppose that VIPs at a certain concert are issued passes for backstage access, and that security guards have been instructed to turn away anyone without a pass. It is a norm that VIPs keep their passes with them, and the sanction for violating the norm is that they should be denied backstage access. The guards enforcing the sanction need not see themselves as enforcing this particular sanction; they only need turn away everyone without a pass. There will still be a meta-norm that it is appropriate for the guards to turn away VIPs if and only if they do not have their passes.

Finally, (SC) states that violations of the norm make it appropriate for the sanction to be applied. This does not mean that the sanction usually will be applied for violating the sanction, only that someone who fails to apply the sanction to a norm-violator has thereby behaved inappropriately. (At least if they have evidence that the norm has been violated, per the above discussion of internalist and externalist views of meta-norms.) One can imagine a law that is systematically neglected but nevertheless remains on the books, so that those who fail to enforce it have violated their own obligations to enforce the law. Suppose (as is arguably true) that police who brutalize citizens are rarely prosecuted and convicted. Legal sanctions for violating laws against police brutality will rarely be applied, but that does not mean that police brutality is legal. It means that errors of impunity are widespread. The rare convictions for police brutality will be legally appropriate sanctions for violations of a norm that is real even if it is rarely enforced.⁷

It is true that some norms may fall into such disuse that they are dead letters, such that it would be inappropriate to apply sanctions for violations of them even though they have some formal status. In Deuteronomy 21:18–21 it is said that a stubborn and rebellious son should be stoned to death by all the men of his town, but at least one Talmudic authority comments “‘A stubborn and rebellious son’ – there never was and never will be such” (Tractate Sanhedrin: Mishna and Tosefta 1919: 107). The punishment for a rebellious son is so disproportionate that under this rabbinical authority it may never be applied. Then there is not a religious norm against being a rebellious son with a punishment of death by stoning, for applying the punishment would be impermissible under the religious authority.

The difference between a dead letter and a neglected sanction-based norm is that applying the sanction for a violation would be appropriate in the case of a neglected norm but not for a dead letter. There may be some borderline cases where it is difficult to judge whether it is appropriate to apply the sanction. However, I will argue below that the credibility norm is not one of these borderline cases. The appropriateness of the credibility sanction follows from the hearer's epistemic obligations. So the credibility norm will be a genuine sanction-based norm, even if it happens that the credibility sanction is not generally applied appropriately.

With the sufficient conditions (SC) for the presence of a sanction in hand, our task is to show that speakers of false testimony ought to lose credibility, and that this loss satisfies (SC) and thus qualifies as a sanction.

4. Credibility and testimony: preliminary sketch

Let us begin with an approximate analysis of testimony. Paradigmatically, testimony that p is an utterance meant to provide the hearer with some reason to believe p, where that reason itself comes from the fact that the utterance has been presented as testimony.⁸ This is not an exceptionless definition; some acts of testimony may not be meant to give a reason for belief, as in complicated double bluffs and entries in private journals (if those are testimony).⁹ Still, these acts can only count as testimony against the background of a practice in which testimony generally is meant to provide a reason for belief.

Here is an outline of how the epistemology of testimony gives rise to a credibility-based norm that satisfies the conditions set forth in (SC). When X issues false testimony, hearers have less reason to believe X's future testimony. Hence, if future hearers evaluate their evidence correctly, they will be less likely to believe X's future testimony. This weakens X's power to induce belief through testimony, as in clause (i) of (SC), and X must intend to use this power if X remains in the business of giving testimony, satisfying clause (ii) of (SC). Thus according to (SC) there is a norm whose content is that testimony ought to be true and whose bite is provided by loss of credibility as a sanction. The meta-norm governing the appropriate application of the credibility sanction is epistemic; a hearer who continues to grant X credibility when X's past testimony has been false, or who diminishes their credibility in X when X's past testimony has not been false, is violating the epistemic norms on evaluating testimonial evidence.

The crucial step is that false testimony from X gives hearers less reason to believe X's future testimony. I will first defend this on a crude analysis of the epistemology of testimony, and then show that it holds even on a more nuanced analysis.

On the crude analysis, part of a hearer's evidence for the truth of a speaker's testimony comes from the speaker's past testimony. By simple induction (this is the crude part), X's current testimony provides better evidence if X has consistently told the truth in the past than if X has often offered false testimony in the past. Thus, for X to offer false testimony now gives future hearers less reason to believe X's future testimony. Future hearers should give X's future testimony less credibility, as long as they follow the epistemic norms governing whether to believe X's testimony.

One complication here concerns whether X's past testimony is part of the hearer's evidence for evaluating X's future testimony. Often a hearer will not be aware of X's past testimony and whether it has been true. One response is that X's past testimony may affect future hearers’ evidence even if the hearers are not aware of the past testimony as such. If J has evidence concerning X's reputation as a testifier, and that reputation has been affected by X's past testimony, then J's evidence will come from X's past testimony even if J is not aware of any particular piece of X's past testimony. Another common piece of evidence concerning X's testimony may be X's institutional position; J ought to be more likely to believe the testimony of a reputable journalist or of a professor in a relevant field than of a crank on YouTube.¹⁰ (Again, this reflects what J ought to believe, not necessarily what J does believe.) Insofar as X's institutional position depends in part on the truth of their testimony, their past testimony will affect J's evidence concerning X's current testimony, without J having to monitor X's past testimony.¹¹

Nevertheless, sometimes X's past testimony will not have been incorporated into J's current evidence. Suppose X has issued past false testimony of which J is unaware either directly or indirectly. Whether we say that J ought to believe X's current testimony will depend on our view of the epistemology of testimony. Consider first a reliabilist view, on which a testimonial belief is justified if and only if it is arrived at through a reliable process. As Shieber (2013) and Goldberg (2010) have argued, on such a view it is plausible that the processes by which X produces testimony ought to be taken to be part of J's belief-forming process. X's past false testimony ipso facto decreases the reliability of X's testimony-producing process, and thus decreases J's justification for believing X's current testimony. Hence, according to the epistemic norms, J ought to deprive X of some of X's power to induce belief through testimony; though we would hold J blameless for violating these epistemic norms by believing X, since J has no access to the facts about X's reliability. This is in accord with an externalist view of the meta-norms on applying the sanction: someone who is reasonably unaware of a norm violation and thus fails to apply the sanction violates the meta-norm, but the violation of the meta-norm is blameless.

Consider next an internalist view, on which J's justification for believing X's testimony depends only on the evidence available to J concerning that testimony. On this view, J will be justified in believing X's testimony, because J's evidence does not include X's false testimony or anything stemming from it. So J is not justified in applying the credibility sanction to X, even though X has in the past violated the norm that testimony should be true. This is as predicted by the internalist view of the meta-norms on applying a sanction. When J is reasonably unaware of a norm violation, it is inappropriate for them to apply the sanction even though the norm has been violated; just as on the internalist view it is inappropriate for judicial actors to convict even a guilty person when there is no evidence of the crime. The sanction-based norm is in effect but inapplicable.

Thus, if we are internalist about both the epistemic norms and the meta-norms, J both follows the epistemic norms by not granting X less credibility and follows the meta-norms by not applying the sanction for X's violation of the norm on testimony, since J lacks evidence of the norm violation. If we are internalist about neither sort of norm, in order to follow the epistemic norms J must grant X less credibility, and in order to follow the meta-norms J must apply a sanction for the norm violation, though in both cases J is blameless if they fail to do so. Either way, the epistemic norms concerning credibility are followed exactly when the meta-norms concerning applying the credibility sanction are followed; as predicted by the analysis on which the epistemic norms just are the meta-norms. A mismatch would arise only if we were internalist about one sort of norm but not the other, and it is not clear what might motivate that.

Another possible objection concerns epistemic injustice. As Miranda Fricker has observed, hearers are often wrong about what is a sign of credibility; there is a distinction between “indicator-properties … which are such that they do in fact reliably indicate rational authority; and what we might call working-indicator properties, which are those properties actually used in a given practice to indicate rational authority” (Fricker 1998: 186).¹² More perniciously, as Fricker argues, working-indicator properties are often associated with high social status, so that hearers are less likely to believe women, African Americans, and members of other marginalized groups, simply because they belong to those groups. This unjustly deprives members of marginalized groups of the power to induce belief through testimony even in the absence of past false testimony. It seems as though these speakers incur the sanction for false testimony regardless of whether they have violated the norm against false testimony.

In fact, epistemic injustice is entirely compatible with the credibility norm on testimony. Epistemic injustice takes place when hearers give members of marginalized groups less credibility than they should. This is compatible with loss of credibility being a sanction for false testimony, which is too often misapplied to members of marginalized groups who have not violated norms on testimony. This sort of misapplication is depressingly common with other sorts of sanction, where members of oppressed groups can be systematically punished for breaking laws that they have not in fact broken.¹³ In this way epistemic injustice is like some forms of legal injustice: sanctions are applied to people who have not violated the relevant norms. In order to explain this sort of injustice we need to be able to point to the law that remains on the books and the testimonial norm that remains in effect; these particular injustices arise from a mismatch between how the norm is applied and how the norm ought to be applied.

5. Beyond crude induction

The previous section sketched an account on which, by simple induction, false testimony from X gives future hearers less reason to believe X's future testimony. But simple induction is too simple; speakers are not black boxes that produce true or false testimony. In this section I sketch a more sophisticated account of the epistemology of testimony and show that on it false testimony from X still will weaken future hearers’ reason to believe X's testimony.

Consider first an internalist view of the epistemology of testimony. Suppose that J is aware that in the past X testified falsely that p. Call a speaker sincere insofar as they tend to testify only to propositions that they believe, and competent on a topic insofar as they tend to form true beliefs on that topic. Either X did not believe that p, demonstrating a lack of sincerity, or X falsely believed that p, demonstrating a lack of competence on the topic of p. If sincerity and competence can be inductively projected, then X's false testimony indicates that in the future either X will be less likely to believe their testimony or their beliefs on that topic will be more likely to be false.¹⁴ Whichever holds, X's future testimony on that topic is more likely to be false.

At first glance, this less crude account narrows the effect of X's testimony on X's future credibility. If X shows a lack of competence on a topic, that impugns their future testimony on that topic, but not all future testimony. This might seem to undermine clause (ii) of (SC), that the sanction on an act of type A is a loss of power that people who perform A typically intend to exercise in the future. If false testimony only ought to weaken the speaker's power to induce belief on the topic of that testimony, it may be less plausible that they typically intend to continue exercising that power as opposed to the power to induce belief through testimony tout court.

One response would be that someone who testifies on a topic typically does intend to testify again on that same topic, so that loss of power on that topic does meet satisfy clause (ii) on (SC). Beyond that, I deny that false testimony should only affect the future credibility of testimony on the same topic. Whether a speaker is willing to offer testimony on a topic need not be independent of their competence on that topic. Some people are more likely to volunteer testimony on topics that they know about and to demur from offering testimony when they do not know about the topic. Let us give the name circumspection to the property of only testifying on topics on which one is competent.¹⁵ To the extent that X has demonstrated circumspection, the very fact that X has testified on a topic provides evidence that X has competence on that topic, and thus strengthens the hearer's reason to believe X's testimony.

Consider again the epistemic effects of false testimony that p by X, which must be either insincere or incompetent. Evidence of X's insincerity would provide evidence against all X's future testimony, so let us focus on the case in which X is incompetent on the topic of the testimony. In this case X's false testimony provides evidence not only of incompetence on this particular topic but also of a lack of circumspection. Evidence that X lacks circumspection weakens hearers’ reason to believe X's future testimony on any topic on which X has not positively demonstrated competence, by weakening the inference from “X has provided testimony on this topic” to “X is competent on this topic.” Thus X's false testimony ought to diminish X's power to induce belief through testimony on a wide variety of topics, which will be a power that X typically intends to continue exercising even if X does not intend to offer testimony on this narrow topic again.

Even the more sophisticated analysis of testimony that I have offered here is a simplification. Sincerity and circumspection will not be monolithic but may sometimes depend on the topic and audience. James is generally honest but insists that he doesn't wear a toupee, while Cody lies to everyone but his mother, and Prof. Roberts is meticulously careful not to exceed her competence on scientific matters but will offer unsupported testimony about history. One can even construct examples in which a particular instance of false testimony ought to increase the credibility of some of the speaker's future testimony. But these cases will be exceptional; in the vast majority of cases false testimony will provide evidence against the speaker's future credibility. (Perhaps with the exception of false but justified testimony, which I will discuss in the next section.)

The analysis of this section has presupposed an internalist epistemology, but reliabilism yields much the same analysis. Again, following Goldberg (2010) and Shieber (2013), it is plausible that reliabilists should take the process of formation of a testimonial belief to comprise the processes by which the speaker produces the testimony. These processes will include processes by which the speaker comes to form beliefs on a certain topic, to choose to offer testimony on a topic on which they are or are not competent, and to offer testimony that does or does not reflect their beliefs. False testimony will lower the actual reliability of the relevant processes, thus weakening the evidence provided by future testimony produced by the same processes. Accordingly, on a reliabilist view false testimony will justify the loss of future credibility, even when the method whose reliability is taken to determine the justification of the future testimonial belief is taken to be not “Believe testimony from this speaker” but “Believe testimony from this speaker, who has come to produce this testimony by these particular processes.”

6. Beyond truth and testimony

I will conclude with some brief remarks about applying the credibility norm to properties other than truth and to assertions that are not testimonial.

In the previous section, I mentioned one possible exception to the claim that the vast majority of false testimony provides evidence against the speaker's future credibility: justified false testimony. A speaker who sincerely expresses a false belief that is supported by their evidence may be demonstrating competence on the topic of the testimony, since forming beliefs that are supported by the evidence is likely to result in true beliefs on the topic in the future. By demonstrating competence, the testimony ipso facto demonstrates circumspection. Conversely, unjustified sincere testimony provides evidence of a lack of competence and thus of a lack of circumspection, even when the particular testimony is true. This suggests that loss of credibility might be a sanction that is made appropriate by unjustified testimony, rather than by false testimony. False testimony would usually make the sanction appropriate, but only because false testimony is usually unjustified. The content of the norm whose bite is provided by the credibility sanction would be that testimony ought to be justified, rather than that testimony ought to be true.

Hence there is a live question about which property of testimony best explains its effect on future hearers’ reasons to believe future testimony. A proponent of a truth norm might respond that future hearers’ evidence depends more on the truth of past testimony than on its justification, because evidence about the truth of past testimony is easier to come by than evidence about what evidence the speaker had at the time. There may be other properties that could be argued to be more predictive than justification or truth, such as the property of having been formed in a reliable way. Resolving this question will require further work in the epistemology of testimony.

But no matter how it is resolved, there will be some property F such that testimony that lacks F ought to be met with the credibility sanction. So there is a norm whose bite is provided by the credibility sanction. Working out the exact content of the norm requires a non-standard argument of the sort provided herein, in which we begin by showing that testimony must be subject to a norm with the credibility bite, and we find a content that is appropriate to that bite.

Finally, a word on extending the credibility norm to non-testimonial assertions. Many assertions are not testimonial in that they are not meant to provide a reason for belief that is rooted in the assertion itself. When someone issues a reminder that the meeting starts in fifteen minutes, the intention is to bring to the foreground something the hearer already had reason to believe. When someone points out a blackbird in a tree that their hearer is looking at, the intention may be to bring the hearer to look for themselves and acquire their own reason for believing that the blackbird is there.

These non-testimonial assertions plausibly affect future credibility much as testimony does. When a speaker says “Look, there's a blackbird” and the bird is a starling, it demonstrates incompetence and lack of circumspection, even though the assertion was not itself meant as a reason to believe. The hearer has less reason to trust the speaker's future testimony about bird identification (and about other topics, as mediated by circumspection).¹⁶ This loss of future credibility will count as a sanction if typical asserters intend to induce belief through testimony in the future. While this may not be as immediate as the fact that typical testifiers intend to continue offering testimony, it is difficult to imagine a human who makes assertions but intends never to offer testimony. Accordingly, our non-standard argument may apply to all assertions: because of the epistemic connections between testimony and other assertions, there is a norm on assertion whose bite is that violating the norm justifies a loss of credibility in the speaker's future testimony, and whose content is suited to that bite, just as for the norm on testimony itself.¹⁷