Testimonial Injustice and Mutual Recognition

1. Introduction

Can failing to accord someone sufficient credibility harm or wrong that person? Much of the recent work on the moral dimensions of testimony suggests that it can. On a widely influential view, a failure to accord sufficient credibility to a speaker, when that failure is prejudicially motivated, can constitute a testimonial injustice against that speaker.

How might we understand the harm or wrong that a speaker stands to incur when she is prejudicially not believed? How, in other words, to understand the nature of the harm or wrong at the heart of testimonial injustice? A natural first answer might involve identifying the various negative consequences a speaker stands to incur if she is deemed insufficiently credible. Consider, for example, a defendant in court who, in virtue of some aspect of her social identity, is prejudicially perceived by the members of the jury to be unreliable, and whose testimony in court consequently fails to gain epistemic uptake. It goes without saying that not taking the defendant to be credible in this context can profoundly negatively affect her in any number of ways. There is a familiar sense, then, in which not being accorded sufficient credibility might harm or wrong a speaker: it can have wide-ranging consequences that stand to leave that speaker significantly worse off.

The negative consequences that a speaker stands to incur by not being accorded sufficient credibility are what Miranda Fricker and others have called the secondary harms of testimonial injustice.¹ These harms are “secondary” insofar as they are caused by, but not intrinsic to, not being taken as credible in a particular context. Many philosophers, like Fricker, who hold that people can be wronged by not being accorded sufficient credibility, do not think that the potential to be wronged in this way is exhausted by the potential to incur the specific negative consequences that, say, a particular defendant stands to incur if her courtroom testimony is not taken seriously. One can experience testimonial injustice in the courtroom, but also in the classroom, as well as in everyday conversation. Depending on context, or luck of circumstance, the actual consequences of not being believed may vary. The thought animating much of the discussion on testimonial injustice is that, apart from the secondary harms that might befall a speaker, there is some primary harm, or wrong, that the speaker stands to incur if she is not accorded sufficient credibility, owing to identity prejudice. It is this primary harm that unifies an otherwise widely diverse range of cases in which speakers are prejudicially not accorded sufficient credibility.

How should we understand the nature of this primary harm? I begin by examining different accounts of the primary harm of testimonial injustice that aim to locate it in some kind of failure of the hearer to appropriately recognize a speaker’s epistemic agency. On Fricker’s epistemic-objectification account of the primary harm, the hearer fails to appropriately recognize a speaker’s epistemic agency insofar as he treats her merely as a passive “source of information,” rather than as an active “informant.”² On what I call conditional-recognition accounts, the hearer fails to recognize a speaker’s epistemic agency insofar as that recognition is conditional on the degree to which the speaker and her testimony conform to certain expectations the hearer has about what that speaker can epistemically contribute.³ I argue that these ways of cashing out the relevant failure of recognition at the heart of the primary harm of testimonial injustice face certain difficulties. Some of these accounts are overly narrow, making it difficult for them to accommodate and fully explain seemingly paradigmatic cases of testimonial injustice. Other accounts are overly broad, counting what I take to be intuitively unobjectionable ways of treating other epistemic agents as morally problematic and leaving the central wrong-making features of cases of testimonial injustice seemingly unexplained.

I then offer an alternative account of the failure of appropriate recognition at the heart of testimonial injustice, one that draws on T. M. Scanlon’s widely influential development and defense of moral contractualism.⁴ On my account, it is often the case that a hearer who fails to accord sufficient credibility to a speaker, owing to an identity prejudice against that speaker, could not justify that failure to that speaker. The relevant sense of justifiability at issue here is to be understood in broadly contractualist terms: such a hearer could not justify her failure to accord sufficient credibility to a speaker to that speaker, in the sense that that speaker can reasonably reject principles that would permit the hearer to accord that speaker insufficient credibility on the basis of some identity prejudice. I argue that the key elements of a contractualist framework can better illuminate the sense in which the primary harm of testimonial injustice centrally involves some kind of failure of appropriate recognition. In contractualist terms, being unable to justify one’s response to a speaker’s testimony entails a kind of failure of mutual recognition, or a failure to acknowledge that speaker as having standing to co-determine the terms on which the hearer and speaker, as epistemic agents, are to engage with one another in testimonial exchanges.

This paper aims to contribute not only to discussions of testimonial injustice, but also the much wider-ranging spate of recent work on whether what we believe about a person can morally wrong them.⁵ Though I do not have the space here to engage directly with this wider debate, some key elements of the contractualist framework I develop in this paper for thinking about the primary harm or wrong of testimonial injustice may be usefully extended to this debate, as a way to explain in virtue of what our beliefs and other doxastic behavior, independently of our actions, might wrong others.

2. The Primary Harm of Testimonial Injustice: Some Preliminaries

A testimonial injustice occurs, according to Fricker, when a hearer accords insufficient credibility to a speaker, owing to negative identity-prejudicial stereotypes the hearer has against that speaker. To put the idea somewhat crudely, a hearer doesn’t believe the speaker, because, e.g., the speaker is a woman, or because the speaker is black, and so on. One case that Fricker uses to bring out the core features of testimonial injustice comes from the original screenplay for The Talented Mr. Ripley.⁶ A moment that nicely crystalizes the phenomenon occurs when Herbert Greenleaf says to Marge Sherwood, in dismissing her testimony about the mysterious disappearance of Greenleaf’s son, Dickie: “Marge, there’s female intuition, and then there’s facts” (Fricker 2007: 9).⁷ Even though Sherwood is epistemically well positioned to provide useful testimony on this matter, Greenleaf’s gendered prejudice against Sherwood leads him to accord her insufficient credibility about his son’s disappearance.

Central to Fricker’s account of testimonial injustice is the concept of a negative identity-prejudicial stereotype. As Fricker defines them, stereotypes are widely held associations between certain social groups or identities and certain traits (2007: 30–32). So understood, the reliance on stereotypes in thought are not as such objectionable, epistemically or otherwise. Many important cognitive tasks depend for their success on our ability to speedily process and respond to new information, and the use of stereotypes as heuristics play a role in facilitating many of those tasks. Making quick decisions about how much credibility to accord a speaker in a given testimonial exchange is just one of those tasks.⁸ Negative stereotypes are stereotypes that link social identities with generally bad or undesirable traits. For our purposes, we’re mainly interested in negative stereotypes that involve undesirable epistemic traits, broadly conceived.⁹ One might, for example, harbor stereotypes that associate women with negatively valenced epistemic traits like being overly emotional, or flighty, or illogical.

What makes a stereotype prejudicial? On Fricker’s account, whether or not a stereotype is prejudicial depends not on the content of that stereotype, but rather on how that stereotype is held. Stereotypes are prejudicial only if they exhibit a certain degree of recalcitrance in the hearer’s psychology—they would persist even in the face of counterevidence that would undermine the relevant association that the stereotype embodies (Fricker 2007: 32–35). Why, in the presence of counterevidence, might prejudicial stereotypes persist? On Fricker’s account, this is often because they are rooted in some kind of “affective investment” on the part of the agent who has them. Someone who harbors prejudicial stereotypes is in some sense attached to the truth of the generalizations those stereotypes embody—perhaps because they cohere with or justify one’s other attitudes or commitments.¹⁰

Our focus going forward will be on how to make sense of the idea that when a hearer’s failure to accord a speaker sufficient credibility is grounded in the operation of negative identity-prejudicial stereotypes the hearer has against the speaker, that hearer stands to harm or wrong that speaker. In particular, our focus will be on the primary harm distinctive of testimonial injustice. Fricker begins her discussion of the primary harm of testimonial injustice by noting that

The primary harm is a form of the essential harm that is definitive of epistemic injustice in the broad. In all such instances, the subject is wronged in her capacity as a knower. The form that this intrinsic injustice takes specifically in cases of testimonial injustice is that the subject is wronged in her capacity as a giver of knowledge. (2007: 44)

Our question is: What is it to be wronged in this way?

It’s worth noting that Fricker often characterizes the primary harm that a speaker incurs in terms of that speaker’s having been wronged. But one may wish to distinguish between harms and wrongs. One might think, for example, that an important difference between being harmed and being wronged consists, very roughly, in something like incurring a setback to one’s interests, on the one hand, and having a right against being treated in a particular way, on the other. One can incur a setback without having been treated in a way to which one had a right not to be treated, just as one can maintain a right against a certain form of treatment without having been left worse off by it.

Since the discussion in Fricker, and in the work of those responding to her work, has largely been put in the language of harms, I will continue to mainly refer to the primary harm of testimonial injustice. But those who want to maintain a distinction between harming and wronging can read my discussion as one that centrally focuses on the primary wrong of testimonial injustice. I am primarily interested in providing a way to explain how a speaker might have a right against having her testimony being prejudicially discounted, in virtue of which that speaker can sensibly be wronged by hearers who do prejudicially discount her. What does a hearer owe a speaker, such that a hearer’s failing to do what is owed can wrong that speaker? That is the central question that guides the paper.

Before we turn to consider different accounts of the primary harm of testimonial injustice, I want to briefly address a worry one might have about the very possibility of any such account. Our assessments of the credibility of speakers are largely automatic and unreflective. Though it is possible for us to deliberate about whether or not acting in some way would harm or wrong someone, and to decide on that basis to act or refrain from acting, our credibility assessments are not similarly guided by deliberation of this sort. If that is so, accounts of the primary harm or wrong of testimonial injustice threaten to be idle.

It is of course true that our credibility assessments are only rarely the product of any explicit deliberation, but that does not mean that they are beyond our capacity for reflection and normative evaluation. We can call into question our credibility assessments and consider whether or not they may have been distorted by our various biases and prejudices, and consequently determine in some cases that we should have heard a speaker out with a more open mind. We can, in hindsight, compare our reactions to different speakers with the same credentials and conclude that we judged some of them unfairly. To put it simply, reflection on how we go about according different speakers varying degrees of credibility has a familiar normative dimension, even when our credibility assessments are themselves automatic.

Note, too, that it is natural for us to experience regret, or to engage in certain forms of self-reproach, upon realizing that we’ve prejudicially assessed a person’s credibility. These reactions make sense, in a way that they wouldn’t with respect to paradigmatically involuntary behavior. One can feel badly for having made a credibility assessment on the basis of identity-prejudice, rather than on the quality of the evidence that the speaker provided. And that sense of regret or self-reproach doesn’t seem to be explained by the observation that we may have judged things inaccurately. When I realize that prejudice has led me not to take a speaker’s word seriously, the source of my regret for having done so is not merely that I missed out on coming to have a true belief. It is that I have somehow failed to give that speaker proper weight in my view of things. I have, in some sense, treated that speaker unfairly.

What might be the source of that unfairness? What did I owe it to him to do? We can sensibly address these questions even if our credibility assessments are not the sort of thing we arrive at on the basis of explicit moral deliberation. One can see the various accounts we will examine below, including my own, as offering different ways to guide normative reflection about our credibility assessments and to help us make sense of these reactions.

3. The Primary Harm of Testimonial Injustice as Epistemic Objectification

The idea that there might be a primary harm intrinsic to testimonial injustice— one distinct from the specific material consequences a particular instance of testimonial injustice might result in—can seem mysterious. If we take all of those away, what could be left? A powerful idea shared among many views of the primary harm of testimonial injustice is that what remains is a hearer’s failure to appropriately recognize a speaker’s epistemic agency, in some important sense. Or, to put it another way, to base one’s assessment of a speaker’s credibility on some issue on an unrelated feature of their identity, rather than on the quality of the reasons or evidence she provides in her testimony on that issue, is to fail to relate to that person as one ought. One does not relate to an epistemic agent as an epistemic agent when one fails to treat that person’s testimony as appropriately reason-giving. It is this failure to recognize or relate to another epistemic agent as such that constitutes a wrong done to that epistemic agent.

There are different ways of clarifying the relevant failure of appropriate recognition at issue. Fricker suggests that the relevant failure to appropriately recognize a speaker in such cases consists in a form of epistemic objectification. What is it to epistemically objectify someone? Consider the distinction Fricker offers between two ways we might epistemically relate to other people. We can treat them as “informants,” as epistemic agents capable of conveying knowledge and information and engaging in the practice of giving and taking reasons for belief. But we can also treat people as “sources of information,” as things from which we can glean information.

Of course, treating others as sources of information is not in itself objectionable, as Fricker readily acknowledges. It is when one’s treatment of another person as a source of information “denies their epistemic subjectivity” or “undermines their general status as a subject of knowledge” that such treatment epistemically objectifies that person (Fricker 2007: 134). In such cases, speakers are treated as mere sources of information.

Are cases of testimonial injustice cases in which hearers treat speakers as mere sources of information? Fricker is clear on this point: “The presence of any significantly identity-prejudicial attitudes against the speaker will always undermine their general status as a subject of knowledge, and so cannot fail to render any resultant epistemic objectification morally bad” (2007: 135). But it’s not clear that even paradigmatic cases of testimonial injustice are cases in which hearers treat speakers as mere sources of information, by failing to treat those speakers as epistemic agents in some way or other. Consider Greenleaf’s gender-prejudicial dismissal of Sherwood’s testimony about her fiancé’s mysterious disappearance. It is not clear why we should think that Greenleaf’s failure to believe Sherwood, objectionable though it is, stems from a denial of Sherwood as an epistemic agent.

Indeed, it is not just that we don’t have reason to think that any particular prejudicially grounded failure to believe a speaker is rooted in some deeper denial of that speaker’s epistemic agency. Rather, a hearer’s prejudicially motivated failure to accord sufficient credibility to a speaker in one domain is often explained by that hearer’s according more credibility to the speaker in some other domain. Our ability to provide this sort of explanation often depends on our ability to see a hearer as treating that speaker as someone who certainly does possess epistemic agency or subjectivity, even if that treatment is objectionably circumscribed. If so, then the primary harm incurred by a speaker in a case of testimonial injustice is not that of having one’s epistemic subjectivity denied.

Cases that bring this point out are cases in which the operative stereotypes that mediate the hearer’s judgment of a speaker are double-sided in an important respect: a person elevated in one domain by a particular stereotype can, by that same stereotype, be devalued in another. For example, familiar stereotypes linking women with emotional intelligence may lead a hearer to elevate a woman’s credibility when she offers up an interpretation of a friend’s perplexing non-verbal cues. But that stereotype can also cost the speaker, insofar as it is taken as evidence of a deficiency somewhere else, as in the ability to coolly and rationally assess a situation. Insofar as perceived superiority in one domain comes at the cost of perceived inferiority in another, then just as a woman may be accorded quite a bit of deference when it comes to matters acquired by empathic attunement, she may, by that same token, be accorded insufficient credibility when it comes to matters acquired through disinterested ratiocination.¹¹

If Fricker’s epistemic-objectification account is right, then even paradigmatic cases of testimonial injustice that involve the operation of prejudicial stereotypes that are double-sided in the relevant sense would seemingly be excluded from that analysis. We should, I maintain, prefer an account of the primary harm that can accommodate cases in which the hearer seemingly does recognize the speaker as an epistemic agent, albeit in overly narrow and domain-specific ways.

Aidan McGlynn has recently argued that we can save the epistemic objectification account of the primary harm of testimonial injustice by developing a broader conception of what epistemic objectification consists in. He notes that Martha Nussbaum’s analysis of objectification, from which Fricker develops her own account, actually affords more resources from which to draw than Fricker’s account might suggest. For Nussbaum (1995), objectification is a kind of cluster concept; there are many ways to treat a person as an object besides treating that person as merely inert, or as a passive thing that lacks agency. Objectification can also consist in treating a person as an instrument for one’s purposes, or as treating a person as fungible or interchangeable with similar others. McGlynn’s suggestion is that we might defend an account of the primary harm that encompasses a wider range of epistemic analogues of Nussbaum’s various forms of objectification—in particular, the epistemic analogues of instrumentality and fungibility.

Epistemic instrumentality, as McGlynn conceives of it, is to treat “what [speakers] say as of epistemic relevance—as potentially yielding worthwhile information—only insofar as one treats it as fodder for [the hearer’s] own information-gathering projects and insinuations …” (in press: 14). Just as one might use another person as a tool for one’s own sexual purposes, so too, McGlynn suggests, might one use another person as a tool for one’s own epistemic purposes. Epistemic fungibility involves “treating members of a certain social group as interchangeable for epistemic purposes; testimony from one member of the group (on the subject at hand, at least) is as good as any other” (in press: 13). Just as one might treat another person as effectively replaceable with others perceived to be equally able to satisfy one’s sexual purposes, so too might one treat another person as effectively replaceable with others perceived to be equally able to satisfy one’s epistemic purposes.

But epistemic instrumentalization and epistemic fungibility, as McGlynn conceives of them, do not appear, as such, to be obviously problematic. If that is so, then the focus on these forms of treatment may leave the central wrong-making element in cases of testimonial injustice unexplained. Consider epistemic instrumentalization. We often treat what people have to say as having relevance only insofar as what they have to say bears on our own epistemic interests—as we do when, for example, we’re engaged in a time-sensitive inquiry, or when we’re undertaking a targeted investigation into some particular matter. I may treat the conversation that a fellow passenger is having about the weather at our destination city as having epistemic relevance only insofar as what he knows can serve as “fodder” for my own “information-gathering projects” (namely, the project of figuring out whether to book a cab or walk to my hotel). I appear to epistemically instrumentalize this person in the way that McGlynn describes, insofar as I treat his “epistemic agency as merely a means for carrying out [my] own epistemic projects, and fail to see [him] as capable of pursuing [his] own epistemic agenda, from which [I] might stand to learn” (in press: 13). In a very minimal sense, I fail to see him as capable of pursuing his own epistemic agenda, insofar as I do not otherwise give thought to what this person’s own epistemic agenda may be, or what I might learn from him. Surely I have not treated my fellow passenger in any sort of objectionable way, even if my interest in him, qua epistemic agent, is strictly limited to the role that his information can play in my own practical deliberation. Of course, I wouldn’t deny that the fellow passenger is capable of pursuing his own epistemic agenda. But that is not required in order to epistemically instrumentalize a person. McGlynn is clear that one can epistemically instrumentalize someone without treating that person as inert, or as someone who does not possess epistemic agency (see especially McGlynn in press: §2). That’s the point, for McGlynn, of expanding the concept of epistemic objectification to include forms of treatment that go beyond treating another person as merely inert.

A similar point can be made about epistemic fungibility. We can perfectly well treat individual epistemic agents as epistemically fungible, in McGlynn’s sense, without thereby wronging them. One might collect testimony from fellow members of an affected group in a preliminary inquiry in order to decide whether to pursue a class-action lawsuit on behalf of that group. The testimony of one affected member is “as good as any other,” in the sense that, relative to the epistemic goals of this particular inquiry, one individual affected member’s testimony is treated as interchangeable with the testimony of any other individual affected member. Of course, no one would deny that each individual affected member has their own story to tell, and their own perspective on the matter. But, again, McGlynn is explicit that such a denial is not necessary in order for epistemic fungibility to be objectionable.

McGlynn’s point that we ought to broaden any epistemic objectification account to include other ways of treating people as less-than-full epistemic agents suggests that we might be able to just drop the focus on epistemic objectification in particular, and redirect attention to other ways of offering epistemic agents only limited forms of recognition. I consider such accounts in the next section.

4. The Primary Harm of Testimonial Injustice as Conditional Recognition

Some philosophers have argued that the primary harm consists in the harm of being offered what we might call conditional recognition of one’s epistemic subjectivity or agency. This is recognition that is conditional on whether the speaker’s epistemic subjectivity or agency somehow conforms to the hearer’s narrow expectations about what that speaker is capable of contributing. I will consider two such conditional-recognition accounts here.

Gaile Pohlhaus Jr. argues, against Fricker, that the primary harm of testimonial injustice does not consist in a hearer’s denial of a speaker’s epistemic subjectivity or agency. Rather, that primary harm consists in the fact that hearer recognizes the speaker’s epistemic subjectivity, but only insofar as the speaker stands to confirm or otherwise support the hearer’s own beliefs and worldview. The speaker is “relegated to the role of epistemic other, being treated as though the range of [the speaker’s] subject capacities is merely derivative of [the hearer’s]” (Pohlhaus 2014: 107). The primary harm that a speaker incurs in a case of testimonial injustice consists, not in epistemic objectification, but in epistemic derivatization.¹²

Pohlhaus characterizes epistemic derivatization in the following way. A hearer epistemically derivatizes a speaker when the speaker is

treated as if her own lived experience from which she draws in order to add to the communal knowledge pool is simply a mirror (or perhaps a shadow) of [the hearer’s] own, but certainly not capable of contributing to our understanding of the world beyond (and in ways that might change the shape of) the scope of the [hearer’s] experienced world. (2014: 106)

Perhaps the most straightforward way of reading this characterization of epistemic derivatization is to see the hearer as presupposing (falsely) that she and the speaker share relevantly similar lived experiences and that they possess a broadly similar epistemic outlook, and so there’s nothing new the speaker might contribute. But it’s not clear how that could be what’s going on in paradigmatic cases of testimonial injustice. Greenleaf, for example, doesn’t take Sherwood to reflect his own lived experiences and to share his epistemic outlook—after all, he characterizes her as, unlike himself, cognitively ruled by her emotions.

There is a second key aspect of epistemic derivatization, as Pohlhaus characterizes it. It’s not just that epistemic derivatization involves something like a presumption of epistemic sameness, in virtue of which the derivatized speaker is treated “as though she has nothing unique to contribute to the intersubjective relations that maintain epistemic practices …” (2014: 106). Epistemic derivatization appears to involve, moreover, a rejection of epistemic difference. As Pohlhaus puts it, the speaker “is not permitted to contribute in ways that would redirect epistemic practices toward those parts of her experienced world that extend beyond or trouble the veracity of the dominantly experienced world” (2014: 107). Broadly speaking, the rejection of epistemic difference involves some kind of failure on the hearer’s part to be disposed to extend epistemic uptake to another person, in virtue of perceived epistemic difference.

But I’m not sure whether this gets us closer toward an explanation of the primary harm of testimonial injustice. One might grant that paradigmatic cases of testimonial injustice are cases in which a hearer is not disposed to grant a speaker epistemic uptake in virtue of perceived epistemic difference, without conceding that it’s those failures to be so disposed that wrong or harm that speaker. This is because we simply do not owe it to others, in general, to be disposed to grant them epistemic uptake. And if we do not generally owe it to others to be so disposed, then it’s difficult to maintain that a speaker can be wronged or harmed (in the primary sense at issue) by a hearer who fails to be so disposed.

Ishani Maitra is clear on this point. She provides cases that demonstrate that failing to be disposed to grant another person epistemic uptake, even when that failure is rooted in an identity prejudice, does not intuitively wrong or harm the speaker (Maitra 2010). One such case involves Zara, a person who is not disposed to grant epistemic uptake to an op-ed writer who explicitly self-identifies as a “committed tea-partier,” owing to an identity prejudice Zara has against Tea Partiers (Maitra 2010: 198). As a result, she dismisses his op-ed. In this case, there is a sense in which, though Zara plausibly recognizes the speaker as “capable of performing epistemic labor,” as Pohlhaus puts it, Zara does not recognize him as in a position to “negotiat[e] the direction of that labor” (Pohlhaus 2014: 107). But Zara does not intuitively wrong the Tea Partier.

Consider, also, how a derivatization account might handle cases in which hearers accord speakers excess credibility. Emmalon Davis argues persuasively that an account of the primary harm of testimonial injustice should be able to accommodate cases in which excess credibility is prejudicially accorded to a speaker. In cases of prejudicial credibility excess, a hearer accords a speaker more credibility than she otherwise would, owing to positive identity-prejudicial stereotypes the hearer holds against the speaker. Borrowing from one of Davis’s examples, imagine that a group of high-school students accord excess credibility to their Asian-American classmate about the answer to a difficult math problem, on the basis of a positive identity-prejudicial stereotype they have that associates Asian-Americans with exceptional mathematical skill (Davis 2016: 487). Cases like this one, in which excess credibility is prejudicially accorded to a speaker, are seemingly not cases in which a speaker is treated as though her outlook were merely derivative of the hearer’s own. Indeed, as Davis points out, the speaker is treated in such cases as though his epistemic outlook were “essentially nonderivable” (2016: 490).

Davis offers a way to amend Pohlhaus’s account to better accommodate cases in which speakers are prejudicially accorded excess credibility. On Davis’s account, the primary harm of testimonial injustice consists in what Davis calls epistemic othering, whereby a hearer recognizes a speaker’s epistemic subjectivity, but only in a way that “bypasses or circumscribes the speaker’s epistemic subjectivity” (2016: 490). This form of conditional recognition manifests differently in different kinds of cases. In cases of prejudicial credibility excess, a hearer recognizes a speaker as an epistemic subject, but only insofar as “the speaker might provide some informational service, where the information in question is perceived by dominant hearers to be inaccessible from their own epistemic position” (2016: 490). In other words, a speaker is invited to epistemically contribute in ways that go beyond, and perhaps even conflict with, the hearer’s own epistemic outlook, but that invitation is extended only because of some aspect of that person’s perceived social identity. It is in virtue of this fact that the speaker is treated as epistemically “fungible or interchangeable with others who share the same social identity” (Davis 2016: 488).

For reasons similar to those I raised in the previous section, it’s not clear to me that recognition that is conditional in this sense need harm or wrong a speaker, even when such recognition is grounded in the operation of (positive) identity-prejudicial stereotypes. Consider, for example, a case in which one is disposed to grant a lot of epistemic uptake to a particular serviceperson, Sierra, about instances of sexism on her military base, precisely in virtue of Sierra’s gender identity. In such a case, a hearer may very well be guided by (positive) stereotypes that link women in predominantly male spaces with being especially epistemically attuned to the reality of sexism in those spaces.¹³ And these (positive) stereotypes may very well be prejudicial, in Fricker’s sense, insofar as those stereotypes are held in ways that are resistant to potential counterevidence.¹⁴ Finally, these stereotypes may be grounded in the hearer’s treatment of a speaker in that context as epistemically fungible, at least insofar as the hearer “perceives no salient differences between this speaker and another of the speaker’s same ‘type’ ” (Davis 2016: 488). For example, someone who visits a website filled with testimony from individual women on a military base may not perceive differences between those individual women, simply because he’s not particularly on the lookout for such differences. He simply wants to get a sense of what it’s like to be a woman in the military, not necessarily what it’s like to be Sierra, rather than Tania, in the military. Even if we think his inquiry here is misguided or naïve, it’s not clear that he plausibly wrongs Sierra, simply because he treats Sierra as epistemically fungible for the purpose of getting a better sense of what it’s like to be a woman in the military (if, indeed, there were such a thing as what it’s like to be a woman in the military, full-stop).

To be clear, this isn’t to say that speakers in cases like this one can’t be wronged; the point is that, if they are, it seems it will be for some further reason. And I don’t want to deny that, in some cases, there may be something deeply objectionable about a hearer who is selectively disposed to grant epistemic uptake to a speaker only in some domains rather than others. But it’s not clear to me that the primary harm intrinsic to a particular case of prejudicial credibility excess could be entirely explained in terms of facts about what that hearer is disposed or not disposed to do in cases other than the actual one in which that degree of credibility is accorded.

The accounts we’ve considered in these last two sections aim to locate the primary harm of testimonial injustice, broadly speaking, in the hearer’s failure to appropriately recognize the speaker. They do so by characterizing this failure of appropriate recognition in terms of some specific form of epistemic treatment, whether that consists in epistemic objectification, derivatization, instrumentalization, or treating agents as epistemically fungible. Some epistemic-objectification accounts of the primary harm, I’ve argued, do not appear to have explanatory power across a sufficient range of cases, while some conditional-recognition accounts locate the primary harm in certain forms of epistemic treatment that appear, in and of themselves, to be part and parcel of living in the world as an epistemic agent whose inquiry is often goal-oriented and whose time is not unlimited. In my view, an account of the primary harm of testimonial injustice should be compatible with the fact that our interest in others as epistemic agents can be relatively narrow and strategic, and not necessarily the worse for that.

5. Testimony and Interpersonal Justifiability

The account I want to offer of the primary harm of testimonial injustice grows out of the following core idea: that many cases in which a hearer prejudicially accords insufficient credibility to a speaker are cases in which that hearer could not justify that credibility assessment to that speaker.¹⁵ Importantly, the relevant notion of justifiability here is not that of the traditional concept of epistemic justification. The suggestion that I want to develop here is that how we respond to the testimony of others ought to be interpersonally justifiable, and that according a speaker too little credibility, owing to identity prejudice, is often not something that a hearer could justify to that speaker. The primary harm or wrong of testimonial injustice consists, I want to suggest, in this failure of interpersonal justifiability.

The idea that moral wrongness consists in a form of interpersonal unjustifiability is one of the hallmarks of the moral contractualist view developed and defended by T. M. Scanlon. On Scanlon’s view, how we act ought to be justifiable to others, and behaving in ways that are interpersonally justifiable is just part of what it is to recognize and relate to others on terms of mutual respect. The goal of this section is to draw out some of the key elements of Scanlon’s contractualism and show how they might be applied to the epistemic domain. In the next section, I will use these elements to provide a more specific account of the primary harm of testimonial injustice.

On Scanlon’s well-known contractualist formula, an action is wrong if any principle that permitted that action is one that other agents, appropriately motivated, would have sufficient reason to reject (1998: 153). Appropriately motivated agents are those engaged in the collective enterprise of trying to find principles governing behavior that others, similarly motivated, could not reasonably reject. Broadly speaking, we can see ourselves, on this picture, as agents aiming to jointly determine the terms on which we all shall live. And we do that, in part, by considering and comparing the various objections that other agents, who share our aim, might raise on their own behalf to candidate principles governing our behavior. When we act in accordance with principles that no one would have sufficient reason to reject, we thereby put ourselves in a position to be able to justify our actions to others.

Objections to candidate principles are not difficult to generate; even an individual who stands to be minorly negatively affected by the actions that a principle would permit would have at least some objection that he could raise on his own behalf to that principle. To determine whether some objection to a candidate principle is strong enough to make a rejection of that principle reasonable, on Scanlon’s account, we have to consider the implications of adopting that principle, as seen from each of the individual perspectives of those who stand to be affected by the behavior that principle licenses. To get a more intuitive feel for this idea, suppose you’re trying to figure out whether or not it would be wrong to break a promise you’ve made to me, because having to fulfill that promise would be slightly inconvenient (even though, let’s suppose, you could perfectly well foresee that inconvenience when you made the promise). You would be in a position to justify breaking your promise to me just in case a principle that would allow promisors to break promises made when keeping them would be slightly inconvenient (and foreseeably so), is a principle that I, the promisee, could not reasonably reject. The intuition that breaking such a promise to me would not be something you could justify to me, under these conditions—and, moreover, that you would wrong me by failing to keep your promise—is explained by the fact that I have stronger reasons to reject a principle that would allow you to break your promise, compared with whatever reasons you would have to reject a principle that would prohibit you from breaking your promise under these conditions.

According to Scanlon, acting in ways that we can justify to others allows for the possibility of standing in a distinctively valuable relationship of mutual recognition (1998: 162). Mutual recognition between agents is realized when we recognize each other as both the legislators of, and the subjects to, the terms by which we govern ourselves. When we treat the objections that others might have to principles that govern our actions as objections that we must consider and weigh against our own, we treat others has having a legitimate say in how we ought to act. When we act wrongly, we have not only failed to consider the interests of those affected by our actions; we have also, as Pamela Hieronymi puts it, “denie[d] their standing to (partly) determine the terms on which we each shall live. [We] have, thereby, acted in a way that fails to accord them a certain form of respect” (2011: 108).

In my view, this normative ideal of standing in a relation of mutual recognition with other agents can be usefully applied to the epistemic domain. In what follows, I want to develop and defend the idea that how we respond to the testimony of other epistemic agents ought to be justifiable to them, and that responding to testimony in ways that are interpersonally justifiable allows us to stand in a valuable relationship of mutual recognition to our fellow epistemic agents. The primary harm or wrong of testimonial injustice, I will argue, can be usefully analyzed as a kind of failure of hearers to be able to justify how they respond to the testimony of speakers—and so, as a kind of failure of mutual recognition.

In the next section, I will focus specifically on using these ideas to provide an account of the primary harm of testimonial injustice. But before we get there, it will help to clarify how the key elements of moral contractualism can be applied to the epistemic domain.

The principles relevant in the present context concern the general permissibility or prohibition of certain ways we might respond to the testimony of speakers by according those speakers varying degrees of credibility. The ways in which we take speakers to be more or less credible in particular circumstances can accord or fail to accord with these general principles. For example, a hearer who accords quite a bit of credibility to an expert in ornithology about the migration pattern of rose-breasted grosbeaks, on the basis of the ornithologist’s expertise, is someone whose response to the ornithologist’s testimony accords with a more general principle permitting hearers to believe what an expert in a particular domain says about things within that domain, on the basis of that expert’s perceived expertise about that domain.¹⁶ What reasons might one have to reject principles that govern how epistemic agents respond to testimony? Broadly speaking, reasons to reject principles concern the various ways that one stands to be affected by the general compliance with such principles. The relevant forms of compliance here include one’s own compliance with a principle about how to respond to testimony, as well as the compliance of other epistemic agents with principles about how to respond to testimony. General compliance with principles governing how to respond to testimony can impose various burdens on and benefits to the relevant affected parties, in just the same way that compliance with principles that permit or prohibit various actions can.

To take a simple example, compliance with a principle that would prohibit an epistemic agent from forming any beliefs on the basis of testimony altogether would impose significant burdens on epistemic agents to rely only on themselves to gather information to form beliefs. Compliance with this principle of prohibition would also have the upshot that epistemic agents would miss out on being able to form beliefs about things to which they do not have direct epistemic access. The sort of reasons that an epistemic agent might have to reject such a principle of prohibition would concern these and other costs of compliance.

Note that objections that one might raise to principles that govern how epistemic agents are to respond to testimony are not difficult to generate. What we’re interested in is whether or not certain of these objections are stronger than others. Staying abstract for the moment, suppose that I respond to a speaker’s testimony in some particular way, and that in so doing, I accord with a principle that would permit an epistemic agent, in my position, to respond to testimony as I did. And suppose that there are reasons of various kinds to reject a principle that would permit an epistemic agent, in my position, to respond in that way to testimony. We can compare those reasons with reasons to reject a principle that would prohibit an epistemic agent from responding to testimony of a speaker in that way. Now suppose that reasons to reject the principle of permission are stronger than whatever reasons there are to reject a principle of prohibition. On the view on offer, that fact would make my response to that speaker’s testimony something I could not justify to that speaker.

Responding to the testimony of speakers in ways that are unjustifiable to those speakers precludes the possibility of standing in a relationship of mutual recognition with those speakers. What does the relation of mutual recognition amount to in this context? It amounts to much the same relation between agents as the moral contractualist envisions—one whereby we acknowledge each other as both subjects to and the legislators of the terms that govern how we are to engage in testimonial exchanges with others. As epistemic agents who depend on each other to know things and to develop our perspective on the world, each of us has an interest in being able to negotiate those terms on which we engage in testimonial exchanges. Some ways of setting up those terms would disadvantage some epistemic agents more than others, and in ways that those disadvantaged epistemic agents would have good reason to reject. One such way would involve permitting hearers to accord speakers too little credibility on the basis of identity-prejudice against them—or so I want to suggest.

6. The Primary Harm of Testimonial Injustice as a Failure of Interpersonal Justifiability

Let’s now consider how we might apply Scanlon’s contractualist framework to analyzing the primary harm of testimonial injustice. To do so, I want to zero in on the case of Greenleaf and Sherwood.

Greenleaf, recall, prejudicially accords Sherwood insufficient credibility when she offers testimony about Dickie’s disappearance. In this case, Greenleaf’s response to Sherwood’s testimony accords with a principle that would permit a hearer to accord insufficient credibility to a speaker, owing to the operation of a negative gender-prejudicial stereotype that the hearer has against the speaker.¹⁷ (Call this a principle of permission.) What I want to suggest is that this principle of permission is something that Sherwood would have good reasons to reject. And these reasons outweigh the strongest reasons that Greenleaf might have on his own behalf to reject a principle that would prohibit a hearer from according insufficient credibility to a speaker, owing to the operation of a negative gender-prejudicial stereotype that the hearer has against the speaker. (Call this a principle of prohibition.)

What reasons are those? And what makes them stronger? Focusing on the case at hand, I want to offer a Fricker-friendly account of the most compelling reasons Sherwood would have to reject the principle of permission in this case.¹⁸ My discussion will draw on a key component of Fricker’s view that comes up relatively early on in Epistemic Injustice, but which recedes from view as Fricker’s focus shifts to developing an account of epistemic objectification.¹⁹

According to Fricker, being in a position to provide knowledge and information, or to otherwise engage in the give-and-take of reasons with other epistemic agents, and on those terms, makes one accountable to others in certain distinctive ways. One needs to make sure that the things one believes, and the reasons for which one believes them, can hold up to critical scrutiny and serve as the basis for the beliefs that others may form on the basis of one’s testimony. Accountability creates a pressure on one to check one’s beliefs, prompting one to “steady the mind” by settling one’s grounds for belief and working to distinguish truth from falsehood (Fricker 2007: 52). For Fricker, this process of steadying one’s mind is not just a bit of doxastic housekeeping, but rather, a process that is central to developing a stable sense of self. Settling my beliefs allows me to determine my overall outlook, which may include my broader values, my commitments, and my own place within the various communities in which I belong. On Fricker’s view, the process by which we “steady the mind” is also “the process by which we may become who we deeply, perhaps essentially, are” (2007: 53).

Drawing on Fricker’s view, we might further specify the sort of reason that Sherwood might have to reject a principle that would permit Greenleaf to fail to believe her testimony owing to gender prejudice. As an epistemic agent like the rest of us, Sherwood has an interest in being able to engage in the mutual give-and-take of reasons with others. A principle that would permit hearers like Greenleaf not to believe her, owing to gendered prejudice, would effectively preempt the possibility that Sherwood be able to participate in this mutual giveand-take of reasons, if that principle allows that hearers may rely on gendered forms of prejudice to determine what exactly the epistemic contributions may be from a speaker in Sherwood’s position.

Fricker says that it would be “melodramatic” to suggest that each individual instance in which a hearer does not believe a speaker, owing to prejudice, is one in which the speaker is somehow inhibited in the formation of her identity (2007: 54). Fricker’s way of handling this point is to suggest that even though many individual instances of epistemic exclusion do not, on their own, diminish one’s ability to develop one’s identity, persistent cases of such a kind would (2007: 54). But note that going this way makes it difficult to analyze the primary harm of testimonial injustice in terms of the negative effects of epistemic exclusion of this sort. After all, the primary harm is supposed to be, on Fricker’s view, something intrinsic to individual instances of testimonial injustice. The fact that a sufficient number of such cases would eventually inhibit a speaker’s ability to form her identity doesn’t explain what goes wrong in any particular case of testimonial injustice.

A contractualist account allows us to draw on the idea that the primary harm of testimonial injustice centrally involves epistemic exclusions that inhibit the formation of one’s identity, while also conceding the point that individual instances of epistemic exclusions may not, on their own, inhibit the formation of one’s identity. It is central to the contractualist account on offer that the relevant reasons are reasons to reject general principles about how we accord credibility to speakers, rather than reasons to reject individual responses to testimony. Sherwood has reason to reject a principle that would allow hearers in Greenleaf’s position to accord insufficient credibility to speakers in Sherwood’s position, owing to gender prejudice, because general compliance with this principle of permission among hearers in Greenleaf’s position would plausibly inhibit speakers in the formation of their identity.

But let’s not leave Greenleaf out. Whether or not some principle of permission can be reasonably rejected depends also on how others, including hearers like Greenleaf, stand to be affected by corresponding principles of prohibition. What might Greenleaf’s reasons be? One kind of reason to reject a principle that would prohibit him from being able to not believe a speaker, owing to gendered prejudice, might include the fact that trying to override a deeply ingrained prejudice is mentally taxing. It’s simply cognitively easier for us to operate with prejudicial stereotypes than to make efforts to neutralize their effect and to focus more attentively on an individual in our judgments of a speaker’s credibility. A principle of prohibition would effectively place certain burdens on Greenleaf to pay attention to things that it would be easier for him to ignore. But these reasons, I want to suggest, are simply not as strong as the reasons Sherwood has to be able to stand in a certain relation with epistemic agents that would allow her to participate in the mutual give-and-take of reasons. Though it may be cognitively easier for Greenleaf to rely on gendered prejudice to help him filter testimony, that fact is comparatively less important than the burdens that the principle of permission in this case stand to impose on Greenleaf.²⁰

To be clear, not all explanations of what secures the primary harm in a case of testimonial injustice will look the same. The question of whether a hearer’s assessment of a speaker’s credibility is something that plausibly harms or wrongs that speaker is to be settled first by articulating the general principles with which that credibility assessment accords, and then by considering the particular objections to those principles that those affected might have on their own behalf. And specific objections to principles may concern very different considerations, depending on what’s at stake. The issues relevant to assessing any particular case are going to be complex and inevitably context-dependent. But in all such cases, the suggestion is that these various objections, and their comparative weight, are to be evaluated relative to a shared commitment: that of being able to relate to other epistemic agents on terms of mutual recognition and respect. The sort of question guiding such assessments is whether an epistemic agent can legitimately demand, as a matter of respect for her as an epistemic agent, that the reasons she puts forward for rejecting a principle governing testimonial exchange be accorded sufficient weight.

I want to close this section by briefly sketching how this contractualist account of the primary harm of testimonial injustice might handle cases of prejudicial credibility excess.²¹ One might think that such an account would have particular trouble on this front. After all, who would object to a principle that allowed hearers to treat speakers like themselves as more credible than they deserve?

Recall the case we considered earlier from Davis (2016), in which a group of high-school students prejudicially accords an Asian-American classmate excess credibility when it comes to solving a math problem. In this case, I think the classmate would have good reason to reject a principle that would permit hearers to accord speakers excess credibility, owing to positive identity-prejudicial stereotypes that link Asian-Americans with superior mathematical skill. Indeed, a speaker’s reasons to reject principles permitting credibility excesses may be closely related to the sort of reasons we considered earlier to reject principles permitting credibility deficits. We each have an interest in being able to participate in the mutual give-and-take of reasons, an interest that is perhaps further grounded, as Fricker suggests, in our interest in being able to settle our beliefs and develop a stable sense of self. In order to develop that stable sense of self, we need not only to not have our testimony dismissed on the basis of our identities. We need, moreover, to not have our testimony uncritically accepted on the basis of our identities. I am not properly held accountable for what I believe when others accord me excess credibility on the basis of some aspect of my identity, rather than on the strength of reasons or evidence I might provide. Just as we might appeal to our interest in being able to settle our beliefs and develop a stable sense of self in order to identify reasons to reject principles permitting credibility deficits, so too might we appeal to those grounds to identify reasons to reject principles permitting credibility excess.

7. Conclusion

All of the accounts considered here of the primary harm of testimonial injustice presuppose a certain normative ideal of how epistemic agents ought to relate to one another. It is in virtue of that ideal that the ways in which we go about according varying degrees of credibility to speakers can fall more or less short. For example, on Fricker’s account, it is in virtue of the normative ideal of recognizing other epistemic agents as potential informants that our treatment of them as mere sources of information can harm or wrong them. Or, on some of the other views we’ve considered, it is in virtue of the normative ideal of recognizing each of our fellow epistemic agents on their own, without overly narrow preconditions or expectations, that merely conditional recognition of those agents can harm or wrong them.

The normative ideal to which my account centrally appeals—that of being able to justify our credibility assessments to speakers, and thereby standing in a relationship to them of mutual recognition—provides an alternative backdrop against which to understand the primary harm of testimonial injustice. What we plausibly owe to other epistemic agents, on my view, is to treat them in ways that they could not reasonably reject. In so doing, we make it possible to relate to them on terms of mutual recognition and respect. In many cases in which we accord a speaker insufficient credibility, owing to identity prejudice, we treat that speaker in ways we could not justify to that speaker, in the sense that that speaker would have sufficient reasons to reject the general principle with which our assessment of their credibility accords. It is in that sense in which our failure to accord a speaker sufficient credibility can plausibly be said to have wronged that speaker.